Philosophy of Technology and Engineering Sciences (Handbook of the Philosophy of Science)

Philosophy of Technology and Engineering Sciences

Handbook of the Philosophy of Science

General Editors

Dov Gabbay Paul Thagard John Woods

AMSTERDAM BOSTON HEIDELBERG LONDON NEW YORK OXFORD PARIS SAN DIEGO SAN FRANCISCO SINGAPORE SYDNEY TOKYO North Holland is an imprint of Elsevier

Philosophy of Technology and Engineering Sciences Volume 9

Edited by

Anthonie Meijers Eindhoven University of Technology, Eindhoven, The Netherlands

AMSTERDAM BOSTON HEIDELBERG LONDON NEW YORK OXFORD PARIS SAN DIEGO SAN FRANCISCO SINGAPORE SYDNEY TOKYO North Holland is an imprint of Elsevier

North Holland is an imprint of Elsevier 30 Corporate Drive, Suite 400, Burlington, MA 01803, USA Linacre House, Jordan Hill, Oxford OX2 8DP, UK Radarweg 29, PO Box 211, 1000 AE Amsterdam, The Netherlands First edition 2009 Copyright © 2009 Elsevier B.V. All rights reserved No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means electronic, mechanical, photocopying, recording or otherwise without the prior written permission of the publisher Permissions may be sought directly from Elsevier’s Science & Technology Rights Department in Oxford, UK: phone ( 44) (0) 1865 843830; fax ( 44) (0) 1865 853333; email: [email protected]. Alternatively you can submit your request online by visiting the Elsevier web site at http://elsevier.com/locate/permissions, and selecting Obtaining permission to use Elsevier material Notice No responsibility is assumed by the publisher for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions or ideas contained in the material herein. Because of rapid advances in the medical sciences, in particular, independent verification of diagnoses and drug dosages should be made British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Library of Congress Cataloging-in-Publication Data A catalog record for this book is available from the Library of Congress ISBN: 978-0-444-51667-1

For information on all North Holland publications visit our web site at elsevierdirect.com

Printed and bound in The Netherlands 09 10 11

11 10 9 8 7 6 5 4 3 2 1

GENERAL PREFACE Dov Gabbay, Paul Thagard, and John Woods Whenever science operates at the cutting edge of what is known, it invariably runs into philosophical issues about the nature of knowledge and reality. Scientific controversies raise such questions as the relation of theory and experiment, the nature of explanation, and the extent to which science can approximate to the truth. Within particular sciences, special concerns arise about what exists and how it can be known, for example in physics about the nature of space and time, and in psychology about the nature of consciousness. Hence the philosophy of science is an essential part of the scientific investigation of the world. In recent decades, philosophy of science has become an increasingly central part of philosophy in general. Although there are still philosophers who think that theories of knowledge and reality can be developed by pure reflection, much current philosophical work finds it necessary and valuable to take into account relevant scientific findings. For example, the philosophy of mind is now closely tied to empirical psychology, and political theory often intersects with economics. Thus philosophy of science provides a valuable bridge between philosophical and scientific inquiry. More and more, the philosophy of science concerns itself not just with general issues about the nature and validity of science, but especially with particular issues that arise in specific sciences. Accordingly, we have organized this Handbook into many volumes reflecting the full range of current research in the philosophy of science. We invited volume editors who are fully involved in the specific sciences, and are delighted that they have solicited contributions by scientifically-informed philosophers and (in a few cases) philosophically-informed scientists. The result is the most comprehensive review ever provided of the philosophy of science. Here are the volumes in the Handbook: Philosophy of Science: Focal Issues, edited by Theo Kuipers. Philosophy of Physics, edited by Jeremy Butterfield and John Earman. Philosophy of Biology, edited by Mohan Matthen and Christopher Stephens. Philosophy of Mathematics, edited by Andrew Irvine. Philosophy of Logic, edited by Dale Jacquette. Philosophy of Chemistry and Pharmacology, edited by Andrea Woody, Robin Hendry and Paul Needham.

vi

Dov Gabbay, Paul Thagard, and John Woods

Philosophy of Statistics, edited by Prasanta S. Bandyopadhyay and Malcolm Forster. Philosophy of Information, edited by Pieter Adriaans and Johan van Benthem. Philosophy of Technology and Engineering Sciences, edited by Anthonie Meijers. Philosophy of Complex Systems, edited by Cliff Hooker. Philosophy of Ecology, edited by Bryson Brown, Kent Peacock and Kevin de Laplante. Philosophy of Psychology and Cognitive Science, edited by Paul Thagard. Philosophy of Economics, edited by Uskali Mäki. Philosophy of Linguistics, edited by Ruth Kempson, Tim Fernando and Nicholas Asher. Philosophy of Anthropology and Sociology, edited by Stephen Turner and Mark Risjord. Philosophy of Medicine, edited by Fred Gifford. Details about the contents and publishing schedule of the volumes can be found at http://www.elsevier.com/wps/find/bookseriesdescription.cws home/BS HPHS/ description As general editors, we are extremely grateful to the volume editors for arranging such a distinguished array of contributors and for managing their contributions. Production of these volumes has been a huge enterprise, and our warmest thanks go to Jane Spurr and Carol Woods for putting them together. Thanks also to Andy Deelen and Arjen Sevenster at Elsevier for their support and direction.

CONTRIBUTORS

Jennifer K. Alexander University of Minnesota, USA. [email protected] Gerhard Banse Institut f¨ ur Technikfolgenabsch¨ atzung und Systemanalyse, Germany. [email protected] Johannes M. Bauer Michigan State University, USA. [email protected] Henk van den Belt Wageningen University, The Netherlands [email protected] Mieke Boon University of Twente, The Netherlands. [email protected] Stefano Borgo University of Trento, Italy. [email protected] Philip Brey University of Twente, The Netherlands. [email protected] Adam Briggle University of Twente, The Netherlands. [email protected] Louis Bucciarelli Massachussets Institute of Technology, USA. [email protected] Richard Buchanan Case Western Reserve University, USA. [email protected] David F. Channell University of Texas at Dallas, USA. [email protected]

viii

Contributors

Kees Dorst Eindhoven University of Technology, The Netherlands, and University of Technology, Sydney, Australia. [email protected];[email protected] Maarten Franssen Delft University of Technology, The Netherlands. [email protected] Pawel Garbacz The John Paul II Catholic University of Lublin, Poland [email protected] Armin Grunwald Karlsruhe Institute of Technology, Germany. [email protected] Sven Ove Hansson Royal Institute of Technology, Sweden. [email protected] Paulien M. Herder Delft University of Technology, The Netherlands. [email protected] Wilfrid Hodges Queen Mary University of London, UK [email protected] Wybo Houkes Eindhoven University of Technology, The Netherlands. [email protected] Jesse Hughes Arlington, USA. [email protected] Christian Illies University of Bamberg, Germany. [email protected] Tarja Knuuttila University of Helsinki, Finland. [email protected] Peter Kroes Delft University of Technology, The Netherlands. [email protected] Bruce MacLennan University of Tennessee, USA. [email protected] Anthonie Meijers Eindhoven University of Technology, The Netherlands. [email protected]

Contributors

Carl Mitcham Colorado School of Mines, USA. [email protected] Roland M¨ uller Switzerland. [email protected] Nancy J. Nersessian Georgia Institute of Technology, USA. [email protected] Paul Nightingale SPRU, University of Sussex, UK. [email protected] Alfred Nordmann TU Darmstadt, Germany [email protected] Kees van Overveld Eindhoven University of Technology, The Netherlands. [email protected] Christopher Patton Georgia Institute of Technology, USA. [email protected] Joseph C. Pitt Virginia Tech, USA. [email protected] Beth Preston University of Georgia, USA. [email protected] Michael S. Pritchard Western Michigan University, USA. [email protected] Hans Radder VU University Amsterdam, The Netherlands. [email protected] Nicholas Ray Cambridge University, UK. [email protected] Eric Schatzberg University of Wisconsin, USA. [email protected] Marcel Scheele The Netherlands. [email protected]

ix

x

Contributors

Joachim Schummer University of Darmstadt, Germany. [email protected] Johnny Hartz Søraker University of Twente, The Netherlands. [email protected] Knut Holtan Sørenssen Norwegian University of Science and Technology, Norway. [email protected] Andreas Spahn Eindhoven University of Technology, The Netherlands. [email protected] Susan G. Sterrett Duke University, USA. [email protected] Patrick Suppes Stanford University, USA. [email protected] Nigel Taylor University of the West of England, Bristol, UK. [email protected] Amie L. Thomasson University of Miami, USA. [email protected] Paul Thompson Michigan State University, USA. [email protected] Ibo van de Poel Delft University of Technology, The Netherlands. [email protected] Pieter Vermaas Delft University of Technology, The Netherlands. [email protected] Laure Vieu IRIT-CNRS, France. [email protected] Marc J. de Vries Eindhoven University of Technology, The Netherlands. [email protected] William H. Wood United States Naval Academy, USA. [email protected]

Contributors

xi

Sjoerd D. Zwart Eindhoven University of Technology, and Delft University of Technology, The Netherlands. [email protected];[email protected]

This page intentionally left blank

CONTENTS

General Preface Dov Gabbay, Paul Thagard, and John Woods List of Contributors

General Introduction Anthonie Meijers, editor

v vii

1

Part I: Technology, Engineering and the Sciences Introduction to Part I Hans Radder, associate editor

23

Defining Technology and the Engineering Sciences Carl Mitcham and Eric Schatzberg

27

Science, Technology and the Science–Technology Relationship Hans Radder

65

The Role of Social Science in Engineering Knut H. Sørensen

93

The Emergence of the Engineering Sciences: An Historical Analysis David F. Channell

117

Coherence and Diversity in the Engineering Sciences Gerhard Banse and Armin Grunwald

155

xiv

Contents

Part II: Ontology and Epistemology of Artifacts Introduction to Part II Wybo Houkes, associate editor

187

Artefacts in Metaphysics Amie L. Thomasson

191

Philosophical Theories of Artifact Function Beth Preston

213

Functional Decomposition and Mereology in Engineering Pieter Vermaas and Pawel Garbacz

235

Artefacts in Formal Ontology Stefano Borgo and Laure Vieu

273

The Nature of Technological Knowledge Wybo Houkes

309

Tacit Knowledge and Engineering Design Paul Nightingale

351

Practical Reasoning and Engineering Jesse Hughes

375

Part III: Philosophy of Engineering Design Introduction to Part III Peter Kroes, associate editor

405

Thinking about Design: An Historical Perspective Richard Buchanan

409

Typologies of Design Practice Kees Dorst and Kees van Overveld

455

Translating Customer Requirements into Technical Specifications 489 Marc J. de Vries Foundational Issues of Engineering Design Peter Kroes

513

Contents

xv

Computational Representations of Function in Engineering Design William H. Wood

543

Rationality in Design Peter Kroes, Maarten Franssen and Louis Bucciarelli

565

Designing Socio-Technical Systems Johannes M. Bauer and Paulien M. Herder

601

Part IV: Modelling in Engineering Sciences Introduction to Part IV Sjoerd D. Zwart, associate editor

633

The Notion of a Model: A Historical Overview Roland M¨ uller

637

Functional Modelling and Mathematical Models: A Semantic Analysis Wilfrid Hodges

665

Models as Epistemic Tools in Engineering Sciences Mieke Boon and Tarja Knuuttila

693

Model-Based Reasoning in Interdisciplinary Engineering Nancy J. Nersessian and Christopher Patton

727

Scale Modelling in Engineering: Froude’s Case Sjoerd D. Zwart

759

Similarity and Dimensional Analysis Susan G. Sterrett

799

Measurement Theory and Engineering Patrick Suppes

825

Technological Explanation Joseph C. Pitt

861

xvi

Contents

Part V: Norms and Values in Technology and Engineering Introduction to Part V Ibo van de Poel, associate editor

883

Why Technologies Are Inherently Normative Hans Radder

887

Artefacts and Normativity Maarten Franssen

923

Professional Standards in Engineering Practice Michael S. Pritchard

953

Values in Engineering Design Ibo van de Poel

973

The Concept of Efficiency: An Historical Analysis Jennifer K. Alexander

1007

Aesthetic Values in Technology and Engineering Design 1031 Joachim Schummer, Bruce MacLennan, and Nigel Taylor Risk and Safety in Technology Sven Ove Hansson

1069

Technology Assessment: Concepts and Methods Armin Grunwald

1103

The Interaction of Ethics and Technology in Historical Perspective Carl Mitcham and Adam Briggle

1147

Part VI. Philosophical Issues in Engineering Disciplines Introduction to Part VI Sven Ove Hansson, associate editor

1195

Contents

xvii

Philosophy of Architecture Christian Illies and Nicholas Ray

1199

Philosophy of Agricultural Technology Paul Thompson

1257

Philosophy of Medical Technology Sven Ove Hansson

1275

Philosophy of Biotechnology Henk van den Belt

1301

Philosophy of Computing and Information Technology Philip Brey and Johnny Hartz Søraker

1341

Index Compiled by Marcel Scheele and Andreas Spahn

1409

This page intentionally left blank

GENERAL INTRODUCTION Anthonie Meijers, editor “Now, if there are no artifacts, then there are no philosophical problems about artifacts.” [Van Inwagen, 1990, p. 128] Not so very long ago most philosophers of science maintained that the subjectmatter of this volume was uninteresting and most ontologists claimed it was nonexistent. It was thought to be uninteresting because technology was taken to be an applied science in which the application itself presented no new philosophical challenges. It was believed to be non-existent, because technological artifacts and systems did not live up to the criteria for being part of the ultimate inventory of the world. A combination of these two views leads to the rather fatal conclusion that the philosophy of technology and engineering sciences is boring stuff about non-existing entities! This volume shows how completely wrong that conclusion is. The fact that most philosophers of science have not regarded technology or engineering as a subject worthy of serious study clearly emerges from various wellknown introductions, companions and anthologies. [Curd and Cover, 1998] and [Curd and Psillos, 2008], for example, do not have a single index entry for ‘artifact’, ‘design’, ‘engineering’ or ‘technology’ in 2000 pages of philosophy of science. There are some exceptions though, such as [Newton-Smith, 2000] which contains a small section on the philosophy of technology.1 In analytic ontology interest in technological artifacts has also been largely lacking.2 If such artifacts are discussed at all it is often in the context of arguments intended to show that they do not really exist. The roots of this attitude lie in the positivist rejection of metaphysics.3 What survived of metaphysics after positivism focused on the fundamental concepts of the natural sciences. Basic social sciences and humanities concepts were ignored, taken to refer to non-existing entities, or thought to be reducible to concepts in physics. Since technological artifacts are 1 More evidence for the lack of interest shown by philosophers of science can be obtained from the Philosopher’s Index (Philosopher’s Information Center 2008), database 1940–2008. A search for the keyword ‘science’ produces 46,250 entries, a search for ‘engineering’ only 450 entries, and a search for ‘technology’ 1250 entries. The keywords ‘artifact’ and ‘design’ generated 300 and 1200 entries respectively. Entries with the subject label ‘ethics’ were excluded, because the focus of the search was on the philosophy of science. 2 A combined search for ‘artifact’ and ‘ontology’ led to only 16 (!) entries in the Philosophers Index database 1940-2008. 3 See Thomasson’s chapter in this Volume, Part II.

Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

2

Anthonie Meijers, editor

human-dependent objects, they do not fit the physicalist mould and are therefore an easy target for eliminativists. There are many reasons why the above conclusion is wrong. The relation between science and technology is infinitely more complex than suggested by the simplistic idea that technology is just an applied science.4 One only has to look at the pervasive role of technology in modern science to see this. Furthermore, the fact that most ontological accounts of artifacts, or medium sized objects in general, are eliminativist can be taken as an indication that there are serious problems with key concepts in metaphysics, such as the concepts of co-location and existence. So instead of simply biting the bullet about the non-existence of artifacts the conclusion might be that we should rethink basic ontological concepts.5 Technology forms a very rich philosophical terrain and the Handbook can be read as a map of the many fascinating issues that can be found here. A number of them have been investigated in depth in the philosophical literature, such as the relation between science and technology,6 the theory of measurement,7 or the role of professional standards in engineering practice.8 Other issues have only been partially explored, such as the types of design problems that engineers solve;9 the epistemic role of models in engineering,10 or the notion of technological explanation as distinct from scientific explanation.11 Many issues, however, have not been addressed at all and that is why there still is a lot of pioneering work to be done in the philosophy of technology and engineering sciences. In what follows I will first define technology and the engineering sciences, which is the subject of this Handbook (Section 1). In Section 2, I will discuss various ways of studying the subject. This will include the approaches taken by historians of technology, by researchers working in the field of Science and Technology Studies (STS), and by philosophers of technology. I will then briefly review highlights in the history of the philosophy of technology and engineering sciences (Section 3). In Section 4, I discuss the architecture of the Handbook which consists of six parts, each covering major aspects of the field. Section 5, the final section, reflects on the nature of the essays in this volume. The philosophy of technology and engineering sciences is a relatively young discipline. In addition to well-established accounts there are explorative essays on a number of areas so far more or less uncharted by philosophers. The Handbook thus also aims to set a research agenda.

4 See Radder’s chapter Science, technology and the science-technology relationship in this Volume, Part I; and Houkes’ chapter, The nature of technological knowledge, in this Volume Part II. 5 See Thomasson’s chapter Artifacts in metaphysics in this Volume Part II. 6 See Radder’s chapter, Science, technology and the science-technology relationship in this Volume Part I. 7 See Suppes’ chapter, Measurement theory and engineering in this Volume Part IV. 8 See Pritchard’s chapter, Professional standards in engineering practice, this Volume Part V. 9 See Dorst and Van Overveld, Typologies of design practices, this volume Part III. 10 See Part IV, this Volume. 11 See Pitt’s chapter, Technological explanation in this Volume Part IV.

General Introduction

3

1 TECHNOLOGY AND THE ENGINEERING SCIENCES The difficulty of delimiting the subject of this volume does not arise from the lack of definitions of technology or engineering as there are dozens of such definitions.12 The problem is rather how to come up with a sensible definition given this multitude. The aim of providing such a definition here would be to mark out the kinds of phenomena this Handbook covers. The purpose would not be to specify the essence of the subject (if any), to describe the linguistic origin of these words or to prescribe what the terms should mean. The English word ‘technology’ comes from the Greek τέχνη, which is usually translated as art, craft or skill.13 For modern language users this needs further clarification, because the Greek notion of τέχνη was intimately connected to the notion of knowledge.14 For the Greeks there was therefore no need to combine the word τέχνη with the word logos (as in technology), because τέχνη already involved logos. In Plato’s early writings there are two types of τέχνη: one requiring a lot of physical work (resulting in paintings or sculptures) and one requiring only minimal physical work (arithmetic, logic, astronomy). In later works the notion of τέχνη became associated with the knowledge and activities aimed at making or producing. The English word ‘engineering’ originates from the Latin ingenera, meaning to implant, generate or produce.15 In the late Middle Ages it was linked to the making and operating of military hardware. The term ‘civil engineering’ was introduced in the 17th century to distinguish non-military applications, such as roads and bridges. Engineering was defined at the beginning of the 19th century as ‘the art of directing the great sources of power in nature for the use and convenience of man’.16 In later definitions ‘art’ was substituted by ‘science and mathematics’: engineering is “the application of science and mathematics by which properties of matter and the sources of energy are made useful to people”.17 These definitions show that technology and engineering cannot be identified exclusively in terms of a body of systematic knowledge. After all they do not aim at knowledge for its own sake, but rather at the development and use of knowledge for practical purposes. Technology or engineering is primarily a practice which is knowledge-based. In this practice scientific knowledge, but also experience-based know-how, codes and standards, customer requirements, organizational, legal and economic constraints, physical circumstances, scarcity of resources, uncertainty and ignorance play an important role. The title of the Handbook seeks to empha12 See Mitcham and Schatzberg’s chapter, Defining technology and engineering science in this Volume Part I. 13 See for an extended discussion [Mitcham, 1994, 114–134]. 14 This excluded those skills that the Greeks took to be solely based on experience, such as cooking or swimming. 15 Ibid., 144–149. 16 This is the classic definition of engineering as a civilian enterprise formulated by Thomas Tredgold for the Royal Charter of the British Institution of Civil Engineers (1828). See also Mitcham and Schatzberg’s chapter in this Volume Part I. 17 Webster’s Third New International Directory (2002).

4

Anthonie Meijers, editor

size both aspects. It refers to the practice of technology and engineering, but also to the engineering sciences as a body of systematic knowledge. Thus defined the philosophy of technology and engineering sciences has a broader scope than most philosophies of the so-called special sciences. It is therefore better to see it as part of the philosophy of technology than as part of the philosophy of science, though these are partly overlapping domains. Carl Mitcham made a useful distinction between four modes of technology:18 • technology as a set of artifacts or systems of artifacts; • technology as a form of knowledge (for the design, production, maintenance and use of technological artifacts and systems); • technology as a range of activities (designing, producing, maintaining and using artifacts); and • technology as an expression of the will of its makers, designers and producers (volition). This distinction shows in another way that the cognitive dimension of technology is important, but does not suffice to define technology.19 It is on the basis of Mitcham’s distinction that the subject-matter of this Handbook can be delimited. It first of all deals with technological artifacts and systems, the objects that technology and the engineering sciences produce. In the second place it covers technology as a body of systematic knowledge. This includes the methodology and epistemology of the engineering sciences as well as the relationship of technology to the natural and social sciences. The Handbook finally addresses technology as a range of activities. The main focus is on the activity of design but the Handbook also looks at other key engineering activities. An important qualification needs to be made at this point. Though Mitcham’s first three modes of technology clearly fall within the scope of the Handbook, the focus is on science-based engineering. The authors of this Volume are mainly interested in the knowledge and activities of modern engineers and in the objects they produce. Users of technological artifacts are only considered insofar as they are relevant to science-based engineering (for example, artifacts are usually designed by engineers with users in mind and they come with a manual). The Handbook only marginally touches on the roles of craftsmen, managers and other professionals involved in the technological domain. This reflects an important decision in the design of the Handbook. The rationale behind this decision is twofold. Firstly, the editors wanted to focus on those aspects that are currently underexposed and ill-understood within the realm of the philosophy of technology. The Handbook thus clearly fills a gap in the field. Secondly, since this is a Handbook in a series 18 See

[Mitcham, 1994]. argues in detail in Part II of this Volume that it is very difficult to distinguish between science and technology solely in terms of this cognitive dimension. 19 Houkes

General Introduction

5

on the philosophy of science, it also seemed appropriate to focus on science-based engineering. Several definitions of technology and engineering given in the Handbook refer to one of these three modes or to a combination of modes. For example, Hans Radder describes technology as “a (type of) artifactual, functional system with a certain degree of stability and reproducibility” (this Volume Part V). Paul Nightingale, on the other hand, defines engineering as “the art of organizing and negotiating the design, production, operation and decommissioning of artifacts, devices, systems and processes that fulfil useful functions by transforming the world to solve recognized problems” (this Volume Part II). The first definition primarily perceives technology as a system of artifacts whilst the second sees technology as a range of activities. Mitcham’s fourth mode of technology, technology as volition, largely extends beyond the scope of this Handbook. It concerns the social, cultural, political and anthropological aspects of technology. The philosophy of technology has a rich tradition of analysing these aspects as testified by authors such as Mumford, Ortega Y Gasset, Heidegger and Ellul. In addition, there has always been a strong emphasis on the ethics of technology, both from the point of view of the user and the professional engineer That the subject-matter of the Handbook is limited to the first three modes of technology reflects once again the desire of the editors to concentrate on those aspects that are currently underexposed. The four modes of technology, however, should not be taken as independent of each other. That is why there is also some discussion of the ethical, social and anthropological aspects of technology in Part V.20

2

VARIOUS APPROACHES

The subject-matter of the Handbook can be studied in many ways. Historians, STS researchers, engineers themselves and philosophers of technology have all contributed to a better understanding of the theory and practice of engineering. They do this from different theoretical and methodological perspectives. Some studies are of an empirical and descriptive nature, others are conceptual and/or normative; yet other studies seek to explain while others aim to evaluate; some studies focus on specific theories and methods of engineering while yet others concentrate on the social and economic forces interacting with technology and the engineering sciences. Obviously, one need not be committed to just one of these approaches. Historians have long been interested in technology as an object of empirical study. Apart from comprehensive overviews of the history of technology [Singer et al., 1954; McNeil, 1996], there are numerous historical case studies of engineers and engineering. For example, there are the biographies of individual engineers, such as Isambard Brunel [Rolt, 1959; Buchanan, 2002], Thomas Edison [Israel, 20 These

aspects of technology are prominent in, for example, [Scharff and Dusek, 2003].

6

Anthonie Meijers, editor

2000] or Vannevar Bush [Zachary, 1997]. Likewise there are the studies of the development of certain artifacts, such as the steam engine [Hills, 1993], the airplane [Constant, 1980; Abzug and Larrabee, 2005] or the atomic bomb [Rhodes, 1995]. There are also inquiries into the nature of technological knowledge that are based on historical cases [Vincenti, 1990]. Increasingly, however, the focus of historical studies has shifted from technology as a subject in its own right to the role of technology in the development of modern societies. Examples are the role of steel in the making of modern America [Misa, 1998], or the role of computers in the transition to an information society [Friedman, 2005]). Landmark studies in this respect are two book series on the role of technology in the formation of Dutch society in the 19th and 20th centuries, edited by Lintsen and Schot [Lintsen, 1995; Schot, 1998]. Researchers engaged in the field of Science and Technology Studies (STS) have always been averse to traditional disciplinary boundaries. They are interested in using social science methods (for example ethnographical methods) to study science and technology. They try to explain their object of study primarily in terms of social action. Science and technology are seen as historically situated social practices that produce knowledge, meaning and impact. Instead of looking at the relation between a theory and the available empirical evidence, STS researchers focus rather on the negotiation processes between actors in the scientific field when explaining the acceptance of a given theory. The primary explanatory objective of STS is to produce “a precise, empirical, multilevel account of the production [of knowledge], influence, and change”.21 One example is the study by Geels en Schot of the various ways in which sociotechnical regimes change.22 The concept of a sociotechnical regime includes here not only the shared cognitive routines in an engineering community but also the social context of policy makers, users and special interest groups. There are two main theoretical positions in STS: the social construction of technology23 and actor-network theory.24 They share a strong empirical orientation in their study of science and technology, whereas their differences concern, among other things, the question of whether all scientific phenomena can ultimately be explained in terms of social action.25 It is fair to say that STS has mainly been a explanatory enterprise. Though the editors of the recent Handbook of Science and Technology Studies believe that the explanatory goal of STS must be “wedded to an agenda of social change, grounded in the bedrock of ethical principles and explicit values (equality, democracy, equity,

21 See

[Hackett et al., 2008, Introduction, p. 4]. Given the large numbers of STS researchers it is probably a simplification to subsume all of them under one explanatory goal. 22 See [Geels and Schot, 2007]. 23 See for example [Bijker, T. P. Hughes and Pinch, 1987] and [MacKenzie, 1993]. 24 See for example [Latour, 1987; 2005]. 25 In his later work Latour opposed Bloor’s so-called ‘strong programme’ in the sociology of scientific knowledge, according to which success and failure in science should be examined and explained symmetrically. Latour saw this as a form of sociological reductionism. See [Latour, 1992].

General Introduction

7

freedom, and others)”, this is largely taken to be an emerging challenge rather than a reflection of current practice.26 Numerous engineers have also contributed to a better understanding of the knowledge, activities and objects of the engineering sciences and technological practices. Examples include David Billington’s work on the role of aesthetic values in structural engineering,27 Larry Bucciarelli’s, Clive Dym’s, John Gero’s and Henry Petroski’s respective work on engineering design,28 Billy Koen’s work on heuristics and the engineering method,29 and Andries van Renssen’s work on an applied ontology for the process industry.30 Engineers-historians have also carried out important studies. Walter Vincenti’s and Edwin Layton’s work on the nature and taxonomy of technological knowledge are exemplary.31 These studies are primarily descriptive and aim at clarifying and giving a systematic account of the practice and science of engineering. The distinctive character of the approach taken in this Handbook cannot be defined in terms of a unique method. There is no such method and in this respect the philosophy of technology and the engineering sciences will always be eclectic. Descriptive studies, historical and social explanations, conceptual analyses and normative evaluations can all be found in this Handbook. There are, for example, historical chapters on the emergence of the engineering sciences (Part I), on the way that conceptions of design have evolved over the course of time (Part III), on the notion of a model (Part IV) and on the concept of efficiency (Part V). What sets the Handbook apart from historical, STS, and engineering approaches, though, is its strong emphasis on conceptual, methodological and normative issues (or combinations of them). For example, in Part I Mitcham and Schatzberg reflect on the very idea of defining technology and the engineering sciences, and on the types of definitions that can be given in relation to explanatory purposes and contexts. Houkes critically examines, in Part II, the epistemological claim put forward by Layton, Staudenmaier, Vincenti and others that technological knowledge forms a category of its own. He concludes that such a strong claim cannot be upheld on the basis of the arguments given but that there is still room for a weaker form of emancipation from scientific knowledge. In Part III Kroes, Franssen and Bucciarelli evaluate to what extent engineering design, which is a creative and social process of decision making, can be called a rational process. In addressing this issue they distinguish between various notions of rationality, such as means-ends rationality, procedural rationality and substantive rationality. Several chapters in Part IV explore the notion of a model, the varieties of models, and the methodological and epistemic roles of models in the engineering sciences. Radder investigates, 26 [Hackett et al., 2008, Introduction, p. 5]. Philosophers of technology inspired by Latour’s actor-network theory, such as Achterhuis and Verbeek, have focused on these moral aspects from the very beginning. See, for example, [Achterhuis, 1995] and [Verbeek, 2000/2005]. 27 See [Billington, 1985]. 28 See [Bucciarelli, 1996; Gero, Tham and Lee, 1992] and [Dym, 1994]. 29 See [Koen, 2003]. 30 See [van Renssen, 2005]. 31 See [Vincenti, 1990] and [Layton, 1974].

8

Anthonie Meijers, editor

in Part V, the normativity of technology and argues that it is not only normative in a contingent sense but also inherently. He thus makes a conceptual claim. Though the emphasis is on conceptual, methodological and normative issues, the editors of the Handbook believe that the philosophy of technology and engineering sciences should be empirically informed. Many chapters therefore refer to specific technologies, engineering theories and engineering practices as cases from which something can be learned. For example, Nersessian and Patton partly base their account of model-based reasoning (in Part IV) on cases drawn from biomedical engineering. Borgo and Vieu, when developing an applied ontology for artifacts in Part II, analyse ways in which these artifacts are represented in information systems. Boon and Knuutilla and Zwart discuss in Part IV the work of Carnot and Froude in order to gain a better grasp of the epistemological roles of models in engineering. The final part of the Handbook, which is devoted to philosophical issues in specific technologies, reflects the empirically informed approach taken here as a whole. 3 A BRIEF HISTORY OF THE FIELD The history of the philosophy of technology and engineering sciences as defined above is not very long. One could alternatively say that it does not yet have a history, only a prehistory. During the last century individual researchers worked on topics such as the nature of technological knowledge, the analysis of design problems, the difference between natural and artificial objects, and the difference between science and technology. Their number has been very small, especially when compared to the number of researchers working on science and technology studies, on the social and ethical problems of technology or on the history of technology. Only after the turn of the century did a community of researchers of a certain size emerge which had a joint interest in the philosophy of technology and engineering sciences.32 From the very beginning to the second half of the 20th century the philosophy of technology (in a broad sense) paid little attention to the topics of this Handbook. Philosophers such as Ernst Kapp, Lewis Mumford, Jos´e Ortega Y Gasset, Martin Heidegger, Jacques Ellul, and Hans Jonas were primarily interested in anthropological, ethical and metaphysical studies of technology. There were exceptions though. The work of Jacques Lafitte, Gilbert Simondon, Tadeusz Kotarbinsky, Alard DuBois-Reymond, and Hendrik van Riessen, to name but a few, contained analyses of the concepts of machine and system, taxonomies of machines and their parts and discussions of the process of invention and technological evolution.33 The eighties witnessed a small wave of publications in the field, largely based on studies carried out in the previous decade. Books were published by Rogers on 32 The American-European Society for Philosophy and Technology (SPT) has a much broader orientation than the philosophy of technology and engineering sciences and has historically been dominated by social and ethical questions concerning technology. 33 See [Mitcham, 1994].

General Introduction

9

the nature of engineering [Rogers, 1983], by Laudan on the nature of technological knowledge [Laudan, 1984], by Bunge on the philosophy of science and technology [Bunge, 1985] and by Staudenmaier on key issues concerning our understanding of technology [Staudenmaier, 1985].34 One landmark study was Walter Vincenti’s book What Engineers Know and How They Know It: Analytical Studies from Aeronautical History [Vincenti, 1990]. The productivity of the nineties did not match that of the eighties though there were a number of articles on the subject in the journal Techn´e and in the book series Research in Philosophy and Technology), edited by Carl Mitcham. Mitcham’s own book Thinking Through Technology; The Path Between Engineering and Philosophy, which was published in 1994, gave a predominantly historical overview of the philosophy of technology. The book was (and still is) influential because it contains an analytic framework for studying technology philosophically (see Section 1). It also made an impassioned appeal to philosophers to engage more intensively in the dialogue with engineers and to take technology much more seriously.35 In so doing, Mitcham paved the way for more research on the subject of this Handbook. On a whole, however, the nineties can be characterized as a period of stagnation. No major studies were published and no major initiatives were taken. The subject was also virtually absent in mainstream philosophy journals. Gradually things started to change. At the turn of the century a lot of new activities were launched in the Netherlands. The research programme The Dual Nature of Technical Artifacts in Delft turned out to be a kernel for much more research in the philosophy of technology and engineering sciences, not only in Delft but also elsewhere. The Dual Nature programme focused on the general concept of a technical artifact, as an entity that can be described in functional-intentional and in physical terms. The results of the programme were published in a special issue of a mainstream philosophy of science journal: Studies in History and Philosophy of Science [Kroes and Meijers, 2006], which was a remarkable deviation from standard publication practices in both the philosophy of technology and the philosophy of science. In parallel developments several philosophies of specific technologies emerged. The electronic journal Hyle started publishing articles on the philosophy of chemistry from 1995 onwards, including articles on chemical technology.36 The philosophy of information technology became a major topic of specialized research in its own right,37 while the philosophy of biotechnology gained prominence.38 Other important developments included the emergence of a philosophy of scientific

34 In his book, Staudenmaier analyzed 25 years of publications in the historical journal Technology and Culture. 35 See [Mitcham, 1994, p. 268]. 36 See http://www.hyle.org. 37 See, for example, [Floridi, 2003] and Brey and Søraker’s chapter in this Volume Part VI. A separate volume in the Handbook series Philosophy of Science is also devoted to the philosophy of information. 38 See Van de Belt’s chapter in this Volume, Part VI.

10

Anthonie Meijers, editor

instrumentation,39 and of a philosophy of risk. The philosophy group at KTH Stockholm played a leading role in the development of the latter.40 In the last decade the community of researchers working on the philosophy of technology and engineering sciences has also become better organized. They have created and organized a portfolio of substantive research programmes, publication projects, regular conferences and workshops. Professional societies such as the SPT showed increased interest by arranging sessions at its biennial conferences on issues such as engineering design, nanotechnology and artifact ontology. Furthermore, the Society’s journal Techn´e started publishing more and more articles on the subject. There were new initiatives as well. Specialized workshops on the philosophy of engineering were organized in Delft (2007) and in London (2008). Finally, in another initiative, the Division of Logic, Methodology and Philosophy of Science of the International Union of History and Philosophy of Science has decided to give the philosophy of technology and engineering sciences much more prominence in its future activities. Despite all these positive developments there is also good reason to be modest about what has been accomplished. In his book The Nature of Engineering: A Philosophy of Technology (1983) Gordon Rogers tried to give a systematic account of the notion of technological explanation, as distinct from scientific explanation and historical explanation. He distinguished between two types of technological explanation. First-order explanations are teleological in nature and are intended to answer questions of the type ‘What is this flywheel for?’ (To reduce the fluctuations in torque which arise from the intermittent nature of the processes in a reciprocating engine), or ‘Why is the spark initiated before the end of the compression stroke?’ (To compensate for the delay in the ignition process).41 More mature technological explanations are of a causal nature and try to answer such questions as ‘Why did this bridge collapse?’, or ‘What causes the ignition delay in an engine?’, or ‘How can one account for the heat transfer in fluid metals in a fast breeder reactor?’. According to Rogers these technological explanations are causal explanations but they differ from each other in that they occupy a different place in a spectrum of causal explanation ranging from scientific explanations to historical explanations.42 Obviously, much more can and should be said about this. Technological explanation is an important issue in the philosophy of engineering. However, it is fair to say that little if any work has been done on the subject in the last twenty-five years.43 Notwithstanding this sobering fact, the Handbook Philosophy of Technology and Engineering Sciences clearly marks a milestone in the history of the field. It brings together for the first time more than fifty scholars who have written extensively 39 See

[Radder, 2003] and [Baird, 2004]. http://www.infra.kth.se/phil/riskpage/index/htm and [Hansson, 2003]. The philosophy of risk focusses on the epistemological as well as the ethical aspects of risks. 41 [Rogers, 1983, p. 42]. 42 Ibid., p. 43. 43 An exception is Jeroen de Ridder’s PhD thesis on the design and explanation of artifacts [de Ridder, 2007] and Joseph Pitt’s chapter “Technical Explanation”, this Volume Part IV. 40 See

General Introduction

11

on such diverse topics as the function theories of artifacts, means-end reasoning, the role of scale models in engineering, notions of computation and efficiency, and philosophical theories of architecture. 4 ARCHITECTURE OF THE HANDBOOK The Handbook consists of six parts, each of which covers a cluster of related issues. These parts correspond to major aspects of the philosophy of technology and engineering sciences. Together they do not exhaust the field though. There is very little on production and maintenance, for example, while the existing parts could easily be expanded (see also the next section). Part I of the Handbook focuses on the demarcation of the object of study: technology and the engineering sciences. Various types of definitions of technology and engineering are discussed as are the aims that these definitions serve. The relationships between technology and the natural sciences and between technology and the social sciences are subsequently analyzed, to situate technology in the disciplinary landscape. Finally, the historical emergence of the engineering sciences together with their diversity and coherence are examined. These issues are central to the field. Part I also contributes to the other parts of the Handbook because it defines their object of analysis. Part II addresses the ontology and epistemology of technological artifacts. It discusses these artifacts in the context of analytic metaphysics and applied ontology. The latter is crucial to the representation of artifacts in information systems. There is also an in-depth discussion of a key property of artifacts: their function. Existing theories of function are evaluated in terms of their ability to account for the functions of technological artifacts. In addition, functional part-whole relationships and their use in engineering are explored. Part II furthermore analyses the notion of technological knowledge. This is a wide concept consisting of many elements. Taxonomies of these elements are presented and the claim that technological knowledge is different from scientific knowledge is carefully evaluated. In addition the role of tacit knowledge in engineering design is discussed. Finally, there is an analysis of means-end reasoning which is central to technological rationality. Part III focuses on a defining activity of engineering: the design of technological artifacts and systems. It includes a historical account of design concepts, a typology of design practices and a discussion of how customer requirements are translated into technical specifications. An analysis is made of the design process in terms of the so-called function-structure relation. A design starts with the specification of an artifact’s desired function. In the process of designing this is transformed into a description of the artifact’s structural properties and a manual for its use. Another topic that is studied is the computational representation of functions in engineering design. Increasingly, engineering design is a computersupported activity and the ability to represent functions in information systems is crucial then. There is also an in-depth analysis of the rationality of design. Design

12

Anthonie Meijers, editor

includes rational problem solving activities but also social and creative processes. Finally, the complexities of the design of socio-technical systems or mixed systems of artifacts and humans, are discussed. Part IV is about methodological issues. Since models are so central to engineering, for example computer models or scale models, most of this part is devoted to the analysis of models and modelling. Firstly, there is an extensive historical account of the notion of a model. This is followed by a semantic analysis of functional modelling and mathematical models. Several case-studies are presented to show how in engineering models are actually used as epistemic and methodological tools. Case studies also form the basis for an account of model-based reasoning. Since scales and dimensions play an important role in modelling, there is also an in-depth discussion of dimensional analysis and measurement theory. Finally, there is an analysis of the notion of explanation, traditionally a key concept in methodology, but in engineering its meaning and use is distinct from that in the natural sciences. Part V investigates the norms and values that are at work in engineering. As discussed in Section 1, technology and engineering aim at the design and production of technological artifacts and systems that are useful or valuable to human beings. They have inherent normativity. This normativity and the normative statements that can be made about artifacts are analysed in great detail in this part. There are also accounts of the various ways in which non-epistemic values and norms play a role in engineering design and engineering practice. In particular the role of aesthetic values is discussed in relation to the design of such diverse artifacts as urban areas, software, and molecules. In addition to this the values of efficiency and safety in engineering are analyzed. Finally, the central concepts and methods of technology assessment are discussed as is the interaction between technology and ethics. Part VI is of a different nature in that it takes specific engineering disciplines as its object of analysis. It thus gives a different cross-section of the landscape of the philosophy of technology and engineering sciences. Some of the analyses presented are strictly specific to certain disciplines, such as the notion of a gene in biotechnology or the notion of computation in information technology. Other analyses are more general. The part also contains discussion about how developments in one discipline can influence developments in another. For example, developments in medical technology have influenced the notion of disease in medicine. The focus is both on more classical engineering disciplines such as architecture, agricultural technology and medical technology, and on more recent disciplines such as biotechnology and information technology. 5 THE HANDBOOK AS A RESEARCH AGENDA The overview of the various parts of the Handbook clearly shows that it does not cover the field of philosophy of technology and engineering sciences in full. As already mentioned, the Handbook can also be read as a research agenda. This

General Introduction

13

will be discussed below using a distinction made at the beginning of the chapter between topics that have been studied in depth, topics that have been partially explored and topics that have not been addressed at all. Topics that fall into the first category are the following: the definition of technology and engineering, the relation between science and technology, the history of design, the translation of customer requirements into technical specifications, the notion of a model, measurement theory, the normative status of artifacts, professional standards in engineering, notions of risk and safety, technology assessment, the philosophy of biotechnology and the philosophy of information technology. There are many more topics that fit into the second category in the Handbook. I have already mentioned the ontology of artifacts which is a topic in need of further elaboration. Similarly, the theories of function in the philosophy of biology appear to be inadequate when it comes to accounting for technological artifacts, while the theories of function that have been specifically developed for these artifacts are still in their early stages.44 Technological knowledge is also an underdeveloped topic. The notion is intrinsically linked to human goals and actions. Many of its elements require further systematic philosophical analysis. For example, the role of practical usefulness (rather than truth-likeness) when validating theories and models in the engineering sciences, or the role of technological rules in engineering practices.45 Other topics that fall into the second category and are included in the Handbook are these: the role of social science in engineering, functional decomposition and mereology in engineering, typologies of design practices, the design of sociotechnical systems, the epistemic roles of models in engineering sciences, similarity and dimensional analysis, technological explanation, the concept of efficiency, the philosophy of architecture and the philosophy of medical technology. A number of topics that belong to the first two categories are unfortunately lacking in the Handbook or only briefly addressed. They are: • the role of technology in experimental sciences (Part I);46 • the dual nature of technical artifacts as functional-intentional and as physical objects;47 the distinction between technical artifacts and natural objects; a structural mereology of artifacts as opposed to a functional mereology; knowledge management in large engineering projects and organizations (Part II); • the evaluation of design methodologies;48 optimization methods in engineering design (Part III); • the relevance of systems theory to engineering; the engineer’s toolbox (finite elements methods, unified modelling language, simulation techniques); 44 See

Preston’s chapter, Philosophical theories of artifact function in this Volume Part II. Houkes’ chapter, The nature of technological knowledge in this Volume Part II. 46 See [Radder, 2003]. 47 See [Kroes and Meijers, 2006]. 48 See journals such as Design Studies and Research in Design. 45 See

14

Anthonie Meijers, editor

the principles of performance measurement; error theory; the foundations of control theory; the role of heuristics and rules of thumb; the role of technical codes and standards; the visual aspects of engineering; the role of idealization in technology; an account of how computers have changed engineering methodologically (Part IV); • the role of aesthetic values in the design of medium-sized objects (Part V). • the philosophy of chemical engineering (Part VI).49 The number of topics falling into the third category is also substantial. They have been on the editors’ lists of issues to explore for a long time. Occasionally we found researchers who were prepared to do serious work in those areas so that their finding could be turned into chapters in the Handbook. But the remaining list of items is still very long. It includes: • an analysis of basic artifact types (Part II); • an account of the trade-offs in design (Part III); • an analysis of the notion of technological rule, as opposed to the notion of scientific law; an investigation into the role and justification of engineering theories; an account of the operational principles of artifacts (Part IV); • epistemic norms in engineering (Part V); • the foundations of nanotechnology, and of classical engineering disciplines such as mechanical engineering and structural mechanics (Part VI). Military technology is also a terra incognita in terms of philosophical analyses, but it cannot be strictly called a discipline. It is rather a collection of technologies used in a certain field of application. The role of military technology is of fundamental importance to the understanding of the development of other technological disciplines. In addition to these topics there is also an entire aspect not yet covered by the Handbook. That has to do with all the issues related to the production, operation and maintenance of technological artifacts and systems as a part of science-based engineering.50 Scientific theories of production and organization such as Taylor’s, theories of multi-agent systems, the role of ISO standards, decision and planning theories and the social context of engineering are among the topics still to be explored in this domain. Thus defined, the research agenda requires substantial effort on the part of philosophers. It calls for a widening of the community of researchers involved. There are indications that this may occur in the near future. Increasingly, philosophers of technology are publishing in mainstream philosophical journals, thus 49 See

the journal Hyle. of these topics are briefly mentioned by Sørensen in Part I and by Radder in Part V.

50 Some

General Introduction

15

reaching a larger audience when discussing topics studied in this Handbook.51 At the same time, philosophers specialized in action theory, ontology, the philosophy of mind and the philosophy of science are turning increasingly to the philosophical problems of artifacts. This has recently resulted in books such as Baker’s The Metaphysics of Everyday Life (2007), Thomasson’s Ordinary Objects (2007) and Margolis and Laurence’s Creations of the Mind: Theories of Artifacts and their Representation (2007).52 To conclude, the Handbook gives the reader an overview of the current state of affairs in the philosophy of technology and the engineering sciences. This field can best be characterized as a field in transition. There are very interesting developments going on and many new topics are being explored. Since this situation will probably continue for some time to come, given the extensive research agenda sketched above, the editors hope that the Handbook will also be made available online in the not too distant future. Ideally it should become a living document that can be improved and extended whenever new or better studies become available. It should give philosophers and engineers easy access to the best and most up-to-date knowledge on the subject. Viewed from this angle the Philosophy of Technology and Engineering Sciences Handbook is merely a step, albeit a step in the right direction.53 6 THE COMPILATION OF THE HANDBOOK AND ACKNOWLEDGEMENTS “Thalassa! Thalassa!” (The sea! The sea!) Xenophon, Anabasis. This Handbook has been a very ambitious project, both in its intellectual and its organizational scope. From the start it was clear that in addition to having chapters on well-researched topics the Handbook would contain explorative chapters on new aspects of the field. The aim was not only to survey the philosophy of technology and engineering sciences in its present state but also to contribute substantially to its development. On the basis of extensive literature searches the first Handbook outlines were produced in the spring of 2004. More than 10 areas were defined, each containing between 5 to 10 topics that were considered to be important. In total 65 topics were chosen that could possibly be turned into chapters. A number of these topics were suggested by the philosophers who later became the associate editors of the 51 See, for instance: [Lelas, 1993; Houkes and Vermaas, 2004; Boon, 2006; Hansson, 2007; Kroes and Meijers, 2006; Zwart and Franssen, 2007; Hughes, Kroes and Zwart, 2007; Vaesen and van Amerongen, 2008], and [Radder, 2008]. 52 See also [Dipert, 1993; Perlman, 2004; Baker, 2004; 2007; Thomasson, 2007] and [Margolis and Laurence, 2007]. 53 The author would like to thank Sven Ove Hansson, Wybo Houkes, Peter Kroes and Hans Radder for their valuable comments on an earlier version of this Introduction.

16

Anthonie Meijers, editor

Handbook. What then followed was a time-consuming effort to find authors who were sufficiently qualified to write on these topics. In some cases that was easy because there were well-known experts in the fields in question. In other cases it turned out to be extremely difficult or sometimes even impossible. The list of authors and topics became relatively stable after two years, though changes were made even in the final phase. Over the five years of the making of the Handbook the number of topics gradually decreased from 65 to 41 and came to involve 51 authors. In January 2007 a workshop was organized at Eindhoven University of Technology to discuss the first versions of the chapters. In parallel sessions in-depth discussions were held between authors writing on similar topics. The aim was to give feedback and improve the chapters but also to create a certain synergy and demarcate the topics. After the workshop a long process followed. The practical limitations of a number of authors and the fact that many chapters addressed new topics requiring a great deal of new research all caused the completion of the Handbook to be delayed. This did not come as a surprise to the editors, given the ambitious nature of the project. We had to walk the tightrope between including chapters on new topics in the Handbook and meeting a certain deadline. In the last phase several chapters had to be omitted so that the final deadline could be honoured. ACKNOWLEDGEMENTS It would be an understatement to say that I am most grateful to my associate editors Sven Ove Hansson, Wybo Houkes, Peter Kroes, Ibo van de Poel, Hans Radder, and Sjoerd Zwart for their intellectual contributions to the Handbook, the enormous amount of work and effort put into the project, and their continued moral support and good spirits. Without them this Handbook would simply not have been possible. We had a common purpose in this project and we shared a belief in its importance. Jane Spurr’s hard work, support and dedication at the London office have been invaluable in the last phase and I am very much indebted to her. I am also grateful to Rianne Schaaf and to Krist Vaesen for organizing the above-mentioned workshop in such a way that the authors of the Handbook were able to develop a sense of a joint mission. Krist Vaesen was also involved in the literature searches, in the development and updating of the Handbook’s website and in the manuscript handling. I would also like to thank the members of the Philosophy and Ethics Section at Eindhoven for their last minute efforts with proof-reading and indexing the chapters of this Handbook. Finally, I would like to thank the editors Dov M. Gabbay, Paul Thagard and John Woods for their sustained support during the development of this Volume for their Handbook Series.

General Introduction

17

BIBLIOGRAPHY [Abzug and Larrabee, 2005] M. J. Abzug and E. E. Larrabee. Airplane Stability and Control: A History of the Technologies That Made Aviation Possible 2nd ed., Cambridge University Press, 2005. [Achterhuis, 1995] H. Achterhuis. De moralisering van de apparaten. Socialisme en Democratie, 52(1), 3-12, 1995. [Baird, 2004] D. Baird. Thing Knowledge: A Philosophy of Scientific Instruments, Berkeley: University of California Press 2004. [Baker, 2007] L. R. Baker. The Metaphysics of Everyday Life: An Essay in Practical Realism, Cambridge (UK): Cambridge University Press, 2007. [Baker, 2004] L. R. Baker. The ontology of artifacts. Philosophical Explorations, 7(2), 99-111, 2004. [Bijker et al., 1987] W. Bijker, T. P. Hughes, and T. Pinch. The Social Construction of Technological Systems: New Directions in the Sociology and History of Technology, Cambridge (Mass.): The MIT Press 1987. [Billington, 1985] D. P. Billington. The Tower and the Bridge, Princeton: Princeton University Press, 1985. [Boon, 2006] M. Boon. How science is applied in technology. International Studies in the Philosophy of Science, 20(1), 27-47, 2006. [Bucciarelli, 1996] L. L. Bucciarelli. Designing Engineers, Cambridge (Mass.): The MIT Press, 1996. [Buchanan, 2002] R. Buchanan. Brunel: The Life and Times of Isambard Kingdom Brunel, New York: Hambledon and London, 2002. [Bunge, 1985] M. A. Bunge. Epistemology and Methodology III: Philosophy of Science and Technology, Dordrecht: Reidel, 1985. [Constant, 1980] E. W. Constant. The Origins of the Turbojet Revolution, Baltimore: Johns Hopkins University Press, 1980. [Curd and Cover, 1998] M. Curd and J. A. Cover, eds. Philosophy of Science: The Central Issues, New York: W.W. Norton, 1998. [Curd and Psillos, 2008] M. Curd and S. Psillos, eds. The Routledge Companion to Philosophy of Science, London: Routledge, 2008. [Dipert, 1993] R. R. Dipert. Artifacts, Art works, and Agency, Philadelphia: Temple University Press, 1993. [Dym, 1994] C. L. Dym. Engineering Design: A Synthesis of Views, Cambridge: Cambridge University Press, 1994. [Floridi, 2003] L. Floridi, ed. The Blackwell Guide to the Philosophy of Computing and Information, Oxford: Blackwell Publishers, 2003. [Friedman, 2005] T. Friedman. Electric Dreams: Computers in American Culture, New York: NYU Press, 2005. [Geels and Schot, 2007] F. W. Geels and J. Schot. Typology of sociotechnical transition pathways. Research Policy, 36(3), 399-417, 2007. [Gero et al., 1992] J. Gero, K. Tham, and H. Lee. Behaviour: a link between function and structure in design. In Intelligent Computer Aided Design. Amsterdam: Elsevier, pp. 193– 225, 1992. [Hackett et al., 2008] E. J. Hackett, et al., eds. The Handbook of Science and Technology Studies. 3rd ed., Cambridge (Mass.): The MIT Press, 2008. [Hansson, 2003] S. O. Hansson. Ethical criteria of risk acceptance. Erkenntnis, 59(3), 291-309, 2003. [Hansson, 2007] S. O. Hansson. What is technological science? Studies In History and Philosophy of Science Part A, 38(3), 523-527, 2007. [Hills, 1993] R. L. Hills. Power From Steam: A History of the Stationary Steam Engine, Cambridge (Mass.): Cambridge University Press, 1993. [Houkes and Vermaas, 2004] W. Houkes and P. Vermaas. Actions versus functions: a plea for an alternative metaphysics of artifacts. The Monist, 87(1), 52-71, 2004. [Hughes et al., 2007] J. Hughes, P. Kroes, and S. Zwart. A semantics for means-end relations. Synthese, 158(2), 207-231, 2007. [Israel, 2000] P. Israel. Edison: A Life of Invention 1st ed., Hoboken: Wiley, 2000.

18

Anthonie Meijers, editor

[Koen, 2003] B. V. Koen. Discussion of the Method: Conducting the Engineer’s Approach to Problem Solving, New York: Oxford University Press, 2003. [Kroes and Meijers, 2006] P. Kroes and A. Meijers, eds. Special Issue: The dual nature of technical artefacts. Studies In History and Philosophy of Science Part A, 37(1), 1-185, 2006. [Latour, 2005] B. Latour. Reassembling the Social: An Introduction to Actor-Network-Theory, Oxford: Oxford University Press, 2005. [Latour, 1987] B. Latour. Science in Action: How to Follow Scientists and Engineers Through Society, Cambridge (Mass.): Harvard University Press, 1987. [Latour, 1992] B. Latour. Where are the missing masses? The sociology of a few mundane artifacts. In Shaping Technology/Building Society: Studies in Sociotechnical Change, ed. W.E. Bijker and J. Law. Cambridge (Mass.): MIT Press, pp. 225-258, 1992. [Laudan, 1984] R. Laudan, ed. The Nature of Technological Knowledge: Are Models of Scientific Change Relevant?, Dordrecht: Reidel, 1984. [Layton, 1974] E. Layton. Technology as knowledge. Technology and Culture, Vol. 15(1), 31-41, 1974. [Lelas, 1993] S. Lelas. Science as technology. British Journal for the Philosophy of Science, 44, 423-442, 1993. [Lintsen et al., 1995] H. Lintsen et al., eds. Geschiedenis van de Techniek in Nederland: de Wording van een Moderne Samenleving, 1800-1890, ’s-Gravenhage; Zutphen: Stichting Historie der Techniek; Walburg Pers, 1995. [MacKenzie, 1993] D. MacKenzie. Inventing Accuracy: A Historical Sociology of Nuclear Missile Guidance, Cambridge (Mass.): The MIT Press, 1993. [Margolis and Laurence, 2007] E. Margolis and S. Laurence, eds. Creations of the Mind: Theories of Artifacts and their Representation, Oxford: Oxford University Press, 2007. [McNeil, 1996] I. McNeil. An Encyclopedia of the History of Technology 1st ed., London: Routledge, 1996. [Misa, 1998] T. J. Misa. A Nation of Steel: The Making of Modern America, 1865-1925, Baltimore: The Johns Hopkins University Press, 1998. [Mitcham, 1994] C. Mitcham. Thinking Through Technology: The Path Between Engineering and Philosophy, Chicago: University of Chicago Press, 1994. [Newton-Smith, 2000] W. Newton-Smith. A Companion to the Philosophy of Science, Malden (Mass.): Blackwell Publishers, 2000. [Perlman, 2004] M. Perlman. The modern philosophical resurrection of teleology. The Monist, 97, 3-51, 2004. [Philosopher’s Information Center, 2008] Philosopher’s Information Center. Philosopher’s Index 1940-2008, New York: Ovid Technologies, 2008. [Radder, 2003] H. Radder, ed. The Philosophy of Scientific Experimentation, Pittsburgh: University of Pittsburgh Press, 2003. [Radder, 2008] H. Radder, ed. Critical approaches to technology: special issue. Social Epistemology, 22(1), 1-124, 2008. [Renssen, 2005] A. Renssen. Gellish: A Generic Extensible Ontological Language Design and Application of a Universal Data Structure, Delft: Delft University Press, 2005. [Rhodes, 1995] R. Rhodes. The Making of the Atomic Bomb, New York: Simon and Schuster, 1995. [de Ridder, 2007] J. de Ridder. Reconstructing Design, Explaining Artefacts; Philosophical Reflections on the Design and Explanation of Technical Artefacts. PhD thesis, Delft, 2007. [Rogers, 1983] G. F. C. Rogers. The Nature of Engineering: A Philosophy of Technology, London: Macmillan Press, 1983. [Rolt, 1959] L. Rolt. Isambard Kingdom Brunel: A Biography, London: Longmans, 1959. [Scharff and Dusek, 2003] R. C. Scharff and V. Dusek, eds. Philosophy of Technology: The Technological Condition - An Anthology, Malden (MA): Blackwell Publishers, 2003. [Schot et al., 1998] J. W. Schot et al., eds. Techniek in Nederland in de Twintigste Eeuw, ’sGravenhage; Zutphen: Stichting Historie der Techniek: Walburg Pers, 1998. [Singer et al., 1954] C. J. Singer et al. A History of Technology, Oxford: Clarendon Press, 1954. [Staudenmaier, 1985] J. M. Staudenmaier. Technology’s Storytellers: Reweaving the Human Fabric, Cambridge (Mass.): MIT Press, 1985. [Thomasson, 2007] A. L. Thomasson. Ordinary Objects, Oxford: Oxford University Press, 2007. [Vaesen and van Amerongen, 2008] K. Vaesen and M. van Amerongen. Optimality vs. intent: limitations of Dennett’s artifact hermeneutics. Philosophical Psychology, 21(6), 779-797, 2008.

General Introduction

19

[van Inwagen, 1990] P. van Inwagen. Material Beings, Ithaca, N.Y.: Cornell University Press, 1990. [Verbeek, 2000] P. Verbeek. De Daadkracht der Dingen: Over Techniek, Filosofie en Vormgeving, Amsterdam: Boom, 2000. [Verbeek, 2005] P. Verbeek. What Things Do; Philosophical Reflections on Technology, Agency, and Design, University Park (Penn.): Pennsylvania State University Press, 2005. Translation of Verbeek (2000). [Vincenti, 1990] W. G. Vincenti. What Engineers Know and How They Know It: Analytical Studies from Aeronautical History, Baltimore: Johns Hopkins University Press, 1990. [Zachary, 1997] G. Zachary. Endless Frontier: Vannevar Bush, Engineer of the American Century, New York: Free Press, 1997. [Zwart and Franssen, 2007] S. Zwart and M. Franssen. An impossibility theorem for verisimilitude. Synthese, 158(1), 75-92, 2007.

This page intentionally left blank

Part I

Technology, Engineering and the Sciences

This page intentionally left blank

INTRODUCTION TO PART I

Hans Radder, associate editor Part I of this Handbook addresses the relationship between technology, engineering and the sciences. On the one hand, this is an interesting and important subject in its own right; on the other, it provides the necessary background to several of the more focused chapters in the other parts of the Handbook. Looking at this subject from a general perspective, three clusters of issues may be distinguished.

1

VARIETIES OF DEFINITION

First, there are questions concerning the appropriate terminology and the definitions of the chosen relata. Why use ‘technology’ and not, for instance, ‘technics’ or ‘techniques’ ? Can the modern concept of engineering be taken to be equivalent to the older notions of the mechanical or industrial arts? Does it make sense to distinguish between engineering and engineering science? What to include among the sciences: the natural sciences, the engineering sciences, the social sciences or even, in the European tradition, the Geisteswissenschaften? A comprehensive discussion of such issues entails the task of providing more elaborate definitions of the terms used. This leads to several further questions. The overarching question is which variety of definition should be used. As is shown in detail in Carl Mitcham’s chapter, there are at least five approaches available: etymological, essential, prescriptive, linguistic, and pragmatic approaches to definition. An important, related issue is this: what can, or should, be achieved by providing a definition? What I mean is the question of the nature of the relationship between the definition and that what is being defined (the definiendum). Clearly, how to answer this question will depend on the chosen variety of definition. A logicist approach requires that the definition specifies a set of necessary and sufficient conditions for the instances to which it applies. However, in the case of wide-ranging and multidimensional notions, such as technology, engineering and science, this proves to be hard, if not impossible, to achieve. Hence a more realistic approach requires that the definition should capture key features or typical patterns of the definiendum, or that it should specify significant family resemblances among its instances. This is the approach taken in the chapters of this part of the Handbook, sometimes explicitly, as in my own chapter, and sometimes more implicitly, as in the chapter by Gerhard Banse and Armin Grunwald. Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

24

Hans Radder, associate editor

2 KINDS OF RELATIONS Next to specifying the relata, we need to study the relationship between technology, engineering and science. A crucial methodological issue is the existence of different kinds of relations that may be studied. Again, from a general perspective the following three approaches can be distinguished. We may look at the empirical relations between technology, engineering and the various sciences. This approach may include an historical account of the actual relations between technology and the natural sciences, as can be found in David Channell’s chapter, or a study of the (problematic) role of the social sciences in engineering and the engineering sciences, as provided in the chapter by Knut Sørensen. Although it is important that philosophers are knowledgeable about the actual practices of technology, engineering and the sciences, philosophy cannot be limited to an empirical study of its subject matter. Hence, a second approach focuses on conceptual relations: it characterizes technology, engineering and the sciences through a conceptual specification of their similarities and dissimilarities. This is the approach taken in the second chapter of this part, which addresses the relationship between technology and natural science, and in the last chapter, where the focus is on a conceptualization of the engineering sciences in relation to natural sciences and technical practices. A subject which is occasionally discussed in this part of the Handbook but which deserves more detailed (empirical and conceptual) study, is the important role of technological instrumentation in the various sciences.1 A third approach focuses on the relationship between technology, engineering and the sciences from an evaluative perspective. How are the various relata evaluated and how should they be evaluated, both in themselves and as compared to each other, and both as regards their epistemic value and in terms of their social and moral value? Although such questions crop up occasionally (for instance, in my own chapter) and although several chapters in Part V of this Handbook include relevant material, exploring these evaluative relations in more detail remains an important task for further research in the philosophy of technology and engineering sciences.2

3 TYPES OF MODELS Finally, different types of models are possible of each of these three kinds of relationship and the corresponding relata. The first type may be called primacy models. In these models, empirical, conceptual or evaluative primacy is given to either technology, to engineering, or to science. Authors who emphasize the practical basis of engineering and science will often give primacy to technology, while 1 For some studies of scientific instrumentation and its philosophical significance, see the contributions by Rom Harr´e, Davis Baird and Michael Heidelberger in [Radder, 2003]. 2 A comprehensive historical discussion is presented in [Forman, 2007].

Introduction to Part I

25

authors who stress the scientific basis of engineering and technology will be inclined to assign primacy to science. The ‘humanities tradition’ in the philosophy of technology frequently endorses the former position, while the ‘engineering tradition’ often advocates the latter.3 For instance, a ‘conceptual primacy of technology model’ can be found in Heidegger’s philosophy of technology. The so-called linear model of the science-technology relationship, discussed in several chapters, exemplifies a ‘primacy of science model’, which may be further specified as empirical and/or conceptual and/or evaluative. A second type of model rejects claims to primacy in favor of a two-way interactive approach, which assumes that technology, engineering and science are independent, yet interacting, entities. For instance, as described in the second and fourth chapters of this part, without denying its interaction with scientific knowledge, historians of technology have often emphasized the independent character of technological knowledge. The third chapter both demonstrates the independence of engineering from social science and implies that a greater interaction between the two would be desirable. The fifth chapter provides an independent characterization of the engineering sciences in terms of its methods and goals, yet it also emphasizes the significance of its interactions with practical technologies and basic sciences. A third type of model is based on the idea of a seamless web: these models assume that technology, engineering and science are so strongly intertwined that they cannot be sensibly distinguished. Because of the claimed seamlessness, proponents of such models often use the notion of technoscience (see the second and fourth chapters). Again, both interactive and seamless-web models may be developed from an empirical, a conceptual or an evaluative perspective. From this brief sketch of a comprehensive conceptual framework for studying the relationship between technology, engineering and the sciences it will be clear that the subject of this part of the Handbook covers a large variety of relevant issues. On several of these issues, substantial work has been done and hence this work is presented and reviewed in the subsequent chapters; on other, less researched issues, the chapters of this part of the Handbook offer more exploratory accounts; discussion of further issues can occasionally be found in other parts of this Handbook (in particular, in Part VI); finally, still other issues have to await the future development of the relatively young area of the philosophy of technology and engineering sciences. BIBLIOGRAPHY [Forman, 2007] P. Forman. The Primacy of Science in Modernity, of Technology in Postmodernity, and of Ideology in the History of Technology History and Technology, 23, 1-152, 2007. [Mitcham, 1994] C. Mitcham. Thinking through Technology. The Path between Engineering and Philosophy. University of Chicago Press, 1994. [Radder, 2003] H. Radder, ed. The Philosophy of Scientific Experimentation. University of Pittsburgh Press, 2003. 3 For

these two traditions, see [Mitcham, 1994].

This page intentionally left blank

DEFINING TECHNOLOGY AND THE ENGINEERING SCIENCES Carl Mitcham and Eric Schatzberg One of the more difficult issues in any regionalization of philosophy often centers on reaching consensus about how best to define the subject under investigation. This is certainly the case with regard to technology and engineering science (often in the plural and also sometimes called technological sciences), the region of reality or human experience on which the philosophy of technology focuses. These phenomena and their key terms are loaded with contested connotations and interpretations, in part because definition and associated conceptualization has implications for other issues, such as the relations between technology, science, and art. The present chapter thus begins (section one) by reviewing various approaches to definition before turning to the issue at hand, that is, defining technology. It then sketches (section two) the etymological and historical background before undertaking to analyze how “technology” has emerged in conjunction with diverse definitional and conceptualizing strategies in science and engineering (section three), in the humanities (section four), and in the social sciences (section five). A conclusion (section six) argues in favor of distinctions that mirror common language use and real-world phenomena while paying special attention to context and implications. A further difficulty with defining technology arises from problems of translation. Most continental European languages use two distinct terms that are commonly rendered in English as “technology,” namely the vernacular forms of the Latin technica and technologia. Although nuances and exceptions exist in each language, in essence the distinction is what one would expect from etymology. The cognates of “technology” generally refer to the science of or discourse about the practical, material arts, while cognates of “technique” are applied to the actual processes and methods of these activities. Through most of the 20th century, “technique” was the dominant term; most philosophical discourse about technology in French, German, Dutch, Spanish, Portuguese, Italian and more is in fact a discourse about “technique”: la technique, die Technik, de techniek, la t´ecnica. (These cognates can also mean “technique” in the conventional English sense of skill or means to an end; contrast Klaviertechnik, piano technique, with Elektrotechnik, electrical engineering. But such usage is distinct from the one generally translated as “technology.”) Since World War II, however, all words rooted in both technica and technologia have regularly been translated into English as “technology,” thus occluding a basic distinction. A subsequent reverberation of English-language discourse about Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

28

Carl Mitcham and Eric Schatzberg

“technology,” often derived from the European discourse on “technique,” when translated into other languages by a cognate of “technology,” has tended to blur the distinction further. In this chapter, “technology” covers both discourses, while nevertheless attempting whenever possible and appropriate to call attention to a distinction between technique and technology.

1

APPROACHES TO DEFINITION

Definitions and how they function are fundamental to philosophy. In the present instance the definitional issue is both nominal and real, that is, it concerns the word “technology” (as well as associated terms) and technical phenomena. The question with regard to technology in the philosophy of technology parallels discussions in the philosophy of science about how properly to demarcate science from nonscience, in the philosophy of religion about how best to describe religion, and in the philosophy of language about how to characterize what constitutes language. Answers have implications as well for defining the engineering sciences. Given its significance, prior to exploring definitional alternatives, it is thus propaedeutic to review different approaches to definition and conceptualization in general theories that are closely related to philosophy of language, theory of meaning, and conceptions of truth. In the present case, however, focus will be restricted to five basic approaches to definition and conceptualization: etymological, essential, prescriptive, linguistic, and pragmatic. 1) Etymological definitions are oriented around the origins of terms. In its study of the historical roots of words, etymology is closely related to philology and linguistics. “Etymology” is itself etymologically a compound of the Greek έτυμον (true) and λόγος (speech or reason). In the European tradition, the approach to meaning through linguistic origins can be found as early as the 4th century BCE in Plato’s Crytalus, where an argument for natural meaning involved the development of numerous fanciful etymologies. The Hebrew Scriptures, as well, included etymologies to explain especially place names. Outside the European tradition, as early as the 7th century BCE Sanskrit scholars employed etymology to examine words of sacred significance. Numerous classical Greek and Roman authors continued to utilize etymology to present meanings, and in the 7th century CE Isadore of Seville’s Etymologiae was written as a general handbook of learning. In its modern (and sometimes scientific) sense, a major contributor to etymology was the linguistic prodigy William Jones who, as a British civil servant in 19th century colonial India, did comparative research on Indo-European words. One of Jones’ discoveries was the existence of a common Greek-Sanskrit stem, tekhn-, meaning “woodwork” or “carpentry,” which is obviously present in “technology.” During this same period Friedrich Nietzsche deployed etymology in his genealogical criticism of bourgeois Christian morality, to be followed in the 20th century by Martin Heidegger, who used etymology to revisit accepted meanings in a variety of contexts. As a method for constructing philosophical definition, however,

Defining Technology and the Engineering Sciences

29

etymology was already in the Crytalus challenged by Socrates, who argued that “knowledge of things is not to be derived from names” and that instead things must “be studied and investigated in themselves” (439b). Calls for a turn from words to things have ever since been a motif in the history of philosophy. 2) Investigating things in themselves aims to produce essential definitions or concepts. In classical theory, definitions are of concepts, which are themselves “structured mental representations that encode a set of necessary and sufficient conditions for their [perceptual] applications” [Margolis and Laurence, 1999, p. 10]. Essential definitions typically take the form of indicative propositions that state the whatness of something. For Aristotle (and many others) essential definitions have a genus and species or differentia structure. One example might be the claim that technology is the systematic human making of physical objects and/or the using of such objects: technology is human behavior (genus) involved with the systematic making or using of artifacts (species). In genus-species definitions, however, questions arise about whether the differentia indicates a structure or reality or is simply a convenient means for controlling behavior and word usage. Biological taxonomies, for instance, employ convenient features of organisms that do not necessarily reflect their most fundamental genetic determinants. But when technology is defined, as it sometimes is, as applied science, this is often taken to express the inner structure of the phenomenon — a view that has also been widely contested as inaccurate, given that such engineering sciences as thermodynamics often follow the invention of technologies, in this case the steam engine. Essential definitions are also often called connotative or intensional, insofar as they specify the necessary and sufficient conditions for something being a member of a class. Connotative definitions are contrasted with denotative or extensional (also enumerative) definitions, respectively, the latter of which simply list all the members of the class. One important form of an extensional definition is an ostensive definition, which points at the thing or things being referenced: “That airplane overhead is a technology.” Denotation has given rise to the prototype theory of concepts, in which concepts are described as “structured representations that encode the properties that objects in their extension tend to possess” [Margolis and Laurence, 1999, p. 31]. 3) Prescriptive definitions have the structure of imperative sentences that instruct or command how a word is to be used. Giving proper names is a prescriptive act: “Call me Ishmael.” Stipulative definitions are another type of prescription, as in geometry: “Let us define a point as a location without any dimensions (height, width, or depth).” Such stipulations may, as in this instance, also have a genus-species structure and indicate some non-linguistic entity, although the non-linguistic entity is usually imaginary, abstract, or ideal. As such examples suggest, prescriptive definitions can be nominal or formal. With prescriptive nominalism, definitions function as semantic rules for word usage; with prescriptive formalism, definitions are syntactic rules for symbol ma-

30

Carl Mitcham and Eric Schatzberg

nipulation. Insofar as it goes beyond proper names, prescriptive nominalism is promoted (as with Francis Bacon) as a way to clarify natural language and avoid confusion in argument. Insofar as this tactic begins with ordinary words and makes them more precise, it has been called precising definition (a method recommended by John Locke). Prescriptive formalism is more associated (as in the work of Alfred North Whitehead and Bertrand Russell) with the construction of artificial languages, especially in logic, mathematics, and other formal systems. This version of prescriptivism has also been described as creating theoretical definitions. Persuasive and genetic definitions, as well, illustrate the prescriptive approach. A persuasive definition is rhetorically structured to appeal to the psychology of the person to whom it is addressed and aims to elicit a positive or negative attitude. For example, to an engineer it might be said: “A good technological design is one that uses materials and energy efficiently.” (Virtually the same definition might also be proposed by a humanities critic of engineering: “Technology is focused only on efficiency.”) Thus persuasive definition is highly context dependent. A genetic definition conceives something in terms of its construction: “An atomic bomb is what you get when you create a critical mass of U-235.” Such definitions are compatible with the neoclassical theory of concepts as “structured mental representations that encode partial definitions, i.e., necessary conditions for their application” [Margolis and Laurence, 1999, p. 54]. Prescriptive definitions come into play in particular in various social and historical contexts; “technology,” that is, has a social history. As historians Ronald Kline [1995], Ruth Oldenziel [2006], Eric Schatzberg [2006], and others have pointed out, “technology” only acquired its current valence in English in the early 1900s — first in the United States and then after World War II by international adoption. As Schatzberg summarizes the issue, when economist Thorstein Veblen transplanted early 20th century German discussions about the social impact of Technik into the North American context, he subtly replaced the more common term “industrial arts” with “technology” in order in part to emphasize the historical power and scientific associations of mechanized forms of production. As Oldenziel further observes, “the ascendancy of technology as a keyword in the United States neatly parallels the emergence of America as a superpower committed to technology as the key tool for development in the rest of the world” [Oldenziel, 2006, p. 485]. Although such social analyses of word usage bear principally on issues related to prescription and persuasion, the focus on rhetorical dimensions is associated with what has been referred to as a linguistic turn in the history of philosophy. 4) The linguistic approach to definition focuses on words more than things and, as a philosophical position, can draw on etymology and reflect a behaviorist rather than essentialist interpretation of genus-species form; as such it is related but not reducible to the construction of lexical definitions (in dictionaries) that simply report usage. In the case of “technology,” for instance, the relevant volume of the first edition of the Oxford English Dictionary (1919) reported, moving from the most to the least common, three usages: (1) “a discourse or treatise on an art or arts; the scientific study of the practical or industrial arts”; and by exten-

Defining Technology and the Engineering Sciences

31

sion, “the practical arts collectively”; (2) “the terminology of a particular art or subject; technical nomenclature”; and (3), obsolete and rare, grammar. The second edition (1989) significantly expanded post-World War II references to the first extended meaning and then introduced the further extension of “a particular practical or industrial art” as well as numerous qualified usages such as “high-” and “low-technology.” It then added a fourth usage in special combinations such as “technology assessment” and “technology transfer.” Remarkably, the OED does not recognize the common sense of technology as artifact, as when Americans are said to have, in comparison to the poor in developing countries, houses filled with technologies (dishwashers, disposals, TVs, computers, etc.). Linguistic philosophy, however, seeks to go beyond lexicography by distinguishing use from usage. For instance, Gilbert Ryle [1953] argued that while “misusage” is not possible (people simply speak the way they speak) “misuse” is. Conceptual analysis can identify logical misuses in common linguistic usage, as when people utter inconsistent propositions. A more subtle effort to distinguish use from usage depends on Ludwig Wittgenstein’s notion of language games. For Wittgenstein, the meaning of a word is dependent on the role it plays in a language game, the rules of which provide the basis for determining proper use; misuse occurs when usage veers beyond the rules of any particular game. With reference to this approach, Fran¸cois Lyotard [1979] described technique (“technology” in the English translation) as a game in which the governing rule is neither truth, justice, nor beauty, but efficiency thus implying that to call an inefficient productive process a “technology” would be to misuse the word. A linguistic approach to definition has deep roots in the conceptual analyses present in many philosophical arguments. One classic example occurred in Plato’s Euthyphro, where Socrates tried to get the eponymous interlocutor to be clear about the meaning of piety. In the tradition of British philosophy that runs from Locke to John Stuart Mill, appeal has also often been made for terminological clarification as necessary to sound argumentation. This approach was strongly promoted by G.E. Moore and subsequently by Wittgenstein, who famously believed that many philosophical difficulties such as free will and the mind-body problem arose from linguistic misuse and could be dissolved by increasing linguistic precision. Richard Robinson [1950], in a philosophical monograph on definition, made an extended argument for a strictly linguistic account of definitions as reports of word usage plus a prescriptive or stipulative element concerning use. In such cases, it may be noted, definition is often the conclusion rather than the beginning of an argument. One example might be: “A careful review of various usages for the term ‘technology’ shows that it is best limited to scientific making and using.” (For one suggestive examination of word usage and thus a linguistic philosophical approach within rather on technology, see [Hollister-Short, 1977].) 5) In an overview of definition on which the present one draws, Raziel Abelson argued that “[e]ssentialists conclude that the knowledge conveyed by definitions is descriptive knowledge of essences, linguistic philosophers conclude that it is descriptive knowledge of language usage, while prescriptivists maintain that defi-

32

Carl Mitcham and Eric Schatzberg

nitions do not convey knowledge of any kind” [Abelson, 1967, p. 321]. In her view, however, none of these approaches, even within their own frameworks, provide sufficient criteria for distinguishing good from bad definitions. There exist a set of rules for the construction of good definitions inherited from scholastic philosophy. Such rules include that a definition should fully cover the definitum (thing being defined), that the definitum should not reappear in the definition, and that definitions should use univocal terms to avoid ambiguity. But even following these rules, definitions can sometimes be inadequate to their contexts. Pragmatic consideration of how a definition fits in or serves the context in and for which it is formulated thus constitutes another approach to definition — one that is able to utilize, as appropriate, etymological, essential, prescriptive, or linguistic approaches. In short, pragmatism seeks definitions that work well in context, denying that there is anything such as a pure definition or a definition that does not serve some purpose. Stipulative definitions work in mathematics. Essential definitions are mistaken in presuming the existence of some essence outside a functioning relationship, but within a context (e.g., biology) may function quite well; one should just not presume that a genus-species definition of a plant or animal has anything to say about (for instance) its inner structure or how the plant or animal can or should be used by human beings. The context for linguistic definitions is general language usage more than language use in any specialized situation, when precising may be more appropriate. To some extent this reflects the anonymously named theory-theory of concepts understood as “representations whose structure consists in their relations to other concepts as specified by a mental theory” [Margolis and Laurence, 1999, p. 47].

2 ETYMOLOGY AND EARLY CONCEPTUAL HISTORY It is not necessary to believe etymology uncovers the true meanings of words to grant that linguistic history often influences current meanings. J. L. Austin, for example, has acknowledged “trailing clouds of etymology” as persistent influences in current meanings [Austin, 1961, p. 149]. Particularly is this so for terms such as “technology,” which were often created with explicit reference to etymology. Understandings of the English “technology” and “engineering/technological sciences” thus benefit from an examination of their links to related terms in other languages. Nevertheless, scholars must take care not to project present meanings onto terms in the past, even the relatively recent past. Just as the present-day German Technologie is not the same as English “technology,” neither are the meanings of “technology” the same in 19th and 21st century English. The English words “technics,” “technique,” and “technology” are all rooted in the Greek τέχνη, commonly translated as “art,” “craft,” or “skill.” The same goes for the German Technik /Technologie and the French technique/technologie. (As already noted, the Greek has behind it a common Indo-European stem.) The root phoneme projects into the Latin texere (to weave) and tegere (to cover). In

Defining Technology and the Engineering Sciences

33

popular classical literature techne and its Latin translation, ars (from which the English “art”), could refer as well to cleverness or deviousness in getting, making, or doing and to specific trades, crafts, and skills of many kinds. Detailed etymologies can be found in [Schadewaldt, 1960; Heyde, 1963; Seibieke, 1968]. The Greek philosophical tradition contained rich discussions on the nature of techne. Carl Mitcham [1994] has noted how techne came to be conceived not only as an activity but as a kind of knowledge. In Plato’s Gorgias, for instance, Socrates argued that every techne involves logoi (words, speech, reason) bearing on the art involved (450b). Additionally, Socrates distinguished two types of techne, one consisting primarily of physical work that required minimal use of conscious reason (such as painting or sculpture) and another depending more intimately on reason that required little physical exertion (such as arithmetic, logistic, or astronomy). Activities such as cooking and persuading were labeled atechnos (nontechnical) — each being a mere tribe (knack or routine) based simply on empeiria (experience) (501a). In the Ion, poets who exercised their poiesis (making) by virtue of divine inspiration were also described as devoid of techne (cf. 533d). The early Plato, then, seemed to adopt what has been called the prototype theory of the concept of technics as coextensive with those human activities that can be talked or reasoned about. Contrary to often repeated characterizations of Plato as looking down on technics, this reading gave techne a nonutilitarian, not to say transhuman, dimension. The later Plato articulated a different but related understanding of techne. In the Philebus, for instance, knowing was divided into that involved with education or upbringing and that engaged with making or producing (55c). Of the second, technical knowledge, there were again two kinds: one sort (exemplified by music, medicine, and agriculture) that proceeded by conjecture and intuition based simply on practice and experience, and another (such as carpentry) that consciously involved the use of numbering, measuring, weighing (55e-56c). The latter possessed greater akribeia (exactness or precision) and this was denominated techne in a primary sense. Thus techne was distinguished from all human activity and knowledge of a political sort (education and, by extension, statecraft) so as to be associated more closely with the activities of physical making or producing. In addition, those making activities were most truly techne that involved some quantitative precision. Aristotle argued a complementary understanding of techne as one in a spectrum of different forms of engagement with reality, moving from sensation through experience to theory (Metaphysics I). According to his formal definition, techne is ἕξις μετὰ λόγου ἀληθοῦς ποιτική (Nicomachean Ethics VI, 4; 1140a11). Translated with slavish literalness: Techne is a habit (or stable disposition to act) with a true logos ordered toward making (human production). This definition of techne was used repeatedly to define art (Latin ars) by later scholars, among them Thomas Aquinas, Ephraim Chambers (in his Cyclopedia of 1728), and more recently by the neothomists Etienne Gilson and Jacques Maritain. Once again the realist, nonutilitarian character of technical knowledge came to the fore; insofar as it was true,

34

Carl Mitcham and Eric Schatzberg

this logos was based on gnosis (mental grasping or cognition) of aitia (causes) (Metaphysics I, 1; 981b6-7). Techne was thus conceived as episteme in that it involved true consciousness of the world and could be taught or communicated (Metaphysics I, 1; 981b8-10); but it was a distinctive form of episteme insofar as it bore on changing rather than unchanging things (cf. Nicomachean Ethics VI, 6; 1141b 31-36). Plato and Aristotle agreed in stressing the “logical” character of techne, even while they disagreed about the kind of logos involved. Yet neither felt drawn to join the two words — to speak of a logos of techne. Techne simply used logos. To put it simply, what can be grasped or known by techne through logos was the eidos (idea or form) in Plato or the aitia (causes) in Aristotle. What was not able to be grasped was the process, the “how to do it” of the poiesis (making). As if to emphasize the point, Aristotle argued further that it was part of techne “to know the form and the matter,” but the matter, hyle, only “up to a point” (Physics II, 2; 194a23). “Matter is unknowable (agnosis) in itself” (Metaphysics VII, 10; 103a9). The form was the idea in the mind of the artisan (Metaphysics VII, 7; 1032a35), but its union with matter was at the mercy of matter and its specific receptivity. The ultimate guide for the making process was not reason but perception, aisthesis (Nicomachean Ethics II,9; 1109b23; cf. II, 2; 1104al-9). In one instance Aristotle went so far as to describe the coming together of form and matter as dependent on a “desire” or “reaching out” in matter (Physics I, 9; 192a18). Reflecting precisely this sensibility, one scholar [Dunne, 1993] has argued at length the strong connections between techne and phronesis or that practical wisdom constituted by recognition of potential ready and able with the right assistance to be actualized in practical affairs — a recognition that was unable to be reduced to some method. In an insightful analysis of technology in Greek and Roman antiquity, the historian Serafina Cuomo [2007] compares Platonic and Aristotelian conceptualizations with those of the Hippocratic corpus, arguing that both philosophical and medical understandings observed a strong connection between techne and tyche or luck. Despite the teachability of techne, the actual making of artifacts or health remained fundamentally stochastic insofar as it depended on the particularity of matter and thus the luck of acting in accord with the kairos or right moment. At the same time this ability to seize the moment introduced into human affairs a clever or devious power that was both necessary and a threat to social order. In a gloss on a passage in the Physics, Thomas noted how Aristotle’s position implied that matter, at least particular matter, was not just privation of form, but a reality in its own right. Although with respect to the object to be made the matter could be spoken of as formless, in reality it was something that “seeks form or further form according to its proper nature” (In libros VIII physicorum I, lec. 15, par. 8). As Thomas argued elsewhere, “Act and form are received into matter according to the capacity of the matter” (Summa theologiae I, qu. 85, art. 7). Absent an artisan’s sensitivity to particular capacities of this ordering toward form, techne or ars would fail to achieve its end.

Defining Technology and the Engineering Sciences

35

Perhaps the limited role granted to logos for techne explains the absence of the Greek term τεχνολογια in both Plato and Aristotle — except in one case, Aristotle’s treatise on the techne of rhetoric, where a form of the word appears four times (Rhetoric I, 1; 1354b17; 1354b27; 1355a19, and I, 2; 1356a110). The exact meaning in each occurrence is debatable. But on one interpretation the connection is reflects how in the case of rhetoric, which deals with the nonsubstantial matter of words, it becomes possible to know not just formal ends of making but also the processes by which the related making takes place. From this imaginative Aristotelian usage it is easy to develop the definition of τεχνολογια as a treatise on (or terminology of) the arts of language, especially grammar and rhetoric, a usage that can be found among Hellenistic and Byzantine authors. Yet the Latin transliteration technologia was unknown in classical or medieval Latin literature. (Cicero used the term once, but only in Greek.) Although Seibicke [1968] concludes that the Greek τεχνολογια had “no direct influence” on scholarly discourse during the Middle Ages or the beginning of the modern era, it is difficult not to suspect some connection between the older Greek usage and the emergence in late 17th century English of “technology” as a treatise on grammar. There is, then, no continuous history of usage linking the classical Greek τεχνολογια with the current meanings of “technology.” Yet the term did reappear in Latin during the Reformation with connotations close to those of its classical roots, in the work of the 16th century French Protestant rhetorician Peter Ramus. Ramus used technologia to refer not to the logos of one techne (with words) but to the logos of relations among all technai. Infected by a passion for method associated with Renaissance humanism, and inspired by a perennial call to move from words to things, Ramist technologia systematically ordered and arranged the arts and sciences. Ramus also coined the term technometria (which occurs in neither Greek nor Latin) as a synonym for technologia. Both terms were taken up by a number of Protestant theologians in the 17th and early 18th centuries, in particular by the English Puritan William Ames. Although in principle Ramist technologia applied to the arts (technai ) as a whole, in practice it referred to the liberal arts, especially as dealt with in higher education. Students at Harvard and Yale, for example, produced a series of theses technologicae from the mid-17th century into the 18th century. This early modern technologia had little to do with the mechanical arts. But by the beginning of the 19th century, “technology” had become strongly linked to the practical arts and modern industry. Perhaps the earliest attestation of this usage is provided by the second or 1661 edition of Thomas Blount’s Glossographia, a dictionary of “Hard Words,” generally of foreign origin, that had become common in “our refined English tongue.” Although the first edition of 1656 did not have an entry for “technology,” the second edition identified the term as Greek and defined it as “a treating or description of Crafts, Arts or Workmanship.” Similarly, John Kersey’s edition of Edward Phillips’s 1706 dictionary, The New World of English Words, defined “technology” as “a Description of Arts, especially the Mechanical.” This new definition also appeared in the work of the German Enlightenment

36

Carl Mitcham and Eric Schatzberg

scholar Christian Wolff. The second volume of his Philosophia rationalis siva logica, that is, Praemittitur discursus praeliminaris de philosophia in genere (1728), defined technologia as scientia atrium et operum artis (science of arts and of works of art) or scientia eorum, quae organorum corporis, manuum potissimum, opera ab hominibus perficuntur (science of what humans produce by the work of the organs of the body, mainly the hands). At both the linguistic and philosophical levels, in England and in Germany, “technology” was beginning to gain a new meaning. Nevertheless, “technology” in any of its forms remained a rare term in the 18th century. For example, Samuel Johnson’s famous English dictionary of 1755 did not include it, nor did subsequent editions of the dictionary well into the 19th century. “Technology” first gained currency not in English, but in German. Johann Beckmann’s Anleitung zur Technologie [1777] was the first work that self-consciously developed the concept of technology as a discipline devoted to the systematic description of handicrafts and industrial arts (see also Part I in this Volume, “Coherence and Diversity in the Engineering Sciences”). Beckmann was a prominent figure in German cameralism, a set of practically-oriented academic disciplines concerned with state administrative organization. In the Anleitung, Beckmann made Technologie a key academic field within cameralism, as an approach to an emerging area of German higher education. In academic discourse, the Anleitung included both craft work and industrial production as aspects of Technologie. Yet this cameralist concept remained distinct from the late 20th century notion of technology. Cameralism was grounded in a natural-historical approach to knowledge and as such focused on classifying rather than explaining. Beckmann’s Technologie rested firmly in the tradition of Bacon’s proposal for a natural history of trades, a project also pursued in Denis Diderot and Jean D’Alembert’s contemporary project of the Encyclop´edie ou dictionnaire raisonn´e des sciences, des arts et des m´etiers. Beckmann, however, sought to make Technologie into a true Wissenschaft by creating a classificatory scheme equivalent to the Linnaean system for plants and animals. In this endeavor he ultimately failed, especially to develop a classification system that could accommodate the new machines and processes that were transforming British industry, such as the steam engine or mechanized spinning, and thereby to alter the practical imagination [Lindenfeld, 1997]. In an insightful assessment of Beckmann’s approach to what the Czech-French philosopher and historian Jan Sebestik terms “technological science,” Sebestik [1983] traces a hundred years of Beckmann’s European influence and a mid-20th century revival of especially French interest in this conception of technology in, for instance, the work of Jean-Claude Beaune [1980]. (In this regard, see also [Guillerme and Sebestik, 1966; Guillerme, 1985; Mertens, 2002].) A work similar to Beckmann’s Anleitung emerged almost half a century later in the United States with Jacob Bigelow’s Elements of Technology [1829]. Subtitled On the Application of the Sciences to the Useful Arts, this was the first English work to use “technology” in its title, and is thus often mistakenly credited with introducing the term into American usage. Bigelow, a physician and Harvard professor, claimed that he had adopted a word “found in some of the

Defining Technology and the Engineering Sciences

37

older dictionaries,” and that “under this title” he included “an account. . . of the principles, processes, and nomenclatures of the more conspicuous arts, particularly those which involve applications of science, and which may be considered useful.” In contrast to Beckmann, who made Technologie into a minor but significant academic field in Germany, “technology” was little more than a pretentious adornment for Bigelow. He only invoked the term once in the preface of this 500page work, and dropped all reference to “technology” in the 1840 edition, which was retitled simply The Useful Arts. Nevertheless, Bigelow’s book was consistent with the classifying spirit of cameralist Technologie, although in substance the Elements was little more than a turgid compilation drawn from published sources. More significant for American usage was the 1861 decision in Boston to name its new engineering school the Massachusetts Institute of Technology. The choice of “technology” rather than “polytechnic” was odd, and may have been suggested by Bigelow [Stratton and Mannix, 2005]. As with Bigelow’s Elements, the “technology” in MIT carried little theoretical weight aside from its usage as a name, although the name did help associate “technology” with engineering education. Yet for the remainder of the 19th century, “technology” remained a secondary term of little philosophical importance. Its meaning was captured well by the American Century Dictionary of 1891: “That branch of knowledge which deals with the various industrial arts; the science or systematic knowledge of the industrial arts and crafts, as spinning, metal-working, or brewing,” a definition in essence little changed from Christian Wolff’s early in the 18th century. Something like the Wolff-Bigelow usage continued to be manifest late in the 19th century in a series of textbooks for correspondence courses that promised students information “such as can be immediately applied in practice” and in which “the knowledge of mathematics required is limited to the simplest principles of arithmetic and mensuration” [International Library of Technology, 1897, pp. iii-iv]. In retrospect, the lack of definitional discourse and critical reflection is surprising, given the central role of industrial arts in the momentous economic, social, and cultural transformations during the second half of the 19th century. When such philosophical and critical discourse did arise, it centered not on “technology” or its cognates but on the German term Technik. This review of the etymological roots of “technology” suggests a more continuous development than is fully warranted. In part, “technology” was repeatedly reinvented by scholars who drew on etymology in new contexts. This construction of meaning continued as the term spread to the various communities of discourse in which the philosophy of technology was to be pursued. “Technology” is not a word or concept that has some pure or univocal meaning outside these contexts, so that its etymology has undergone multiple adaptations. Although there are many ways to describe these contexts, for the purposes of the philosophy of technology and the engineering sciences, they can be divided into three broad scholarly communities: natural sciences and engineering, humanities (and humanistic discourse in general), and the social sciences, all of which emerged as distinct fields of knowledge in the 19th century. Adopting a combined pragmatic and linguistic approach

38

Carl Mitcham and Eric Schatzberg

to definition, it is thus appropriate to consider in more detail how the term functions and related phenomena are understood in different but related manners in these three communities — beginning with that community to which etymology most immediately points, science and engineering. 3 IN SCIENCE AND ENGINEERING A sophisticated philosophical discourse on technology first emerged among German engineers in the late-19th century. Yet instead of the cameralist term Technologie it adopted the more common Technik, which was nevertheless interpreted in new ways. Germany was rapidly industrializing during the second half of the 19th century, in part on the basis of new, science-shaped industries such as those producing synthetic dyes and using electrical power. Although Technologie never disappeared completely, it was not embraced by German engineers and industrialists, for several reasons. Among engineering educators, the new theory-centered approaches to engineering education were incompatible with the encyclopedic, taxonomic approach of Technologie. Nor was the field of Technologie, intended for training administrators, of much use to practicing engineers. Finally, cameralism itself became a discredited doctrine in liberal economic though of the late-19th century, as did Technologie by association. Rather than Technologie, German engineers embraced Technik — and “the notion of Technik cannot be considered the heir of Technologie” [Frison, 1998, p. 119]. Technik and Technologie were the focus of independent discourses and rarely discussed together or compared. Like Technologie, Technik entered German through modern Latin, mainly in the late-18th century. In its broadest sense Technik indicated the practical rules and methods used to achieve any given end. This usage is similar to the English “technique,” in the sense that one speaks of the technique of a painter or musician. By the mid-19th century, however, Technik had also become firmly associated with the industrial arts. Used without modification, Technik encompassed all the arts of material production, conceived as a whole. Thus the meanings of Technik split into two related strands, a narrower one referring to the material aspects of industry, and a broader one encompassing the rules, procedures, and skills for achieving any goal. In this context the German words are appropriately translated as “engineering” and “engineering sciences” (sometimes “technological sciences”). In the second half of the 19th century, Technik became central to the selfunderstanding of the German engineering profession. According to its 1856 constitution, for instance, the primary purpose of the pan-German engineering association Verein Deutscher Ingenieure (VDI) was the advancement of German Technik more than promotion of the interests of engineers. The organization defined membership almost completely in terms of Technik or things technische open to practicing Techniker (engineers), teachers of Technik or Technikwissenschaften, along with the owners and managers of technical establishments. Engineers were only part of a larger technical professional community.

Defining Technology and the Engineering Sciences

39

As Mitcham [1994] has argued, it was from the reflections of German engineers that a sustained philosophy of technology initially emerged, a philosophy nonetheless centered on the concept of Technik. The first work in this spirit was not actually from an engineer, but from the left-Hegelian Ernst Kapp, whose two decades of exile on the Texas frontier gave him considerable experience with tools and machinery. In his Grundlinien einer Philosophie der Technik [1877], the first book to explicitly call itself a philosophy of either Technik or technology, Kapp conceptualized Technik in terms of “organ projection,” that is, as an extension of the human body. More significant than Kapp were the philosophical writings of a small cadre of late 19th century Techniker. In the rapid industrialization that followed German unification, engineers as part of a technical community sought status within the German cultural world in which pride of place was associated with Bildung, understood roughly as educational development through the principles of high culture. In pursuit of recognition through a distinctly technical Bildung, engineers developed an engineering-based philosophy of technology. Contributions to such a philosophy had become extensive by the end of the century, when the German-speaking Russian engineer Peter von Engelmeyer published a series of 12 articles in a German engineering journal on the “Allgemeine Fragen der Technik ” (or general questions of technology). Engelmeyer discussed dozens of authors who addressed fundamental issues regarding Technik, including other engineers, such as Max-Maria von Weber, Franz Releaux, and Josef Popper-Lynkeus. These engineer-philosophers, including Engelmeyer himself, grappled with the relationships between science and Technik, culture and Technik, and progress and Technik, as well as the social status of Techniker and the nature of Technik itself — all issues that have remained central to the philosophy of technology. This engineering philosophy continued into the 20th century, with engineers emphasizing the creative and spiritual aspects of Technik to defend themselves against attacks from humanities intellectuals. Some of this work contributed to what Jeffrey Herf [1984] has termed “reactionary modernism,” an approach to technology common among German national socialists. In English “technology” received little attention within scientific and engineering communities before World War II. When the term did emerge as a focus of discussion, it combined two sets of meanings: its 19th -century definition as the science of the industrial arts and connotations borrowed from the German Technik. By the 1960s, “technology” with these merged meanings appeared promiscuously among scientists and engineers in reference to objects (mostly products and devices, less to structures), processes (from skills to systems or networks of production, transportation, and communication), and knowledge (of both how to make and use) — with a bias toward products and processes. When pushed for specificity, both scientists and engineers tended to conceive technology in essentialist terms as applied science. This conception was derived from the idea present in Bigelow of technology as the application of science to the useful arts. But the genus and species for applied science can take two different forms: for scientists,

40

Carl Mitcham and Eric Schatzberg

the genus is more likely to be science, with application as the species differentia; for engineers the genus is more commonly the industrial or mechanical arts, with their scientific pursuit as differentia. The former takes the results of science and transposes them into forms that can be of use in the design and creation of structures, products, processes, or systems; the latter adopts and adapts the methods of science to transpose traditional technical practice into more systematic or scientific forms (see also Part I in this Volume: “Models of the Science-Technology Relationship”). In what has become a minor classic of interpretation, historian Edwin Layton summarized the emergence of both forms of technology as applied science in the “mirror-image twin” communities of science and engineering. It is nevertheless the second of these transpositions that is most notable. During the 19th century, Technological knowledge was uprooted from its matrix in centuries-old craft traditions and grafted onto science. The technological community, which in 1800 had been a craft affair but little changed since the middle ages, was reconstructed as a mirror-image twin of the scientific community. The artisan was replaced in the vanguard of technological progress by a new breed of scientific practitioner. For the oral traditions passed from master to apprentice, the new technologist substituted a college education, a professional organization, and a technical literature patterned on those of science. Equivalents were created in technology for the experimental and theoretical branches of science. As a result, by the [early 20th century], technological problems could be treated as scientific ones; traditional methods and cut-and-try empiricism could be supplemented by powerful tools borrowed from science. This change was most marked in the physical sciences and civil, mechanical, and electrical engineering, [the result of which] might be termed “the scientific revolution in technology”. [Layton, 1971, p. 562] Paradoxically, the first transposition — of the content of science into engineering — was more difficult ; but since this seemed the more ready interpretation, there readily emerged a common resistance to the definition of technology as applied science. The example of James Clerk Maxwell is illustrative. Although “Maxwell was one of those scientists who consciously attempted to contribute to technology,” it was necessary for the “engineer Oliver Heaviside to translate [Maxwell’s] electromagnetic equations into a form usable by engineers” [Layton, 1971, p. 577]. Scientific laws of nature do not immediately function as engineering design principles. Implicitly present in Layton’s exposition are thus two different conceptualizations of technology: technology as craft or technics that can be transformed by science, and technology as the result of such a transformation. In the latter case, technology is closely identified with engineering, so that it is appropriate to consider as well the meanings of this commonly associated term.

Defining Technology and the Engineering Sciences

41

Like German engineers in the late-19th century, subsequent engineers have commonly identified their profession in some sense with technology, but in ways influenced by the history of engineering. The term “engineering” is itself rooted in the classical Latin ingenera (to implant, generate, or produce). As the name for a kind of maker, however, “engineer” and its cognates originated in the late Middle Ages to designate builders and operators of battering rams, catapults, and other “engines of war.” This military focus remained primary as late as Noah Webster’s American Dictionary of the English Language (1828), which defined the engineer as “a person skilled in mathematics and mechanics, who forms plans of works for offense or defense, and marks out the ground for fortifications.” From classical times, planning or designing in the civilian realm, for peace rather than for war, was the work of the architect — as illustrated by Vitruvius’s De architectura (1st century CE), which dealt with urban planning, options in building materials, aesthetic principles, general construction strategies, hydraulics, geometry, mechanics, and so forth. In the 18th century, as the Industrial Revolution began a militarylike exploitation of nature, it was the Englishman John Smeaton who coined what it would have been reasonable to take for an oxymoron, the term “civil engineer.” (In English, civil engineering has become restricted to the design, construction, and maintenance of roads, bridges, water supply and sanitation systems, railroads, and such, although in some continental European contexts civil engineering continues to cover all non-military engineering.) Importantly, the classic definition of engineering as a civilian enterprise, formulated by Thomas Tredgold for the royal charter of the British Institution of Civil Engineers (1828), used standard genus-species form to distinguish engineering, not within the genus of science but of art: “Engineering is the art of directing the great sources of power in nature for the use and convenience of man.” Interestingly, almost a hundred years later an engineering professor at a British technical college opened a textbook with a definition of technology that simply shifted the genus: technology is “that branch of knowledge which deals with the processes and apparatus employed in the conversion of the raw products of nature into finished articles of utility” [Charnock, 1916, p. 1]. In English these definitions of engineering (and sometimes of technology as well) have found repeated echoes without and within the technical community. Webster’s Third New International Dictionary (2002), for instance, defines engineering as “the application of science and mathematics by which properties of matter and the sources of energy in nature are made useful to people.” The McGraw-Hill Dictionary of Scientific and Technical Terms (10th ed., 2007) describes engineering as “the art of directing the great sources of power in nature for the use and the convenience of humans.” Supplementing this general definition, many engineers claim that design for efficiency is the essence of engineering. Ralph J. Smith, an influential engineering educator, has argued that “the conception and design of a structure, device, or system to meet specified conditions in an optimum manner is engineering.” Furthermore, “it is the desire for efficiency and economy that differentiates ceramic engineering from the work of the potter, textile engineering from weaving,

42

Carl Mitcham and Eric Schatzberg

and agricultural engineering from farming.” “In a broad sense,” Smith concluded, “the essence of engineering is design, planning in the mind a device or process or system that will effectively solve a problem or meet a need” [Smith et al., 1983, pp. 10-12]. Complexities of usage are nevertheless manifested by the ways “engineer” can refer, in a more restricted sense, to one who operates engines, as in the expression “railroad engineer.” But the operating engineer is such by virtue of using skills rather than any systematic body of knowledge. Railroad engineering insofar as it involves more than the skill of using engines, is not expressed in the operating of locomotives but in the designing of railroad systems. Engineering is thus identified as a profession possessing the systematic knowledge of how to design structures, products, or processes, a profession that (as the standard engineering curriculum illustrates) includes some pure science and mathematics, the “applied” or “engineering sciences” (e.g., strength of materials, thermodynamics, electronics), and aims to meet some social utility. The existence of engineering or technological sciences presents another definitional challenge closely related to that of technology. To begin, engineering/technological sciences (usually in the plural) are to be distinguished from technological science (in the singular). Technological science exists external to engineering and functions as a synonym for Technologie rather than Technikwissenschaften. By contrast, “engineering science” or Technikwissenschaft (which can legitimately be used in the singular) is constituted by a knowledge production activity internal to engineering. Two useful efforts to deal with the challenge of defining this activity and its cognitive products, which include as well reference to other significant discussions of the topic, can be found in the work of G¨ unter Ropohl and Sven Ove Hansson. Drawing on earlier analyses [especially the largeonig, 1995], Ropohl [1979] argues for distinguishscale illustrated examination of K¨ ing the engineering sciences from natural sciences in terms of objectives, objects, methodology, characteristics of results, and quality criteria. (Complementary contributions to this discourse on distinguishing Technik and Technikwissenschaften in the German scholarly community can be found in [Lenk and Moser, 1973] and [Friedrich, 1999].) Hansson [2007], in turn, argues for understanding the engineering sciences as efforts to bring engineering into the academic world by adaptation of the methods of science oriented not toward explaining the facts of nature but artifacts. However, rather than formulating a definition in terms of genus and species (as a specific application of science), Hansson argues for identifying the engineering or technological sciences in terms of a set of six major characteristics: a focus on the human-made, attention to design practice, use of functional analysis, evaluation with category-specified value judgments, utilization of restricted idealizations, and the eschewing of precise mathematical solutions in favor of close approximations. As Hansson also interestingly notes, there are parallels between the development of the engineering and medical sciences. Both emerged from context specific efforts to transpose the methods of the natural sciences into a realm of technical professional practice.

Defining Technology and the Engineering Sciences

43

The most careful effort to define engineering as a profession rather than as an art or knowledge is that of philosopher Michael Davis. As he rightly notes, “Defining a field is more than semantics” [Davis, 1998, p. 31]. According to Davis, semantics is subservient to social organization, and engineering is defined in relation to engineers: engineering is what distinguishes that self-organizing professional group called engineers. Despite reservations, he ultimately favors a modified version of the definition from the Canadian Engineering Qualifications Board: “The ‘practice of professional engineering’ means any act of planning, designing, composing, evaluating, advising, reporting, directing or supervising, or managing any of the foregoing that requires the application of engineering principles, and that concerns the safeguarding of life, health, property, economic interests, the public welfare or the environment” [quoted in Davis, 1998, pp. 203204, note 6]. Davis himself avoids the mistake of including the definitum in the definition with the following reformulation: Being a professional engineer requires being accepted in a self-organizing professional community on the basis of “(1) specific knowledge and (2) commitment to use that knowledge in certain ways” [Davis, 1998, p. 37]. This approach reduces the centrality of design in favor of professional community engagement. For many professional engineers, engineering and technology are coterminous [see, e.g., Petroski, 1996]. But beginning in the 1960s, a new, subordinate category of “engineering technology” emerged within American engineering to describe the skills and instruments of those who function as support personnel or follow the directions of engineers. This category arose as a result of new educational programs established to produce technical workers operating at an intermediate level between skilled workers and engineers. Typically taught at the community college level, programs in engineering technology emphasize the hands-on aspects of engineering and require less advanced coursework in mathematics and basic sciences. Many introductory engineering textbooks (e.g., [Kemper, 2001; Wright, 2002]) now describe a spectrum of activities that run from those of the engineer (more knowledge-based) through technologist and technician to craftsperson or artisan (more hand-skill based). Any philosophy of technology emerging within the engineering and scientific communities will need to be sensitive to the tensions between these different linguistic usages. 4

IN THE HUMANITIES

In marked contrast to the discourse of science and engineering stands humanities discourse. Construed narrowly, the “humanities” — German Geisteswissenschaften or Kulturwissenschaften, as distinct from Naturwissenschaften and Technikwissenschaften — enacts a categorization of knowledge that emerged in higher education in the late 19th and early 20th centuries. The humanities category consists of a heterogeneous set of disciplines that remained after the newly professionalized natural sciences laid claim to their own spheres of knowledge and initially included the social sciences (German Sozialwissenschaften or Gesellschafts-

44

Carl Mitcham and Eric Schatzberg

wissenschaften), which were subsequently separated in part by their own trajectory of professionalization. For the U.S. National Endowment for the Humanities, for instance, the humanities include the study of language, linguistics, literature, history, jurisprudence, philosophy, archeology, comparative religion, ethics, history and criticism of the arts, and non-quantitative social sciences. But for present purposes, humanities discourse is taken to be part of a continuous tradition of critical reflection on what it means to be human, a tradition much older than the 19th century and manifest prominently in philosophy, literature, religion, the fine arts, and similar fields. Philosophical reflection about making and using activities has deep roots in the European intellectual tradition, but there is no typical humanities approach to technology. Certainly most humanities discourse is produced by intellectuals who experience technology primarily from the outside. Until recently, much of this discourse displayed significant ignorance if not hostility toward the industrial arts. The American pragmatist John Dewey often criticized this attitude of “profound distrust of the arts” and “disparagement attending the idea of the material,” which was expressed philosophically in “the sharp division between theory and practice” [Dewey, 1929, pp. 2-3]. At the same time there exists an alternative humanistic tradition that has affirmed the dignity of the mechanical arts, from Hugh of St. Victor through F. Bacon to Karl Marx. Theorists of the fine arts such as the British critic John Ruskin condemned modern technology, while others such as the Italian futurist Filippo Marinetti embraced the “machine age.” Some conservative German intellectuals of the early 20th century attacked the soullessness and materialism of modern technology, while others romanticized technological creativity, sometimes using the term “technological eros,” or linked an explosion of productivity and power variously with capitalist, national socialist, and communist ideologies. In early 21st century contexts, democratic and non-democratic alike, there continues to be a divide between humanist public intellectuals taking critical versus promotional stances; one need only consider the conflicting views with regard to post- or transhumanist technologies between, for example, the American Leon Kass and the Swede Nick Bostrom. Despite some differences of opinion, perhaps the most persistent strain in the European philosophical tradition has been a recurring theme of critical distrust in the practical arts and technology. Along with many others, Cuomo [2007] has traced this uneasiness to a criticism of techne in classical Greek philosophy. Reflecting a common interpretation, she argues that in the classical humanities techne, being teachable, was seen to pose a threat to a social order based on birth, because technicians could use acquired skills for social mobility. She notes, for example, Plato’s concern in the Republic with keeping technicians in their place, subordinate to the guardians, by promulgating the “noble lie” that one’s position in society is determined by an innate metallic constitution: gold for guardians, silver for soldiers and bronze for farmers and artisans. Similarly, Plato stressed the moral neutrality or ambiguity of techne, which thus depends on external ends for its moral value. Aristotle drew a similar distinction between poiesis and praxis,

Defining Technology and the Engineering Sciences

45

the former interpreted as production that is properly subordinated to the latter understood as moral action (Nicomachian Ethics VI, 4;1140a2). In effect, suggests Cuomo, such claims about techne justified a rejection of the idea that technicians either could or should exercise “control of their own activities” [Cuomo, 2007, p. 30] and became the font of long tradition of ethical and political concern about technics. In the 20th century, however, this tradition was given new life in response to the rise of an effort at the rationalization of technics and other forms of practice in industrial manufacture, consumer economics, behavioral psychology, business management, and totalitarian politics. In each case, the claim was that forms of manipulation and control, which had once depended on personal experience and insight, could be replaced and enhanced by their more conscious and systematic development. This effort was criticized by Jacques Ellul [1954], for instance, as an effort to turn everyone into technicians who at the same time failed to recognize the extent to which technique is inherently unable fully to master its own consequences. For Ellul, there are always unintended consequences to human action, even and perhaps especially when action becomes enlarged in power and scope through technology. Exploring this tradition of distinctly modern humanities criticism, Joseph Dunne places Aristotle’s analysis of techne and phronesis into dialogue with John Henry Newman, R. G. Collingwood, Hannah Arendt, Hans-Georg Gadamer, and J¨ urgen Habermas, all of whom have tried to rethink the nature of practical activity in the face of pressures toward technization. Stimulated especially by efforts to transform pedagogy into an efficient technique, Dunne’s study questioned “the attainability of technical mastery” in many areas of life by arguing that “practical knowledge . . . [is] a fruit which can grow only in the soil of a person’s experience and character” [Dunne, 1993, p. 358]. Despite this tradition of concern about what is now called technology, within humanities discourse there is a curious paradox with regard to the English term. In Dunne, for instance, it does not even occur in the index. Indeed, from the early 19th century forward, poetry and fiction — and especially modernist poetry and fiction — increasingly dealt with modern forms of technics, yet there were few if any significant poems, short stories, plays, or novels in which “technology” appeared in title or text. One can find closely related terms such as “machine,” “industrialization,” “invention,” and “applied science,” along with particular technologies (such as trains and bridges) in titles and many texts. The whole genre of science fiction might more accurately be labeled “technology fiction.” Even beyond that mislabeled genre, imaginative literature dealing with technology has favored terms such as “science” or “machine”; think of Mary Shelley’s Frankenstein (1818), Samuel Butler’s Erewhon (1871), Mark Twain’s A Connecticut Yankee in King Arthur’s Court (1889), Henry Adams’ The Education of Henry Adams (1901), Aldous Huxley’s Brave New World (1932), and Kurt Vonnegut’s Player Piano (1952). In each case, although the focus of interest was the problematic relations between advanced technical artifacts and human affairs, the term “technology” is conspicuous by its absence. On the one hand, the absence of the term

46

Carl Mitcham and Eric Schatzberg

even in technology-themed fiction of the 1950s supports the idea that “technology” only became a keyword relatively recently. On the other, as a term covering general phenomena, “technology” apparently lacked the specificity required by literature, even when that literature dealt with some aspect of industrialized making and using. Appropriately enough, then, the word “technology” made its initial sustained appearance in the humanities in the 1960s via works of literary criticism — criticism that called attention to a second curious and perhaps ironic feature: Literature, especially modernist literature, manifested a paradoxical attitude toward technology. While such literature more likely than not criticized technological culture, it also often implicitly integrated technology into culture. Thomas Reed West’s Flesh of Steel: Literature and the Machine in American Culture, for instance, documented “a dominant attitude toward the machine and its disciplines [as] one of repudiation” [West, 1967, p. 133]. Wylie Sypher’s Literature and Technology: The Alien Vision [1968], by contrast, demonstrated a modernist obsession with a technological aesthetic, and Cecelia Tichi’s Shifting Gears: Technology, Literature, Culture in Modernist America similarly explored the extent to which “the culture of the gear-and-girder technology was a collaborative effort of the engineer, the architect, the fiction writer, and the poet” [Tichi, 1987, p. 16]. Two other classics of literary criticism further explored related issues: Leo Marx’s The Machine in the Garden: Technology and the Pastoral Ideal in America [1964] and Herbert Sussman’s Victorians and the Machine: The Literary Response to Technology [1968]. Note how even in both Marx and Sussman the primary term was “machine,” the secondary one “technology.” Only in subsequent decades did humanities studies increasingly foreground the term “technology” while occasionally moving a particular technology into the subtitle — as in, for instance, Carolyn Marvin’s When Old Technologies Were New: Thinking about Electric Communcation in the Late Nineteenth Century [1988], Tim Armstrong’s Modernism, Technology and the Body: A Cultural Study [1998], or Pamela Thurschwell’s Literature, Technology and Magical Thinking, 1880—1920 [2001]. (See also [Greenberg and Schachterle, 1992].) In her World War I novel, One of Ours (1922), Willa Cather described the reaction of newspaper readers in the central United States to reports of the fall of the forts at Li`ege, which were reduced in a few hours by siege guns. . . which evidently could destroy any fortifications that ever had been, or ever could be constructed. Even to these quiet wheat-growing people, the siege guns before Li`ege were a menace; not to their safety or their goods, but to their comfortable, established way of thinking. They introduced the greater-thanman force which afterward repeatedly brought into this war the effect of unforeseeable natural disaster, like tidal waves, earthquakes, or the eruption of volcanoes.

Defining Technology and the Engineering Sciences

47

An “unprecedented power of destruction had broken loose in the world” for which “none of the well-worn words descriptive of human behavior seemed adequate” (Cather, One of Ours, Book II, chapter 9, paragraphs 2 and 3). The emergence “technology” to fill the terminological gap noted by Cather can be documented in two manifestations of synthetic humanities scholarship: libraries and encyclopedias. With regard to libraries, note how in 1876 Melvil Dewey’s decimal classification system grouping number of the 600s “Useful Arts” (which included medicine, engineering, agriculture, domestic economy, ordnance, etc.) was sandwiched between the 500s “Natural Sciences and Mathematics” and the 700s “The Arts.” Originally “technology” only occurred in the compound “chemical technology,” as one of several useful arts along with pyrotechnics, “wines, liquors and ales,” and metallurgy. In other words, “technology” was not a central conceptual category in 19th century English [Dewey, 1876, pp. 18-20]. “Useful Arts” remained one of the nine fundamental categories in the Dewey system until after World War II, when it was replaced first by “Applied Science” and then in 1958 by “Technology.” The same year witnessed the change of name of the H. W. Wilson Company’s Industrial Arts Index of technical literature (created in 1913) to the Applied Science and Technology Index. By contrast, the original U.S. Library of Congress classification system, which began to be conceptualized in 1897 and was, like other reference frameworks, quite conservative in its principles, had by 1904 already made “technology” one of its 20 top-level categories. But this was not true from the beginning. Rathar than “technology,” a 1901 draft grouped together “Useful Arts, Agriculture, [and] Manufactures” in one category [Miksa, 1984, p. 24]. With development of the LC system, “technology” was not only stabilized as a primary category that included all forms of engineering, engineering science, manufactures, handicrafts, and home economics, but also became distributed in subject headings across any number of other categories, from philosophy to social sciences and fine arts. Although these subject headings appeared as additions to the original structure, it is nevertheless significant that “technology” became salted throughout in so many classificatory combinations. With regard to encyclopedias, consider as a representative case the Encyclopaedia Britannica: A Dictionary of Arts, Sciences, Literature and General Information, the first edition of which was published in 1771 in three volumes in pale imitation of the great French Encyclop´edie (1751-1772). In none of the first ten editions does the word “technology” even appear. Because the term was indeed extant in English, the failure of the editors to include it demonstrates their judgment of its relative unimportance. “Technology” finally appeared in the classic 11th edition (1910-1911), but just twice in subsidiary roles. Only in the reconfigured 15th edition (1974 et seq.) does “technology” become a major theme — a process that takes place in conjunction with a generally intensified appreciation of its presence [see Oldenziel, 1999]. When “technology” arrived on the encyclopedic stage, it did so in a big way, and no doubt under the influence of developments in other fields of the humanities.

48

Carl Mitcham and Eric Schatzberg

Under tutelage of neoaristotelian philosopher Mortimer Adler, the 15th edition included what was termed a Propaedia, which classified knowledge into ten subject areas: (1) matter and energy, that is, the physical sciences, and (2) the earth, moving on through (3) the sciences of living things, then (4) human life, and finally (5) human society, (6) art, (7) technology, (8) religion, (9) history, and (10) knowledge itself, including science and philosophy. Although its proximity to art is revealing, technology was easily the most anomalous of these major categories; it was, for instance, the only one that did not appear at all in the first edition of the Britannica. (It was also the only category not closely linked to one of Adler’s 101 great ideas, as surveyed in The Great Ideas: A Syntopicon of Great Books of the Western World of 1952.) As part seven, technology was approached from three main perspectives: its historical development and social impact (particularly on work), its internal divisions (energy conversion, tools, measurement and control, extraction of raw materials, industrial production), and its major fields of application (agriculture, industrial production, construction, transportation, information processing, the military, the city, earth and space exploration). In a kind of echo, “The Technological Sciences” were considered again in part ten as a seventh and last subdivision, with a four-part analysis in terms of history, professional branches (civil, aeronautical, chemical, electrical, mechanical, etc., engineering), agricultural sciences, and interdisciplinary technological sciences (bionics, systems engineering, cybernetics). One of the closely related areas of activity in the humanities that surely influenced this encyclopedic scheme had been taking place in conjunction with a series of international expositions or world fairs that began in 1851 with the Crystal Palace in Hyde Park, London, under the title “Great Exhibition of the Works of Industry of All Nations.” (The 1933-1934 Chicago World’s Fair had the famous motto, “Science Finds, Industry Applies, Man Conforms” — implicitly conceiving of industry as applied science.) It was in conjunction with the 15th exhibition, held in 1900 in Paris (in response to which The Education of Henry Adams included a chapter on “The Dynamo and the Virgin”), that a proposal developed to create a parallel intellectual inventory of the types of knowledge being exhibited. This Congress of the Arts and Science took place in association with the St. Louis exhibition of 1904 and resulted in an eight-volume assessment of the state of knowledge across seven broad areas [Rogers, 1905-1906]: normative sciences, historical sciences, physical sciences, mental sciences, utilitarian sciences, regulative sciences, and cultural sciences. The fifth of these areas, utilitarian sciences, was composed of three divisions: medical, economic, and technological sciences. The technological sciences in turn included such disciplines as civil engineering, mechanical engineering, electrical engineering, and more. (It is noteworthy that in volume one, which provided a conceptual overview, these disciplines were titled mechanical technology, electrical technology, etc., whereas when reviewed at length in volume six they were denominated as branches of engineering.) There was thus emerging during the same period as Cather’s novel a new term broader than machines to refer to an “unprecedented power” — creative as well

Defining Technology and the Engineering Sciences

49

as destructive — that “had broken loose in the world” and for which “none of the well-worn words . . . seemed adequate.” The need in the humanities for a new term, less particular than machines but inclusive thereof, has also been analyzed by L. Marx [1997]. For Marx the need was two-fold: ideological and substantive. Ideologically, new ideas about progress, in the 1800s so closely associated with discovery and invention, increasingly called for a more expansive and abstract objective correlative. Substantively, the emergence of systems or networks of machines also demanded a more abstract and general terms than “machines” or “industry” to reference phenomena such as linkages of canals, railroads, telegraphs, radio, and more — which were themselves becoming interconnected. The idea of tools or machines or, for that matter, any other material artifacts did not begin to convey the complex, quasiscientific, corporate character of the new sociotechnical formations that emerged at that time. The curious fact is that the discursive triumph of the concept of technology is in large measure attributable to its vague, intangible, indeterminate character — the fact that it does not refer to anything as specific or tangible as a tool or machine. [Marx, 1997, p. 981] In the humanities, then, during the last half of the 20th century, “technology” increasingly began to reference the complex of industries, industrial products, and technical infrastructures that were variously enmeshed with but also distinct from modern science. Humanities scholars struggled to identify appropriate language to reference these new forms of making and using that increasingly transformed not just agriculture, industry, transportation, and communication, but elements of higher culture as well. Looking back, scholars such as Lewis Mumford argued for greater exploration of the interactions between Technics and Civilization [1934], pointing out how the mechanical clock transformed experiences of time and the printing press literacy. Looking forward, scholars such as Walter Benjamin (in an essay written in 1935-1936 but not published until after his death) analyzed the interaction of technology and culture in “Das Kunstwerk im Zeitalter seiner technischen Reproduzierbarkeit” [1961]. The original English title for this influential essay was “The Work of Art in the Age of Mechanical Reproduction” — rendering the German technischen as “mechanical” rather than the more literal “technical.” Had Benjamin’s concerns been articulated during the last third of the 20th century, they would have considered a manifold of digital reproduction and image manipulation that exceeded anything merely technical or mechanical, so that the pivotal English adjective would have been “technological.” Likewise, by the last quarter of the century, in fictions as different as Thomas Pynchon’s Gravity’s Rainbow (1973) and Robert Pirsig’s Zen and the Art of Motorcycle Maintenance: An Inquiry into Values (1974), “technology” had become a term of sufficiently definite reference — and seemed to include not only industrial making activities and their social or institutional extensions but the knowledge involved with such activities, the products both physical and organizational, and the use of such products as well as the diverse intentions and motivations bound

50

Carl Mitcham and Eric Schatzberg

up therewith. Technology had become a way of life, a form of consciousness, an attitude toward the world — although Don Delillo’s White Noise (1985) continued to eschew the term in favor of a plethora of more precisely imagined objective correlatives such as toxic events, televisions, traffic, and drugs. Nevertheless, late in his life, Dewey acknowledged the rise of this new term by suggesting a preference for “technology” over “instrumentalism” for his analysis of human experience: “It is probable that I might have avoided a considerable amount of misunderstanding if I had systematically used ‘technology’ instead of ‘instrumentalism’ in connection with the view I put forth regarding the distinctive quality of science as knowledge” [Dewey, 1944, p. 285n]. Dewey’s connection of “technology” with “instrumentalism” suggests a link between the concept of “technique” in continental European languages and the English “technology.” A number of humanities scholars have apparently sought to preserve the continental emphasis by embracing the term “technics.” Across a long and productive life, Mumford, a literary and social critic, always preferred the term “technics” to “technology.” (For an extended defense of Mumford’s usage, see Fores [1981].) In some early work so did the political theorist Langdon Winner and the philosopher Don Ihde. Winner merged the two terms in Autonomous Technology: Technics-Out-of-Control as a Theme in Political Thought [1977]. Ihde maintained a subtle distinction in Technics and Praxis: A Philosophy of Technology [1979], where “technics” was identified with the instrumental use, but then shifted in Lifeworld and Technology: From Garden to Earth [1990] to “technology” as an inclusive term. To some extent, then, English usage of technics instead of technology appears to be a distinction without a difference. Nevertheless, it is reasonable to postulate “technics” as less expansive than “technology,” which is typically perceived in the humanities as having become a distinctive way of being in the world, a form of consciousness, and thus — like religious belief or democracy — a phenomenon deserving of expressive critical assessment and careful reflective analysis. Philosophy of technology, arising within the context of this humanities community of discourse, thus initially took on a decidedly moral tone — often censorious or negative, sometimes celebratory and affirmative, always parsing and distinguishing. 5

IN THE SOCIAL SCIENCES

As a distinct area of knowledge, the social sciences emerged during the 19th century, based on an Enlightenment faith in the applicability of reason to human affairs, but refracted through the lenses of the Industrial and French Revolutions. Only in the early 20th century, however, did the social sciences become securely institutionalized as a stable set of disciplines. Their complexities and contested origins played out in shifting constellations of economics, sociology, anthropology, and political science, as well as in debates about the nature of social science research, theory construction, relations to the natural sciences and to the humanities, and relevance to practice and policy. Throughout all these originating struggles,

Defining Technology and the Engineering Sciences

51

there existed an effort to bring science to bear on concerns traditionally associated with the humanities, reflecting a desire to be more directly relevant to human interests than the physical sciences, yet with a methodological effectiveness that transcended the often apparently impotent rhetoric typical of the humanities. Yet just as with the humanities, there is no canonical social science understanding of or approach to technology, particularly across disciplines. The social sciences have nevertheless directed more sustained attention to the definition of technology than any other area of scholarly inquiry. In general, “technology” has a broader meaning in the social sciences than it does in engineering, but one that is more determinate than in the humanities. To some extent, the social-science usage includes both the expansive and the restrictive connotations found in engineering, while including a critical stance typical of the humanities. It emphasizes not so much the study of any internal features of technical processes or engineering sciences but of those processes and their sciences as socially effective givens, with an interest directed toward how such processes and forms of knowledge arise from and influence society. The first sustained use of a “technology” cognate to cover industrial processes, Beckmann’s Technologie, occurred (as already noted) in the context of an early form of German social science known as cameralism. Indeed, questions posed by technological change, in particular the social dislocations associated with industrialization, provided much of the early focus in the social sciences. In the 19th century, the “machinery question” was central to British political economy and social reform movements, as Maxine Berg [1980] has argued. From Smith’s analysis of pin manufacture to K. Marx’s dissection of the factory, political economists incorporated industrial technology into their analyses. Yet for all the attention paid to industry, 19th century social sciences rarely made technology itself the focus of analysis. As Frison [1998] has observed, the terms “technology” and “technique” are largely absent from the treatises by Smith, J. S. Mill, and other key works of classical British political economy. Furthermore, these authors show little evidence of having had a general concept of technology and technological change. For example, one key debate in the early 19th century arose over claims by Thomas Malthus that diminishing marginal productivity of land would inevitably produce stagnation. David Ricardo, among others, contested Malthus by arguing that the tendency toward stagnation was “more than counterbalanced by the improvements in machinery, by the better division of labor, and by the increasing skill, both in science and art, of the producers” (On the Principles of Political Economy and Taxation, ch. 5, paragraph 4). Present-day economists would, of course, make the same argument in terms of technological change, but “technology” did not become a key term in economic theory until the 1930s. The most sustained discussion of industrial technology in classical political economy is found in Marx’s Das Kapital, volume one, especially the chapter on “Machinery and Large-scale Industry.” Even Marx, however, did not deploy a general concept of technology, although he combined cameralist Technologie with a focus

52

Carl Mitcham and Eric Schatzberg

on machinery drawn from British political economy. Marx’s most general description of Technologie occurs in a lengthy footnote to his discussion of the concept of the machine. This note began by calling for “a critical history of technology” that would demonstrate the collective basis of invention. It then constructed an analogy to Charles Darwin’s interest in the history of “natural technology, i.e., . . . the formation of plant and animal organs as productive instruments for the lives of plants and animals.” Marx proposed a similar “developmental history of the productive organs of social humans,” which led to one of his most well-known statements: “Technologie discloses [enth¨ ullt] the active relations of humans with nature, the immediate production processes of their lives, and thereby also their social life relations and the mental presentations that flow from them” [Marx, Kapital I, ch. xiii, n4]. One can certainly interpret this use of “technology” as a direct reference to material processes of production, a usage comparable to meanings. Yet elsewhere in Das Kapital Marx made it clear that he interpreted Technologie along cameralist lines as a Wissenschaft concerned with directing the process of production. In the note Marx was apparently conceptualizing Technologie as a potential (social) science of production that could help reveal the material basis of social relations and ideologies. Although this view was not developed further by Marx or his followers, subsequent German social scientists were influenced by Marx’s insistence that the technical process of production was central to human history. Much of the work of Max Weber and Werner Sombart, for example, can be read as a response to the perceived determinism of Marxist theory. Yet when these scholars addressed Marx’s analysis of production, they did so using the term Technik rather than Technologie. Technik entered German social science through the discourse generated by German engineers in the late 19th century, as the phenomenon it referenced was attended to by the work of Gustav Schmoller, Weber, Georg Simmel, and especially Sombart. In the first edition of his Der Moderne Kapitalismus [1902], Sombart discussed at length die neuen Technik (modern technology). He expanded on this theme the following year with an analysis of Technik in 19th century Germany, stressing the shift from empirical to scientific Technik. Sombart and Weber elaborated their views in 1910 at the first conference of the Deutsche Gesellschaft f¨ ur Soziologie (German Society for Sociology), where Sombart presented a paper on “Technik und Kultur.” Both Sombart and Weber rejected Technik as an independent variable in human affairs, instead insisting that Geist (spirit or mind) played a more fundamental, causal role. Weber briefly discussed Technik in his famous analysis of types of action in Wirtschaft und Gesellschaft [1914], where he drew a sharp distinction between Technik and Wirtschaft (or technology and economy). Weber restricted Technik to the question of the optimal means to a given end, arguing that any consideration of costs implied choices among ends; such choices, for Weber, belonged to the sphere of economic action. This understanding in effect removed Technik from the domain of culture, reducing it to an almost mechanical application of

Defining Technology and the Engineering Sciences

53

scientific principles. In France between the two world wars a number of social scientists — among them Marcel Mauss, Ignace Meyerson, And´e Leroi-Gourham, and Lucien Febvre — developed a similar discourse about la technique. In comparison to the Germans, this French discourse focused more on anthropology, social history, and workers ([Long, 2005]; see also [Salmon, 1984]). The English term “technology” became significant in American social sciences in the early 20th century largely through its use to translate the discourse of “technique” from continental European languages. Before 1900, “technology” remained of marginal significance in the social sciences, used mainly as a classificatory heading. For example, the 1882 constitution of the Anthropological Society of Washington made “Technology,” defined as “the science of the arts,” one of the four main sections of the society, alongside Somatology, Ethnology, and Philology. In this same time period, however, English-speaking social scientists began to take notice of the German discourse of Technik, especially in the United States, where German universities had become models for higher education. The most suitable translation the central term in German social-science discourse on Technik was “industrial arts,” a phrase that in the process of being appropriated for programs of manual instruction in American secondary schools, U.S. social scientists gravitated instead toward “technique” or “technology” or occasionally “technics” (e.g., [Seligman, 1902]). The result was a terminological muddle. One scholar who played a pre-eminent role in transposing the German discourse into the American context was the iconoclastic social scientist Thorstein Veblen. Veblen took the concept of Technik directly from the works of Schmoller and Sombart and merged it into the existing meaning of “technology” as the science of the industrial arts, thus creating a new concept that transcended received meanings. He frequently equated technology with the “state of the industrial arts,” drawing a parallel with the “state of the art,” a concept used to determine priority in American patent law. In his hands technology came to be understood as a universal characteristic of human cultures and an alternative to the older idea of art, which by that time had become thoroughly aestheticized. Veblen [1906] provided the first explicit discussion of the relationship between technology and science, one that posited the two as fundamentally distinct spheres of human culture linked through a materialist sociology of knowledge. Veblen [1908] undertook the first significant analysis of the economic role of technology, conceptualized as productive knowledge and skills belonging collectively to an entire community. Given its novelty, Veblen’s concept of technology was perhaps a bit too subtle for the time. Although his works were widely read, he had few students and no true disciples. A number of influential American social scientists adopted his usage, but some of the finer points fell away as the term became more common. In particular, a dialectic understanding of the science-technology relationship was replaced by the assumption that technology equaled the application of science and technology, which became firmly linked to a dominant belief in material progress. Such meanings were strongly present in the work of two leading American social scientists who embraced the term before World War II, the historian and political scientist

54

Carl Mitcham and Eric Schatzberg

Charles A. Beard [1927] and the sociologist William F. Ogburn [1938]. A fairly sophisticated concept of technology did find a home in institutional economics, a dissenting branch of the field that drew on Veblen’s ideas, but the dominant neoclassical paradigm generally viewed technology as exogenous to the economic system [Lower, 1987]. Yet the German discourse on Technik continued to affect American social science, as can be seen in the work of Talcott Parsons, one of the pre-eminent American social theorist of the 20th century. Parsons earned his doctorate from the University of Heidelberg in the late 1920s and in the process absorbed many ideas from European social theorists such as Weber and Sombart. His early publications examined recent German writings on capitalism, many of which dealt significantly with Technik, especially Sombart’s. Parsons’ earliest discussions of technology were framed primarily in terms of “technique,” often translated directly from the German. But he gradually altered his terminology from “technique” to “technology” while importing into English the two major German meanings of Technik as practical arts and as means to an end (e.g., [Parsons, 1935]). His sociologist colleague at Harvard, Robert K. Merton, underwent a similar German-influenced transition from “technique” to “technology” in the late 1930s [Merton, 1935]. This view identified technology primarily with productive machinery, as in the sevenpage entry in the Encyclopaedia of the Social Sciences of 1934, which lacked a clear definition and remained heavily dependent on references to German and French literature (see [Lederer, 1934]). After World War II, “technology” assumed a secure if somewhat marginal place in anthropology, economics, and sociology. The definition of technology as applied science, which was common in postwar science and engineering, had much less currency in the social sciences. Social scientists generally embraced a broad understanding of technology. This broad definition received explicit sanction in the authoritative 17-volume International Encyclopedia of the Social Sciences. The relevant entry, which was twice the length of its predecessor, offered the following explicit definition: Technology in its broad meaning connotes the practical arts. These arts range from hunting, fishing, gathering, agriculture, animal husbandry, and mining through manufacturing, construction, transportation, provision of food, power, heat, light, etc., to means of communication, medicine, and military technology. Technologies are bodies of skills, knowledge, and procedures for making, using and doing useful things. They are techniques, means for accomplishing recognized purposes [Merrill, 1968, pp. 576-577]. Some social scientists have nevertheless continued to contest the boundaries of technology. For example, the short, three-page entry in a second edition of the International Encyclopedia of the Social Sciences retreats to a definition of technology as “the underlying production methodology through which inputs or resources are converted into output (goods and services)” along with the further

Defining Technology and the Engineering Sciences

55

claim that at any point in time “there is one best way to produce a good or service” ([Goel, 2008, p. 302]; see also [Tantoush, 2001, p. 15503]; and [Mokyr, 2008, p. 217]). Others continue to limit it to modern industry or to distinguish between “technics” and “technology,” letting the former stand for prescientific arts and crafts and the latter for more sophisticated engineering (see, e.g., in Spanish, [Medina, 1985]). More characteristic, however, has been an expansive view associated with the history of technology. In history, which tends to bridge the social sciences and the humanities, technology has, from the point at which it became a theme for narrative analysis, been defined broadly. According to the preface to the first comprehensive internalist history, technology would be construed as “how things are commonly done or made,” including “what things are done or made” [Singer et al., 1954, p. vii]. This view was modestly criticized by the editors of a subsequent social history as “so broad and loose [as to encompass] many items that scarcely can be considered as technology” [Kranzberg and Prusell, 1967, p. 5]. Yet a philosophical comparison of definitions in these and basic histories of technology in French and other languages, including various historiological debates, continued to favor an inclusive over a more restricted definition [Mitcham, 1979]. Indeed, one historian and management theorist even defended the idea that technology should include not just “how things are done or made” but “how [the human being] does or makes,” interpreted not so much in terms of human nature trying to control the environment as a cultural extension of the processes of biological evolution oriented toward transcending human limitations [Druker, 1959, p. 28] From this perspective, technology includes not only successful but failed human undertakings, insofar as they are oriented (consciously or not) toward making and using — so that the history of technology includes a history of work, invention, economics, politics, science, and so forth. This is also an approach characteristic of socially oriented philosophers such as Larry Hickman who, building of the philosophy of Dewey, argues for naturalizing technology “as a cognitive activity within the evolutionary history of complex organisms” [Hickman, 2001, p. 21]. 6 CONCLUSION AND IMPLICATIONS By the last quarter of the 20th century, “technology” had become a well established keyword, although one influential treatment of 131 such terms placed it in the bottom rung with regard to analytic attention space [Williams, 1983]. At the same time, the meanings of technology varied significantly across communities of discourse, as revealed by the preceding review of understandings and definitions in science and engineering, in the humanities, and in the social sciences. Such variations suggest the need for context-dependent approaches to the philosophy of technology as understood within each of the three basic scholarly communities. Within and across each community, one definition or definitional strategy may well be more appropriate than another, resulting in complementary philosophies of technology and the engineering sciences.

56

Carl Mitcham and Eric Schatzberg

With regard to the possibility of complementary philosophies, two points can be argued. First, although phenomena can be parsed in an indefinite number of ways, sorting out, classifying, and defining nevertheless remain fundamental to human acting and thinking. A context dependency of definitions does not mean they can be dispensed with or treated with indifference. Metaphysically, definitional contingency can be interpreted as disclosing the extent to which all presence is limited or partial, but not for that reason unreal (see [Harman, 2005]). Epistemologically, to define and classify are inherently human activities. Second, any act of classifying and sorting has manifold implications, from the practical and cognitive to moral and aesthetic. To paraphrase one argument relative to this point, we live in, on, and around definitions, which reflect and shape our practical, epistemic, ethical, and aesthetic choices and experiences. Both epistemologically and morally, it is important “to produce flexible classifications whose users are aware of their political and organizational dimensions and which explicitly retain traces of their constructions” [Bowker and Star, 1999, p. 326]. Comparison has already been made to the pursuit of definitions in other regionalizations of philosophy such as the philosophy of science, of religion, and of language. In each case, context functions to help specify definitional strategies, which in turn set the stage for regionally distinctive philosophical discussions. In the first case, especially in the view of scientists themselves, science is commonly argued to be distinguished by its method. This argument is based on a self-understanding within the rather tightly coupled scientific community about the importance of method, and the fact that science aspires to be a progressively unfolding cognitive activity. Scientists themselves often see the scientific method as essential and prescriptive, a view endorsed by many philosophers. For example, as interpreted by Karl Popper [1963], science involves the advancement of knowledge claims that are able (in principle if not in current practice) to be falsified by some test and have in fact survived at least one such test (although they may in the future fail to satisfy other tests). Yet debate about the adequacy of any one description of the scientific method has become a basic theme in the philosophy of science, while social scientists have even challenged the self-understanding among scientists (and some philosophers) of science as a method-based community (see [Pinch and Collins, 1993]). In a manner like scientists, some engineers have proposed to define engineering in terms of method [Koen, 2003], leading again to discussions about the adequacy of various proposals — as well as to social science questioning of the extent to which engineering lives up to its own methological ideals. Additionally, there are reasons to question whether the engineering community is as tightly knit or well formed as that of science. In the philosophy of religion, by contrast, there is no equivalent to the scientific or engineering communities. The many religions that scholars so designate fail to see themselves as a well formed or unified community. Within each religion there are groups such as Catholic Christians or Theravada Buddhists that appear more or less well established, but Christians and Buddhists only in a weak sense see themselves as engaged in a common enterprise or dialogue. Indeed, the first

Defining Technology and the Engineering Sciences

57

“Parliament of Religions” did not take place until 1893 in Chicago, and has been much less an on-going enterprise than the transnational “republic of science” that emerged during the 1700s and achieved a relatively stable professional identity during the following century [Harrison, 2008]. As a result, philosophers have been forced to define religion more in terms of a selective set of possible key characteristics or what Wittgenstein called family resemblances than anything like a method. Taking such an approach, Ninian Smart [1998] proposed a set of seven possible features for religion: ritual practices, experience or emotion, doctrines, morality, narratives or myths, social or institutional organization, and material culture or art — all oriented toward some kind of transcendence. In any one religion the set will be present in different forms and balances. In some cases one or more features might be quite attenuated or absent. But in aggregate these dimensions serve to differentiate religious from non-religious phenomena and thus to mark out in rough form that complex which is the subject of critical reflection in the philosophy of religion. In like manner there is considerable difference between, at least, technics and technology understood as the making activities in premodern and modern form, respectively. Insofar as this is the case, it argues for a key features approach to definition, one that has in fact been adopted by a number of philosophers (e.g., [Radder, 1996]). In the philosophy of language there tends to be less debate about the definition of the subject matter than with regard to either science or religion. Because there exists a science of language, linguistics, that has defined the phenomenon, philosophy can take this as a given. There is simply not as much debate about what constitutes language as there is about what constitutes science or religion. Something close to a definition commonly used in the science of linguistics is generally accepted: language is a system of communication using a finite set of arbitrary symbols (semantics) and structuring rules (syntax) for their manipulation. Nevertheless, within linguistics there are debates about such questions as the relation between language and speech, semantics and syntax, meaning and reference, and more, all of which have become issues for philosophical scrutiny. (It is important to note, as well, that the philosophy of language as a distinct regionalization of philosophy is much more prominent in analytic than in phenomenological traditions, where it tends to be subsumed within philosophical anthropology or hermeneutics and semiotics.) Considering analogies between the philosophy of technology and the philosophy of language, is it possible to conceive of engineering or the engineering sciences as like linguistics and thus containing sufficiently well established definitions of subject matter that these could be taken as given? To some extent higher educational institutions of technology would seem to be based on assumptions about technology as constituted by sciences of application — application that differentiates into multiple levels and fields of engineering. Since whatever might be given in linguistics is taken in for scrutiny by the philosophy of language, on this analogy the philosophy of technology could be described as critical reflection on the manifold of engineering. That is, the philosophy of technology would become the

58

Carl Mitcham and Eric Schatzberg

philosophy of engineering and the engineering sciences. Another possibility would be to take technology in a more etymologically literal sense as a science of technics or technique and divide the philosophy of technology in two: the philosophy of the science of technique and the philosophy of technique. Yet the extent to which technology is sufficiently captured by either engineering or technics remains an open question. A related issue concerns whether any of these three approaches — defining technology in terms of method, a flexible set of key features, as engineering, or as technique — would fit sufficiently with common linguistic use, coordinate with questions of rhetorical influence, and work well for integrating philosophies of technology in engineering, in humanities, and in social science contexts into something that might be thought of as the philosophy of technology in general. This final implication is important. As mentioned at the beginning of the chapter, definition is closely related to a number of basic philosophical questions of more general interest. But definition, when considered at the shallow level of sorting out the phenomenon or phenomena to be named and classified, does not of itself determine in a deeper or interpretative sense. This is at least part of what Martin Heidegger [1954] meant when asserting that the essence of technology is not anything technological. In pursuing any deeper, interpretative definition — whether in philosophies of technology arising within the contexts of science and engineering, the humanities, or the social sciences — it would be appropriate to consider at least the following ten basic but non-ordered and not mutually exclusive questions: 1. Does technology have an inner or essential distinguishing feature? 2. If technology does have an essential or necessary feature or features, how might such be distinguished from accidental or contingent features? 3. What is the relation between technology and nature? 4. What is the relation between technology and human action? (Insofar as technology can be defined as a type of human action, then a host of issues in the philosophy of action, ethics, and political theory become relevant.) 5. Is technology one or many, a unity or plurality? That is, is it more accurate to speak of “technologies” than of “technology”? If a plurality, what are the best ways to understand it as such? 6. What, if any, are the “parts” of (or divisions within) technology (technologies)? 7. Is there historical continuity in the development of technology (or technologies)? 8. What is the relation between technology (technologies) and science (sciences)?

Defining Technology and the Engineering Sciences

59

9. What is the relation between technology (technologies) and engineering (engineerings)? 10. What is the relation between technology (or technologies) and other aspects of human life (culture, language, religion, art, society, politics, economics, etc.)? Possible responses to these questions will be strongly influenced by how technology is sorted out from other aspects of the world. Responses to these questions in turn will have substantive implications for other questions, such as whether or not technology is neutral, autonomous, good, beautiful, and more. Adopting the pragmatist approach to definition implies that any definition will need to be undertaken in dialogue with or through critical reflection on its implications. Definition is not something that can take place independent of context ACKNOWLEDGEMENTS This chapter has drawn on previous and forthcoming work by each co-author. Some sections from Mitcham’s Thinking through Technology (1994) have been revised or extended. Some research from Schatzberg’s work in progress on the history of the term “technology,” supported in part by National Science Foundation grant #0646788, has also been utilized. Interested readers should consult both for complementary expositions of the argument presented here. BIBLIOGRAPHY All references to classical or standard works are handled with standard page references or by means of textual divisions that are independent of particular editions; as a result such works are not listed here and references in the text are in parentheses rather than brackets. All references in brackets are to works listed here. All quotations not originally in English are in our own or highly adapted translations. [Abelson, 1967] R. Abelson. Definition. In Encyclopedia of Philosophy, Paul Edwards, ed., vol. 2, pp. 314-324. New York: Macmillan, 1967. [Armstrong, 1998] T. Armstrong. Modernism, Technology and the Body: A Cultural Study. Cambridge, UK: Cambridge University Press, 1998. [Austin, 1961] J. L. Austin. “A Plea for Excuses.” In Philosophical Papers, J.O. Urmson and G.J. Warnock, eds., pp. 123-152. Oxford: Clarendon Press, 1961. [Beakley et al., 1982] G. C. Beakley, H. W. Leach, J. K. Hedrick, and R. E. Lovell. Engineering: An Introduction to a Creative Profession. New York: Macmillan, 1982. [Beard, 1927] C. Beard. “Time, Technology, and the Creative Spirit in Political Science.” American Political Science Review, vol. 21 (February), pp. 1-11, 1927. [Beaune, 1980] J.-C. Beaune. La technologie introuvable: Recherche sur la d´ efinition et l’unit´ e ecles. Paris: J. Vrin, de la technologie a ` partir de quelques mod` eles du XVIII e et XIX e si` 1980. [Beckmann, 1777] J. Beckmann. Anleitung zur Technologie, oder zur Kentniss der Handwerke, Fabriken und Manufakturen. G¨ ottingen: Vandenhoeckschen, 1777. [Benjamin, 1968] W. Benjamin. “Das Kunstwerk im Zeitalter seiner technischen Reproduzierbarkeit,” in Illuminationen Ausgeweahlte Schriften, pp. 148-184, Frankfurt a.M.: Suhrkamp, 1961. First English version in Illuminations, trans. Harry Zohn, New York: Harcourt, Brace, and World, 1968.

60

Carl Mitcham and Eric Schatzberg

[Berg, 1980] M. Berg. The Machinery Question and the Making of Political Economy, 18151848. Cambridge: Cambridge University Press, 1980. [Bigelow, 1840] J. Bigelow. Elements of Technology: On the Application of the Sciences to the Useful Arts. Boston: Hilliard, Gray, Little and Wilkins, 1829 (Second edition, The Useful Arts: Considered in Connection with the Applications of Science, 2 vols. Boston: Marsh, Capen, Lyon, and Webb, 1840.) [Bowker and Star, 1999] G. C. Bowker and S. L. Star. Sorting Things Out: Classification and Its Consequences. Cambridge, MA: MIT Press, 1999. [Charnock, 1916] G. F. Charnock. Mechanical Technology: Being a Treatise on the Materials and Preparatory Processes of the Mechanical Industries. New York: Van Nostrand, 1916. [Collins and Pinch, 1993] H. Collins and T. Pinch. The Golem: What Everyone Should Know about Science. New York: Cambridge University Press, 1993. [Cuomo, 2007] S. Cuomo. Technology and Culture in Greek and Roman Antiquity. Cambridge: Cambridge University Press, 2007. [Davis, 1998] M. Davis. Thinking Like an Engineer: Studies in the Ethics of a Profession. New York: Oxford University Press, 1998. [Dewey, 1876] M. Dewey. A Classification and Subject Index, for Cataloguing and Arranging the Books and Pamphlets of a Library. Amherst, MA: 1876. [Dewey, 1929] J. Dewey. The Quest for Certainty. New York: Minton, Balch, 1929. [Dewey, 1944] J. Dewey. “By Nature and by Art.” Journal of Philosophy, vol. 41, pp. 281-292, 1944. [Drucker, 1959] P. F. Drucker. “Work and Tools.” Technology and Culture, vol. 1, no. 1 (Winter), pp. 28-37, 1959. [Dunne, 1993] J. Dunne. Back to the Rough Ground: “Phronesis” and “Techne” in Modern Philosophy and in Aristotle. Notre Dame, IN: University of Notre Dame Press, 1993. [Ellul, 1964] J. Ellul. La Technique, ou l’en jeu du si` ecle. Paris: Colin, 1954. English version: The Technological Society, trans. John Wilkerson. New York: Knopf, 1964. [von Engelmeyer, 1899] P. K. von Engelmeyer. “Allgemeine Fragen der Technik,” Dinglers Polytechnisches Journal (Berlin-Stuttgart), vol. 311, 21-22, 69-71, 101-103, 133-134, 149-151; vol. 312, 1-2, 65-67, 97-99, 129-130, 145-147; and vol. 313, 17-19, 65-67, 1899. [Fores, 1981] M. Fores. “Technik : Or Mumford Reconsidered.” History of Technology, vol. 6, pp. 121-137, 1981. [Friedrich, 1999] K. Friedrich. Technik, Technologie, Technikwissenschaften: Philosophische und sozialwissenschaftliche Probleme. Cottbus: Brandenburgische Technische Universit¨ at Cottbus, Fakult¨ at f¨ ur Mathematik, Naturwissenschaften und Informatik, 1999. [Frison, 1998] G. Frison. “Some German and Austrian Ideas on Technologie and Technik between the End of the Eighteenth Century and the Beginning of the Twentieth.” History of Economic Ideas, vol. 6, pp. 107-133, 1998. [Greenberg and Schachterle, 1992] M. L. Greenberg and L. Schachterle, eds. Literature and Technology. Bethlehem, PA: Lehigh University Press, 1992. [Goel, 2008] R. K. Goel. “Technology.” In International Encyclopedia of the Social Sciences, second edition, William A. Darity Jr., ed., vol. 8, pp. 302-305. Detroit: Macmillan Reference, 2008. [Guillerme, 1985] J. Guillerme. “Technologie.” In Encyclopaedia Universalis, vol. 17, pp. 764767, 1985. (?) [Guillerme and Sebestik, 1966] J. Guillerme and J. Sebestik. “Les commencements de la technologie.” Thal` es, vol. 12, pp. 1-72, 1966. [Hansson, 2007] S. O. Hansson. “What Is Technological Science?” Studies in History and Philosophy of Science, Part A, vol. 38, pp. 523-527, 2007. [Harman, 2005] G. Harman. Guerrilla Metaphysics: Phenomenology and the Carpentry of Things. Chicago: Open Court Press, 2005. [Heidegger, 1954] M. Heidegger. “Die Frage nach der Technik.” In Vortr¨ age und Augs¨ atze. Pfullingen: Neske, 1954. [Herf, 1984] J. Herf. Reactionary Modernism: Technology, Culture, and Politics in Weimar Germany and the Third Reich. New York: Cambridge University Press, 1984. [Heyde, 1963] J. E. Heyde. “Zur Geschichte des Wortes ‘Technik’.” Humanismus und Technik, vol. 9, no. 1, pp. 25-43, 1963. [Hickman, 2001] L. Hickman. Philosophical Tools for Technological Culture: Putting Pragmatism to Work. Bloomington, IN: Indiana University Press, 2001.

Defining Technology and the Engineering Sciences

61

[Hollister-Short, 1977] G. Hollister-Short. “The Vocabulary of Technology.” History of Technology, vol. 2, pp. 125-155, 1977. [Ihde, 1979] D. Ihde. Technics and Praxis: A Philosophy of Technology. Boston: Kluwer, 1979. [Ihde, 1990] D. Ihde. Lifeworld and Technology: From Garden to Earth. Bloomington, IN: Indiana University Press, 1990. [International Library of Technology, 1897] International Library of Technology. Strength of Materials; Applied Mechanics; Machine Design. Scranton, PA: International Textbook Company, 1897. [Kapp, 1877] E. Kapp. Grundlinien einer Philosophie der Technik: Zur entstehungsgeschichte der Cultur aus neuen gesichtspunkten. Braunschweig: Westermann, 1877. [Kemper, 1993] J. D. Kemper. Introduction to the Engineering Profession. Second edition. Ft. Worth, TX: Saunders, 1993. [Kline, 1995] R. Kline. “Constructing >Technology= as ‘Applied Science’: Public Rhetoric of Scientists and Engineers in the United States, 1880-1945.” Isis, vol. 86, pp. 194-221, 1995. [Koen, 2003] B. V. Koen. Discussion of the Method: Conducting the Engineer’s Approach to Problem Solving. New York: Oxford University Press, 2003. [K¨ onig, 1995] W. K¨ onig. Technikwissenschaften: Entstehung der Elektrotechnik aus Industrie und Wissenschaft zwischen 1880 und 1914. Chur, Switzerland: Gordon and Breach Verlag Fakultas, 1995. [Kranzberg and Pursell, 1967] M. Kranzberg and C. W. Pursell, Jr. “The Importance of Technology in Human Affairs.” In Technology in Western Civilization, Melvin Kranzberg and Carroll W. Pursell, Jr., eds., vol. 1, pp. 3-11. New York: Oxford University Press, 1967. [Layton, 1971] E. Layton. Mirror-Image Twins: The Communities of Science and Technology in 19th Century America, Technology and Culture, vol. 12, no. 4 (October), pp. 562-580, 1971. [Lederer, 1934] E. Lederer. “Technology.” In International Encyclopedia of the Social Sciences, Edwin R.A. Seligman and Alvin Johnson, eds., vol. 14, pp. 553-560. New York: Macmillan, 1934. [Lenk and Moser, 1973] H. Lenk and S. Moser, eds. Techne, Technik, Technologie: Philosophische Perspektiven. 3 vols. Pullach bei Munich: Dokumentation, 1973. [Lindenfeld, 1997] D. F. Lindenfeld. The Practical Imagination: The German Sciences of State in the Nineteenth Century. Chicago: University of Chicago Press, 1997. [Long, 2005] P. A. Long. “The Annales and the History of Technology.” Technology and Culture, vol. 40, no. 1 (January), pp. 177-186, 2005. [Lower, 1987] M. D. Lower. “The Concept of Technology within the Institutionalist Perspective.” Journal of Economic Issues, vol. 21, pp. 1147-1176, 1987. [Lyotard, 1984] F. Lyotard. La condition postmoderne. Paris: Editions de Minuit, 1979. (English version: The Postmodern Condition: A Report on Knowledge, Geoff Bennington and Brian Massumi, trans. Minneapolis, MN: University of Minnesota Press, 1984.) [Margolis and Laurence, 1999] E. Margolis and S. Laurence. Concepts: Core Readings. Cambridge, MA: MIT Press, 1999. [Marvin, 1988] C. Marvin. When Old Technologies Were New: Thinking about Electric Communcation in the Late Nineteenth Century. New York: Oxford University Press, 1988. [Marx, 1964] L. Marx. The Machine in the Garden: Technology and the Pastoral Ideal in America. New York: Oxford University Press, 1964. [Marx, 1997] L. Marx. “Technology: The Emergence of a Hazardous Concept.” Social Research, vol. 64, no. 3 (Fall), pp. 965-988, 1997. [McDonald, 1914] H. McDonald. [Presidential] Address at the Annual Convention, in Baltimore, MD., June 2D, 1914. American Society of Civil Engineers Transactions, vol. 77, pp. 17371759, 1914. [Medina, 1985] M. Medina. De la techne a la tecnolog´ıa. Barcelona: 5Tirant Lo Blanch, 1985. [Merrill, 1968] R. S. Merrill. “The Study of Technology.” In International Encyclopedia of the Social Sciences, David L. Sills, ed., vol. 15, pp. 576-589. New York: Macmillan, 1968. [Mertens, 2002] J. Mertens. “Technology as the Science of the Industrial Arts: Louis-S´ebastien Lenormand (1757-1837) and the Popularization of Technology.” History and Technology, vol. 18, no. 3, pp. 203-231, 2002. [Merton, 1935] R. K. Merton. “Science and Military Technique.” Scientific Monthly, vol. 41 (December), pp. 542-545, 1935.

62

Carl Mitcham and Eric Schatzberg

[Miksa, 1984] F. Miksa. The Development of Classification at the Library of Congress, Occasional Papers, University of Illinois, Graduate School of Library and Information Science, no. 164. University of Illinois, 1984. [Mitcham, 1979] C. Mitcham. “Philosophy and the History of Technology.” In The History and Philosophy of Technology, George Bugliarello and Dean B. Doner, eds., pp. 163-201. Urbana: University of Illinois Press, 1979. [Mitcham, 1994] C. Mitcham. Thinking through Technology: The Path between Engineering and Philosophy. Chicago: University of Chicago Press, 1994. [Moaveni, 2005] S. Moaveni. Engineering Fundamentals: An Introduction to Engineering. Second edition. Toronto, Ontario: Nelson, 2005. [Mokyr, 2008] J. Mokyr. “Technology.” In The New Palgrave Dictionary of Economics, second edition, Steven N. Durlauf and Lawrence E. Blume, eds., vol. 8, pp. 217-223. New York: Palgrave Macmillan, 2008. [Mumford, 1934] L. Mumford. Technics and Civilization. New York: Harcourt Brace, 1934. [Ogburn, 1938] W. F. Ogburn. “Technology and Sociology.” Social Forces, vol. 17 (October), pp. 1-8, 1938. [Oldenziel, 1999] R. Oldenziel. Making Technology Masculine: Women, Men, and the Machine in America, 1880-1945. Amsterdam: Amsterdam University Press, 1999. [Oldenziel, 2006] R. Oldenziel. “Signifying Semantics for a History of Technology.” Technology and Culture, vol. 47, no. 3 (July), pp. 477-485, 2006. [Parsons, 1935] T. Parsons. “Sociological Elements in Economic Thought,” Quarterly Journal of Economics, vol. 49 (May 1935), pp. 414-453. [Petroski, 1996] H. Petroski. Invention by Design: How Engineers Get from Thought to Thing. Cambridge, MA: Harvard University Press, 1996. [Popper, 1963] K. R. Popper. Conjectures and Refultations: The Growth of Scientific Knowledge. London: Routledge and Kegan Paul, 1963. [Radder, 1996] H. Radder. In and About the World: Philosophical Studies of Science and Technology. Albany, NY: State University of New York Press, 1996. [Robinson, 1950] R. Robinson. Definition. Oxford: Clarendon Press, 1950. [Rogers, 1905–1906] H. J. Rogers, ed. Congress of Arts and Science. 8 vols. Boston: Houghton Mifflin, 1905-1906. [Rophol, 1997] G. Ropohl. “Knowledge Types in Technology.” International Journal of Technology and Design Education, vol. 7, pp. 65-72, 1997 [Ryle, 1953] G. Ryle. “Ordinary Language.” Philosophical Review, vol. 62, no. 2 (April), pp. 167-186, 1953. [Salmon, 1984] J.-J. Salmon. “What Is Technology? The Issue of Its Origins and Definitions.” History and Technology, vol. 1, no. 2, pp. 113-156, 1984. [Schadewaldt, 1960] W. Schadewaldt. “Die Begriffe ‘Natur’ und ‘Technik’ bei den Griechen.” In Natur, Technik, Kunst, pp. 35-53, G¨ ottingen: Musterschmidt, 1960. (English version: “The Concepts of Nature and Technique According to the Greeks,” Willima Carroll, Carl Mitcham, and Robert Mackey, trans., Research in Philosophy and Technology, vol. 2 (1979), pp. 159171.) [Schatzberg, 2006] E. Schatzberg. “Technik Comes to America: Changing Meanings of Technology before 1930.” Technology and Culture, vol. 47, no. 3 (July, pp. 486-512, 2006. [Sebestik, 1983] J. Sebestik. “The Rise of Technological Science.” History and Technology, vol. 1, no. 1, pp. 25-26, 1983. [Siebicke, 1968] W. Seibicke. Technik: Versuch einer Geschicte der Worffamilie um τέχνη in Deutschland vom 16.Jahrhundert bis etwa 1830. D¨ usseldorf: VDI-Verlag, 1968. [Siebicke, 1983] W. Seibicke. Duden, wie sagt man anderswo?: Landschaftliche Unterschiede im deutschen Sprachgebrauch. Mannheim: Bibliographisches Institut, 1983. [Seligman, 1902] E. R. A. Seligman. “The Economic Interpretation of History II.” Political Science Quarterly, vol. 17 (March), pp. 71-98, 1902. [Singer et al., 1954] C. Singer, E. J. Holmyard, and A. R. Hall. “Preface.” In A History of Technology, Charles Singer, E.J. Holmyard, and A.R. Hall, eds., vol. 1, pp. v-viii. New York: Oxford University Press, 1954. [Smart, 1998] N. Smart. The World’s Religions. Second edition. New York: Cambridge University Press, 1998. [Smith et al., 1983] R. J. Smith, B. R. Butler, and W. K. LeBold. Engineering as a Career, 4th edition. New York: McGraw-Hill, 1983.

Defining Technology and the Engineering Sciences

63

[Snow, 1959] C. P. Snow. The Two Cultures and the Scientific Revolution. Cambridge: Cambridge University Press, 1959. [Sombart, 1902] W. Sombart. Der Moderne Kapitalismus. 2 vols. Leipzig: Dunker und Humbolt, 1902. Revised and expanded in 1916, 1917, and 1927. [Stratton and Mannix, 2005] J. A. Stratton and L. H. Mannix. Mind and Hand: The Birth of MIT. Cambridge, MA: MIT Press, 2005. [Sussman, 1968] H. Sussman. Victorians and the Machine: The Literary Response to Technology. Cambridge, MA: Harvard University Press, 1968. [Sypher, 1968] W. Sypher. Literature and Technology: The Alien Vision. New York: Random House, 1968. [Tantoush, 2001] T. M. K. Tantoush. “Technology and Organization.” In International Encyclopedia of the Social and Behavioral Sciences, Neil J. Smelser and Paul B. Baltes, eds., vol. 23, pp. 15502-15506. Amsterdam: Elsevier, 2001. [Thurschwell, 2001] P. Thurschwell. Literature, Technology and Magical Thinking, 1880—1920. Cambridge, UK: Cambridge University Press, 2001. [Tichi, 1987] C. Tichi. Shifting Gears: Technology, Literature, Culture in Modernist America. Chapel Hill, NC: University of North Carolina Press, 1987. [Veblen, 1906] T. Veblen. “The Place of Science in Modern Civilization.” American Journal of Sociology, vol. 11 (March), pp. 585-609, 1906. [Veblen, 1908] T. Veblen. “On the Nature of Capital.” Quarterly Journal of Economics, vol. 22 (August), pp. 517-542, 1908. [Weber, 1922] M. Weber. Wirtschaft und Gesellschaft: Grundriss der verstehenden Sozialogie. T¨ ubingen: J.C.B. Mohr, 1922. [West, 1967] T. R. West. Flesh of Steel: Literature and the Machine in American Culture. Nashville, TN: Vanderbilt University Press, 1967. [Williams, 1983] R. Williams. Keywords: A Vocabulary of Culture and Society. Revised edition. New York: Oxford University Press, 1983. [Winner, 1977] L. Winner. Autonomous Technology: Technics-Out-of-Control as a Theme in Political Thought. Cambridge, MA: MIT Press, 1977. [Wright, 2002] P. H. Wright. Introduction to Engineering. Third edition. New York: John Wiley, 2002.

This page intentionally left blank

SCIENCE, TECHNOLOGY AND THE SCIENCE–TECHNOLOGY RELATIONSHIP Hans Radder

1

INTRODUCTION

This chapter discusses the principal philosophical questions concerning the relationship between science and technology. As for science, the discussion is meant to cover a variety of disciplines, even if the examples show some emphasis on the natural sciences.1 As for technology, in the present chapter this notion will be used in a broad sense. That is to say, technology is taken to embrace the technological sciences, while the technological sciences include several disciplines in addition to the engineering sciences, such as information science, medical science, and agricultural science. Making such a direct link between technology, more broadly, and the technological sciences makes sense in view of the fact that these sciences aim to contribute towards realizing contemporary or future technologies. Accordingly, the chapter includes discussions and illustrations of a broad range of technological activities, such as research, design, production, use and maintenance. This also fits the comprehensive approach to technology and the engineering sciences that is taken in this Handbook. The prime subject of this chapter is the relationship between science and technology. That science and technology have been, still are, and can be expected to remain, ‘related’ hardly needs to be argued. Rather, the important questions concern, first, the empirical features of this relationship (including its historical development) and, second, its theoretical conceptualization in relation to our philosophical understanding of both science and technology. As will be demonstrated in this and the other chapters in this part of the Handbook, these two questions may receive quite different answers. The layout of the chapter is as follows. Section 2 discusses some important methodological issues that naturally present themselves to a reflexive philosophical approach. Since any account of the science-technology relationship presupposes some characterization of both science and technology, the question is how to acquire a plausible characterization. As to the relationship between science and technology, we face the related methodological question of how to study this relationship. The sections that follow then review several important views of science, 1 For a review of the role of the social sciences in technology and engineering, see Sørensen’s chapter in this Volume, Part I.

Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

66

Hans Radder

technology and their relationship: the idea of technology as applied science (Section 3); the conception of the technological finalization of science (Section 4); the claim that experimentation constitutes the central link between science and technology (Section 5); and the account of science as technology (Section 6). Section 7 sums up the main conclusions about the science-technology relationship and, in particular, about the uses of science in technology. Overall, I follow the common philosophical practice of presenting both an exposition and a critical assessment of the views discussed. Where appropriate, references to other chapters of this Handbook, both in this and in the other parts, are provided.

2

PRELIMINARY METHODOLOGICAL ISSUES

A reflexive philosophical study of the relationship between science and technology needs to confront some preliminary methodological issues. Since making claims about the nature of this relationship presupposes some characterization of science and technology themselves, there is the question of how to acquire a plausible specification of these notions. Next, there is the closely related question of how to investigate the science-technology relationship itself and how to obtain a fitting account of it. The question of how to characterize science and technology is often addressed through a specification of their respective aims. Many authors claim that the aim of science is epistemic, and in particular, the acquisition of knowledge. The aim of technology, in contrast, is said to be the construction of things or processes with some socially useful function. Many other authors, however, claim that such a conceptual-theoretical specification of science and technology does not do justice to the richness and variety of actual scientific and technological practices. By way of alternative they advocate a nominalistic-empirical approach: go and see, and define science (respectively technology) as the practical activity that is called science (respectively technology). These two points of departure — either a conceptual-theoretical definition or a nominalistic-empirical account of science and technology — differ greatly. Both lead to several further questions. Consider first the view of science as the search for knowledge. Since there is also nonscientific knowledge, some authors add that science is the activity that systematically strives for theoretical and explanatory knowledge. However, a strict application of this definition would exclude many activities that are usually, and rightly, seen as part of science. Quite a few scientists aim at observational or experimental knowledge and scientific knowledge can also be non-explanatory, for instance in the case of taxonomical knowledge. A possible solution might be to distinguish between primary and subsidiary aims. Accordingly, the search for theoretical, explanatory knowledge would be the primary aim of science, while other types of knowledge are always subsidiary to this aim. This solution is rather questionable, however. It is, for instance, difficult to reconcile with the many

Science, Technology and the Science–Technology Relationship

67

studies that have convincingly shown that experimental practice has an extensive and worthwhile life of its own.2 Furthermore, defining science as the search for theoretical, explanatory knowledge presupposes a specific philosophical interpretation of science, which will not be generally acceptable. Thus, Bas van Fraassen [1980] sees explanation as merely a pragmatic aspect of science and he puts forward the empirical adequacy of theories, instead of their truth, as the aim of science. Patrick Heelan [1983] also emphasizes the primacy of perception, although his notion of perception differs significantly from van Fraassen’s account. Clearly, for these philosophers and their followers a plausible characterization of science, and a fortiori of the contrast between science and technology, cannot be based on the explanatory nature of theoretical science. What about the definition of the aim of technology as the construction of things or processes having some socially useful function? Although this definition seems to be intuitively plausible, two qualifications are in order. First, many authors claim that it is too narrow because technology is not limited to the making of useful material things or processes. Technology, as the etymology of this term suggests, also involves the generation and utilization of knowledge ([Layton, 1974]; see also the chapters on artifact epistemology in Part II of this Handbook). More specifically, it is design knowledge that is claimed to have a prominent place in technology. Moreover, in the engineering or technological sciences, this design knowledge is often of a quite general nature [Kroes, this volume, Part III]. Second, this definition of technology (with or without the addition of design knowledge) is not of much help in clarifying the science-technology relationship. After all, designing and constructing material things or processes, including the generation and utilization of design knowledge, is common business in the practices of observational and experimental science.3 Both the overall observational or experimental setup and their component devices, apparatus or instruments often require an extensive process of design and construction (see, e.g., [Rothbart, 2007]). Such observational and experimental practices constitute a major part of scientific disciplines. Hence, in contrast to what Layton [1974], Kroes [1992] and many others claim, design (knowledge) and construction do not demarcate technology and engineering from science. What to conclude from this discussion of the conceptual-theoretical approach? The only tenable intuitive distinction seems to be the relation to social usefulness. In contrast to science, technology would be oriented towards potential usefulness for society at large. Even this suggestion needs to be qualified, however. First, should this social usefulness be explicit and immediately visible, right at the start of a technological project? In this case, some of the research carried out in industrial laboratories may not qualify as technological. For instance, the research done 2 See, e.g., [Hacking, 1983; Gooding, 1990; Galison, 1997; Lange, 1999; Radder, 2003; Baird, 2004]. 3 Even computational science has a material side and hence it involves some design of material things or processes. See the analysis of the simulation laboratory in [Petersen, 2006, Chap. 2].

68

Hans Radder

between 1947 and 1972 at the Philips electronic laboratories did not always aim at immediate technological applications [De Vries, 2005]. But if social usefulness is permitted to emerge in the course of a project, then quite a few projects in prima facie scientific research will also count as technological. After all, basic research is often supported by funding agencies because of its contribution to the ‘knowledge base’ of a society, and hence this research can be seen as practically useful in the long run (cf. [Tiles and Oberdiek, 1995, Chap. 4]). For this purpose, present-day applications for basic research projects have to be routinely justified also in terms of their possible technological and societal payoff. Let us now have a closer look at the nominalistic-empirical strategy. This involves the empirical investigation of whichever activities that present themselves as scientific or technological. As will be clear from the preceding comments on the conceptual-theoretical approach, this nominalistic-empirical strategy certainly has its place. In particular, it constitutes a healthy antidote against those philosophers who simply proclaim a specific aim for science or technology, without any further evidence or reflection. Yet, although this strategy may initially seem straightforward, on closer inspection it appears to have its problems as well. First, any empirical identification of either science or technology requires some pre-understanding. Suppose we visit a site called ‘Institute for Biomedical Science’. We may, then, safely conclude that this is a site of scientific activity. But many different activities take place in this institute: the toilets are cleaned, the board of directors holds meetings, the catering service provides lunches and the PhDs write articles. When we focus on the writing of articles in studying science, we apparently apply a certain pre-understanding of what counts as (the core activities of) science. Thus, [Latour and Woolgar, 1979] characterize laboratory science through its production of ‘inscriptions’ (and not, to mention another option, through its catering procedures). More precisely, they focus on a specific subset of the inscriptions produced in the laboratory and disregard other inscriptions, such as the receipts generated by the PhDs through having their lunch in the lab canteen. Hence, the nominalistic-empirical approach presupposes some conceptual-theoretical interpretation of what constitutes science and technology, and the question of whether we can make this pre-understanding more explicit, or even define it, is still with us. A second problem of the nominalistic-empirical approach is that different languages and cultures use different names for activities that might be quite similar. Anglo-Saxons distinguish sciences and humanities, which in Germany are both called Wissenschaft. In earlier centuries, natural philosophy denoted what is now called physical science. And nowadays we speak of computer science and information technology as being roughly equivalent. In order to see whether or not such types of activities might be essentially, basically, or to a large extent similar, we again need a conceptual-theoretical clarification of those activities. My conclusion from this preliminary discussion is that we need both the theoretical and the empirical approach. We have to start from some interpretive perspective on what we take to be basic aspects of science and technology. Next,

Science, Technology and the Science–Technology Relationship

69

we articulate and test this interpretation on the basis of the empirical study of the activities thus defined. And we try to determine its scope by examining its possible applicability to natural philosophy, humanities, information technology, and the like. Once we have established a plausible interpretation of science and technology, this interpretation will acquire some normative force. Activities that do not conform to the established characterization of science or technology should not be named scientific or technological. We stick to a particular interpretation as long as it enables us to cover (what we take to be) the interesting and important cases and dimensions of both science and technology. Thus, the theoretical and the empirical approach should not be separated. On the one hand, a plausible conceptual model should be backed up by empirical studies of the practice of science and technology. On the other, an empirical investigation presupposes an interpretive pre-understanding of both science and technology, and an appropriate empirical model of the science-technology relationship needs to be based on a plausible interpretive pre-understanding. In this chapter, the emphasis is on conceptual-theoretical accounts of the relationship between science and technology, but I will also pay attention to the empirical support of those accounts and refer to empirical studies of this relationship. David Channell’s contribution [this volume, Part I] provides more detailed discussions of several important aspects of the empirical relationship between science and technology. 3

TECHNOLOGY AS APPLIED SCIENCE

A still current view of the relationship between science and technology is phrased by means of the formula ‘technology is applied science’. A classic account of this view has been presented by Mario Bunge. He makes the following distinction between technology as applied science and pure science. The method and the theories of science can be applied either to increasing our knowledge of the external and the internal reality or to enhancing our welfare and power. If the goal is purely cognitive, pure science is obtained; if primarily practical, applied science. Thus, whereas cytology is a branch of pure science, cancer research is one of applied research. [Bunge, 1966, p. 329] Thus, it is the distinct aims which differentiate (pure) science from technology. In Bunge’s view, these aims pertain to the outlook and motivation of the scientific and technological researchers. Bunge develops this view as follows. Scientists strive for empirically testable and true theoretical laws, which accurately describe (external or internal) reality and which enable us to predict the course of events. The technologist, in contrast, uses scientific laws as the foundation of rules which prescribe effective interventions in, and control of, this reality for the purpose of solving practical problems and realizing social objectives. Taken together, science and technology (the latter in the sense of applied science) should be distinguished from those practical techniques and actions that are not based on scientific theories

70

Hans Radder

or methods. Thus, in this view, Roman engineering and medieval agriculture are practical arts and crafts rather than technologies. Since experimentation is a basic method for testing scientific theories, Bunge distinguishes experimental action from both technological and purely practical action. Bunge [1966, p. 330] claims that the different aims of science and technology are inferred from alleged differences in outlook and motivation of their practitioners. If this were meant in a literal sense, he should have provided us with the results of empirical investigations, such as surveys, interviews, or other evidence about the attitudes or self-images of scientists and technologists. Apparently, this is not Bunge’s intention. Instead, his discussion suggests that he thinks that these outlooks and motivations can in some way be ‘derived’ or ‘reconstructed’ on the basis of the activities of scientists and technologists. Hence, the discussion in this section focuses on these (alleged) differences in scientific and technological activities. A further characteristic of this account of the science-technology relationship is its hierarchical nature. In particular, Bunge postulates an epistemological hierarchy between science and technology. If true, scientific laws can provide a justification of technological rules. The converse is not possible, however: a working technological rule, which is merely practically effective, can never justify a scientific law. By way of example, he discusses the technology of making an optical instrument, such as the telescope. In designing and constructing such a device we do not exclusively employ wave optics, the most truthful theory of light in this context, but make ample use of the false theory of geometrical optics, which conceives of light as propagating along straight lines. Moreover, usually such construction work requires specific craft skills (such as the grinding of the lenses or mirrors) which do not employ scientific theories but are based on effective practical knowhow and procedures. Bunge concludes that a practically working artifact, such as the telescope, cannot justify the scientific laws employed in its construction. In addition to the epistemological primacy of science over technology, Bunge’s view entails a temporal ordering. If technology is the result of applying science, it follows that temporally prior scientific research constitutes the driving force of technological development and innovation. This idea of ‘science finds — industry applies’ is often called the linear model of the science-technology relationship. More or less similar hierarchical views of the science-technology relationship can also be found outside of philosophy, for instance among scientists, policy-makers, and the public at large. Sometimes such views include an even stronger hierarchical evaluation in that science is seen as an exciting, creative quest for truth, while technology would merely involve the routine application and exploitation of the fruits of this quest. In the remainder of this section, I discuss and evaluate this view of technology as applied science.4 First, several scholars have criticized Bunge’s approach on 4 In doing so, the focus will be on the ‘substantive’ theories of scientists and engineers, that is, theories about the technological objects themselves, thus leaving aside the ‘operative’ theories of social scientists and technologists, that is, the social theories about technological action and

Science, Technology and the Science–Technology Relationship

71

historical grounds. They claim that historical studies show that many important technological inventions and innovations came about independently from scientific research and scientific theorizing. Well-known illustrations include steam engines, water power devices, mechanical clocks, metallurgical techniques and the like (e.g., [Laudan, 1984]; see also Channell’s chapter in this Volume, Part I). Although these criticisms seem basically correct, they do depend on the precise interpretation of Bunge’s version of the linear model of the science-technology relationship.5 A flexible interpretation of Bunge’s model would permit the following replies. First, many of the historical counterexamples date back long ago, often to the eighteenth century and before. Hence, they need not be taken as a refutation of the account of technology as applied science, but might be seen as limiting the scope of this account. Put differently, Bunge’s account might be construed as a definition of technology and as such it would be immune to empirical counterexamples. If a certain case does not fit the account of technology as applied science, then it is, by definition, not a technology. The remaining issue, then, pertains to the usefulness and relevance of Bunge’s definition. In view of the great significance of modern, science-based technology, the usefulness and relevance of his definition seems obvious enough. Second, one might note that, in Bunge’s view, technology may also result from applying the method of science (see the above quotation) and that one could make a case for the claim that (some of) the counterexamples did apply scientific methods, even if they were not based on available scientific theories. However this may be, I will not pursue this debate any further here but instead develop a different assessment of Bunge’s technology-is-applied-science account. For this purpose, it is important to realize that this account implies two distinct claims. The first is that there is a clear ‘kinship’ between science and technology, in the sense that technology is based on scientific theories and methods. The historical criticisms are aimed at this claim. They seem to accept Bunge’s characterization of science as a quest for true knowledge of laws and theories (e.g., [Layton, 1974]), but they object that technology has often developed independently from these laws and theories. That is to say, they claim that the differences between science and technology are larger than Bunge assumes. Secondly, however, Bunge advocates the claim that science and technology also display essential differences, in the sense that scientists aim at truth and technologists at practical effectiveness and usefulness. I will assess this second claim by analyzing, like Bunge, sciencebased technology and by showing that its contrast to science is much smaller than Bunge assumes. Consider the claim that scientists aim for truth by constructing testable, fundamental theories and by accepting or rejecting these theories according to their match to the empirical data. This account suggests that separate, fundamental theories can be confronted more or less directly with the empirical data. In fact, organization (for the latter, see Sørensen’s chapter in this Volume, Part I]). 5 For extensive, critical discussions of the linear model, see the contributions to [Grandin et al., 2004].

72

Hans Radder

however, scientific practice is much more complex. Fundamental theories as such, for instance quantum mechanics or the theory of evolution, do not tell us much about empirical reality. To become empirically applicable they first have to be developed and specified with a view to particular domains of empirical phenomena. The point can be illustrated by the case of nonrelativistic quantum mechanics. The basic structure of this theory was developed between 1925 and 1927. Since those days, this theory has been, and is being, ‘tested’ in many different domains, including atomic and nuclear physics, quantum chemistry, solid state physics, and so on. Within each of these domains we find a diversity of subfields, such as the study of electrical conductivity in crystals within solid state physics. Moreover, there are overlapping research areas, such as laser physics which combines insights from both atomic and molecular physics and from quantum electrodynamics. Hence, we are confronted not with two types of activities (theoretical and experimental) but three: the construction of fundamental theories; their development and specification to enable actual empirical tests; and the design and performance of experiments to test the theories. The second type of activity requires the articulation of the fundamental theories, usually through extensive calculation and substantial model building.6 Two aspects of these processes of development and articulation are particularly relevant to the present comparison between science and technology. First, even within one subfield one often finds a large variety of different models and methods of calculation, each of them specific for and appropriate to particular types of experiment. Nancy Cartwright [1983, pp. 78-81] discusses the example of laser physics and documents the use of at least six different models of the natural broadening of spectral lines. She emphasizes that the scope of each of these models is often quite small, namely limited to a few types of experimental phenomena. Moreover, scientists do not see these different models as competing but rather as complementing each other since each serves a specific purpose. Second, a major function of model building is to bridge the large distance between the relatively schematic and simple fundamental theories and the mostly complex experimental world [Morgan and Morrison, 1999]. Because of this distance, bridging cannot succeed on the basis of the fundamental theories alone. Hence, what we see in practice is the use of a diversity of methods and approaches that cannot be rigorously justified from a theoretical perspective. Frequent use is made of convenient rules of thumb, intuitively attractive models, mathematically feasible approximations, and computationally tractable computer simulations. Often the test also depends on other experiments, for instance when the value of parameters that cannot be calculated theoretically, is determined through tuning them to the results of other experiments. Thus, the variety of experimental domains and the large distance between fundamental theories and experimental phenomena require the indispensable use of 6 See [B¨ ohme et al., 1983; Cartwright, 1983; 1999]. For the sake of argument I have, with Bunge, assumed the availability of a fundamental theory. In actual practice, calculation and model building may just as well precede the construction of such a theory.

Science, Technology and the Science–Technology Relationship

73

‘workable methods’ in testing the theories. Scientific practice includes the regular application of a variety of convenient rules of thumb and intuitive models for solving different problems, the making of approximations based on mathematical or computational feasibility and the blackboxing of (part of) a system through tuning to experimentally determined parameters. The crucial point is that these are exactly the kinds of procedures that are typical of technology, also according to Bunge. Thus, on the basis of an analysis of their testing activities, there is no reason to assume a fundamental contrast in outlook and motivation between scientists and technologists.7 A test of quantum mechanics by a laser physicist is not essentially different from the test of a design of a specific acoustic amplifier by an engineer [Cartwright, 1983, pp. 107-112]. Thus far, I have focused on Bunge’s account of the relationship between science and technology as applied science. Apart from this, there is his claim that both science and technology should be clearly distinguished from skillful, practical action. This claim suggests that practical craft skills play no (or no significant) role in science and in science-based technology. However, if we — in contrast to Bunge — take full account of the practice of scientific and technological observation and experimentation, it is immediately clear that this suggestion makes no sense. After all, as every observer or experimentalist knows, skillful action is an essential aspect of observational and experimental science and technology (just think of the grinding of the lenses in the case of constructing a telescope).8 The reason for the importance of skillful action is straightforward. In contrast to what generations of empiricists have claimed, the typical way of obtaining scientific experience is not through passive sensation but through active observation and experimentation. As we will see in more detail in Section 5, the stability and reproducibility that scientific observers and experimenters try to establish is almost never given by nature, but has to be realized through a difficult and laborious process of intervention and control. For this purpose, skillful practical action is indispensable (see, e.g., [Ravetz, 1973; Collins, 1985; Gooding, 1990; Radder, 1996]). The discussion in this section does not claim to provide an exhaustive assessment of the view of technology as applied science.9 Yet it should suffice to demonstrate that Bunge’s hierarchical approach is questionable. A reconstruction of their cognitive activities does not support the attribution of essentially different aims to scientists and technologists. Consequently, this way of demarcating science from technology proves to be difficult, if not impossible, and the same applies to sub7 Another relevant argument, which I will not pursue here, is that scientists do not aim at truth simpliciter but at significant truths, where the criteria of significance may be both epistemological and social (see [Kitcher, 2001]). 8 In a later publication, Bunge admits that ‘even the scientific inventor is a bit of a tinkerer (bricoleur) and — like the scientist — he possesses some tacit knowledge, or know-how, that cannot be rendered fully explicit’ [Bunge, 1985, p. 220]. In spite of this, he immediately adds that it is only explicit, science-based technology that is philosophically interesting and worth studying. 9 For further discussions and assessments, see [Tiles and Oberdiek, 1995; Cuevas, 2005; Boon, 2006; Koningsveld, 2006]. See also Channell’s chapter in this Volume, Part I.

74

Hans Radder

stantiating the epistemological subordination of technology to science. To avoid misunderstanding, I should like to emphasize that the argument of this section is not that there are no differences at all between science and technology. But it does imply that, generally speaking, these differences will be a matter of degree and that they do not add up to an unambiguous contrast between science and technology in terms of singular and essentially different aims. In the concluding section of this chapter I will come back to this issue and address the question how science and technology may be distinguished and related on the basis of their similarities and dissimilarities. 4 TECHNOLOGY AS FINALIZED SCIENCE During the 1970s a group of German scholars, also called the Starnberg group, published an impressive series of articles and books about the finalization of science (see [Sch¨ afer, 1983], and further references therein). ‘Finalized science’ denotes a particular stage of scientific development that is, more or less consciously, oriented towards external social goals and interests. The authors themselves see their finalization theory as an improvement of the theory of technology as applied science. Thus, in their account of agricultural chemistry Wolfgang Krohn and Wolf Sch¨afer state: Our aim here is not to introduce a distinction between agricultural chemistry as a finalized science and applied science, but rather to offer a more precise meaning for the vague notion of ‘applied science’. The term ‘applied science’ gives the misleading impression that goaloriented science simply involves the application of an existing science, rather than the creation of a new theoretical development. This in turn feeds the misconception that pure science is superior to applied science. [Krohn and Sch¨ afer, 1983, p. 46] One of the main aims of the finalization theory is to establish at which stages of scientific development finalization is possible and fruitful. For this purpose, it includes an account of scientific development that makes use of, but also substantially expands on, Thomas Kuhn’s model of scientific development. Although it is not generally realized, Kuhn advocates a strongly internalist view. For a scientist, the solution of a difficult conceptual or instrumental puzzle is a principal goal. His success in that endeavour is rewarded through recognition by other members of his professional group and by them alone. The practical merit of his solution is at best a secondary value, and the approval outside the specialist group is a negative value or none at all. [Kuhn, 1970, p. 21] The finalization theory also starts from a rather strict internal-external distinction, but then goes beyond a Kuhnian internalism by arguing that an interaction

Science, Technology and the Science–Technology Relationship

75

between external or societal goals and interests and internal or cognitive goals and interests is possible, and even to some extent necessary, at a certain stage of the development of scientific disciplines. The theory focuses on the disciplines of the natural sciences and claims that these sciences pass through three successive stages. First, there is the explorative stage, which bears some resemblance to Kuhn’s preparadigmatic stage. At this stage, a well-developed domain-structuring theory is not (yet) available, and research methods are primarily empirical and classificatory rather than theoretical and explanatory. The next, the paradigmatic stage is guided by a general theory that structures the field of phenomena and directs the way they should be investigated. As in Kuhn’s normal science, the aim is the empirical and conceptual articulation and validation of the central theoretical ideas. These second-stage developments may lead to ‘closed theories’, a notion adapted from physicist Werner Heisenberg and explained as follows: In general three things can be said of a closed theory ...: firstly, it will possess sufficient conceptual material to capture a particular field of phenomena; secondly, its validity will at least be proven for a number of instances; and thirdly, there are good reasons to expect that its validity extends to the whole category of phenomena in question. [B¨ ohme et al., 1983, p. 148] Because these are quite demanding criteria, which will not always be met in actual scientific practice, the authors introduce the weaker notion of theoretical maturity for cases where the theories are not strictly closed but still possess a substantial measure of comprehensiveness and stability. Hence the claim of the finalization theory is that, from an internal-scientific perspective, theoretically mature disciplines are more or less complete. Nevertheless, they can be developed further into a third, or postparadigmatic, stage, in which they become oriented towards external goals and interests through the development of ‘special theories’ (sometimes also called ‘theoretical models’) for the purpose of realizing certain technological applications. It is at this stage that science becomes finalized. In contrast to Kuhn, at this stage the ‘practical merit’ and the ‘approval outside the specialist group’ are primary values, and yet realizing this merit requires the development of genuinely new theoretical knowledge. The finalization theory has been developed in close interaction with case studies of important episodes in several disciplines (see Part I of [Sch¨afer, 1983]). In physics, the articulation of classical hydrodynamics into a variety of special theories of fluid dynamics for the development of airplanes has been studied. In chemistry, the relationships between nineteenth-century organic chemistry, the special area of agricultural chemistry and the production of artificial fertilizers has been investigated. And in biomedical science, the development of molecular biochemistry into special theories of carcinogenic processes with a view to the production of appropriate drugs has been scrutinized. The authors themselves conclude that their theory applies best to the discipline of physics. Its appropriateness for the other

76

Hans Radder

disciplines is judged to be (far) less straightforward, the major problem being the applicability of the notion of theoretical maturity. The theory of finalization was proposed more or less simultaneously with, though independently of, the strong program in the sociology of scientific knowledge.10 Although both approaches share an emphasis on the significance of external factors, there are also important differences between the finalization theory and the sociology of scientific knowledge. First of all, the former, in contrast to the latter, does not claim that scientific truth depends on external goals and interests. Furthermore, the finalization theory focuses on conscious or intentional external influences in a science policy context. Hence, the theory includes an explicit evaluative and normative component: although orientation towards external goals and interests is feasible in the explorative and, to some extent, even in the paradigmatic stage, the best and most fruitful way to exploit the technological potential of the sciences is through the finalization of mature scientific theories in their postparadigmatic stage. During the 1970s and early 1980s, the finalization theory sparked an extensive and at times acrimonious debate (see the bibliography in [Sch¨afer, 1983, pp. 301306]). This debate was both philosophical and political in nature,11 but it was primarily restricted to Germany.12 Thus far, in Anglo-Saxon philosophy of science the relationship between science and technology has been a neglected issue anyway (cf. [Ihde, 1991; 2004]). Within the recently rising philosophy of the technological sciences, however, the theory of finalization constitutes a worthwhile topic for studying the intersections between science, technological science and technology. In the remainder of this section, I will discuss the merits and problems of this theory. A first merit of the theory is that it provides a significant extension of Kuhn’s account of the development of science. It shows that older paradigms are not, or not necessarily, discarded after the advent of a successor, since they may be further developed through processes of finalization. Furthermore, the theory takes into account the obvious importance of external goals and interests, especially since the second half of the nineteenth century, and thus goes beyond Kuhn’s inadequate internalist approach. What is particularly insightful is the subtle way in which these internal and external factors are shown to be interwoven. Even if finalized science is not autonomous, the external goals and interests do not operate as purely extrinsic impositions. Instead, they are transformed and internalized as cognitive constraints on, or specifications of, the special technological theories that need to be developed on the basis of a mature scientific theory. For instance, in nuclear fusion research scientists try to develop a special theory of plasma physics that will ultimately enable the construction of a stable and reproducible nuclear fusion 10 For

a detailed exposition of this program, see [Barnes et al., 1996]. the proponents of the finalization theory were accused of promoting socialist state regulation and criticized for advocating the societal steering of science at the expense of its academic freedom. 12 The philosophical claims of the finalization theory have also been widely discussed in The Netherlands. See, e.g., [Nauta and De Vries, 1979; Zandvoort, 1986]. 11 Politically,

Science, Technology and the Science–Technology Relationship

77

reactor (see [B¨ ohme et al., 1983, pp. 154-156]). Technically, this means that only such processes are considered for which the product τ of the containment time and the temperature of the plasma exceeds a certain minimum value τ0 . Thus, the external technological goal of providing nuclear fusion energy in a controlled, safe and economically efficient way has been transformed and internalized as a specific guideline for scientific theorizing. It tells the researchers to focus their theoretical work only on such constellations of plasma and container for which τ > τ0 . Furthermore, the finalization theory convincingly demonstrates that technological science develops genuinely original knowledge, a point that is also emphasized in many recent contributions to the philosophy of the technological sciences (see, e.g., [Boon, 2006]). Technological knowledge is not, as seems to be implied in Bunge’s view of technology as applied science, a mere application of existing scientific knowledge. Another important aspect of the finalization theory is the attempt to provide a differentiated account of the relationship between external-societal and internalcognitive factors in the development of the sciences. Whether fully successful or not, the theory at least attempts to make explicit the specific conditions under which external steering of science is possible and fruitful. In this respect, it favorably contrasts to some more recent approaches, in particular to the now fashionable idea of a linear historical succession of a ‘Mode 1’ science, which is largely autonomous and disciplinary, followed by a ‘Mode 2’ science, which is primarily focused on, and guided by, technological, economic and socio-political contexts of use.13 Finally, at least some of the proponents of the finalization theory foster a commitment to a science ‘in the public interest’. Finalized science, they claim, should not evolve in a power-driven, Darwinist way, but be guided by procedures of explicit and democratic deliberation about the rational acceptability of the means and ends of proposed technological developments. Again in contrast to the Mode 1/Mode 2 approach mentioned above, this acknowledgment of normative issues is important, even for those who do not share the specific position of the advocates of the finalization theory. Moreover, given the problematic consequences of the rapidly increasing commercialization of science over the past twenty-five years, the notion of a science in the public interest is still as timely as ever (see, e.g., [Krimsky, 2003]). Next to these merits, however, the finalization theory has several problematic characteristics and implications. As we have seen, the authors themselves already confronted the problem of the definition of a closed theory and especially its application to the history of science. They concluded that the applicability of the theory to disciplines other than physics is unclear. Thus, in the case of nineteenth-century agricultural chemistry, there was no closed theory available and the authors of the case study fall back on watered-down notions such as ‘relative theoretical maturity’ and ‘methodological maturity’ [Krohn and Sch¨afer, 1983]. But even cases 13 See

[Gibbons et al., 1994]; for a critical review, see [Weingart, 1997].

78

Hans Radder

from physics are not straightforward. An interesting case would be to investigate the recent ‘finalization’ of climate science in the face of the human-induced greenhouse effect. It is by no means obvious that this research is building on a closed, or mature, theory of the dynamics of the entire climate system (see [Petersen, 2006, Chaps. 5 and 6]). The finalization theory rightly claims that technological science develops genuinely new knowledge. But whether its characterization of this knowledge exhausts the knowledge generated in the technological sciences is another matter. According to the finalization theory, technological knowledge is developed on the basis of closed or mature scientific theories. In general, however, such knowledge will only be a part of the knowledge required for the design, production, use or maintenance of technological artifacts or systems (see also Houkes’ chapter in this Volume, Part II). For instance, a fluid dynamics model of the boundary layer and the concepts of lift and circulation — as discussed in [B¨ ohme, 1983] — does not yet permit the design and manufacture of a real airplane, let alone the realization of the entire technological system of air transportation.14 This obviously limits the value of the finalization theory for a philosophy of technology and the technological sciences. indent Related to this is a theory-dominant view of (natural) science. Although the significance of experimentation is acknowledged in principle, the finalization theorists’ view of the technological sciences is still thoroughly theory-biased. It is theory formation which is seen as the core of scientific development and as the royal road to the fruitful exploitation of science for practical purposes. In the meantime, however, many authors in the philosophy of scientific experimentation (see note 2) have demonstrated that experimentation has a life of its own and is not limited to the testing of pre-existing theories. For this reason, it is also incorrect to identify the notion of a paradigm with that of a theory (see also [Rouse, 1987, Chap. 2]). Moreover, seeing observational and experimental science as merely preparadigmatic overestimates the role of explanatory scientific theories, especially in the technological sciences. Finally, the finalization theory exhibits certain questionable modernist characteristics. It entails a belief in the possibility of a universally valid model of scientific development. As such, it cannot do justice to the diversity and richness of the actual development of the (technological) sciences. Moreover, the theory strongly suggests an overoptimistic belief in social progress through the employment of science. As such, it does not show great awareness of the fact that (technological) science may itself be a source of social problems. One does not need to be a radical postmodernist to see the problematic character of these two beliefs.

14 For

7].

more on the systemic character of technology, see [Hughes, 1987; Radder, 1996, Chap.

Science, Technology and the Science–Technology Relationship

5

79

EXPERIMENT AND THE SCIENCE-TECHNOLOGY RELATIONSHIP

As we have seen, the finalization approach represents a form of theory-dominant philosophy of science. In fact, however, a focus on experimentation provides a quite natural starting-point for studying the science-technology relationship. To mention just one example: the method of systematic parameter variation pioneered in the eighteenth century by John Smeaton to scrutinize and test the working and efficiency of waterwheels [Channell, this volume, Part I] plays an important part in both experimental science and in engineering and technological research. Hence, in this section I will review some philosophical accounts of experimentation as a crucial link between science and technology. In his early philosophy, J¨ urgen Habermas has discussed the relation between technology and the natural sciences in some detail (see [Habermas, 1971; 1978]). He conceives of these sciences as intrinsically related to technology. Like logical positivism, Habermas sees observation as the basis of science, but he emphasizes that what counts in science is never the single, isolated observation but only the observation that can be reproduced by other scientists. Thus, his actual focus is on reproducible observations and, more generally, on predictive empirical laws. Such laws, Habermas claims, cannot be interpreted as reflecting a human-independent reality, since their universal validity depends on the possibility of active intervention and control of the empirical situation by human beings. Put differently, the epistemic warrant for the empirical law ‘whenever x, then y’ is provided by the practical result that ‘whenever we do x (under controlled conditions c), then we can bring about y’. This intervention and control is enabled through human, instrumental action. In this way, a ‘technical interest in prediction and control’ guides the production of natural scientific knowledge. The very constitution of experience on the basis of instrumental action orients science towards the technological application of the knowledge acquired. Prediction and control through intervention are the essential characteristics of the empirical laws of science and as such these characteristics foreshadow its technological application. In science, instrumental action takes the form of experimental action. Hence, experiment constitutes the basic link between science and technology. Following Charles Peirce, Habermas explains the notion of a scientific experiment as follows: In an experiment we bring about, by means of a controlled succession of events, a relation between at least two empirical variables. This relation satisfies two conditions. It can be expressed grammatically in the form of a conditional prediction that can be deduced from a general lawlike hypothesis with the aid of initial conditions; at the same time it can be exhibited factually in the form of an instrumental action that manipulates the initial conditions such that the success of the operation can be controlled by means of the occurrence of the effect. [Habermas, 1978, p. 126] This quotation clearly expresses the intrinsic relation between predictive scientific knowledge and controlled technological action and production that is char-

80

Hans Radder

acteristic of Habermas’s early philosophy. In his further development, however, Habermas changed his views on this subject, in particular by incorporating the theory-ladenness of observation and more in general by acknowledging the relative autonomy of theoretical argumentation in science. Thus, the focus of his philosophy shifted to the subjects of argumentation and communication. As a consequence, he did not develop his rather schematic view of experimentation as a significant link between science and technology. Hence, it is worthwhile to take a closer look at this subject on the basis of a more detailed account of scientific experimentation.15 The purpose of this discussion is to employ this account to illuminate important aspects of the relationship between science and technology. A characteristic feature of experimental science is that access to its objects of study is mediated through apparatus (in the form of instruments and/or other equipment or devices).16 In an experiment, we (try to) bring about a correlation between an object of study and some apparatus, and to draw conclusions about that object on the basis of a ‘reading’ of some features of the apparatus. As Habermas correctly argues, scientific experiments are meaningful only to the extent that our intervention and control produces a correlation between object and apparatus which is stable and reproducible. An important necessary condition of experimental stability and reproducibility is the appropriate control of the actual and possible interactions between the experimental (or object-apparatus) system and its environment.17 It is useful to distinguish three types of such interactions: the required interactions, which enable the object-apparatus system to behave according to its design; the forbidden interactions, which might disturb the intended experimental processes; and the allowed interactions, which are neutral with respect to the planned course of the experimental system and thus neither enabling nor disturbing. To realize a stable and reproducible experimental system, the required interactions need to be produced and maintained, the forbidden interactions need to be eliminated or prevented from taking place, while the allowed interactions do no harm and hence do not need to be controlled. For instance, if a particular experimental design requires a low temperature of, say, 100K, then we need to produce a starting temperature of 100K and we need to control the heat flow between experimental system and environment in such a way that the system stays at this temperature during the entire course of the experiment. Furthermore, if an impact of electromagnetic waves could disturb the intended experimental processes, we have to prevent such waves from interfering with the object-apparatus system during all experimental runs. Finally, if the gravitational interaction between system and environment does no harm, we do not have to control for it. The presence of required and allowed interactions 15 The present sketch of this account draws on analyses in [Radder, 1988, Chapter 3; 1996, Chapter 6; 2003]. Additional detail, including a characterization of the implied notion of ‘technology’, can also be found in Radder’s chapter in this Volume, Part V. 16 For discussions and classifications of scientific apparatus, see [Harr´ e, 2003; Baird, 2003; Heidelberger, 2003]. 17 Of course, this control is not sufficient, since the object-apparatus system itself may be internally unstable and irreproducible.

Science, Technology and the Science–Technology Relationship

81

implies that successful experimentation does not necessitate a completely isolated system, that is, a system that does not at all interact with its environment. Materially realizing such a system would be very difficult and probably even impossible, given the ubiquity of gravitational and/or electromagnetic interactions. Of course, in actual scientific practice we may not always, or not yet, know which interactions are required, forbidden or allowed; or we may be wrong in our assessment of these interactions. Anyway, an important part of the aim of experimentation is to get to know which interactions are enabling, disturbing or neutral. Two features of such processes of acquiring experimental knowledge are directly relevant to the issue of stability and reproducibility. First, what is seen to be required, forbidden or neutral will depend on the theoretical interpretation of the experiment in question. Types of interaction that are claimed to be theoretically impossible (e.g., telepathic influences or signals traveling faster than light) will be irrelevant and do not need to be taken into account. The same applies to interactions that are possible (and may be present) but are claimed to be inconsequential to the plan and aim of the experiment (e.g., the ‘impact’ of daylight in measuring the temperature of a fluid) and hence classified as ‘allowed’. Yet, we should note that such claims may be contested by other experimenters or overthrown by later developments. Second, controlling the relevant interactions is, in practice, not only a matter of exercising the required material control, but it also demands a social discipline and control of all the people that have, or might have, an impact on the material realization of the experiment. After all, it is these people who play, or might play, a critical role in the processes of producing or securing the enabling conditions and eliminating or preventing the disturbing conditions. In addition to these two features, there may also be social or ethical reasons for the need to control further interactions between an experimental system and its environment. For instance, impacts of an experimental system on the environment that could endanger the safety of the experimenters or of other human beings are generally seen to be undesirable and hence they need to be prevented. Thus, the necessary control of the (desirable and undesirable) influences and disturbances between the object-apparatus system and its environment exhibits important theoretical, material and social features of scientific experimentation. Next, this analysis may be used to discuss and assess the science-technology relationship in two different ways. Just like experiments, working technologies need to be stable and reproducible, while the control of the relevant interactions between the technological system and its environment constitutes a necessary condition for achieving this goal. Again, we may distinguish between required, forbidden and allowed interactions. Thus, in a conceptual-theoretical sense, the successful realization of a technological system poses similar requirements as the successful realization of an experimental system. The system-environment interactions that enable the technological system to behave according to its design need to be produced and maintained, the interactions that might disturb the intended technological processes need to be eliminated or prevented from taking place, while

82

Hans Radder

the interactions that are inconsequential to the stable and reproducible working of the technological system may be ignored. Furthermore, in an empirical sense, materially realized experimental substances, devices or processes may be, and often are, exploited as (part of) technological systems. A particular piece of experimentally developed electrical circuitry may be used to fulfill a certain function as part of a larger technological system, for instance a computer. Or an organism that has been genetically modified in a biology laboratory may get exploited in particular agricultural technologies. As in the case of their scientific counterparts, such ‘experimental technologies’ are supposed to exhibit a certain measure of stability and reproducibility, and hence the relevant system-environment interactions need to be controlled. Materially and socially, however, experimental systems and the corresponding experimental technologies will usually be quite different for two reasons. First, technologies are typically required to remain stable and reproducible for a much longer period and in many more places. That is to say, the technology is supposed to function properly on a much larger spatiotemporal scale than its laboratory counterpart. Second, and related to the first reason, the environments in which the experimental technologies are expected to function may be quite different from the average laboratory environment. For these reasons, we cannot assume that a successfully realized experiment guarantees the success of the corresponding experimental technology.18 A nuclear fusion device that works well in the laboratory by no means provides us with a stable and reproducible fusion reactor that can be effectively exploited for controlled energy production. Similarly, a successful in vitro test of experimental AIDS vaccines does not necessarily entail a successful in vivo therapy for AIDS patients.19 Time and again, however, scientists from all kinds of disciplinary backgrounds have made such unwarranted leaps, either because of their inadequate view of the relation between science and technology or simply to flatter their funding agencies for the purpose of acquiring additional financial support. In this respect, it is interesting to look back briefly at the finalization theory. According to this theory, during the paradigmatic stage so-called ‘transfer research’ is possible. This research includes the systematic ‘scaling-up’ of laboratory experiments into industrial processes. Apparently, this scaling-up is seen as the unproblematic application of existing knowledge and as not requiring specific further research. Hence it is claimed that, in the paradigmatic stage, science policy can only promote research, but it cannot substantially guide it in novel directions [B¨ ohme et al., 1983, pp. 152-153]. As my more detailed examination of the relations between experimental and technological science has shown, however, these ‘scaling-up’ processes are by no means straightforward. They require a 18 Hence, the twofold meaning of ‘experimental technology’ as ‘resulting from experimental research’ and as ‘still being tentative’. See also the notion of ‘society as a laboratory’ in [Krohn and Weyer, 1994]. 19 See [Radder, 1996, Chaps. 6 and 7], where these issues and relevant cases, such as nuclear power production, entomological pest control and agricultural biotechnology, are examined in detail.

Science, Technology and the Science–Technology Relationship

83

substantial additional study of the processes that will, or may, occur at the larger temporal and spatial scales and of the new environments in which the technologies are expected to function. An important aim of such studies is to generate new knowledge about the stable and reproducible working of these technologies at the required scales and in the intended environments. The account of the science-technology relationship discussed in this section engenders two critical questions, both of which are crucially important regarding the social governance and normative assessment of scientific and technological projects. First, there is the factual question of whether an intended extension of a successful experiment to a stable and reproducible experimental technology can be reasonably believed to be feasible. The larger the spatial or temporal extension of the intended technological system, the more pertinent this question will be. Second, there is the normative question of whether the controlled material and social world that is needed to guarantee the stability and reproducibility of the technological system is a normatively desirable world. If one or both of these questions are answered in the negative, the only reasonable option is not to realize this particular technology. In my chapter in this Volume, Part V I will come back to these questions and discuss them more fully.

6 SCIENCE AS TECHNOLOGY The fruitfulness of seeing experimentation as a central link between science and technology might tempt us to conceptualize science and technology as substantially, basically, or even essentially, similar. And, indeed, philosophical accounts of the science-technology relationship repeatedly advocate such a conception of ‘science as technology’. Illustrations can be found in the work of Martin Heidegger, (the early) J¨ urgen Habermas, Peter Janich and Srd-an Lelas. More recently, comparable views in terms of the notion of technoscience have been developed by Donna Haraway, Bruno Latour, Don Ihde and Karl Rogers, among others. This notion of technoscience is claimed to capture the crucial similarities between science and technology. First, it posits the primacy of practice: both scientists and engineers or technologists are centrally involved in practical processes of intervention, negotiation and construction. Furthermore, in contrast to more traditional accounts of the science-technology relationship (such as Bunge’s applied-science account), a technoscientific approach highlights the importance of materiality — that is, the material artifacts, interactions and procedures — for both science and technology. Finally, this approach emphasizes the fact that, in the course of the twentieth century, science has increasingly become ‘big science’ and as such it has acquired — and it does require — the format of an industrial organization. By way of example, consider Bruno Latour, who rejects any basic distinction between science and technology by emphasizing the constructive and adversarial nature of both.

84

Hans Radder

It is now understandable why, since the beginning of this book, no distinction has been made between what is called a ‘scientific’ fact and what is called a ‘technical’ object or artefact. The problem of the builder of ‘facts’ is the same as the problem of the builder of ‘objects’: how to convince others, how to control their behaviour, how to gather sufficient resources in one place, how to have the claim or the object spread out in time and space.20 In this section I address the views of Srd-an Lelas [1993; 2000], who has developed the science-as-technology account in more philosophical detail. Lelas opposes his account to contemplative, or theoria, views of science. Such views, he claims, separate epistemology from ontology and semantics. That is to say, observation and experiment may be required for ascertaining the truth of theories but as such they are taken to be mere means. After all, whether or not theories are true is supposed to be exclusively a matter of their correspondence to a human-independent reality. Hence, when theories are true, all traces of the way we have found them, through interacting with and intervening in the world, become irrelevant and should be erased. That is to say, ultimately observation and experiment are eliminable. From his science-as-technology perspective, Lelas raises two kinds of objections to such theoria views of science. First, he argues that experimentation, as the design and production of artifacts, involves an interaction and interference with nature, and he notes that scientific observation shares a number of crucial features with experiment [Lelas, 1993]. Through processes of experimentation and observation, which involve the making of artifacts through implementing an idea, science discovers because it invents. In Lelas’s Heideggerian phrase, ‘nature is at once revealed and produced’. The two sides of this process — revealing and producing nature — cannot be separated, as it is done in the theoria account. Lelas concludes that the productive activity of observing and experimenting, which is essentially technological in nature, constitutes an indispensable element of the ontology of science. For this reason, the significance of observation and experimentation goes far beyond their role as instruments for testing the truth of theories. The second objection to theoria views has to do with the function and meaning of theories. Like Janich and Latour, Lelas claims that the meaning of theories cannot be divorced from their function in experimental or observational processes. Theories should be experimentally testable and this requires that the route from theory to experiment should be mapped out by the theory itself. Theory [cannot] be treated as a mere instrument for calculation and prediction of the experimental outcome. It is much more than that. It is an instrument of design, and being that, it encompasses both ontology and technology. A theory can be considered as a condensed set of instructions of how to build an experimental apparatus, or, better, 20 [Latour, 1987, p. 131]. He does, however, allow for some differences in degree, in the sense that scientists more often focus on new and unexpected procedures or objects, while technologists are more often engaged in coordination and consolidation of existing activities or artifacts.

Science, Technology and the Science–Technology Relationship

85

how to guide the production of experimental artefacts. [Lelas, 1993, p. 442] Hence, the essence of scientific theories is not to be found in their abstract conceptual or mathematical structures as such, but rather in the translations and interpretations which connect theoretical concepts or statements to the practice of observational and experimental action and production. In his book Science and Modernity, Lelas develops these views about science and technology and embeds them in a comprehensive and (broadly) naturalist theory of the processes of human cognition, of the rise of (modern) science and of the nature of scientific knowledge. For instance, from an evolutionary, biological perspective, humans prove to be ‘prematurely born, retarded and unspecialized mammals’. In order to survive they need to be able to adapt to a large variety of selection environments. For this purpose, technology is seen to be particularly important. Artefact making is not the only component of human existence; it covers only one aspect of the relationship between humans and nature. Mind/brain, language and institutions are the others. Together they constitute what we usually call culture. But technology is the essential part of it; it is the part that completes the physical exchange between humans as living systems and their physical environments. [Lelas, 2003, p. 112] Lelas goes on to explain the rise of science as having been enabled by the ‘urban revolution’ in ancient Egypt, the Middle East, India, China and the Americas. Yet modern science, which emerged from the sixteenth and seventeenth century onwards, required two important further developments: first, the economically motivated doctrine and practice of the human mastery of nature; and second an ever increasing transfer of human activities and functions to technological artifacts. This leads him to the aforementioned claims that experimentation constitutes the most important innovation of modern science and, more specifically, that even scientific theory is, ultimately, about making. In concluding this section, I will briefly assess Lelas’s science-as-technology account. His general theory of science and modernity primarily deals with the natural and cultural preconditions and contexts of (modern) science. The theory is thoughtful and intriguing, and Lelas’s book contains a wealth of interesting discussions, but a more detailed review is really beyond the scope of the present chapter (for this, see [Radder, 2002]). Hence, I will limit myself to some more specific remarks on the relationship between science and technology. On the basis of the discussion in the previous sections, in particular Section 5, we may conclude that Lelas’s emphasis of the significance of the action and production character of experimentation is fully justified. Moreover, extending this account from experimentation to scientific observation has much to recommend it. As we have seen, Lelas endorses the more specific claim that theory plays a role not just in making predictions of experimental results but much more generally

86

Hans Radder

as an instrument guiding the entire process of the production of experimental artifacts. Although some authors have claimed that theory-free experimentation is possible and regularly occurs in the development of science, a closer look at scientific practices reveals that Lelas’s claim can be maintained, if it is more specifically construed as stating that the performance and understanding of experiments depends on a theoretical interpretation of what happens in materially realizing the experimental process [Radder, 2003]. In spite of this, the general reductionist view that science is, basically, technology cannot be upheld. Consider the claims that there is a ‘full continuity between high scientific theory and the skills of the experimenter’ and that ‘a theory can be considered as a condensed set of instructions of how to build an experimental apparatus’ [Lelas, 1993, pp. 441-442]. In this respect it is important to make a distinction between the ‘high theory’ of the object under study and the theoretical interpretation of the entire experimental process. Generally speaking, the former tells you something about the experimental process, but in no way can it be said to guide the production of experimental artifacts. For instance, as we have seen in Section 3, the high theories of quantum physics do not even suffice to construct and use theoretical models of laser phenomena, let alone tell us how to build such devices. A further problem of Lelas’s science-as-technology account is the fact that scientific theories have a meaning that transcends the meaning of the particular experiments that have thus far been used to test these theories. In as far as this account overlaps with the operationalist theory of meaning, it is vulnerable to the well-known criticism that this theory entails an unfruitful proliferation of theoretical concepts and that it neglects the systematic significance of theoretical frameworks [Hempel, 1966, pp. 88-100]. That theories have such a ‘surplus’ meaning can also be seen by analyzing the notion of experimental reproducibility in more detail. In Section 5 I employed the notion of reproducibility in an undifferentiated way. In fact, however, reproducibility is a rather complex notion. First, it is important to distinguish between the actual reproductions and the (claimed) reproducibility of an experiment; in addition, we need to ask what has been reproduced, or is (claimed to be) reproducible, and by whom? [Radder, 1996, Chaps. 2 and 4]. In the present context, the relevant distinction is that between the (claimed) reproducibility of the entire experimental process and the (claimed) reproducibility of the result of this process. An important point of this distinction is that the latter notion, which is also called replicability, implies the reproducibility of the result through a number of possibly radically different experimental processes. Both notions play an important role in scientific practice. On the one hand, if an entire experimental process is reproducible, this fact will facilitate its technological use. For instance, the reproducible procedures of Justus von Liebig’s experiments in organic chemistry definitely facilitated the technological production of artificial fertilizers (even if the full implementation of this agricultural technology, in line with the discussion in Section 5, required further research and additional knowledge). On the other hand, if the result of an experimental process is replicable, it may be considered

Science, Technology and the Science–Technology Relationship

87

in abstraction of the original experimental process through which it was produced thus far. This kind of abstraction constitutes a first step towards a wider theoretical treatment and understanding of the meaning and implications of this result. Suppose, for example, that certain reproducible experimental processes in a ruby crystal result in the production of a laser beam. If this result is replicable, it will make sense to abstract it from the specific processes in ruby crystals and to study the phenomenon of lasing from a more general, theoretical perspective. This argument may be summarized by saying that theoretical concepts possess a nonlocal meaning, that is to say, a meaning that essentially transcends the meaning they have as interpretations of the local experimental processes to which they have been applied thus far. I conclude that the meaning and function of theories cannot be reduced to their guiding function in producing particular experimental artifacts. This conclusion undermines the core of Lelas’s science-as-technology view, as well as the similar views of other philosophers, such as Latour, the early Habermas, Heidegger and Janich.21

7

CONCLUSION

In this chapter I have addressed the relationship between science and technology, primarily from a conceptual-theoretical perspective but with a keen eye for their actual practices. As we have seen in Section 2, strict definitions of (the aims of) science and technology, in the sense of one or two characteristics that constitute necessary and sufficient conditions, are hard to come by. All attempts to provide essentialist definitions of science and technology prove to be questionable (cf. [Mitcham, 1994] and Mitcham and Schatzberg’s chapter in this Volume, Part I). What results from the preceding discussions is a more differentiated account in which science and technology exhibit both similarities and dissimilarities. Starting from an intuitive pre-understanding that needs to be qualified or modified by empirical studies, science, technology and their relationship may be characterized by these similarities and dissimilarities, or more precisely by certain patterns that they share and by further patterns that are more typical of the one than of the other. Thus, as explained in Section 2, the intuitive idea that the design of material things and processes might constitute an essential contrast between science and technology needs to be adjusted to a pattern of similarity and dissimilarity: since design is a pervasive characteristic of observational and experimental science, the contrast merely applies to theoretical science. Section 5 shows the significance of controlling the interactions of both experimental and technological systems with their environment. At the same time, the typical dissimilarities in spatiotemporal scale and in the nature of the environment entail a number of important cognitive, 21 For an extensive historical review and an intriguing cultural critique of the science-astechnology interpretation, see [Forman, 2007], who argues that the sudden rise of this interpretation (circa 1980) is a major sign of a general turn from modernity towards postmodernity.

88

Hans Radder

material and social differences between science and technology. Similarly, Section 6 demonstrates that the notion of reproducibility applies both to science and technology. But again, an important dissimilarity arises as well, since technology focuses primarily on the reproducibility of the entire technological process while scientific practice exhibits an additional emphasis on replicability and abstraction. Thus, this line of reasoning goes against the reduction of science to technology and argues for the legitimacy of a theoretical science that is not, or at least not immediately, technologically useful.22 Section 3 shows that Mario Bunge’s account of technology as applied science is fundamentally flawed. The claimed epistemological subordination of technology to science and the alleged insignificance of practical craft work do not fit exemplary episodes of scientific and technological development. A remaining dissimilarity is a greater emphasis (in technology) on realizing external, societal objectives. Yet, even this claim needs a twofold qualification. First, such objectives are, so to speak, the distal, collective aims that need not have an immediate impact on the proximate aims (and hence on the ‘outlook and motivation’) of the individual technologists. Furthermore, as I emphasized in Section 2, basic science — in particular contemporary basic science — may just as well be oriented towards such distal aims. More generally, in agreement with the finalization theory discussed in Section 4, the notion of ‘application’ has become too closely linked to views similar to those of Bunge. Hence, to keep using this notion seems to be ill-advised. Instead, I suggest the locution ‘the uses of science’. Of course, simply replacing ‘applying science’ by ‘using science’ is not very helpful either. We need to specify this phrase in a fourfold way. That is to say, we need to pose and answer the following questions: which aspects of science are used, with which further means, with which technological results, and for which purposes? As for the different ‘aspects of science’, we have seen that not just fundamental laws may be used, but also more local models, and not only theoretical tools but experimental or observational results and techniques as well. What we have also seen, especially in Sections 4 and 5, is that using science requires ‘further means’ in the form of substantial additional work to bridge the gaps between scientific and technological problems, results and contexts. Major examples of such further means are the development of genuinely new technological knowledge and the substantial research needed to transpose the results of successful laboratory experiments to stable and reproducible technological systems. This immediately implies a differentiation in ‘technological results’, which may be technological knowledge, technological methods and procedures, or technological artifacts and systems, including the social knowledge and social conditions needed for their stable and reproducible realization. Finally, there are the ‘purposes of using science’ in tech22 Since patentable technologies need to be ‘industrially applicable’, that is, technologically useful, the argument has significant implications for the justifiability of current practices of academic patenting [Radder, 2006, Chap. 16]. See also van den Belt’s chapter in this Volume, Part VI.

Science, Technology and the Science–Technology Relationship

89

nological projects. These purposes may be broad, societal aims, but there may also be more limited, scientific ends. Since the advancement of science is often dependent on the availability of cutting edge technological instrumentation, the end of making new instrumentation may be to feed it immediately back into the development of science itself.23 Of course, science is also often used with a view to ‘broader societal aims’. A satisfactory account of the nature and legitimacy of such aims would require much more differentiation. After all, there is a big difference between the case of a single firm wishing to produce a specific artifact for enhancing its own profit or the case of the World Health Organization urging biomedical scientists to develop more medical knowledge and technology for the purpose of a worldwide struggle against malaria. Thus, philosophical accounts of the relationship between science and technology, as discussed in this chapter, should be complemented by equally differentiated accounts of the social and normative issues that are intrinsic to the uses of science in technology.24

ACKNOWLEDGMENTS In discussions about the topics of this chapter, I have received useful feedback from several audiences. It is a pleasure to thank the participants at the preparatory Eindhoven workshop on the Handbook project (in particular, Peter Kroes), the audiences at the Bielefeld/ZiF seminars of the research group ‘Science in the Context of Application’, and the members of the research group ‘Philosophy of Science and Technology’ at VU University Amsterdam.

BIBLIOGRAPHY [Baird, 2003] D. Baird. Thing Knowledge: Outline of a Materialist Theory of Knowledge. In The Philosophy of Scientific Experimentation, H. Radder, ed., pp. 39-67. University of Pittsburgh Press, 2003. [Baird, 2004] D. Baird. Thing Knowledge. A Philosophy of Scientific Instruments. University of California Press, 2004. [Barnes et al., 1996] B. Barnes, D. Bloor, and J. Henry. Scientific Knowledge. A Sociological Analysis. Athlone Press, 1996. [B¨ ohme, 1983] G. B¨ ohme. Autonomization and Finalization: A Comparison of Fermentation Research and Fluid Dynamics. In Finalization in Science, W. Sch¨ afer, ed., pp. 53-91. Reidel, 1983. [B¨ ohme et al., 1983] G. B¨ ohme, W. van den Daele, and R. Hohlfeld. Finalization Revisited. In Finalization in Science, W. Sch¨ afer, ed., pp. 131-172. Reidel, 1983. [Boon, 2006] M. Boon. How Science is Applied in Technology. International Studies in the Philosophy of Science, 20, 27-47, 2006. [Bunge, 1966] M. Bunge. Technology as Applied Science. Technology and Culture, 7, 329-347, 1966. 23 Just think of the impact of particle accelerators and detectors on the history of twentiethcentury microphysics documented in [Galison, 1997]; other examples are the scientific uses of multi-purpose research technologies, such as the ultra-centrifuge, discussed in [Shinn and Joerges, 2002]. 24 See the chapters on normativity and values in Part V of this Handbook.

90

Hans Radder

[Bunge, 1985] M. Bunge. Treatise on Basic Philosophy. Volume 7: Philosophy of Science and Technology, Part II. Reidel, 1985. [Cartwright, 1983] N. Cartwright. How the Laws of Physics Lie. Clarendon Press, 1983. [Cartwright, 1999] N. Cartwright. The Dappled World. A Study of the Boundaries of Science. Cambridge University Press, 1999. [Collins, 1985] H. M. Collins. Changing Order. Replication and Induction in Scientific Practice. Sage, 1985. [Cuevas, 2005] A. Cuevas. The Many Faces of Science and Technology Relationships. Essays in Philosophy, 6, no. 1, 2005. Available at http://www.humboldt.edu/∼essays/cuevas.html. [De Vries, 2005] M. J. de Vries. 80 Years of Research at the Philips Natuurkundig Laboratorium (1914-1994). Amsterdam University Press, 2005. [Forman, 2007] P. Forman. The Primacy of Science in Modernity, of Technology in Postmodernity, and of Ideology in the History of Technology. History and Technology, 23, 1-152, 2007. [Galison, 1997] P. Galison. Image and Logic. A Material Culture of Microphysics. University of Chicago Press, 1997. [Gibbons et al., 1994] M. Gibbons, C. Limoges, H. Nowotny, S. Schwartzman, P. Scott, and M. Trow. The New Production of Knowledge. The Dynamics of Science and Research in Contemporary Societies. Sage, 1994. [Gooding, 1990] D. Gooding. Experiment and the Making of Meaning. Kluwer, 1990. [Grandin et al., 2004] K. Grandin, N. Wormbs, and S. Widmalm, eds. The Science-Industry Nexus. History, Policy, Implications. Science History Publications, 2004. [Habermas, 1971] J. Habermas. Toward a Rational Society. Heinemann, 1971. [Habermas, 1978] J. Habermas. Knowledge and Human Interests, 2nd edition. Heinemann, 1978. [Hacking, 1983] I. Hacking. Representing and Intervening. Cambridge University Press, 1983. [Harr´ e, 2003] R. Harr´ e. The Materiality of Instruments in a Metaphysics for Experiments. In The Philosophy of Scientific Experimentation, H. Radder, ed., pp. 19-38. University of Pittsburgh Press, 2003. [Heelan, 1983] P. A. Heelan. Space-Perception and the Philosophy of Science. University of California Press, 1983. [Heidelberger, 2003] M. Heidelberger. Theory-Ladenness and Scientific Instruments in Experimentation. In The Philosophy of Scientific Experimentation, H. Radder, ed., pp. 138-151. University of Pittsburgh Press, 2003. [Hempel, 1966] C. G. Hempel. Philosophy of Natural Science. Prentice-Hall, 1966. [Hughes, 1987] T. P. Hughes. The Evolution of Large Technological Systems. In The Social Construction of Technological Systems, W.E. Bijker, T.P. Hughes, and T. Pinch, eds., pp. 51-82. MIT Press, 1987. [Ihde, 1991] D. Ihde. Instrumental Realism. The Interface between Philosophy of Science and Philosophy of Technology. Indiana University Press, 1991. [Ihde, 2004] D. Ihde. Has the Philosophy of Technology Arrived? A State-of-the-Art Review. Philosophy of Science, 71, 117-131, 2004. [Janich, 1978] P. Janich. Physics — Natural Science or Technology? In The Dynamics of Science and Technology, W. Krohn, E.T. Layton, and P. Weingart, eds., pp. 3-27. Reidel, 1978. [Kitcher, 2001] P. Kitcher. Science, Truth, and Democracy. Oxford University Press, 2001. [Koningsveld, 2006] H. Koningsveld. Het verschijnsel wetenschap, revised and expanded edition. Boom, 2006. [Krimsky, 2003] S. Krimsky. Science in the Private Interest. Rowman & Littlefield, 2003. [Kroes, 1992] P. Kroes. On the Role of Design in Engineering Theories; Pambour’s Theory of the Steam Engine. In Technological Development and Science in the Industrial Age, P. Kroes and M. Bakker, eds., pp. 69-98. Kluwer, 1992. [Krohn and Sch¨ afer, 1983] W. Krohn and W. Sch¨ afer. Agricultural Chemistry. The Origin and Structure of a Finalized Science. In Finalization in Science, W. Sch¨ afer, ed., pp. 17-52. Reidel, 1983. [Krohn and Weyer, 1994] W. Krohn and J. Weyer. Society as a Laboratory: The Social Risks of Experimental Research. Science and Public Policy, 21, 321-334, 1994. [Kuhn, 1970] T. S. Kuhn. Logic of Discovery or Psychology of Research? In Criticism and the Growth of Knowledge, I. Lakatos and A. Musgrave, eds., pp. 1-23. Cambridge University Press, 1970.

Science, Technology and the Science–Technology Relationship

91

[Lange, 1999] R. Lange. Experimentalwissenschaft Biologie. Methodische Grundlagen und Probleme einer technischen Wissenschaft vom Lebendigen. K¨ onigshausen & Neumann, 1999. [Latour, 1987] B. Latour. Science in Action. How to Follow Scientists and Engineers through Society. Open University Press, 1987. [Latour and Woolgar, 1979] B. Latour and S. Woolgar. Laboratory Life. The Social Construction of Scientific Facts. Sage, 1979. [Laudan, 1984] R. Laudan. Introduction. In The Nature of Technological Knowledge, R. Laudan, ed., pp. 1-26. Reidel, 1984. [Layton, 1974] E. T. Layton Jr. Technology as Knowledge. Technology and Culture, 15, 31-41, 1974. [Lelas, 1993] S. Lelas. Science as Technology. British Journal for the Philosophy of Science, 44, 423-442, 1993. [Lelas, 2000] S. Lelas. Science and Modernity. Toward an Integral Theory of Science. Kluwer, 2000. [Mitcham, 1994] C. Mitcham. Thinking through Technology. The Path between Engineering and Philosophy. University of Chicago Press, 1994. [Morgan and Morrison, 1999] M. S. Morgan and M. Morrison, eds. Models as Mediators. Perspectives on the Natural and Social Sciences. Cambridge University Press, 1999. [Nauta and De Vries, 1979] L. Nauta and G. de Vries, eds. Maatschappij of methode. Bijdragen tot het internalisme/externalisme-debat. Special issue of Kennis en methode, 3, 4-198, 1979. [Petersen, 2006] A. C. Petersen. Simulating Nature. A Philosophical Study of ComputerSimulation Uncertainties and Their Role in Climate Science and Policy Advice. Het Spinhuis, 2006. [Radder, 1988] H. Radder. The Material Realization of Science. Van Gorcum, 1988. [Radder, 1996] H. Radder. In and about the World. State University of New York Press, 1996. [Radder, 2002] H. Radder. The Origin and Nature of Modern Science. International Studies in the Philosophy of Science, 16, 291-295, 2002. [Radder, 2003] H. Radder, ed. The Philosophy of Scientific Experimentation. University of Pittsburgh Press, 2003. [Radder, 2003] H. Radder. Technology and Theory in Experimental Science. In The Philosophy of Scientific Experimentation, H. Radder, ed., pp. 152-173. University of Pittsburgh Press, 2003. [Radder, 2006] H. Radder. The World Observed/The World Conceived. University of Pittsburgh Press, 2006. [Ravetz, 1973] J. R. Ravetz. Scientific Knowledge and its Social Problems. Penguin Books, 1973. [Rothbart, 2007] D. Rothbart. Philosophical Instruments. Minds and Tools at Work. University of Illinois Press, 2007. [Rouse, 1987] J. Rouse. Knowledge and Power. Cornell University Press, 1987. [Sch¨ afer, 1983] W. Sch¨ afer, ed. Finalization in Science. Reidel, 1983. [Shinn and Joerges, 2002] T. Shinn and B. Joerges. The Transverse Science and Technology Culture: Dynamics and Roles of Research-Technology. Social Science Information, 41, 207251, 2002. [Tiles and Oberdiek, 1995] M. Tiles and H. Oberdiek. Living in a Technological Culture. Routledge, 1995. [Van Fraassen, 1980] B. C. van Fraassen. The Scientific Image. Clarendon Press, 1980. [Weingart, 1997] P. Weingart. From ‘Finalization’ to ‘Mode 2’: Old Wine in New Bottles? Social Science Information, 36, 591-613, 1997. [Zandvoort, 1986] H. Zandvoort. Models of Scientific Development and the Case of Nuclear Magnetic Resonance. Reidel, 1986.

This page intentionally left blank

THE ROLE OF SOCIAL SCIENCE IN ENGINEERING Knut H. Sørensen In an era that seemingly celebrates interdisciplinarity [Nowotny et al., 2001] where technology is no longer the exclusive preserve of engineers one might imagine that it would be rewarding to review research into the influence that social science has had upon engineering. For a long time many have also argued that social science issues should be given more prominence in engineering curricula. More to the point, social studies of technology have repeatedly observed how important the understanding of the social world is to successful engineering. This emanates especially from the consistent reconceptualisation of technology as seamlessly sociotechnical, as an outcome of combining so-to-speak nature and culture (see, e.g., [Bijker et al., 1987; Latour, 1988]). Such observations raise questions about the modes of appropriation of social science that one expects to find among engineers. Broadly speaking, there seem to be two options. One is to adopt transdisciplinary collaboration so that engineers and social scientists work together as specialists from distinct professional fields. This may take the form of teamwork but it can also give social scientists the role of consultants or advisors. I will term this the transdisciplinary mode of appropriation since it involves combining knowledge from different recognised disciplines or professions. Transdisciplinary modes may also be confrontational in the sense of social scientists representing a critique of engineers’ proposals and vice versa. Exchanges across disciplines and professions bring conflict as well as consensus [Sørensen, 2008]. The second mode of appropriation is where social science knowledge and competence is assimilated by the profession of engineers to become part of an increasingly hybrid form of engineering knowledge. Such appropriation may occur during the education of engineers, through what engineers read, from interaction with social scientists, etc. This I will call the profession-based mode of appropriation, since it takes place within an ecology of knowledge production characterised by the professional autonomy of engineers in which engineers remain the active and controlling party. The relative importance of these two modes, together with details about their features, will be discussed in this chapter. The major barrier when assessing these issues is the fairly limited amount of research on engineers’ appropriation and use of social science. The comprehensive literature on social studies of technology emphasizing social aspects of engineering has not been particularly concerned with the actual sources of knowledge on Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

94

Knut H. Sørensen

the social aspects and the way in which such knowledge may be appropriated. Moreover, the field of technology studies have engaged in case studies of specific examples of technological development rather than in a broader examination of the education and work practices of engineers. For example, Trevelyan and Tilli [2007, pp. 305-306] conclude their review on research into engineering work by stating [t]hat there has been no recent, comprehensive investigation into the processes involved in engineering work as it is actually practiced. Few researchers have asked engineers what they do and none has asked where they acquired the skills they use; nor have we found any systematic research on the links between what is taught in engineering institutions, what graduates learn early in their careers, what training engineers undertake while in the workforce, and how any of this contributes to producing competent engineers. While they overstate the problem, the main thrust of their argument is justifiable. The volume of research on engineers and engineering is not extensive and the available information about, and analysis of, the role of social science is even more limited. The dominant focus in the literature is the social status and professionalism of engineers, not their actual practice with respect to technology [Sørensen, 1998]. In addition, we face inherent theoretical and methodological difficulties when studying the role of social science knowledge in engineering. In Section 1, I shall briefly examine some of them because they are important for clarifying how the issue in hand — the influence that social science has on engineering — may be understood and discussed. This will serve as a backdrop to the exploration of the two main areas of information pertaining to the role of social science in engineering: education and work or design. Section 2, on education, will explore the place ascribed to social science in engineering curricula. I shall draw especially on a Norwegian study but will also review some programmatic papers about the kinds of skills engineers require. Then, in Section 3, I will briefly discuss the use of social science in the field of information systems design before turning to engineering work and design more broadly in Section 4. Section 5 summarizes the main arguments.

1 THEORETICAL AND METHODOLOGICAL ISSUES Should we expect to be able to detect social science input in engineering and engineering design? Clearly, this a complex issue. How would we unambiguously identify a fact or an insight that identifies design or engineering as originating from the social sciences? A main achievement of historians of technology has been to show that engineering is an independent science, or rather a set of sciences, and not just applied natural science (see, e.g. [Layton, 1971]) or applied social

The Role of Social Science in Engineering

95

science. Clearly, natural science does have a prominent role in the development of modern technology but new technology or solutions to engineering problems are not usually produced through straightforward applications of established natural science facts, theories and discoveries. The main point has been succinctly summarised in the critique of the so-called linear model of innovation (see Radder and Channell, both in this volume, Part 1, for a more detailed discussion). It is a critique that emphasizes that innovation involves autonomous creative acts directed at assembling a mix of relevant kinds of knowledge. Science may play a part as a supplier of such knowledge but not in every case, and the links between scientific knowledge and design choices have proven to be difficult to trace [Kline and Rosenberg, 1986]. Analysing the role of social science knowledge in innovation and engineering may be an even more complex task than in the case of natural science. Firstly, there are many different ways in which results from social research may be implemented, and such use may not always be constructive. For example, Weiss [1979] distinguishes five ways in which policy-makers make use of social science: (1) instrumental use where research results are used in problem-solving, (2) political, conflict-related use where research is used as an argument or weapon in a political controversy, (3) enlightenment because social science research leads users to conceptually re-orient or change their ways of thinking, (4) interactive use, where research is applied in combination with other information to provide a knowledge base for policy purposes and (5) tactical applications where research is used to create a feeling of change or where it becomes part of an “avoid or delay” strategy. The role of social science in engineering could display a similar variety of practices. Secondly, social science knowledge claims, which are often controversial and unstable, are characterised by disagreement and are thus difficult to apply in a setting where one does not wish to take a stand on social issues. At a more basic level, social science representations may interact with and transform the very phenomena that are to be represented (see, for example, [Suchman, 2007]). This is the problem aptly characterised by Giddens [1976] as the double hermeneutic circle: social scientists interpret the world, but the world also interprets social scientists. Thirdly, in general, the social sciences have not given priority to research that aims to be relevant to engineers and to their efforts to design and innovate. Thus, the availability of off-the-shelf social science knowledge applicable to engineering work, together with a bank of social scientists interested in interacting with engineers may prove to be a greater limitation than in the case of natural science. However, these problems should not be overestimated. Many social science-related issues or questions emerging from engineers’ work may be tackled by drawing upon well-established social science knowledge or skills. For example, many engineers base their work on too simplistic assumptions about human behaviour, assumptions that may be disputed or rectified on the basis of fairly standard knowledge about decision-making, consumption or phenomena like cognitive dissonance. Overall, there is a range of situations related to technological design and other forms of engineering work that raise questions relevant to social scientists;

96

Knut H. Sørensen

some that can be quickly answered and some that make it necessary to co-produce new social science knowledge with new engineering knowledge. Fourthly, social scientists are often perceived by engineers to be critical conversationalists engaging in ‘philosophy’, a thing that is not appreciated in a profession which values, above all else, hands-on problem-solving engagement. From interacting professionally with engineers over a long period of time I know from experience that social scientists may be seen as unhelpful sceptics rather than as constructive team-workers. Similarly, many social scientists are critical of engineers whom they tend to find difficult to communicate with and insufficiently reflexive with respect to the effects of their work. Thus, collaboration is not easy. We could, of course, sidestep the challenges discussed in this section by basing the analysis of the impact of social science on engineers’ own accounts of how and to what extent they make use of such knowledge. However, this would at best tell us about the instances where such use was explicit. It seems more probable to assume that if engineers really used knowledge gained from social science, this fact would tend to be rendered invisible in their accounts of developing new technologies because the social sciences are less prestigious than natural science. Social science contributions might consequently become prone to being overlooked or hidden because acknowledging the value of social science input might ultimately damage reputations. Such acknowledgement might even be thought to endanger the scientific status of engineering. It is possible to have a good intuitive understanding of the social conditions of engineering performance without consciously drawing on social science research. For example, such knowledge may be mediated through the mass media, thus being appropriated from journalists’ reports rather than scientific accounts. Alternatively it may become a part of the standard secondary school curriculum. The point is that engineers may be affected by social science without being aware of that form of appropriation. The underlying problem is aptly summed up in the arrogant claim made by a Norwegian research director after a seminar when he stated: “We are all sociologists. We all read newspapers!” In many accounts of successful inventions and innovation a good understanding of user needs and the social context is vital [Freeman, 1982; Bijker et al., 1987; Latour, 1987]. A classic example is Thomas Edison’s assessment of the competitive situation concerning gas when he embarked on the invention of electrical lighting. The system was designed by optimising the cross-section of copper power cables in relation to the price of gas and copper so that electrical lighting could be made cheaper than gas lighting [Hughes, 1985]. Similarly, the Norwegian effort to develop a technology to extract nitrogen from air in order to produce synthetic fertilizer was based on the engineer-entrepreneur Sam Eyde’s comprehensive study of the international fertilizer market and the decline in supply of guano [Sørensen and Levold, 1992]. One way of interpreting such observations emerges from Callon’s [1987] study of an early French effort to develop fuel cells for cars. He observed how engineers produced fairly complex social scenarios to support their project which led him

The Role of Social Science in Engineering

97

to make the — admittedly polemical and overstated — claim that engineers are better at sociology than sociologists. While Callon correctly reminds us that engineers need to possess an awareness of social needs and interests where social scientists may be unable to offer useable knowledge, it is nevertheless something different from what concerns us here. Callon’s claim raises the question of how engineers may learn to act as competent producers of social scenarios. Is this a skill inherent to engineering or is it a qualification that is developed through a mix of experience and exposure to social science observations? A good place to start studying such issues is in the education of engineers. First of all, engineering curricula provide evidence of the extent to which social science is incorporated in the training of engineers. What is of equal importance is the fact that the education of engineers frequently gives rise to debates about what engineers need to know. Such debates would be interesting topics of study because they represent good opportunities to voice the need for change in the education of engineers. 2

WHAT ENGINEERS NEED TO KNOW

Comparative studies of engineers have shown substantial variations in a multitude of dimensions such as status, professional orientation, placement in industrial hierarchies and the relative importance given to theoretical and practical competence [Maurice et al., 1986; Meiksins and Smith, 1996; Sørensen, 1998]. This reflects differences in the roles of engineers in national division of labour systems as well as in relation to historical traditions. Still, it seems that there are some similarities, like the strained relationship between theory and practice, between the perception of engineering as science-based as opposed to growing out of practical, industrial concerns (see also Banse and Grunwald’s chapter in this Volume, Part I). This issue partly emerges from concerns about the social status of engineering. In most countries, having a scientific background is more prestigious than being versed in practical skills. The matter of establishing what engineers need to know has been a controversial issue. Efforts in the late 19th and early 20th centuries to provide engineers with a professional status resulted in the combined challenge of acquiring scientific standing as well as being recognised as educated. This challenge was met in three main ways. In the French tradition, in which the Ecole Polytechnique is the paradigmatic institution, academic status could be achieved by heavily emphasising science and mathematics. The North-American tradition placed greater weight on making liberal arts courses a compulsory part of the engineering curriculum to achieve recognition as educated, while in Germany endeavours to turn engineering into a science with an academic status were successful [Kranakis, 1989; Lundgreen,1990; Manegold, 1978; Noble, 1979; Shinn, 1984]. The sociology of professions has compared engineering to professions like law and medicine and found engineers wanting with respect to autonomy and social status. It has been asked if engineers really constitute a profession. Within the con-

98

Knut H. Sørensen

text of this chapter another concern emerges from the consistent differentiation of engineers into specialist fields like mechanical engineering, civil engineering, chemical engineering, and so on, specialisations that often have their own professional societies. Moreover, as Bucciarelli [1994] shows, these specialists — trained in distinctly different ways — see technologies and define problems in specific ways that cannot easily be communicated across the educational divides. Nevertheless, for the purposes of this chapter, it seems appropriate to regard engineering as a single profession, at least in a national context, partly because engineers are educated similarly in similar types of institutions and partly because they are also organised into general engineering societies designed to help provide a common engineering identity. What is gained from employing the concept of professions is not just to remind of the phenomenon of education-based enclosures within the labour market but also to point to the existence of a regime of organising knowledge that differs from discipline-based patterns. Professions are characterised by outspoken theory-practice concerns involving efforts to strike a balance between the respective importance of professionals’ experience and research-based knowledge. It may be argued that the education of professionals mixes discipline-based knowledge with specialised professional knowledge. In engineering education, this mix has resulted in curricula that contain mathematics and natural science subjects in combination with several kinds of engineering science. Such broad input has provided what may be called polytechnical competence in individual engineers. The term polytechnical is used to highlight the tradition of giving engineers a fairly broad-based education, involving introduction to the basic competence of several fields of engineering and with finally cultivating an area of specialisation. We may recognise this as a particular form of interdisciplinary education. Such interdisciplinary, polytechnical training of for example mechanical engineers would involve teaching them basic knowledge of civil engineering, electrical engineering, chemical engineering, and so on. This corresponded to the demand for less specialised, broadly competent engineers, which dominated in most countries in the early modern period. Social science and humanities subjects could be part of the broad knowledge base, in which case these forms of knowledge were integrated into the hybrid framework created along the lines of the polytechnical knowledge strategy. The result was an individualised polytechnical type of interdisciplinarity which was innately different from the type of interdisciplinarity characterised by specialists collaborating in teams [Sørensen, 1996]. Science, mathematics and the liberal arts have been used to strengthen the social status of engineers, as outlined above. Insofar as social science subjects were made part of engineering curricula, the underlying reasoning seems to have diverged. It was believed that engineers needed to know something about management and business. Subjects like economics, law and business administration were thus taught at many if not most institutions of engineering education. However, these fields of study were seen as peripheral even though many, if not most, engineers tended to embark on management careers, at least until the end of the 1970s [Sørensen, 1998].

The Role of Social Science in Engineering

99

A recent study on the changes in educational ideology with respect to the training of engineers at the Norwegian University of Science and Technology (NTNU, previously the Norwegian Institute of Technology) that took place between 1910 and 2006, provides detailed insight into these underlying considerations [Amdahl and Sørensen, 2008]. The higher education of engineers in Norway was initially particularly shaped by the German tradition adhered to at the Technische Hochschulen but after 1945 it was the US that was turned to for inspiration. The source material — containing the matriculation speeches of the Presidents of the Institute and the reports made by committees engaged in curricular reform — clearly indicates that there was ongoing dialogue with institutions devoted to the education of engineers in other countries. The observations made in this study should therefore have more general validity, even in view of the fact that Norway is a small country with fairly small industrial enterprises. This undoubtedly explains why the polytechnical ideal was probably dominant here for longer than in most other advanced economies. Large companies make better use of specialised engineers than small companies. However, in Norway just like in most other countries, the polytechnical ideal is in decline. The traditional outlook on specialisation among those responsible for educating engineers in Norway was aptly described by the Institute’s President Olav Heggstad in his matriculation speech of 1932: Here, you [the engineering students] will not be educated as specialists but will receive a comprehensive education in a broader professional field. For a time, there was a strong mood for specialisation across the institutes of technology. But this idea has increasingly been departed from among other things because, after the education has ended, it is not certain that the graduate engineers will find employment in their field of specialisation [Amdahl and Sørensen 2008, p. 55]. Later Presidents and curricular reviews emphasized that Norwegian industry consisted largely of small companies in need of engineers with broad areas of competence — polytechnical as defined above — rather than specialised. It was not until the 1970s that the importance of specialised knowledge among engineers was fully recognised. The notion of the engineer as a general kind of practitioner was an indication that individually based polytechnical interdisciplinarity was the dominant mode of education. If the humanities or social sciences were to become integrated in engineering practice, this would have to involve adding such topics to the engineering curriculum. At the Norwegian Institute of Technology there were some social science topics in the curriculum when the Institute was established in 1910 but the scope was modest and the main focus was on certain aspects of economics and a little bit of law. For a long time, the presidents of the Institute mentioned the need for more such topics in their matriculation speeches. Usually they concluded — occasionally with remorse — that such needs could not be catered for. The need for social science-related subjects was not considered large enough to merit

100

Knut H. Sørensen

curricular space. Obviously the various presidents presented their considerations in different ways. Firstly there was a set of responses that dismissed the need for more social science in the engineering curriculum and emphasized that technology represented a prominent cultural element in itself. President Alfred Getz eloquently formulated this point of view in his 1916 matriculation speech: And nevertheless, technology is truly a means of education. It is pure and ideal. Like the artist, the creative power of accomplishment of engineers resides in inner vision and, like the artist, the engineer also has to grapple with the fabric in order to fulfil the spiritually envisaged reality. From this perspective, engineers could easily cope on their own without professional input from the social sciences. Their own cultural capacity would be sufficient to respond to the need of understanding social issues and concerns. Similar ideas have resurged from time to time as counter-arguments to accusations that engineers are narrow-minded cultural dupes (see, e.g., [Florman, 1976]). Secondly, many presidents emphasized that later in their careers, often as managers, engineers would need additional skills like a knowledge of foreign languages, psychology, organisation theory, etc. Students should therefore seek to acquire such skills, but — unfortunately — they would have to do that in their spare time. There was no room for such enlargement of the engineering curriculum. This was also indicative of the perceived low status of social science. A third set of concerns related to the impact of engineering on society and to the social responsibilities of engineers. Such issues were voiced from time to time from 1910 onwards but the plea became more persistent after 1970, clearly in response to discourse that saw technology as a potential social and environmental threat. In his matriculation speech of 1991 President Karsten Jacobsen went so far as to argue that the future of engineering would be shaped by the tension between technology and human concerns: It is no longer sufficient to know one’s discipline; the technologist of the future has to enter the playing field with quite a different and more general value base and outlook than before, with an ability, will and training to face the consequences of this in practice — to see actions from a broader view — what we could call a holistic perspective — technological-ecological-human-aesthetic-economic. [Amdahl and Sørensen, 2008, p. 61] However, at this point in time, specialisation had become the dominant theme underlying the education of engineers. It was not really believed that Norwegian engineering students should receive a much broader and comprehensive training in the social sciences and humanities to be able to act on the challenges outlined by Jacobsen. Rather than asking for new types of knowledge, Jacobsen and indeed later presidents, spoke about the issue in terms of interdisciplinary collaboration.

The Role of Social Science in Engineering

101

The students needed educational reform but this reform had to focus on instilling new virtues rather than on gaining new knowledge. The new virtues called for included, of course, a broader outlook on engineering and greater sensitivity to the social and environmental impacts of the work of engineers. However, even more prominent in the presidents’ speeches was the ideal that engineering students should be encouraged to become skilled in and willing to engage in interdisciplinary activities that also involved graduates from the humanities and the social sciences, and vice versa. As President Eivind HiisHauge put it in his matriculation speech of 2004: No single person is able, with sufficient depth, to be interdisciplinary alone: it is the capacity to perform constructively and to be committed to working together with others in teams which gives results. It was an appeal to students from all academic fields to be prepared for interdisciplinary collaboration. If we return to the two modes of appropriation of social science in engineering proposed in the introduction, we can see that both are emphasised in the above quotation. For a long time, the presidents’ speeches together with all the curricular reviews voiced the opinion that engineers should be able to cope with a broad spectrum of challenges including certain social science concerns on their own. In that way, they were emphasising what I have termed the profession-based mode of appropriation. Social science input were seen as potentially important by some, but ultimately it was given little curricular space and assimilated into the broad polytechnical education of the individual engineer. The switch to a more collective, heterogeneous kind of interdisciplinarity observed in the last decade has accentuated the transdisciplinary mode of appropriation of social science. In the end it may pave the way for a greater number of social scientists working together with engineers — and thus for social science in a distinct and visible form — within the framework of technological development. The critical question is, of course, whether this latter alternative has been realised? What modes of appropriation are most important to engineering education and the professional development of engineers? The US-based Accreditation Board for Engineering and Technology (ABET) is important when it comes to the world-wide setting of standards for engineering education. Their lists of criteria do emphasize teamwork — with or without the prefix ‘multidisciplinary’ — the social sciences though tend to be referred to in an indirect and imprecise manner, like in the mention of the need for students to acquire ‘a broader outlook’ and understanding of the social, economical and political constraints on engineering work, together with an awareness of the importance of social responsibilities.1 Both the social sciences and the humanities are mainly seen to contribute insight into the ethical, legal and social aspects of engineering — all the so-called ELSA concerns also present in the Norwegian situation. The 1 http://www.abet.org/forms.shtml#Applicable

to All Programs (downloaded 30 August 2008).

102

Knut H. Sørensen

ABET criteria therefore invite both modes of appropriation of social science without placing any great emphasis on either mode. Recent discussions on engineering education include contributions that argue in favour of the importance of profession-based appropriation. For example, Vesilind [2001] claims that the traditional view of the encyclopaedic engineer should be maintained but changed — engineering should not just be seen as ‘applied natural science’ but also as ‘applied social science’. Facets of this view seem fairly widespread in the international literature on the skills that are essential to practising engineers and ideas on how engineering education could be changed. Ravensteijn et al. [2006] suggest, for example, the need for engineers to be more communicative. Nguyen [1998] lobbies for communication skills in combination with many other competences related to the various business demands. Grimson [2002] and Robinson et al. [2005] emphasise that engineers need to possess a broader set of non-technical skills. In line with this, Jones [2003] argues that the image of the renaissance engineer might actually be an appropriate educational reform goal; but the main issue is this: what should be the components of appropriate encyclopaedic knowledge? On the other hand, Russell and Stouffer [2005] show how U.S. undergraduate civil engineering education is overwhelmingly dominated by technical subjects, with little indication that profound changes are taking place. This supports the impression given by the ABET list of criteria for the accreditation of such programmes. The U.S. National Academy of Engineering has carried out extensive reviews into the situation of engineers in 2020. These discussions only relate vaguely to the potential role of social sciences with respect to engineering education and work. Interdisciplinarity is signalled as important and social skills are claimed to be important, which means that both modes of appropriation of knowledge of the social sciences are inherently present. However, the overriding impression made by the reports is the assumption that the engineering profession will continue to be largely self-sufficient, thus demonstrating that the profession-based mode of appropriation is given priority [National Academy of Engineering, 2004; 2005]. A different perspective is provided by Williams [2002; 2003]. She argues that during the last few decades, the former close link between technology and engineering has been broken. Technology is no longer exclusively the domain of engineers; engagement with technology has far outgrown any single professio. Williams’ assertion has important implications, not just for engineers but also for social scientists who need to reflect much more about the implications of this change in relation to their own practices. Probably neither group yet fully grasps where this will lead; they may not even have discovered the ongoing change. If Williams is correct, the future development of technology will include interdisciplinary encounters of many kinds but the present developments, as enumerated above, are not promising in this respect. The education of engineers and of social scientists seems to be well entrenched in the established perception that the technical and the social aspects of modern society are worlds apart.

The Role of Social Science in Engineering

103

Williams identifies as a serious challenge what she perceives as the way in which the engineering profession is currently developing along two main lines. One camp is concerned with doing real engineering by designing and building useful things that actually work. The other camp, she observes, advocates a new emphasis on large technological systems and management: Both the design movement and the systems engineering movement seek to reclaim a distinctive identity for engineering: to proclaim that here is something engineers do that scientists and businessmen do not do. In the end, however, the reclamation efforts only underscore engineering’s loss of identity. In both design and systems work, many people other than engineers are in on the act. In design today, engineering, programming, science, language, and art converge. In dealing with technological systems, it is even more obvious that engineers have to collaborate with political scientists, economists, lawyers and managers [Williams, 2003. p. B12]. On the one hand, it seems that Williams is correct to note that late modern technology is embracing an increasingly wider body of disciplines and professions. Consequently, engineering will become more and more engaged in broad transdisciplinary collaboration, also with social scientists and, for that matter, humanities scholars as well. Engineers seem to put considerable effort into preserving the boundaries of their profession and their professional influence, for example by giving priority to the profession-based rather than to the transdisciplinary appropriation of social science. How may such prioritising of profession-based appropriation be achieved and how may that shape the intake of social science knowledge and skills?

3

MAKING SOCIAL SCIENCE THEIR OWN? THE EXAMPLE OF COMPUTER SYSTEMS DESIGN

The area of computing, in particular the sub fields involving the design of information systems, may provide interesting insight into the way in which social science knowledge may be appropriated in engineering. To begin with, we should note that computing — in particular with respect to areas like information systems design, information systems analysis, and software engineering — does draw on methods as well as research findings from social sciences. It is commonly acknowledged that practitioners in these areas need to possess broader-based knowledge. Early examples are Vitalari [1985] and White and Leifer [1986] who argue that the knowledge base of systems analysis should be broad and should include a variety of technical and non-technical skills. For such reasons, some argue that software or information systems engineering is not really engineering at all, but something quite different (see, e.g. [Davis, 1998, pp. 31-40]). Given the existing diversity within the loosely defined profession of

104

Knut H. Sørensen

engineering, such boundary work seems less fruitful. However, software or information systems engineering probably struggles more explicitly than most other fields of engineering with issues similar to those that concern social scientists. In this respect, engineers working with software and information systems will probably find that they are particularly engaged in the appropriation of knowledge and skills gained from social science, thus allowing us a better understanding of the processes and content of such intake. Recent research confirms the early arguments to the effect that software and information systems engineers need broad skills, both technical and non-technical. Iivari et al. [2004] maintain that the distinctive competence of information system experts lies (1) in their expertise of aligning IT artefacts with the organizational and social context in which the artefact in question is to be used, (2) in identifying and specifying the needs of people who are supposed to use the system, (3) in organizational implementation, and (4) in the evaluation/assessment of these artefacts and related changes (see also Radder in this volume, Part V). While the emphasis may vary, such observations of the need to combine technical and non-technical skills seem commonplace [Goles et al., 2008; Lee, 2005; Litecky et al., 2004; Turley and Bieman, 1995]. In fact, the requirements are seen as quite comprehensive. Lee [2005, p. 90] summarizes this succinctly when he remarks that organizations ‘expect their systems analyst to become all-round athletes who play every corner of the field’. Such ideas are also used to argue that a broad education is required if IT specialists are to be properly prepared to combine technical and non-technical challenges in their professional practice [Brookshire et al. 2007, Dahlbom and Mathiassen 1997]. However, Brookshire et al. end up by proposing a fairly conservative curriculum comprising predominantly technical courses. Dahlbom and Mathiassen, on the other hand, suggest taking a much more radical step. They argue that [s]tudents of computing should develop the ability to ask serious questions about the social impact of computing and to evaluate proposed answers to those questions, and they must be able to anticipate the impact of introducing a given product to a given environment. [1997, p. 84] What we observe in this literature is that there is a clear tendency to want to promote individualised polytechnical interdisciplinarity, which is similar to what we observed in Section 2. Even if these authors present their arguments in different ways, their main thrust is that computer professionals should be self-sufficient in terms of the competencies required to carry out IT work. Social science may be a resource, but it should be presented as something that is integrated into systems design or software engineering. In the final instance, the origin of social science input is rendered unclear through the insistence on profession-based appropriation. This is not, of course, a problem in itself. If indeed attainable, individualised polytechnical interdisciplinarity of such comprehensive scope is extremely demanding and therefore also risky with respect to the quality of the outcome.

The Role of Social Science in Engineering

105

Research areas like Human Computer Interaction (HCI) and Computer Supported Collaborative Work (CSCW) are definitively meeting places for scientists with diverse disciplinary backgrounds, including social science. The shared concern for developing more useful computer systems also produces a shared interest in the ways humans interact with computers, what strategies evolve from such interaction and how greater benefits can be achieved from computers. In early participatory computer system design efforts and trade union involvement, social scientists collaborated with computer scientists to organise participation and to establish methods that were helpful for workers to articulate their needs and requests [Ehn, 1988]. However, a main contribution that social scientists have made to HCI and CSCW has been to underline the complexities of human actions and the deep-rooted problems involved in predicting and stabilising human interaction with machines [Suchman, 1987; 2007]. Such critical interventions seem to have been made notice of, but their actual appropriation is less clear, probably because these kinds of insights are difficult to integrate into computer science methodology which, to some extent, depends on achieving some level of prediction and stability. The reviewed literature about software and information systems engineering shows the presence of both modes of appropriation in the case of social science. However, it is particularly the discussion about the training of such engineers and the emphasis placed on a broad knowledge base that also includes strong non-technical components proving that even in this case where concerns for social issues are so prominent, profession-based appropriation dominates. Furthermore, what is appropriated from social scientists is, above all else, methodology. Some software and information systems engineers seem to want to perform their own social science type of investigations but mainstream approaches to the design of computer systems seem less aware that parts of their work — for example the modelling work practices of customers — could just as well be construed as a job for social scientists. When practitioners are asked about the competences they need to carry out their work, they tend to emphasize skills like communication and the understanding of people’s needs. When asked how they can acquire such skills, they tend to point to their experience — not to any form of social science input [Sørensen et al., 2007]. The potential complexity of the information system designing required to facilitate decision-making and to access the knowledge needed to make proper decisions, is considerable. In the long run, this may make the profession-based appropriation of social science inadequate. However, the challenges involved in establishing good practices based on transdisciplinary appropriation may be substantial, not least because of the fairly tight professional collaboration required to achieve accurate forms of knowledge integration. As Lagesen and Sørensen [2008] demonstrate, it is common to assume that communication practices may be separated from computer-related practices like programming, making the first a task for social scientists and the second a job for software specialists. However, Lagesen and Sørensen found that the claim made by software specialists to the effect that a knowledge of computers and programming is needed if they are to communicate

106

Knut H. Sørensen

properly with customers or product users was convincing. If social scientists are to be part of such a process they, like the software specialists, will need to engage in reciprocal profession-based appropriation. Thus, the issue is not perhaps so much that of replacing one strategy of appropriation with another as that of combining them. However, at present, there is little doubt that the pursuit of profession-based appropriation dominates thanks to the prevailing position of individualised polytechnical knowledge performance. Is this situation more generally characteristic of engineers? Our analysis of engineering education does point in that direction but what may be observed from engineers working to develop technology? 4 WHAT SOCIAL STUDIES OF ENGINEERS AND DESIGN TELL US ABOUT THE RELATIONSHIP WITH SOCIAL SCIENCE Probably the most comprehensive studies of engineers and engineering work are those that have emerged from the history and sociology of professions. A main finding was that the professional behaviour of engineers is characterised by less autonomy and a larger degree of collectivist culture than that which applies, for example, to medical doctors or lawyers [Gerstl and Hutton, 1966; Perruci and Gerstl, 1969; Ritti, 1971; Hutton and Lawrence, 1981; Zussman, 1985; Whalley, 1986]. Still, engineers have considerable autonomy as ‘trusted workers’ [Whalley, 1986], even if they work under managerial control, in particular with respect to resources and deadlines [Meiksins and Watson, 1989]. According to these studies, engineers are mainly engaged in a diversity of technical and non-technical work, unless they move into management, which used to be a common career move [Sørensen, 1998]. Since of the focus of these studies is professional behaviour, there is little substantive discussion of the actual content of engineering knowledge and thus of the modes of appropriation of, for example, social science. This even applies to most ethnographic studies on engineering work (see [Bucciarelli, 1994; Downey, 1998; Forsythe, 2001; Vinck, 2003]). Vincenti [1990], whose main concern is engineering knowledge, does not provide insight into such appropriation processes either (see also [Downey and Lucena, 1995]). As already mentioned, management, business administration and economics are long-standing professional interests of engineers. Some such knowledge is a standard part of engineering curricula and many engineers become further educated in these fields. A combination of engineering knowledge, computer skills and knowledge about finance or management has become the basis of careers in consulting [Williams, 2002]. However, this trajectory substantially extends the idea of what constitutes engineering and more to the point, what is normally taken to constitute social science. Rather than being a good example of how the social sciences affect the work of engineers, it points to the development of a set of practices centred on the construction of mathematical or computer models in which technology and social issues tend to be represented only in a very abstract and oversimplified fashion.

The Role of Social Science in Engineering

107

For the purposes of this chapter, design remains a more interesting topic and one in which social science might be expected to be a potentially useful resource. In the previous section, we observed how some such influences could be traced in information system design. However, we also saw that this influence was largely attributable to computer scientists’ profession-based appropriation of social scientific knowledge and methods. This raises interesting questions about the nature of such processes of appropriation in engineering design as well as about the kind of knowledge designers want. It is tempting to assert that social scientists could present engineers with clearly defined design criteria related to user needs, the cultural conditions of domestication of new products, etc. To counter such ideas, Williams et al. [2005, p. 102] warn us about what they call the design fallacy, ‘the presumption that the primary solution to meeting user needs is to build ever more extensive knowledge about the specific context and purposes of an increasing number and variety of users into technology design’. They base this warning on the problems encountered with linear thinking as discussed in Section 2 of this chapter. Still, designers tend to base their notions on certain ideas about future use and users [Akrich, 1992]. Where do such ideas come from? There are many sources that can provide designers with information about users and with ideas about how products may be shaped [Walsh et al., 1992; Williams et al., 2005]. Arguably, social scientists are skilled at analysing the use and users that could produce insights that would be conducive to the development of new products. Increasingly, companies are using market research to inform their efforts with respect to design and innovation. However, such research seems above all to be used to identify potential groups of users/customers, for example according to gender, age, etc., which means that its impact on the work of engineers tends to be a point of departure in terms of design rather than in terms of informing concrete problem-solving efforts [Cockburn and Ormrod, 1993; Chabaud-Rychter, 1994]. Engineers have to use other resources to interpret what is needed to design technology that can be considered appropriate to targeted groups of users, like kitchen appliances for women or microwave ovens for young men. While there are ways to translate user requirements into design specifications (see the contribution of De Vries in this volume, Part III), these methods tend to be vague when it comes to defining what it is that users want. The most common resource used to interpret user requirements is not social science but the personal experience, knowledge and taste of the designing engineers. For example, engineers involved in design are frequently observed to implicitly model the intended user after themselves or to invoke stereotypes [Berg, 1994; Oudshoorn et al., 2004; Williams et al., 2005]. Usability trials may be employed to test how a given design matches user needs and tastes, but most technologies are developed without such testing. Moreover, usability tests are generally undertaken by engineers. Arguably, the study of users could amount to an area of interaction between social science and engineering knowledge. Experiments in participatory design

108

Knut H. Sørensen

have demonstrated the potential to provide better computer systems [Ehn, 1988] but such approaches are expensive and the results have not been convincing. A leading drawback is the widespread perception among designers that users are conservative, the implication being that participation needs to be limited to the shaping of the user — technology interface, while decisions about technology choice and emerging new practices are seen as the prerogative of designers [Hatling and Sørensen, 1998]. Williams et al. [2005] identify a range of other difficulties, for example the fact that user groups may change during the lifecycle of a product. A major challenge is the inherent instability of needs and tastes — what users want, may change as the technology in question develops. It may thus seem more tempting to employ a kind of trial and error approach than to carry out a comprehensive study of users as the backbone of design. In turn, this may make it less interesting for engineers to collaborate with social scientists, partly because they cannot provide the well-defined answers that engineers are looking for and partly because the experiments with potential users that could be undertaken to obtain some relevant input on the design process appear to be too expensive. Trial and error may actually be cheaper, at least if it can be organised on a small scale. Clearly there are huge challenges attached to translating the — already existing or specially produced — knowledge gained from social science into design criteria, shapes and functions. What does the fact that an artefact is easy to use, that a system is efficient or that a machine is flexible actually mean? The challenge is well illustrated by Cockburn and Ormrod [1993]: how can a microwave oven be made attractive to young men? For example, why should one assume, as did the company, that the colour brown is more ‘gender authentic’ to young men than the colour white? These challenges could be viewed as the appropriate tasks of interdisciplinary teams, combining the skills and competencies required to fulfil the relevant problemsolving. Such interdisciplinary practices combining the social and engineering sciences [Sørensen et al., 2008] are not, however, widespread. Instead, as was observed with computer systems designers in the previous section, engineers seem to prefer a professional mode of operation which enables them to opt to independently access information and knowledge about social and cultural aspects mainly by drawing on experience and only occasionally by delving into material related to social science. For example, it is evident that books like those written by Suchman [1987; 2007] or Norman [1988] are being discussed in various design communities. Probably this has resulted more from engineers adhering to a profession-based appropriation of the work rather than to them engaging in interdisciplinary collaboration to combine ideas. Is that a problem? Social scientists might be inclined to think so but if we try to favour one interpretation over the other we will run into difficulties. In the first place there is often dissent among social scientists themselves about how precisely social scientific research should be interpreted. Secondly, and more importantly, what is really at stake here is: the correctness of the interpretation or the quality of the resulting design? On whose premises do we base our decisions — those of the so-

The Role of Social Science in Engineering

109

cial scientists’ or those of the engineers’ ? Does it matter if social scientists claim that engineers have made an incorrect interpretation? From the point of view of engineers — and probably most people — what lies at the heart of the matter is the quality of the resultant design. To prove that a social scientist’s interpretation is ‘better’ than an engineer’s, we need to show that better interpretation leads to better design. Or, to put it more generally, that transdisciplinary appropriation is a more fruitful mode than the profession-based mode. There is no research available to help us to settle this point, so in a sense social scientists are left to prove their own worth. Perhaps such proving is easier in areas like economics and accounting? Financial constraints and economic motives definitely underlie technological design but it is not easy to unambiguously translate technology into economic potential. To a certain extent costs may be predicted and much research has been done into project management, cost control, etc. inside and outside the field of engineering but this aspect of technological development also involves considerable risk as evidenced by the frequent overspending in many projects. According to Thomas [1994], making cost calculations that are perceived by management to be realistic is a prerequisite to starting projects on the designing of new or improved technology. However, whether or not these calculations prove correct is a different matter; they often do not but by then the engineers who initiated the project, tend to have moved on to other projects. Thomas therefore concludes that the creating of new projects is more fundamentally a politicalrhetorical matter, demanding skills in providing the right arguments and making convincing economic calculations, rather than something shaped by what might be termed the strict application of economic knowledge. While it is not well described in the literature reviewed in this chapter, there is no doubt that many companies and laboratories put considerable effort into achieving cost control as far as technological projects are concerned. There are many different methods and tools available to support such efforts. Most engineers are no doubt concerned with economic issues, but ultimately they prefer a professionbased appropriated version, an engineering economics, to the skills and knowledge represented by economists and MBAs. In engineering stories, economists and MBAs are troublemakers rather than helpful parties. The actual influence on technological design and on the engineering work of professional economics may not therefore be so strong. Costs are important, economists are not. When many engineers perceive economists and MBAs to be too conservative and too control-oriented, this probably reflects the different perceptions of the economic dynamics of technological development. In his study of wind turbine development in Norway, Solli [2007] shows how economists evaluated the economic prospects of wind energy on the basis of the notion that production costs were known and would be similar to the costs measured at any given time. The advice of the economists was therefore to say no to wind energy projects. Engineers involved in this technology favoured a more dynamic approach, arguing from the so-called learning curve effects that a considerable drop in production costs would

110

Knut H. Sørensen

occur because accumulated engineering experience would lead to much cheaper installations. The fact that there has actually been a large drop in the cost of producing energy would suggest that, at least in this case, the engineers had a better grasp of the economic dynamics than the economists. However, the matter is more complicated than that, given the restrictions of the available national economic resources. The case demonstrates that also with respect to economics, engineers prefer profession-based appropriation, where some arguments go in their favour but do not necessarily win the day. For a thorough evaluation of the initial decision not to support wind energy technology, we would probably need the kind of transdisciplinary appropriation that results from engineers and economists debating the issue together and appreciating each other’s arguments — whilst perhaps also taking on board other bodies of knowledge.

5 SOCIAL SHAPING VERSUS SOCIAL SCIENCE SHAPING OF TECHNOLOGY Technology is always a social achievement, a material or mental representation of human activity. In principle, this makes the development of technology as much a challenge to social sciences as to engineering. If anything, social science research constitutes an effort to provide representations of human activities. However, there is no guarantee that such representations are useful or will indeed be used by engineers when they engage in design and technological development. As we have seen in this chapter it is rather the case that the relationship between the social sciences and engineering is problematic and unclear. Most studies of engineering work reveal little about how engineers appropriate and use social science. Engineers are probably not very concerned about this because the use of such knowledge is normally implicit and mediated. Social science is most commonly appropriated in a profession-based way, providing professional self-sufficiency and resulting in what I have termed individualised, polytechnical interdisciplinarity. The acknowledgement that technology is socially shaped raises interesting questions about how the social dimensions are represented and mediated in engineering work and design processes. From accounts found in technology and engineering studies, it seems clear that the dominant form of such mediation is in the experience, knowledge, outlook, etc. of the involved engineers In a sense, it is the engineering body that is the main instrument of observing, learning and mediating of social aspects relevant to engineering work. The social sciences have a subdued and much less visible role which is difficult to assess. Some of the features that produce this somewhat paradoxical situation have already been reviewed. Firstly it should be noted that most engineers show rather little interest in the social sciences, with the exception of the areas of management and economics which are perceived to be career-enhancing types of knowledge. Compared to the natural sciences, with which engineers engage heavily, the social

The Role of Social Science in Engineering

111

sciences have less prestige and, probably, are more difficult to translate into useful design criteria. Secondly, today, engineers and natural scientists tend to professionally dominate the development of technology. As a consequence, engineers (and probably also the natural scientists) tend to prefer to abide by their own profession-based appropriation of the social sciences — including economics — rather than to be involved in interdisciplinary collaboration with social scientists. As noted above, this appropriation process seems to be largely about the accumulation of experience — often from interacting with customers and user communities over a fairly long period of time [Sørensen et al., 2007]. Presumably, there is also a kind of ‘citizen effect’ in the sense that some social science knowledge seeps in from news media and similar sources but this phenomenon is also difficult to access and assess in an empirical fashion. Thirdly, as suggested several times throughout the chapter, the social sciences have not particularly set out to be relevant and useful to engineers. Teaching social science to engineering students never had much status; moreover, such teaching has tended to focus on ethical and other social concerns related to the possible negative effects of new technology. In that way the engagement of social science with technology and engineers has been doubly marginalised. It has remained external to the core social science concerns while possessing a kind of policing role that is not particularly appreciated by most engineers. However, it should also be recognised that substantial efforts have been made by groups of social scientists to actively engage in collaboration with engineers and to produce potentially relevant knowledge about many aspects of technology. Nevertheless, it remains a problem that there is little if any empirical research that actually examines such efforts and investigates the role of social science with respect to engineering and the design and development of technology more generally. There is even a danger that the interpretations made in this chapter may underestimate or misjudge the influence of social science. To some extent, this is a methodological problem attributable to the dominant position of the professionbased mode of appropriation in the enactment of social science knowledge among engineers. This form of appropriation tends to reduce the impact of social science. Another challenge is the argument that to engineers (and probably also to social scientists) the insight gained from social science may appear difficult to apply to the design and development of new technologies. When social scientists accentuate the complexity and instability of human cultures, which is what they tend to do (and with good reason), they provide explanations that engineers do not usually find helpful because they in their activities are more concerned about reducing complexity and constructing stable technological standards that will instigate development and problem-solving. Obviously, there are challenges and opportunities behind finding better ways to co-produce new social science and engineering science that may be integrated. My own experience indicates that there is increasing interest among engineers to collaborate with social scientists to find ways to manage challenges that engi-

112

Knut H. Sørensen

neers experience as problematic and outside their professional area of expertise. New technologies may be rejected or may meet with resistance while their actual effects may be quite different from what was originally intended. When social scientists enter into such collaboration, they will probably discover a range of ways of interacting with engineering knowledge. Some problems, like policy conditions or public views about new technologies may be dealt with in a fairly isolated manner while others, like the need to analyse user requirements, may call for more integrated ways of working. The difficulties involved should not be underestimated. For example, the tendency among engineers to find doing more important than reflection and to demand constructive input rather than critical opposition, may mean that social scientists have to adjust their normal mode of operation or vice versa. To communicate and collaborate, one needs some insight into the knowledge of the other party if one is to acquire what Collins and Evans [2007] call interactional expertise — the capability to interact constructively with the experts with whom you are supposed to collaborate. It may thus prove fruitful — at least for some period — for engineers to pursue a profession-based appropriation of relevant social science, while social scientists concentrate on a profession-based appropriation of relevant engineering science, so paving the way for a productive transdisciplinary appropriation of both kinds of knowledge. BIBLIOGRAPHY [Akrich, 1992] M. Akrich. The Description of Technological Objects. In Shaping Technology/Building Society, W.E. Bijker and J. Law, eds., pp. 205-224. MIT Press, 1992. [Amdahl and Sørensen, 2008] E. Amdahl and K. H. Sørensen. Den polytekniske kunnskapsideologien: Fra viten til dyd. In Vitenskap som dialog - kunnskap i bevegelse. Tverrfaglighet og kunnskapskulturer i forskning, K. H. Sørensen, H. J. Gansmo, V. A. Lagesen and E. Amdahl, eds., pp. 49-69. Tapir Akademiske Forlag, 2008. [Berg, 1994] A.-J. Berg. A Gendered Socio-technical Construction: The Smart House. In Bringing Technology Home. Gender and Technology in a Changing Europe, C. Cockburn and R. F¨ urst-Dili, eds., pp. 165-180. Open University Press, 1994. [Bijker et al., 1987] W. E. Bijker, T.P. Hughes, and T. Pinch, eds., The Social Construction of Technological Systems. MIT Press, 1987. [Brookshire et al., 2007] R. G. Brookshire, R. Yin, S. Hunt and T. B. Crews. An End-User Information System for the 21st century. The Journal of Computer Information Systems, 47 (3), 81-88, 2007. [Bucciarelli, 1994] L. L. Bucciarelli. Designing Engineers. MIT Press, 1994. [Callon, 1987] M. Callon. Society in the Making: The Study of Technology as a Tool for Sociological Analysis. In The Social Construction of Technological Systems, W.E. Bijker, T.P. Hughes, and T. Pinch, eds., pp. 83-106. MIT Press, 1987. [Chabaud-Rychter, 1994] D. Chabaud-Rychter. Women Users in the Design Process of a Food Robot: Innovation in a French Domestic Appliance Company. In Bringing Technology Home. Gender and Technology in a Changing Europe, C. Cockburn and R. F¨ urst-Dili, eds., pp. 77-93. Open University Press, 1994. [Cockburn and Ormrod, 1993] C. Cockburn and S. Ormrod. Gender and Technology in the Making. Sage, 1993. [Collins and Evans, 2007] H. Collins and R. Evans. Rethinking Expertise. University of Chicago Press, 2007. [Committee on the Fundamentals of Computer Science, 2004] Committee on the Fundamentals of Computer Science. Computer Science. Reflections on the Field. Reflections from the Field. National Academies Press, 2004.

The Role of Social Science in Engineering

113

[Dahlbom and Mathiassen, 1997] B. Dahlbom and L. Mathiassen. The Future of Our Profession. Communications of the ACM, 40 (6), 80-89, 1997. [Davis, 1998] M. Davis. Thinking Like an Engineer. Studies in the Ethics of a Profession. Oxford University Press, 1998. [Downey, 1998] G. L. Downey. The Machine in Me. An Anthropologist Sits among Computer Engineers. Routledge, 1998. [Downey and Lucena, 1995] G. L. Downey and J.C. Lucena. Engineering Studies. In Handbook of Science and Technology Studies, S. Jasanoff, G.E. Markle, J.C. Petersen and T. Pinch, eds., pp. 189-204. Sage, 1995. [Ehn, 1988] P. Ehn. Work-Oriented Design of Computer Artifacts. Arbetslivscentrum, 1988. [Florman, 1976] S. C. Florman. The existential pleasures of engineering. St. Martin’s Press, 1976. [Forsythe, 2001] D. E. Forsythe. Studying Those Who Study Us. An Anthropologist in the World of Artificial Intelligence. Stanford University Press, 2001. [Freeman, 1982] C. Freeman. The Economics of Industrial Innovation. The MIT Press, 1982. [Gerstl and Hutton, 1966] J. E. Gerstl and P. Hutton. Engineers: The Anatomy of a Profession. Tavistock, 1966. [Giddens, 1976] A. Giddens. New rules of sociological method. Hutchinson, 1976. [Goles et al., 2008] T. Goles, S. Hawk and K. M. Kaiser. Information Technology Workforce Skills: The Software and IT Services Provider Perspective. Inf Syst Front 10,179-194, 2008. [Grimson, 2002] J. Grimson. Re-engineering the Curriculum for the 21st Century. European Journal of Engineering Education, 27 (1), 31-37, 2002. [Hatling and Sørensen, 1998] M. Hatling and K.H. Sørensen. Social Constructions of User Participation. In The Spectre of Participation. Technology and Work in a Welfare State, K.H. Sørensen, ed., pp. 171-188. Scandinavian University Press, 1998. [Hughes, 1985] T. P. Hughes. Edison and Electric Light. In The Social Shaping of Technology, D. A. MacKenzie and J. Wajcman, eds., pp. 39-52. Open University Press, 1985. [Hutton and Lawrence, 1981] S. Hutton and P. Lawrence. German Engineers: The Anatomy of a Profession. Clarendon Press, 1981. [Iivari et al., 2004] J. Iivari, R. Hirschheim and H. K. Klein. Towards a Distinctive Body of Knowledge for Information Systems Experts: Coding ISD Process Knowledge in Two IS Journals. Info Systems J, 14, 313-342, 2004. [Jones, 2003] M. E. Jones. The Renaissance Engineer: A Reality for the 21st Century? European Journal of Engineering Education, 28 (2), 169-178, 2003. [Kline and Rosenberg, 1986] S. J. Kline and N. Rosenberg. An Overview of Innovation. In The Positive Sum Strategy. Harnessing Technology for Economic Growth, R. Landau and N. Rosenberg, eds., pp. 275-305. National Academy Press, 1986. [Kranakis, 1989] E. Kranakis. Social Determinants of Engineering Practice: A Comparative View of France and America in the Nineteenth Century. Social Studies of Science, 19, 5–70, 1989. [Lagesen and Sørensen, 2008] V. A. Lagesen and K.H. Sørensen. Re-locating software engineers: From technologists to communicators? Submitted to Engineering Studies, 2008. [Latour, 1988] B. Latour. How to Write ‘the Prince’ for Machines as well as Machinations. In Technology and Social Process, B. Elliot, ed., pp. 20-43. Edinburgh University Press, 1988. [Latour, 1987] B. Latour. Science in Action. Open University Press, 1987. [Layton, 1971] E. T. Layton. The Revolt of the Engineers: Social responsibility and the American Engineering Profession. Press of Case Western Reserve University, 1971. [Lee, 2005] C. K. Lee. Analysis of Skill Requirement for Systems Analysts in Fortune 500 Organizations. The Journal of Computer Information Systems, 45 (4), 84-92, 2005. [Litecky et al., 2004] C. R. Litecky, K. P. Arnett and B. Prabhakar. The Paradox of Soft Skills versus Technical Skills in IS Hiring. The Journal of Computer Information Systems, 45 (1), 69-76, 2004. [Little et al., 2002] S. Little, P. Quintas and T. Ray, eds. Managing Knowledge. Sage, 2002. [Lundgreen, 1990] P. Lundgreen. Engineering Education in Europe and the U.S.A., 1750–1930: The Rise to Dominance of School Culture and Engineering Professions. Annals of Science, 47:33–75, 1990.

114

Knut H. Sørensen

[Manegold, 1978] K. H. Manegold. Technology Academised. Education and Training of the Engineer in the Nineteenth Century. In The Dynamics of Science and Technology. Social Values, Technical Norms and Scientific Criteria in the Development of Knowledge, W. Krohn, E.T. Layton and P. Weingart, eds., pp. 137-158. Reidel, 1978. [Maurice et al., 1986] M. Maurice, F. Sellier and J.-J. Silvestre. The Social Foundations of Industrial Power. MIT Press, 1986. [Meiksins and Smith, 1996] P. Meiksins and C. Smith, eds. Engineering Labour. Technical Workers in Comparative Perspective. Verso, 1996. [Meiksins and Watson, 1989] P. F. Meiksins and J.M. Watson. Professional Autonomy and Organizational Constraint: The Case of Engineers. The Sociological Quarterly, 30 (4), 561-585, 1989. [National Academy of Engineering, 2004] National Academy of Engineering. The Engineer of 2020. Visions of Engineering in the New Century. National Academies Press, 2004. [National Academy of Engineering, 2005] National Academy of Engineering. Educating the Engineer of 2020. Adapting Engineering Education to the New Century. National Academies Press, 2005. [Nguyen, 1998] D. Q. Nguyen. The Essential Skills and Attributes of an Engineer: A Comparative Study of Academics, Industry Personnel and Engineering Students. Global Journal of Engineering Education, 2 (1), 65-75, 1998. [Nobel, 1979] D. Noble. America by Design. Science, Technology and the Rise of Corporate Capitalism. Oxford University Press, 1979. [Normal, 1988] D. A. Norman. The Psychology of Everyday Things. Basic Books 1988. [Nowotny et al., 2001] H. Nowotny, P. Scott and M. Gibbons. Re-thinking Science. Knowledge and the Public in an Age of Uncertainty. Polity Press, 2001. [Oudshoorn et al., 2004] N. Oudshoorn, E. Rommes and M. Stienstra. Configuring the User as Everybody: Gender and Design Cultures in Information and Communication Technologies. Science, Technology, & Human Values, 29 (1), 30-63, 2004. [Perruci and Gerstl, 1969] R. Perruci and J.E. Gerstl. Profession Without Community. Random House, 1969. [Ravensteijn et al., 2006] W. Ravensteijn, E. de Graaff and O. Kroesen. Engineering the Future: The Social Necessity of Communicative Engineers. European Journal of Engineering Education, 31 (1), 63-71, 2006. [Ritti, 1971] R. R. Ritti. The Engineer in the Industrial Corporation. Columbia University Press, 1971. [Robinson et al., 2005] M. A. Robinson, P. R. Sparrow, C. Clegg and K. Birdi. Design Engineering Competencies: Future Requirements and Predicted Changes in the Forthcoming Decade. Design Studies, 26, 123-153, 2005. [Russell and Stouffer, 2005] J. S. Russell and W. B. Stouffer. Survey of National Civil Engineering Curriculum. Journal of Professional Issues in Engineering Education and Practice, 131 (2), 118-128, 2005. ´ [Shinn, 1984] T. Shinn. Reactionary Technologists: The Struggle Over the Ecole Polytechnique, 1880-1914. Minerva, 22 (3–4), 329–345, 1984. [Solli, 2007] J. Solli. Bærekraftige kalkyler? Utviklingen av økonomiske argumenter om vindkraft. In Mellom klima og komfort: Utfordringer for en bærekraftig energiutvikling, M. Aune og K.H. Sørensen, eds., pp. 125-142. Tapir Akademisk Forlag, 2007. [Suchman, 1987] L. Suchman. Plans and Situated Action: The Problem of Human-Machine Communications. Cambridge University Press, 1987. [Suchman, 2007] L. Suchman. Human-machine reconfigurations. Plans and situated action 2 nd edition. Cambridge University Press, 2007. [Sørensen, 1996] K. H. Sørensen. Den nye tverrfagligheten. Fra polyteknisk generalist til polyvalent spesialist. In Perspektiver p˚ a tvers. Disiplin og tverrfaglighet p˚ a det moderne forskningsuniversitetet, T. Dahl and K.H. Sørensen, eds., pp. 19-28. Tapir Akademiske Forlag, 1996. [Sørensen, 1998] K. H. Sørensen. Engineers Transformed: From Managers of Technology to Technological Consultants? In The Spectre of Participation. Technology and Work in a Welfare State, K.H. Sørensen, ed., pp. 139-160. Scandinavian University Press, 1998.

The Role of Social Science in Engineering

115

[Sørensen, 2008] K. H. Sørensen. Dialog og tverrfaglighet som kunnskapspolitikk. In Vitenskap som dialog — kunnskap i bevegelse. Tverrfaglighet og kunnskapskulturer i forskning, K. H. Sørensen, H. J. Gansmo, V. A. Lagesen and E. Amdahl, eds., pp. 9-26. Tapir Akademiske Forlag, 2008. [Sørensen et al., 2007] K. H. Sørensen, V.A. Lagesen and N. Levold. Flytende profesjoner? Om organisering av kunnskap. In Arbeid, kunnskap og sosial ulikhet. Festskrift til Olav Korsnes, J. Hjellbrekke, O. J. Olsen and R. Sakslind, eds., pp.197-220. Fagbokforlaget, 2007. [Sørensen and Levold, 1992] K. H. Sørensen and N. Levold. Tacit Networks, Heterogeneous Engineers, and Embodied Technology. Science, Technology & Human Values. 17 (1), 13-35, 1992. [Sørensen et al., 2008] K. H. Sørensen, H.J. Gansmo, V.A. Lagesen and E. Amdahl, eds. Faglighet og tverrfaglighet i den nye kunnskapsøkonomien. Tapir Akademiske Forlag, 2008. [Thomas, 1994] R. J. Thomas. What Machines Can’t Do. Politics and Technology in the Industrial Enterprise. University of California Press, 1994. [Trevelyan and Tilli, 2007] J. Trevelyan and S. Tilli. Published Research on Engineering Work. Journal of Professional Issues in Engineering Education and Practice, 133 (4), 300-307, 2007. [Turley and Bieman, 1995] R. T. Turley and J. B. Bieman. Competencies of Exceptional and Nonexceptional Software Engineers. J. Systems Software, 28, 19-38, 1995. [Vesilind, 2001] P. Vesilind. A. Engineering as Applied Social Science. Journal of Professional Issues in Engineering Education and Practice, 127 (4), 184-188, 2001. [Vincenti, 1990] W. Vincenti. What Engineers Know and How They Know It: Analytical Studies from Aeronautical History. Johns Hopkins University Press, 1990. [Vinck, 2003] D. Vinck, ed. Everyday Engineering. An Etnography of Design and Innovation. MIT Press, 2003. [Vitalari, 1985] N. P. Vitalari. Knowledge as Basis for Expertise in Systems Analysis: An Empirical Study. MIS Quarterly, 9 (3), 221-241, 1985. [Walsh et al., 1992] V. Walsh, R. Roy, M. Bruce and S. Potter. Winning by Design. Technology, Product Design and International Competitiveness. Blackwell, 1992. [Weiss, 1979] C. H. Weiss. The Many Meanings of Research Evaluation. Public Administration Review, 39 (5), 426-431, 1979. [Whalley, 1986] P. Whalley. The Social Production of Technical Work. Macmillan, 1986. [White and Leifer, 1986] K. B. White and R. Leifer. Information Systems Development Success: Perspectives from Project Teams Participants. MIS Quarterly, 10 (3), 215-223, 1986. [Williams, 2002] R. Williams. Retooling. A Historian Confronts Technological Change. MIT Press, 2002. [Williams, 2003] R. Williams. Education for the Profession Formerly Known as Engineering. The Chronicle of Higher Education, 49 (20), B12, 2003. [Williams et al., 2005] R. Williams, J. Stewart and R. Slack. Social Learning in Technological Innovation. Experimenting with Information and Communication Technologies. Edward Elgar, 2005. [Zussman, 1985] R. Zussman. Mechanics of the Middle Class. University of California Press, 1985.

This page intentionally left blank

THE EMERGENCE OF THE ENGINEERING SCIENCES: AN HISTORICAL ANALYSIS David F. Channell

1

INTRODUCTION

The engineering sciences, which provide a basis of knowledge for the understanding and design of humanly constructed artifacts, emerged during the 18th and 19th centuries when new economic, political, social and intellectual forces were creating new relationships between science and technology [Channell, 1989]. Throughout history three main models have played a role in understanding the relationship between science and technology — the independent model, the dependent model and the interdependent model. According to the independent model, science and technology are independent realms of knowledge with little interaction between them. The dependent model sees either technology to depend upon applications of science or science to depend upon applications of technology. Finally the interdependent model argues that the two areas form a symbiotic relationship so that the distinguishing characteristics of the two areas become blurred. Although examples of all three models of the relationship between science and technology often can be found in any given historical period, the independent model tended to be dominant in the ancient and medieval periods, the dependent model was more dominant from the early modern period through the 19th century, and the interdependent model came into dominance during the 20th century. The changing relationships between science and technology were themselves based on changing definitions of the two areas of activity and the emergence of a more or less modern notion of what we today label as science and technology [Oldenziel, 2006]. As such the history of the engineering sciences reflects the development of the more or less modern concepts of science and technology and their interactions [Laudan, 1984; Kline, 1995]. Throughout much of history the activities that we now label as science and technology followed different social and intellectual traditions and, except for a few instances, they had little impact on each other [Layton, 1974] (for an argument that science and technology are linked in a conceptual-theoretical way, see [Radder, this volume, Part I]). Beginning in ancient times the activity we see today as science was more correctly labeled natural philosophy. As a branch of philosophy it focused on asking and answering questions concerning the ultimate nature of the physical world and the universe. Traditionally natural philosophy Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

118

David F. Channell

placed little emphasis on solving individual or practical problems. Rather, natural philosophy used mathematical, geometrical, and other rational methods in order to idealize problems so their solutions could be universally applied to descriptions of nature as a whole. Since natural philosophy relied primarily on contemplation, it became closely associated with the elites and after the founding of universities during the Middle Ages, most natural philosophers had university educations. Finally, because of natural philosophy’s close association with formal education, knowledge of science was gained and disseminated through a written tradition of treatises, textbooks and journals. Since ancient times, the activity we see today as technology was more often labeled the manual, mechanical or technical arts. As a practical activity it was concerned with practical problems that occurred in specific situations. Mechanics and artisans traditionally used cut-and-try empiricism or rule-of-thumb techniques that were applicable to individual situations but usually could not be generalized to a wider class of problems. The mechanical arts were the province of an artisan class and before the 18th century few mechanics had any formalized education or university training. Since most knowledge of mechanical arts was the result of observation and experience, it was primarily gained through an apprenticeship system and disseminated through direct contact between master and pupil. Very little technical knowledge was recorded and when it was it was usually in a visual rather than a written form. During the later medieval and early modern periods some of the intellectual and social differences dividing natural philosophy from the mechanical arts began to be overcome, and by the 18th and 19th centuries social, economic, political and philosophical changes led to the emergence of the more or less modern concepts of science and technology [Kuhn, 1977, pp. 31-65]. With the development of a conception of science that was based on both a mechanical and experimental philosophy and the development of a concept of technology, or engineering, which was based on a science of the practical arts, the door was opened to a new set of interactions between science and technology [Layton, 1971]. This possibility was brought to fruition during the 18th and 19th centuries by the new demands resulting from industrialization and with it the establishment of new institutions, such as engineering schools, technical institutes, industrial research laboratories and professional engineering societies. These new institutions created new research and methodologies that resulted in a new harmony between theory and practice, which led to the emergence of the engineering sciences which served as an intermediary mode of knowledge that linked together science and technology. As an intermediary mode of knowledge, the engineering sciences facilitated the transfer and transformation of knowledge and methodologies between science and technology. As such, the engineering sciences emerged not simply as applications of science to technology but equally as applications of technology to science. By the late 19th century the rise of science-based industries, particularly the chemical and electrical industries, caused the engineering sciences to take on the character of what some have called industry-based sciences [K¨onig, 1996, p. 100]. Finally, dur-

The Emergence of the Engineering Sciences: An Historical Analysis

119

ing the 20th century the rise of the military-industrial complex began to completely blur the distinctions between science and technology and helped to transform the engineering sciences into technoscience. 2

2.1

THE ANCIENT, MEDIEVAL AND EARLY MODERN PERIODS

The ancient period

Although the engineering sciences did not emerge until the 18th and 19th centuries, earlier developments prepared the way for the establishment of the engineering sciences [Landels, 1978]. The dominant philosophies and social systems that influenced ancient Greece and Rome created an ideology that made it difficult for there to be anything more than very limited interactions between natural philosophy and the mechanical arts. The widespread Platonic philosophy in ancient Greece placed the highest intellectual value on an ideal world that could be grasped with the mind and not with the senses. Because of the influence of such a system it was much more important to apply natural philosophy and mathematics to an understanding of the ideal world of the forms rather than to apply that knowledge to the material world which was considered a mere shadow of the ultimate reality. Some of the limitations of Plato’s philosophy were overcome by Aristotle’s philosophy, which placed more value on the material world and the role of the senses, but Aristotle’s influence on an interaction between natural philosophy and the technical arts was limited by social barriers between the two areas. Ancient Greece was a society built on slavery. As such, activities associated with manual labor or the mechanical arts were seen as vulgar and ignoble while activities associated with philosophy were seen as liberal and enlightening. Although interactions between natural philosophy and the mechanical arts were limited during the ancient period there were a few examples of some interactions that would play a role in the future development of the engineering sciences [Clagett, 1963]. Many of these interactions took place in the Hellenistic period during which time the Alexandrian state began to actively support the study of natural philosophy through the establishment of such institutions as the Museum of Alexandria. With state support natural philosophers began to address some problems that were more practical and which reflected the needs of the state. Some contemporaries claimed that one reason for establishing the Museum was so that natural philosophers could help to improve weapons of war [Hacker, 1997]. Beginning as early as the 4th century B.C., the Greeks were undertaking a study to reduce mechanics to mathematical principles. An early work in this tradition was the Treatise on the Balance (c. 300 B.C.), attributed to Euclid who was associated with the Museum of Alexandria. In this work Euclid put forward a geometrical analysis of the lever, but the most extensive mathematical study of mechanics during the Hellenistic period was found in the works of Archimedes, who may have visited Alexandria as a youth [Drachmann, 1962]. In his work, On the Equilibrium of Planes, Archimedes provided a formal mathematical proof of

120

David F. Channell

the law of the lever using a framework of postulates and propositions similar to the approach used by Euclid in geometry. An important element of Archimedes’ approach was his idealization of all elements of the lever so that the weights on each end became forces acting at single points and the lever became a geometrical line resting on an idealized fulcrum. Although Archimedes’ approach would play a major role in the later development of the application of science to the study of mechanics, there is evidence that the purpose of Euclid’s and Archimedes’ work was to gain insights into mathematical principles and propositions through the use of physical examples, and it may be more correct to see their approach as an application of the technical arts to natural philosophy rather than the other way around. Another important figure in Hellenistic mechanics was Hero of Alexander (1st century C.E.), who followed in the tradition of Archimedes and probably had some connection to the Museum of Alexandria [Drachmann, 1948]. In his treatise the Mechanica, he provided a geometrical analysis of the so-called five simple machines in terms of how they could move a weight using a given force. By analyzing each of the simple machines in terms of a single principle Hero provided the basis for understanding much more complex machines, which could be seen as made up of combinations of simple machines. In his most famous book, the Pneumatica, Hero used the Aristotelian principle of the impossibility of a vacuum to describe a large number of machines and devices that operated with water, air or steam [Hall, 1973]. Again, the Mechanica and the Pneumatica appear to be examples of the application of science to technology, and they would influence later engineers who helped establish the engineering sciences, but several scholars have argued that Hero’s work, like Archimedes’, was intended to show physical examples designed to provide a concrete understanding of some scientific law, such as the principle that nature abhors a vacuum. As such, the development of Hero’s mechanics may be more representative of the application of technology to science.

2.2 The medieval period The development of natural philosophy and the technical arts during the Middle Ages was almost a mirror image of that of the period of the Greeks. During the period of the ancient Greeks natural philosophy and mathematics flourished and the mechanical arts were neglected, but during the medieval period the technical arts began to flourish while natural philosophy became the “handmaiden to theology” [Clagett, 1959; Lindberg, 1992]. Although natural philosophy and the mechanical arts were, for the most part, still considered distinct independent activities, changes that took place during the period began to break down some of the intellectual and social barriers that divided the two areas. The rise of Christianity, which dominated the social and cultural life of the Middle Ages, brought about significant changes in the attitudes toward natural philosophy and the technical arts. The decline in the population during the chaos that resulted from the fall of Rome and discouragement from the Christian Church

The Emergence of the Engineering Sciences: An Historical Analysis

121

led to a decline in slavery. Along with this, the rise and spread of the monastic system helped to overcome some of the social barriers that had separated natural philosophy from the mechanical arts. The fact that the monks were literate scholars with high social standing and yet were actively engaged in manual labor established the idea of the dignity of labor. The rise of medieval Christianity also provided a new philosophical framework that helped to overcome the intellectual barriers that divided natural philosophy from the mechanical arts [White, 1978]. A number of scholars began to treat the technical arts more like the liberal arts. During the 12th century Domingo Gundisalvo, in his De divisione philosophiae (c. 1150), introduced an Arabic interpretation of Aristotle which argued that the mechanical arts were simply the operative part of theoretical knowledge. At the same time Hugh of St. Victor, master of the abbey of St. Victor near Paris, wrote the Didascalicon, in which he argued that there were seven mechanical arts — weaving, weapon foundry, navigation, agriculture, hunting, medicine and drama — which paralleled the seven liberal arts [Taylor, 1991]. For Hugh the seven mechanical arts could help restore humans to the physical conditions that existed before the Fall. The fact that the mechanical arts were associated with the restoration of a prelapsarian physical state and that they were essentially imitative meant that the mechanical arts had to have some connection to the world of nature and therefore to natural philosophy. He noted that humans were given the capacity of reason so they could study nature and create what they needed through the act of invention. By the 13th century Robert Kilwardby, in his De ortu scientarum, synthesized the ideas of Gundisalvo and Hugh of St. Victor and questioned the still widely held distinction between theoretical and practical knowledge. He argued that theoretical knowledge could often be practical and practical knowledge could be theoretical. This meant that each of the mechanical arts had some relationship to some specific form of theoretical knowledge, such as the relationship between navigation and astronomy. At the same time that some medieval scholars were arguing that the mechanical arts could be seen as more theoretical, other scholars argued that natural philosophy could be seen as more practical. The shortage of Bibles and literate people to read them led many in the Church to turn to another book — the book of nature [White, 1978]. Such a book could be “read,” using a natural theology, for signs or signatures that indicated God’s divine plan. Eventually people came to believe that nature not only contained signs pointing towards salvation but that nature contained the necessary elements to achieve salvation. This led to a reconceptualization of nature that made it much more practical and open to the mechanical arts. Nature, which had been the focus of natural philosophy was no longer seen as an abstract mathematical ideal; rather it was now seen as a storehouse of potentials that could not only be studied for their own sake, but which could be exploited in order to play out some drama of salvation. Instead of a purely contemplative approach to nature the Middle Ages took a more activist approach.

122

David F. Channell

The reconceptualization of natural philosophy was also influenced by the battle over faith versus reason. Ironically the medieval Church’s attempt to place faith above reason helped to introduce a new more empirical view of natural philosophy [McEvoy, 1982, p. 206]. During the 13th century the Bishop of Paris issued a condemnation of a number of Aristotelian philosophical propositions, such as the argument that God could not create a vacuum or multiple worlds. While the condemnations have been seen as an attack on natural philosophy, several scholars have argued that they opened up new possibilities for natural philosophy. One important result of the condemnations was a new focus on God’s omnipotence. But if God could create the world any way He wanted, the world that humans experienced was contingent rather than necessary. For some medieval scholars this meant that natural philosophy was useless since nature was unpredictable, but others argued that natural philosophy must examine the world directly and empirically in order to discover what kind of world God actually had created. One of the most influential natural philosophers of the late Middle Ages was Robert Grosseteste, who introduced the concept of experimenta into natural philosophy [McEvoy, 1982, pp. 207-208]. Grosseteste’s concept did not include the modern notion of systematic and controlled experiments, but he did argue that a study of the natural world should be based on experience since knowledge of the actual world could never attain the certainty one could gain through the mathematical study of ideal objects. This new empirical view of natural philosophy brought it closer to the mechanical arts as can be seen in the work of Roger Bacon, who was a great admirer of Grosseteste [Easton, 1952; Lindberg, 1983]. But Bacon went beyond Grosseteste and argued that experiments could not just confirm knowledge gained through reason but could be used to discover some practical uses for that knowledge [Lindberg, 1983, p. 226]. In his Secret Operations of Nature Bacon put forward some ways in which natural philosophy could have been of practical use, such as self-propelled carriages and flying machines. While the medieval period began to overcome some of the social and intellectual barriers that separated natural philosophy from the mechanical arts, there were still other barriers that had to be overcome. Many of the medieval scholars who dealt with the mechanical arts focused only on the intellectual or theoretical aspects of those arts and ignored the actual skills and labor that went into them. In doing so they often continued the distinctions between intellectual activity and the manual crafts. More important was the barrier created by Aristotelian philosophy. While there were some challenges to Aristotle in the late Middle Ages, his philosophy, and particularly his organic ontology, still dominated the medieval period. As long as Aristotelian philosophy dominated natural philosophy it was difficult to create a natural philosophy based on mechanical principles that would provide a common ontological framework that could incorporate both science and technology.

The Emergence of the Engineering Sciences: An Historical Analysis

2.3

123

The early modern period

During the early modern period, which included what some have labeled the Scientific Revolution, the social and intellectual barriers that divided the mechanical arts from what was sometimes being labeled science continued to be overcome. Building on inventions, such as the printing press, gunpowder, and new navigational techniques that had originated in the late Middle Ages, the early modern period saw a dramatic expansion of world trade and commercial activity that some have labeled a commercial revolution. This commercial activity, along with almost constant warfare, led many of the European nobility to patronize engineers because of their potential contributions to the military, the mercantile system, and because many of the engineers also had skills in painting, sculpture and architecture. The emergence of the Renaissance artist-engineer, such as Leonardo Da Vinci, Leon Battista Alberti, Filippo Brunelleschi and Francesco di Giorgio Martini helped to bridge the social and intellectual divide that often separated the mechanical arts from natural philosophy [Prager and Scaglia, 1970]. Most artist-engineers had a background in the manual arts but they usually were not illiterate, and because of their connections to the nobility as patrons they had access to the upper classes of society. The Renaissance artist-engineers helped to transform the mechanical arts into a systematic form of knowledge that could be abstractly studied and shared. During the Renaissance, artists discovered the idea of fixed point perspective. While the techniques provided artists with a new style in which to paint, it provided engineers with a powerful new tool with which they could think about technical devices and a new efficient way to share technical information with other engineers [Ferguson, 1977]. Fixed point perspective provided a method to represent a three dimensional machine or structure in two dimensions. In doing so, the technique provided engineers with a way to visualize and analyze machines and structures without going through the time and expense of building the machine or structure. By drawing and analyzing a wide variety of machines, Leonardo Da Vinci was able to discover the concept of a mechanism [Reti, 1974]. Before Leonardo most engineers thought of each machine as unique and something that had to be designed as a whole. But by analyzing drawings of a wide range of machines, Leonardo appeared to understand that different types of machines have a number of elements, or mechanisms, in common, and that machines and structures could be fundamentally reduced to universal components, such as mechanisms or frames. This opened the possibility of a systematic study of machines and structures. At the same time that the mechanical arts were becoming closer to a science, natural philosophy underwent some significant transformations that brought it closer to the mechanical arts [Jacob, 1988]. Many of the significant developments of the ancient Greeks in natural philosophy and mathematics had not been available to the Latin West during the Middle Ages, but during the 15th and 16th centuries new translations of the works of Plato, Euclid, Archimedes and the Greek atomists began to appear in Europe and led to what some have called the Scientific

124

David F. Channell

Revolution [Lindberg and Westman, 1990]. During this period three approaches to natural philosophy were combined to transform it into something we might label science. First, in some universities, such as Padua, the medical faculty came to dominance over the divinity faculty. As a result more attention was paid to Aristotle’s natural philosophy than to his logical works. This in turn led to new interests in observation and experimentation. Second, the new translations of Plato and Archimedes led to a new interest in developing mathematical theories of natural philosophy. Third, these two different approaches were brought together by the emergence of a magical, or Hermetic, approach to nature [Yates, 1979]. Hermetic ideas strongly influenced a number of leading natural philosophers and led to the view that a combination of theoretical and experimental knowledge could be used to manipulate the forces of nature for some practical purposes [Newman, 2004]. John Dee, a possible model for Shakespeare’s Prospero, argued that both natural philosophy and such mechanical arts as hydraulics, mechanics and navigation, were all governed by geometry, which he saw as a form of mathematical magic [Dee, 1975]. Religious changes in the early modern period, especially those associated with the Protestant Reformation, also played an important role in establishing a more practical view of natural philosophy. The Puritans’ and Anglican Latitudinarians’ emphasis on good works and a millenarian belief in perfecting the world, led to the idea that knowledge, particularly natural philosophy, had to be seen as useful. The idea of a new practical view of natural philosophy was popularized in the 17th century by Francis Bacon, who argued that the goal of natural philosophy should not be simply knowledge, but rather a form of knowledge that would allow humans to exert power and control over nature [Rossi, 1970; Stewart, 1992]. Bacon put forward his idea of a new practical natural philosophy in his New Atlantis (1628), where he described a utopian society which had an institution, called Salomon’s House, in which groups of researchers used knowledge of natural philosophy to improve engines, machines, cannons, clocks and ships [Farrington, 1951; Martin, 1992]. Although Bacon’s goal of a natural philosophy based on the mechanical arts was not realized in his lifetime, his ideas influenced the founding of the Royal Society of London in 1662 and foreshadowed the emergence of the industrial research laboratory in the 20th century. While Bacon put forward the idea that natural philosophy could be applied to the mechanical arts, others began to argue that the mechanical arts could be used to explain natural philosophy. Beginning in the 16th century with Ren´e Descartes, and continuing with Pierre Gassendi, Thomas Hobbes, Robert Boyle, G.W. Leibniz and Isaac Newton, natural philosophers began to develop a mechanical philosophy in which all of nature was seen as functioning like a gigantic machine or clockwork mechanism. The fact that much of the natural philosophy that emerged from the Scientific Revolution was conceived in terms of mechanical analogies with machines or clocks made that natural philosophy appeared to be much closer to the mechanical arts [Dijksterhuis, 1961].

The Emergence of the Engineering Sciences: An Historical Analysis

125

The systematic study of the mechanical arts brought about by the artistengineers and the development of a new practical view of natural philosophy brought about by the Baconian and mechanical philosophies, resulted in a number of developments that would contribute to the later establishment of the engineering sciences [Drake and Drabkin, 1969]. One of the most important of these developments was the work of Galileo. Although he is best known for his support of the Copernican system which resulted in his subsequent trial, Galileo had gained practical experience working in the Venetian Arsenal, which led to a revolutionary analysis of machines which he put forward in his On Mechanics (1600) [Cardwell, 1995, pp. 83-89]. Before Galileo most machines were judged by mostly qualitative standards, such as the quality of construction, and machines were often seen as ingenious devices that somehow functioned by cheating nature, but Galileo argued that machines simply took some natural force and transformed it for a useful purpose. Using Archimedes’ principle of the lever, Galileo showed that in a perfect, frictionless machine, the forces that set the machine in motion were the same as the forces required to keep it in a state of equilibrium. This geometric approach allowed Galileo to calculate how an ideal machine transformed the forces and motions applied to it, and by comparing an actual machine with this ideal machine, he was able to quantitatively evaluate that actual machine in terms of something that would later be called efficiency. After his trial Galileo returned to the study of mechanics and helped to establish some of the basic principles of what would become the engineering sciences through his work Discourse on the Two New Sciences. In the Discourse, which he set in the Venetian Arsenal, Galileo addressed a number of practical problems. First, he analyzed the scale effect and showed why it was not possible to build a machine or structure twice as big as a given machine or structure by simply doubling all of the dimensions of the original. Galileo’s study of the scale effect led him to the first of his new sciences — the strength of materials. Using Archimedes’ geometrical approach, he calculated the amount of weight a beam fixed at one end to a wall could support if loaded at the other end. He was also able to analyze how the shape of the beam would affect its strength. Galileo also used an Archemedian approach in his second new science — the study of motion. Using geometrical principles he was able to show that a projectile fired from a cannon would follow a parabolic path, and he calculated the comparative range of a projectile fired at different angles of elevation. 3

THE ORIGINS OF THE ENGINEERING SCIENCES IN THE 18th AND 19th CENTURIES.

By the 18th century a number of things were coming together that would allow for the emergence of the engineering sciences. The social and intellectual barriers that had divided natural philosophy from the mechanical arts continued to be overcome. During the beginning of this period there were significant developments in technology, especially in Great Britain, that many have called the Industrial Rev-

126

David F. Channell

olution. The production of iron using coke rather than charcoal, the invention of the steam engine, the mechanization of the textile industry, the centralization of production in the factory system, and the development of railways changed 18th and early 19th century society. With the development of steam engines, railways, ocean-going iron-hulled ships, and large scale iron bridges, it became impractical and uneconomical for engineers to use traditional rule-of-thumb or trial-and-error techniques. At the same time, scientists who were interested in the practical application of science were learning what engineers already knew, that many of the newly discovered laws of science were not directly applicable to technology. Newtonian mechanics might explain the forces acting between two atoms but it did not help in determining how an iron beam might act under a complex load. Boyle’s law explained the relationship between pressure and volume in an ideal gas but was of little use in describing how steam acted in a working steam engine. The Bernoulli equation of classical fluid mechanics had limited application in describing real fluids under non-laminar flow. In response to the needs of the Industrial Revolution a number of institutions arose whose primary goal was to develop sciences that were more technological. Oftentimes these institutions reflected the intellectual and social values of the cultures in which they arose which led to different approaches to the creation of the engineering sciences. In England the new business and industrial classes encouraged the establishment of a number of institutions inspired by the Baconian ideology of a practical application of science. Through Masonic lodges, coffee house lectures, Dissenting Academies, Mechanics’ Institutes, and local provincial societies, such as the Lunar Society of Birmingham and the Manchester Literary and Philosophical Society, scientific ideas, particularly Newtonian natural philosophy, were disseminated and popularized to the newly emerging business and industrial classes [Schofield, 1963]. England’s more democratic approach to the development of a practical science led to an emphasis on empiricism and experimentation which were seen to be less elitist than mathematics. Bacon’s ideology of a practical application of science also influenced France, but social and political differences between England and France led French institutions to become more strongly shaped and more closely aligned with the State [Artz, 1966; Weiss, 1982]. As early as the 17th century Jean-Baptiste Colbert, chief finance minister to Louis XIV, argued that French manufacturers and mercantilists would benefit from educational reform and he proposed establishing a number of academies to teach the scientific and practical basis of such subjects as navigation, bridge building, and manufacturing. By the 18th century the French government’s involvement in military and mercantile projects led to the creation ´ of the Corps des Ponts et Chauss´ees in 1716 and the establishment of the Ecole des Ponts et Chauss´ees in 1747 [Brunot and Coquand, 1982]. Soon after, military ´ schools at La F`ere and at M´ezi`eres were established along with the Ecole des Mines [Alder, 1997]. It was at the artillery school at La F`ere where Bernard Forest de B´elidor first used the term engineering science in his book entitled La science des

The Emergence of the Engineering Sciences: An Historical Analysis

127

ing´enieurs [1729]. One of the most important contributions to the emergence of ´ the engineering sciences was the establishment in Paris of the Ecole Polytechnique in 1794-1795 [Fourcy, 1987; Shinn, 1980]. The school, created by Napoleon to train both military and civil engineers, represented the new recognition that these two branches of engineering rested on the same principles. ´ Since in France science had been closely associated with the elite, the Ecole Polytechnique took a very theoretical and mathematical approach to the development of the application of science to technology. The school also helped to spread the new scientific approach to engineering to other countries. The school’s organization and curriculum influenced the Austrian polytechnics at Vienna and Prague, the German Technische Hochschulen at Karlsruhe, Munich, Dresden, Stuttgart and Hannover, and the Military Academy at West Point and Rensselaer Polytech´ nic Institute in America. Although influenced by the Ecole Polytechnique, the Austrian polytechnics and the German Technische Hochschulen developed their own model of the interaction of science and technology [Gispen, 1989]. Prechtl, the founder of the Vienna Polytechnic Institute, which served as a model for the German Technische Hochschulen, combined the French idea that science and mathematics served as a common basis for the study of technology, with the German university ideal that education should be based on some inner principle or unity of knowledge called Bildung [Fox and Guagnini, 1998-1999, pp. 100-107; Gispen, 1989, p.42]. The result was to go beyond the French idea of the engineering sciences as simply applied science and to develop a true and autonomous engineering science that could synthesize scientific theory with technological practices [Manegold, 1992, p. 142]. Scotland played one of the most significant roles in the emergence of the engineering sciences [Marsden, 1992]. Scotland shared cultural values with both England and France and was able to bring together the empirical/experimental traditions of the English with the theoretical/mathematical traditions of the French. The leading figure in the development of what would be called engineering science was W.J.M. Rankine at Glasgow University [Channell, 1982]. As one of the first professors of engineering at a British university, he faced the challenge of not duplicating what was being taught by the science faculties but not interfering with the practical education given through the apprenticeship system. Rankine’s solution was to create an autonomous branch of knowledge, which he labeled engineering science, which would be an intermediary between pure theory and pure practice. The creation of a new “harmony of theory and practice” would be the result of bringing together the practical observations and experiences of the properties of materials with the theoretical laws governing the action of machines and structures, and treating them as a science. In doing so the harmony of theory and practice would not duplicate existing sciences but instead would establish new engineering sciences. Rankine spread his idea of engineering science through the publication of a series of Manuals that became the standard textbooks for university trained engineers throughout Europe, America and even Japan. On the Continent, Ferdinand Redtenbacher at the Technische Hochschule at Karlsruhe

128

David F. Channell

played a role similar to Rankine’s by using practice to inform theory so as to create autonomous engineering sciences, which he disseminated through a series of textbooks [B¨ ohme, Van den Daele and Krohn, 1978, pp. 237-238]. During the 19th century there was a struggle in America between the “shop culture” and the “school culture” which wanted to emphasize the links between engineering and science which could be taught through formal education [Calvert, 1967]. By the end of the century the “school culture” came into dominance, and was reflected in the work of the leading American exponent of engineering science, Robert Henry Thurston who taught at the U.S. Naval Academy, Stevens Institute of Technology and Cornell University. Although influenced by Rankine, Thurston developed his own philosophy of engineering science while at Stevens and at Cornell. Rather than apply the laws of science to technology, Thurston argued that the methodology of science, which he saw as essentially Baconian, should be applied to technology. By collecting facts through observation and then inductively developing laws, Thurston believed that new laws of technology could be developed that were independent of scientific laws. Although like Rankine and Redtenbacher, Thurston also produced a number of textbooks; his most significant contribution was the establishment at Stevens of the first mechanical engineering laboratory in America [Durand, 1929, pp. 68-71].

3.1 The strength of materials During the 18th and 19th centuries the emergence of the engineering sciences can be seen in the development of a number of fields. With the demand for larger more complex machines, ships and structures brought about by industrialization and the military’s demand for larger and stronger fortifications, new approaches to the strength of materials and the theory of elasticity were needed [Timoshenko, 1953; Todhunter, 1886-1893]. Some of the earliest work on experimental testing of a variety of materials was carried out by the Dutch natural philosopher Pieter van Musschenbroek and reported in 1729. Using a series of machines he conducted tests on small scale samples to determine when various materials would fail under tension, compression and bending. A number of French engineers argued that small scale tests would be of little use in large scale construction projects. In response Emiland Gauthey, of the Corps de Ponts et Chauss´ees and Jean Rodolphe ´ Perronet, of the Ecole de Ponts et Chauss´ees each built machines that tested large scale samples of stone used in bridge building [Kranakis, 1997, pp. 106-107]. In the later part of the 18th century Charles Coulomb, while serving as an engineer on Martinique, conducted a series of experimental studies which allowed him to relate tensile strength to shear strength and to develop a mathematical analysis of bending [Gillmor, 1971; Heyman, 1972]. During the 19th century Peter Barlow in England built a series of machines at the Royal Dockyard in order to test timber for ships. The introduction of iron as a building material stimulated a number of additional experimental studies on the strength of materials. In England Barlow, William Fairbairn and Eaton Hodgkinson all conducted extensive tests on cast

The Emergence of the Engineering Sciences: An Historical Analysis

129

and wrought iron. On the Continent, Johann Bauschinger established one of the earliest academic materials testing laboratories at the Polytechnic Institute of Munich, while in America Robert Thurston established materials testing laboratories at Stevens and Cornell. This work helped establish what Edward Constant has called a “tradition of testability” [Constant, 1980, p. 20]. While a number of experimental studies were being conducted on the strength of materials, important advancements were taking place in the theory of elasticity. During the late 17th and early 18th century Jacob and John Bernoulli applied Leibniz’s calculus to the study of the elastic deflection of beams, and soon after Jacob’s son Daniel developed differential equations describing vibratory motion in elastic materials. This work was further extended by Leonhard Euler, one of Daniel’s students. During the 19th century Augustin Cauchy, Gabriel Lam´e, C.L.M.H. Navier, Sim´eon Poisson and Adhemar Barr´e de Saint-Vanant made significant contributions to the theory of elasticity. Particularly important was Cauchy’s formulation in 1822 of the concept of stress, which incorporated elements of both science and technology. In Germany Johann Bauschinger and Hermann Zimmermann applied the technique of graphical analysis to the analysis of stresses. The development of systematic testing, the creation of new concepts, and the use of graphical analysis would all become hallmarks of the emergence of an engineering science.

3.2

The study of structures

Closely connected, and often overlapping with the study of the strength of materials and the theory of elasticity, was the development of new experimental and theoretical approaches to the study of structures. Again, the increase of scale of structures, such as bridges, and the introduction of new building materials, such as iron, led to new demands for ways to analyze such structures [Billington, 1983]. During the 18th century Charles Coulomb used both mathematical theories and practical data to develop theories of retaining walls and arches [Gillmor, 1971; Heyman, 1972]. At the beginning of the 19th century the American James Finley conducted a series of experiments in order to acquire the knowledge that he needed to develop design principles for the construction of some of the earliest suspension bridges [Kranakis, 1997]. Finley’s work influenced British engineers, such as Thomas Telford, who used similar experimental techniques to design some of the first suspension bridges in Britain. About mid-century William Fairbairn, Eaton Hodgkinson and Robert Stephenson conducted experimental studies that contributed to the design of the Conway and the Britannia tubular railway bridges [Vincenti and Rosenberg, 1978]. While the Americans and the British were taking a more experimental approach to the design of structures, the French were developing a more theoreti´ cal approach. C.L.M.H. Navier, who studied at the Ecole Polytechnique and at ´ the Ecole de Ponts et Chauss´ees, developed a mathematical theory of suspension bridges [Kranakis, 1997]. This theory led engineers to see that a connection existed between the role of the strength of materials and the theory of elasticity and

130

David F. Channell

that both were essential for understanding the stability of structures. By the mid-19th century, some American, British and German engineers began to develop an approach to the study of structures that was in between a purely empirical and a purely theoretical approach. In Britain W.J.M. Rankine and James Clerk Maxwell developed the idea of parallel projection and reciprocal figures which established a graphical manner in which complex structures were related to simple structures and provided new ways to study the forces acting in these complex structures. In America, Squire Whipple developed a graphical approach to the analysis of truss bridges, while in Germany Karl Culmann, a graduate of the Technische Hochschule at Karlsruhe, developed another graphical method for the analysis of truss bridges [Timoshenko, 1953, pp. 190-197]. Again, by creating a middle ground between the highly mathematical approach of natural philosophers and the purely empirical approach of many mechanics, graphical approaches became one of the characteristics of the engineering sciences.

3.3

The study of machines

During the 18th and 19th centuries, developments in the study of machines contributed to the emergence of the engineering sciences. With the new demands for manufacturing brought about by mercantilism and industrialization, scientists and engineers began to focus their attention on understanding and improving machines [Reynolds, 1983]. Throughout much of the 18th century the study of machines focused on the waterwheel. Building on Galileo’s idea that machines should be analyzed in terms of their ability to apply the forces of nature in the most efficient way, the Frenchman Antoine Parent used the differential calculus to calculate the maximum efficiency of undershot waterwheels in 1704. A few years later Henri Pitot calculated the optimum number of blades an undershot waterwheel of a given size should have and presented his calculations in tabular form. By mid-century a debate arose whether the undershot or overshot waterwheel was the most efficient. Antoine de Parcieux, a member of the French Academy of Sciences, conducted a series of experimental tests using scale models and concluded that the overshot wheel was more efficient. At about the same time, John Smeaton in Britain began conducting an extensive series of experiments in which he systematically varied the type of wheel, the quantity of flowing water, the head of water, and the load on the wheel, and concluded that the overshot wheel was more efficient. His experiments would help establish a new “tradition of testability” [Constant, 1980, p. 20]. Smeaton’s technique would later be called “parameter variation,” and would become an important methodology of the engineering sciences [Vincenti, 1990, pp. 146-151]. The experimental research on waterwheels encouraged new theoretical work. In France, Jean Charles Borda, a member of the Academy of Sciences, analyzed waterwheels in terms of vis viva (mv2 — which is related to the modern concept of kinetic energy) and concluded that the inefficiencies of the undershot wheel were the result of vis viva being lost to turbulence. In 1783 Lazare Carnot, in

The Emergence of the Engineering Sciences: An Historical Analysis

131

his Essai sur les machines en gen´eral extended Borda’s analysis of waterwheels and developed a theory that was applicable to all machines [Gillispie, 1971]. By applying the idea of vis viva to machines Carnot concluded that all impacts and shocks needed to be avoided. By the 19th century the combination of experimental and theoretical studies of machines led to new design principles. In 1824 Jean Victor Poncelet designed a new waterwheel based on the work of Borda and Carnot. In America during the middle of the 19th century James B. Francis began a series of extensive experiments on a water turbine designed by Benoit Fourneyron [Layton, 1979]. Combining both the experiments and theoretical work based on Newtonian mechanics, Francis developed a series of design principles for turbines. At about the same time, G.G. Coriolis in France and William Whewell and Henry Moseley in Britain employed the idea of work to analyze machines. By comparing the transmitted work to the wasted work, they analyzed machines in terms of efficiency which would become another fundamental concept of the engineering sciences. Machines like waterwheels and turbines could be seen as devices that transmitted and modified force or work but machines could also be seen as devices that transmitted and modified motion. If one neglected the action of dynamic forces, a machine could be analyzed in terms of what became known as the theory of mechanisms [Ferguson, 1962; Hartenberg and Denavit, 1964], an idea that went back to Leonardo Da Vinci. By the beginning of the 18th century the Swedish natural philosopher Christopher Polhem created a Laboratorium mechanicum by collecting mechanisms from across Sweden. He put forward the idea that all machines could be created from a “mechanical alphabet” with the five ancient machines serving as vowels and other mechanisms serving as consonants. Throughout the 18th and 19th centuries several institutions, such as the Conservatoire des Arts et M´etiers in Paris, the Royal Institution in London and the Franklin Institute in Philadelphia, established large collections of mechanisms in order to provide a basis for the understanding and improvement of machinery. At the end of the ´ 18th century and the beginning of the 19th century Gaspard Monge at the Ecole Polytechnique argued that machines could best be understood by analyzing the elements of a machine that converted one type of motion into another. In order to do so, Monge and his followers, Agustin de Betancourt, Pierre Hˆachette and Phillipe Lanz, developed a system to classify mechanisms that was similar to Carl Linnaeus’ binomial classification system for plants. This work was extended by Andr´e-Marie Amp`ere, who coined the term cin´ematique, later called kinematics. By the middle of the 19th century Robert Willis at Cambridge began studying mechanisms in terms of the relationship of motions created by the mechanism, seeing that the action of a mechanism was independent of the given motion that was applied to the mechanism. By the second half of the century, Franz Reuleaux, at Charlottenberg, further extended the theory of mechanisms by moving away from the study of individual mechanisms and toward the analysis of mechanisms as part of an integrated system. Particularly important work was also done by

132

David F. Channell

James Clerk Maxwell in Britain in 1868 with his analysis of mechanical governors which led to one of the earliest theories of feedback control [Mayr, 1970].

3.4 Thermodynamics One of the most significant developments in the emergence of the engineering sciences was the establishment of thermodynamics [Cardwell, 1971]. With the demand for new sources of power brought about by the Industrial Revolution, and with the invention of the Newcomen, Watt, and high pressure engines to meet some of those new demands, there was an increased interest in improving the efficiency of these engines by gaining a better understanding of the scientific principles which lay behind heat engines. Beginning in 1768 John Smeaton conducted a series of experimental studies on a model steam engine using the same method of parameter variation that he used in his study of waterwheels. At the end of the 18th century James Watt conducted a number of experiments in order to improve his understanding of his engines. The fact that Watt allowed steam to expand in his engines made it difficult to calculate the work done by an engine without knowing how the pressure in the cylinder changed throughout the stroke. In 1796 one of Watt’s assistants developed a simple device that would make an “indicator diagram” which would record the drop in pressure by marking a piece of paper throughout the stroke. During the 19th century the pressure-volume (PV) diagram would become a fundamental element of thermodynamics. The invention in England of the high pressure steam engine at the beginning of the 19th century further stimulated the development of thermodynamics. When French engineers first saw the new high pressure steam engine, they began a program to understand the reasons behind its increased efficiency. The leading figure ´ in this effort was Nicholas L´eonard Sadi Carnot, who was trained at the Ecole Polytechnique. Still believing in a material theory of heat, Carnot used the work done on water power by 18th century French engineers to develop an analogous set of conditions for the maximum efficiency of a heat engine. The results, published in his R´eflexions sur la Puissance Motrice du Feu (1824), outlined what became known as the Carnot cycle which Emile Clapeyron later showed could be represented as a PV diagram. Carnot showed that his cycle represented the most efficient heat engine and that the cycle was independent of any particular theory as to the nature of heat, making his theory a completely general theory that could be applied to any type of heat engine. The development in the 1840s and 1850s of the mechanical theory of heat by J.P. Joule and others led to attempts by William Thomson (later Lord Kelvin) and W.J.M. Rankine in Britain and Rudolf Clausius in Germany to reconcile Carnot’s theory, which argued that work was produced by the transfer of heat from a higher temperature to a lower one, with Joule’s theory that heat was converted into work [Smith, 1998]. Clausius argued that the two theories could be reconciled if during a Carnot cycle some of the heat was converted into work and another proportion was simply transferred to a lower temperature. By mid-century Thomson and

The Emergence of the Engineering Sciences: An Historical Analysis

133

Rankine began to reformulate thermodynamics in terms of the new concept of energy. Near the same time Clausius and Rankine independently developed a formula for the heat dissipated in a Carnot cycle. Clausius would later introduce the term entropy to describe the “equivalence value” of this dissipated heat. Using the new concepts of energy and entropy Clausius was able to formulate two laws of thermodynamics: the first law being that the energy in the universe, or in any closed subsystem, was constant and the second law being that the entropy of the universe, or any closed subsystem, always tends towards a maximum. Although the theory of thermodynamics arose out of a study of the steam engine it was soon realized that the concepts of energy and entropy were not limited to the phenomena of heat but were universal concepts that could be applied to a wide range of scientific and technological phenomena, making thermodynamics a true engineering science.

3.5

Fluid mechanics

During the 18th and 19th centuries, developments in fluid mechanics also contributed to the emergence of the engineering sciences [Anderson, 1997; Rouce and Ince, 1957]. New demands to improve water power, to design better ships, and to better understand ballistics led to new theoretical and experimental studies of the behavior of fluids and objects moving through them. As early as 1673, in response to Colbert’s attempt to rationalize shipbuilding, Ignace-Gaston Paridies developed a theoretical study of bodies moving through fluids at varying speeds, and in 1697 Paul Hoste published an analysis of how the form of a ship influenced its speed and stability. In 1738 Daniel Bernoulli published his book Hydrodynamica in which he applied Newtonian mechanics and the Leibnizian concept of vis viva to the study of fluids. The book contained, although somewhat obscurely, what has come to be known as the Bernoulli principle which states that an increase in velocity of a fluid will result in a decrease in pressure of that fluid. The theoretical work done by Bernoulli was continued by Jean d’Alembert, who derived a differential equation that governed the motion of a fluid element of fixed mass. This theoretical work in the 18th century reached its high point with the work of Leonhard Euler, who placed the work of Bernoulli and d’Alembert into a consistent mathematical framework. In three papers published between 1752 and 1755 Euler formulated the basic equations that explained much of the behavior of fluids. Bernoulli, d’Alembert and Euler each used their theoretical studies to analyze the relationship between fluid mechanics and ship design [Pritchard, 1987]. While theoretical studies of fluid mechanics were taking place on the Continent, the British were conducting some important experimental studies, particularly on the movements of bodies through the air. During the middle of the 18th century Benjamin Robins used a pendulum-like device and a whirling arm mechanism to measure how air resistance affected cannon balls [Steele, 1997, pp. 145-180]. Euler would later use Robins’ data to develop a mathematical theory of ballistics. Not long after Robins’ work, John Smeaton conducted a series of experiments on

134

David F. Channell

windmills. His experiments led him to conclude that air moving over a curved surface created more force than air moving over a flat surface. At the beginning of the 19th century George Caley extended Smeaton’s work and demonstrated how a wing moving through the air could give rise to aerodynamic lift. During the 19th century engineers and scientists extended the work that had been done in the 18th century [Anderson, 1997]. Euler’s work had led to a significant advance in the understanding of the motion of fluids but his work neglected the effects of friction which was a particularly important effect when fluids were moving near a solid surface, like a hull or a wing. Between 1822 and 1845 the Frenchman C.L.M.H. Navier and the Irishman George Gabriel Stokes independently derived equations, known as the Navier-Stokes equations, which took into account the role of friction in the flow of fluids. As in other areas of the engineering sciences, a few 19th century engineers, especially in Britain, began to develop a more graphical approach to the problem of fluid mechanics that would provide a bridge between mathematical theories and practical experimental data. One area where this approach played an important role was in naval architecture. Traditionally ships’ hulls were designed with smooth surfaces but this was based on experience rather than theory. With larger and faster ships, naval architects needed to relate the shape of a ship’s hull to some laws of the motion of fluids in order to be able to rationally design a hull. In the 1840s John Scott Russell developed a graphical approach that based the shape of a hull on the shape of natural waves in water [Emmerson, 1977]. This idea was further developed by W.J.M. Rankine in 1870 when he introduced the new concept of streamlines, which would become a fundamental part of the new conceptual framework of the engineering sciences. 4 THE ENGINEERING SCIENCES IN THE ERA OF INDUSTRIAL RESEARCH: 1850-1925 During the late 18th and early 19th centuries the engineering sciences began to ´ emerge, primarily from academic institutions, such as the Ecole Polytechnique, Glasgow University, Stevens Institute of Technology, and the German Technische Hochschulen. But during the late 19th and early 20th centuries the engineering sciences more and more became associated with industrial research laboratories [Fox and Guagnini, 1998-1999]. This helped to reshape the engineering sciences into what some have labeled industry-based sciences [K¨onig, 1996, p. 100]. Beginning in the late 18th century there were significant new advancements in science, particularly in the understanding of chemistry and the understanding of electricity and magnetism. Some have called these developments a second scientific revolution. A result of this new scientific understanding was the emergence of new large-scale science-based industries which some have called a second industrial revolution. The new understanding of chemistry that emerged at the end of the 18th century led to new discoveries, such as coal tar dyes, the Leblanc process, the Solvay process, celluloid and plastics. These served as the basis for companies,

The Emergence of the Engineering Sciences: An Historical Analysis

135

such as Du Pont, BASF, Bayer, I.G. Farben and Kodak. The discovery of phenomena such as electromagnetism and electromagnetic induction during the early 19th century were quickly developed into inventions, such as the telegraph, the telephone, electric motors, electric lights and dynamos. These inventions in turn served as the basis for new large scale companies, such as Western Union, American Bell Telephone, Edison General Electric, Westinghouse, Telefunken, and Siemens. With the emergence of these science-based industries came the realization that industrialists could no longer rely on the Romantic ideal in which inventions appeared as the result of some flash of inspiration from a lone inventor. Instead there was the recognition that discovery and invention could be the result of a rational and planned process undertaken by a group of researchers. As a result, the establishment of the industrial research laboratory staffed by a multidisciplinary team of scientists and engineers helped to transform the engineering sciences into industry-based sciences. In addition to industrial research laboratories, universities began to establish experimental engineering laboratories and research stations, which often became closely linked to industries [Seely, 1993].

4.1

The chemical industries

Many of the developments in the chemical industries [Clow and Clow, 1952] during the 19th century can be traced to the work of the German chemist Justus ´ von Liebig, who after studying at the Ecole Polytechnique returned to Germany and established a chemical research laboratory at the University of Giessen [Beer, 1959]. The laboratory reflected the dual goals that the German educational system should teach both existing knowledge and create new knowledge. Although the laboratory’s original purpose was to train pharmacists, its emphasis slowly shifted to research in organic chemistry. In the process, Liebig created a number of instruments and techniques that allowed his students to analyze a large number of organic compounds. Most importantly, in 1840 he introduced a new approach to research in his laboratory. He began to organize a significant number of his students into teams to conduct research on fatty acids. This idea of systematic group research focused on a specific problem would become a hallmark of industry-based science. Liebig’s laboratory served to train an entire generation of academic and industrial chemists. One of the most important of Liebig’s students was August Wilhelm von Hofmann, who became professor of chemistry at the Royal College of Chemistry in England in 1845. As part of a systematic search for useful properties of organic compounds, Hofmann’s student, William Perkins in 1856 discovered that aniline, a byproduct of coal tar, had the ability to dye textiles a bright purple. The commercial success of this first chemical dye led to the search for other dyes in England and France. Through a chemical analysis of the first aniline dyes Hofmann provided a method by which chemists could systematically create new color dyes. Although the first chemical dyes were produced in England and France, by the 1870s Germany began to dominate the field by breaking down the social and edu-

136

David F. Channell

cational barriers that separated the universities from the Technische Hochschulen and thus creating a means by which theoretical knowledge could be combined with systematic experimental work. This allowed a bridging of the gap between academia and the factory. The factory-centered industrial research laboratory emerged in response to the demand for continual innovation which resulted from the commercial success of chemical dyes and the pressure to produce new and cheaper dyes. Some scholars have argued that the roots of industrial research in Germany can be traced to 1868 when BASF hired Heinrich Caro to become technical director of the company. Over the next few years Caro recruited a number of chemists from universities and Technische Hochschulen and created a central laboratory in 1889. Other companies followed BASF’s lead. In 1891 the Bayer Company centralized its research in a new laboratory. A distinguishing characteristic of the industrial research laboratory was group research. The discovery of new azo dyes opened the possibility of more than one hundred million new dyes. With such a large number of possible dyes the era of the lone inventor using trial and error was over. In its place came what Caro labeled “scientific mass work” (wissenschaftliche Massenarbeit) [Meyer-Thurow, 1982, p. 378]. The creation of the industrial research laboratory also led to a new relationship between academia and industry. As industrial research became more autonomous chemical firms began to rely more heavily upon university and Technische Hochschule graduates, especially those with Ph.D.s. In turn the success of industrial research laboratories put pressure on universities and Technische Hochschulen to adapt their curricula to the needs of industry. Wolfgang K¨ onig has argued that since a significant number of the faculty at German universities and Technische Hochschulen had gained their experience in industrial research laboratories, much of the work they produced should be classified as industry-based science [K¨ onig, 1996, p. 87]. The development of industrial research in the chemical dye industry quickly spread to other chemical industries. The pressure for continual innovation in chemical dyes soon led to the development of other organic chemicals, especially pharmaceuticals, celluloids and plastics. During the 19th century the new understanding of chemistry also led to the industrial development of heavy chemicals, such as alkalies, acids, fertilizers and explosives. In all of these areas chemical firms, such as Bayer, Agfa, Kodak and Du Pont established industrial research laboratories modeled after those in the chemical dye industry.

4.2 The electrical industries Along with the chemical industries, the leading area that contributed most to the transformation of the engineering sciences into industry-based sciences was the development of the electrical industries and the emerging field of electrical engineering [Reich, 1985]. The dependence of the electrical industries upon new scientific discoveries led them to follow a path similar to the chemical industries

The Emergence of the Engineering Sciences: An Historical Analysis

137

and to establish industrial research laboratories [Israel, 1992]. A key figure in this movement was Thomas Edison who established his private research laboratory at Menlo Park, New Jersey in 1876 [Israel, 1998]. Although Menlo Park differed from the typical industrial research laboratory in that it was independent of any specific company, it would serve as a prototype for company-based laboratories. Like the chemical laboratories, a key characteristic of Edison’s laboratory was the use of group research. As with the chemical laboratory, this was because a lone inventor using trial and error could no longer solve the problems that arose in the new science-based industry. The electrical industries were not based on single inventions; rather they were systems of inventions. Edison’s electrical lighting system involved generators, wiring, circuits, bulbs, switches, and meters, all of which had to function together. Developing such a system was not something that could be carried out by a single inventor, but instead it required the efforts of a group of engineers, scientists and entrepreneurs [Hughes, 1983]. By the beginning of the 20th century the intense competition and consolidation that had taken place during the 1880s and 1890s resulted in the domination of the industry by a few firms, such as Western Union, American Telephone and Telegraph, General Electric, Westinghouse, the Marconi Company and National Telephone in Britain, Telefunken and Siemens in Germany, and Philips in the Netherlands [de Vries, 2005]. But many of these firms were facing an uncertain future since many of the original patents were set to expire. In order for the electrical firms to be able to develop continual innovations and to protect their market share through control of patents, they needed to establish a way to control and manage the process of invention and innovation. In response to these needs, several of the leading firms in the electrical industry established industrial research laboratories during the early years of the 20th century [Reich, 1985]. The electrical industries were influenced by the German chemical industry but there were important differences in the industrial research laboratories that arose in the electrical industries. First, the electrical laboratories took a multidisciplinary approach that involved physicists, chemists, metallurgists, mechanical engineers and electrical engineers. Second, while the goal of the chemical laboratories was to discover new products, a significant goal of the electrical laboratories was defensive. That is, much of the research went into establishing a patent position and developing patent interferences in order to give the company a dominant, or almost monopolistic, role in the market. Third, the integration of pure and applied research played an important role in the electrical laboratories. In the German chemical industry much of the fundamental research was still left to the universities, but because of the newness of scientific research in the field, the electrical industries had to carry on both pure and applied research. As such the lines between pure and applied research, and between scientists and engineers, began to disappear. This new type of industrial research laboratory first emerged in 1900 with the establishment of the General Electric Research Laboratory in Schenectady, New York [Reich, 1985]. By this time Edison had left the company, his original patents

138

David F. Channell

on the light bulb had expired and GE was facing competition from metal filament bulbs invented in Europe. From the beginning, Willis Whitney, the director of the GE laboratory who had a Ph.D. in chemistry from Leipzig, focused the laboratory on the problem of improving the light bulb [Wise, 1985]. Through the work of teams of researchers led by William Coolidge, a Ph.D. from Leipzig, and Irving Langmuir, a Ph.D. from G¨ ottingen, GE developed a new argon filled bulb with a tungsten filament that gave the company a dominant position in the market. It is impossible to classify Coolidge’s and Langmuir’s work as simple applications of science to technology. While both conducted basic scientific research, that research was always done in the context of solving practical problems so that they simultaneously created new scientific knowledge and solved practical problems. During the early 20th century a number of other companies, such as AT&T, Siemens & Haske, Philips, and Westinghouse faced the same market forces that faced GE and they responded in a similar manner by establishing industrial research laboratories [de Vries, 2005]. In these laboratories research did not follow the pure science model. Rather these laboratories developed new technological theories and design methodologies that blurred the distinctions between science and technology and might be better labeled as industry-based engineering science.

5 THE ENGINEERING SCIENCES IN THE ERA OF GOVERNMENT SPONSORED RESEARCH: 1900-1945 Overlapping the emergence of the industrial research laboratories was the establishment of government sponsored research. The interdependence of science and technology was strongly influenced by the creation of industrial research laboratories, but during the first half of the 20th century the role of government — and especially military — directed scientific and technological development would play a significant role in shaping the engineering sciences. While governments have played a role in supporting science and technology ever since the rise of the modern nation-state in the 15th and 16th centuries, what distinguished the governments’ role in the 20th century was the breadth, scale and explicit nature of such support. Also, rather than supporting science for its own sake, governments in the 20th century began to see science as a form of knowledge, similar to technology, that could be manipulated for political power. For many scholars the key factor in establishing a new relationship between knowledge and power was the role played by warfare in the 20th century, especially World Wars I and II. Unlike previous wars, these were total wars which involved much of the globe. By erasing the distinctions between civilian and the military, the total wars of the 20th century required that all elements of society, including science and technology, be harnessed as part of the war effort. The two world wars helped to shape the engineering sciences by helping to establish the roots of what President Dwight Eisenhower called the military-industrial complex (or more correctly the military-industrial-academic complex) [Leslie, 1993].

The Emergence of the Engineering Sciences: An Historical Analysis

139

During the 19th century the increase in population, the rise of industrialization and the invention of new weapon systems led to the potential for a new scale of warfare. In particular the late 19th century naval arms race that led to the development of steam powered, armor plated ships with large, long-range armaments showed the advantage of having links between the military and industries, such as Krupp in Germany and Elswick Ordinance Company in Britain.

5.1

Government sponsored research in World War I

Soon after World War I began it devolved into a stalemate of trench warfare and great loss of life, especially on the Western Front. In response, governments on both sides of the conflict began to consider ways to break the stalemate and they began to mobilize scientists into the war effort [Hartcup, 1988]. Chemistry played a particularly important role during World War I, which was often called the chemists’ war. The demands for substitutes for blockaded goods, new high explosives, poison gases and the means to defend against them led governments on both sides to establish new relationships between government research institutes, universities, and chemical firms. For example, in Germany Fritz Haber helped to convert the Kaiser Wilhelm Institute into the chemical warfare arm of the military and was able to call on the I.G. Farben cartel to produce the new chemical weapons discovered at the Institute. In Britain, research on chemical warfare was conducted at Imperial College and the chemical firm of Castner-Keller began production of poison gas for the military. In America, the National Research Council established a laboratory at American University in Washington, D.C. to conduct research on chemical warfare. By 1918 President Woodrow Wilson established the Chemical Warfare Service in order to conduct chemical research for military needs and to build factories to produce those chemicals. Although World War I was known as the chemists’ war, governments and the military also encouraged research into other areas of science and technology. Wireless communications played a significant role in World War I, especially in naval operations. The Marconi Company provided equipment for the British Royal Navy as did the Telefunken Company for the German Navy. The American government encouraged research programs into wireless communications and protected companies, such as General Electric, Western Electric and AT&T from patent infringement litigation so that they could spend more money on research and development. Concern over U-boat attacks in the Atlantic led the American Naval Consulting Board to establish an experimental research station which brought together distinguished scientists, such as Irving Langmuir, with companies, such as GE and AT&T, to conduct research on submarine detection. Even before the war governments and the military encouraged research into aeronautics because of the potential use of airplanes as tools of war. Soon after a demonstration flight by the Wright Brothers in 1909 the British government established an Advisory Committee on Aeronautics and the National Physical Laboratory (NPL) built a wind tunnel in order to conduct aeronautical research.

140

David F. Channell

On the Continent significant work was being done on the theory of aeronautics [Anderson, 1997]. Nikolai Jouwski in Russia and Wilhelm Kutta in Germany developed a new theory of lift, while Ludwig Prandtl in Germany revolutionized the study of drag with his boundary layer theory. At the same time in France Gustave Eiffel began a series of experimental tests on lift using a wind tunnel. In the United States, Congress established the National Advisory Committee for Aeronautics (NACA) which contracted with William Durand and Everett Lesley at Stanford University to conduct wind tunnel tests on airfoils and propellers using parameter variation [Vincenti, 1990, pp. 137-169]. With the exception of Anthony Fokker’s use of Prandtl’s work to design aircraft for the Germans, aeronautical research played little role in aircraft design during World War I, but it would play a much larger role in revolutionizing aircraft design during the inter-war period. Britain’s NPL and NACA’s newly completed Langley Laboratory conducted new experimental tests that led to the streamlining of airplanes. By the 1930s NACA built a large scale wind tunnel which it used to design new cowlings for engines and a new series of airfoils [Anderson, 1997]. The new improvements in aircraft design and advances in aerodynamic theory led some individuals in Britain and Germany to contemplate attaining high speed, high altitude flight [Constant, 1980]. This led Frank Whittle in England and Hans von Ohain and Ernst Heinkel in Germany to begin work on turbojet engines. Much of this work was done with the support of the Aeronautical Research Council in Britain and the Air Ministry in Germany.

5.2 Government sponsored research in World War II While governments began to encourage and direct interactions between science and technology during World War I, governmental and military sponsored research began to take off during World War II [Hartcup, 2000]. While chemical warfare, wireless communications and airplanes had not proved decisive in World War I, a number of officials began to recognize that the outcome of World War II might hinge on weapon systems not yet even invented and therefore there needed to be a way to harness science, technology and industry for military needs. Because most of Europe was under the threat of aerial bombardment, much of the most successful wartime research was conducted in the United States. In 1940, even before America entered the war, a group of academic and industrial scientists and engineers convinced President Franklin Roosevelt to establish the National Defense Research Committee (NDRC) in order to direct wartime research. A year later the government established a broader organization, the Office of Scientific Research and Development (OSRD), to oversee the NDRC and medical research, and to actually develop that research into the production of new weapons. Since time was of the essence, the NDRC decided not to establish its own laboratories but to contract with universities and industry for the use of their laboratories and staff. This new organizational structure linking universities, industry and the

The Emergence of the Engineering Sciences: An Historical Analysis

141

military produced a large number of weapon systems that helped the Allies win the war — the most important of which were radar and the atomic bomb. It is often said that while the atomic bomb ended the war it was radar that won the war [Brown, 1999; Buderi,1996]. Although some early work on radar had been done in Germany and America during the 1920s, most historians give credit for the invention of radar to Robert Watson-Watt, head of the NPL’s Radio Research Station, whose work led to the first large-scale defensive radar network. Watson-Watt recognized that pulsed signals and short wavelengths were the key to a practical radar system. While British radar helped to win the Battle of Britain, traditional vacuum tube technology limited the development of smaller microwave radar which could be carried on planes and ships for offensive purposes. Work in England at the Clarendon Laboratory at Oxford and at the General Electric Laboratory in Wembley led to the development of the cavity magnetron in 1940 which could produce microwaves. Because of the difficulty producing the devices under wartime conditions in Britain, the new radar technology was shared with the Americans. Soon after, the Microwave Committee of the NDRC established a central Radiation Laboratory (Rad Lab) at M.I.T. With a staff that grew to more than four thousand, annual budget of $43 million and links with Bell Labs, Raytheon Corporation, Western Electric, General Electric and Westinghouse, the Rad Lab developed airborne radar systems that could detect other planes and Uboats and the laboratory created the Long Range Aid to Navigation (LORAN). At the same time, Section T of the NDRC, working with the Sylvania Corporation, developed and produced the proximity fuse which was a miniaturized radar system that could fit inside an anti-aircraft shell, which proved crucial in shooting down planes in the Pacific. All of this work brought together academic and industrial researchers in a way that blurred the distinctions between science and technology and would be better classified as engineering science. One of the most important developments that emerged from the alliance between the military, universities and industry was the atomic bomb [Rhodes, 1986]. Again most of the initial research occurred in Europe, beginning with the discovery of fission by Otto Hahn and Fritz Strassmann in Berlin in late 1938. A key step in developing an atomic bomb was work done by Otto Frisch and Rudolf Peierls in 1940 at the University of Birmingham in England. They recognized that if the isotope Uranium-235 could be separated from the more common Uranium-238 the critical mass for a nuclear chain reaction would be in the order of pounds rather than tons and the material could be used to produce a bomb. Other researchers in America indicated that Plutonium-239, a material that could be produced by bombarding Uranium-238 with neutrons, might also make a suitable material for a bomb. Again wartime conditions in Britain made it unlikely that the necessary research and development could be done there so the British shared their research with the Americans. Between 1940 and 1942 much of the research for the atomic bomb was done under contract from the NDRC at Columbia University, the Metallurgical Laboratory (Met Lab) at the University of Chicago, Iowa State University and the

142

David F. Channell

University of California at Berkeley. This research culminated in 1942 with the first controlled nuclear chain reaction at the Met Lab under the direction of Enrico Fermi. The nuclear reactor served both as a test-bed for a uranium atomic bomb and as a means by which material could be produced for a plutonium bomb. Shortly before the success of the Chicago reactor the OSRD decided that the Army Corps of Engineers was better suited to oversee the combination of research in industrial production that would be needed to actually create an atomic bomb. The newly created Manhattan Engineering District (or Manhattan Project) established three major research and industrial facilities: one at Oak Ridge, Tennessee to separate U-235 through electromagnetic separation or gaseous diffusion; a second at Hanford, Washington to produce plutonium using large-scale nuclear reactors; and a third at Los Alamos, New Mexico to design the two types of bombs. These facilities brought together academic and industrial researchers. The separation projects at Oak Ridge were run by Tennessee Eastman and Union Carbide based on research done at the University of California at Berkeley, and the Hanford facility was run by Du Pont based on research done at the University of Chicago. Again, the governmental/military directed research that led to the development of the atomic bomb transcended traditional distinctions between basic and applied research, or between science and technology and is better classified as engineering science. 6

THE ENGINEERING SCIENCES IN THE ERA OF TECHNOSCIENCE: 1945-2000

The Allied victory in World War II was also seen as a victory of the concept of the military-industrial-academic complex that produced the winning weapons. When the Cold War soon followed the end of World War II this led to a continuation, and even a dramatic increase of governmental support for science and technology. During the Cold War a number of new governmental agencies were established to support and direct scientific and technological research [Geiger, 1992]. In the United States the Navy established the Office of Naval Research (ONR), the Air Force founded the RAND Corporation, and the Congress created the National Science Foundation (NSF), the National Aeronautics and Space Administration (NASA) and the Atomic Energy Commission (AEC). In this period, these agencies distributed millions of dollars to industry and universities, especially for doctoral research. In doing so, the government helped shape the direction of research in science and technology by focusing funding on research into nuclear weapons, solid state electronics, rocketry, computer science, biotechnology and nanotechnology. At the same time Europe followed a similar path with the establishment of the European Center for Nuclear Research (CERN), the European Space Agency, and the French National Center for Scientific Research (CNRS). The Cold War governmental agencies helped to reshape the relationship between science and technology in such a way that it would become difficult to make meaningful distinctions between the two areas. In fact, a number of the new fields

The Emergence of the Engineering Sciences: An Historical Analysis

143

of research, such as solid state electronics, computer science and biotechnology, contained one term drawn from science and one from technology [Channell, 1991]. World War II demonstrated that very little research could be considered “pure.” Something as esoteric and seemingly unpractical as nuclear physics led to a weapon that would contribute to the end of World War II and come to define the Cold War. Philosopher Stephen Toulmin has noted that after World War II the basic focus of scientific research was no longer nature itself; rather the focus became some unit of technology, such as a reactor, missile, or computer [Capshew and Rader, 1992, p. 9]. This elimination of any real distinction between science and technology has led to the emergence of the concept of a single integrated realm of knowledge that some have labeled technoscience [Latour, 1987]. In such a realm technology and science are no longer simply dependent upon one another; rather they are interdependent so that technology cannot exist without science and science cannot exist without technology [Sassower, 1995, pp. 4, 24](for an argument that distinctions still exist in a conceptual sense between technology and theoretical science, see [Radder, this volume, Part I]).

6.1 Nuclear weapons One of the first areas to reflect the new concept of technoscience was the postwar development of nuclear weapons and nuclear power [Rhodes, 1986]. The success of the atomic bomb in contributing to the end of World War II and the new tensions of the Cold War led the United States and the Soviet Union to focus on developing more efficient and more powerful nuclear weapons. By the 1950s both countries, using academic, industrial and governmental supported research, had developed the more powerful H-bomb. About the same time new research began on nuclear reactors, especially in the United States. Much of the early research, supported by the AEC, was aimed at breeder reactors that could produce plutonium for weapons, or at compact reactors to power submarines, but by the mid-1950s research was also being done to produce commercial power reactors in response to President Eisenhower’s Atoms for Peace program. In all of this work the distinction between science and technology all but disappeared. Stuart W. Leslie has argued that this new research transformed the nature of physics into something closer to technology [Leslie, 1993, p. 133]. For example, in 1946 Harvard University changed its Department of Engineering Science into the Department of Engineering Science and Applied Physics and shortly after Cornell established a Department of Engineering Physics. Technology began to change nuclear physics in other ways. New experimental equipment, such as particle accelerators emerged from wartime microwave research and new detectors, such as spark and bubble chambers arose from technology originally developed for nuclear weapons or to detect missiles. This dependence upon technology began to influence the development of theories in nuclear physics. Peter Galison has argued that this complicated technology led laboratories to rely on “in house” theorists who developed theories more directly related to a

144

David F. Channell

specific machine being used in an experiment and outside theorists who developed theories that could only be tested on some specific machine [Galison, 1997, p. 43]. In addition these machines were so large, complex and expensive that the experiments conducted on them could only be managed by teams of researchers at national or even international laboratories.

6.2 The space race Closely related to the work on nuclear weapons was the development of missiles to deliver them. The modern idea of a guided ballistic missile began to emerge in the late 19th and early 20th centuries with the work of the Russians Konstantin Tsiolkovsky and Sergei Korolev and the German Hermann Oberth and the American Robert H. Goddard [Von Braun and Ordway, 1976]. But it was the successful development of the V-2 rocket by the Nazis during World War II that stimulated the postwar development of ballistic missiles and the space race between the United States and the Soviet Union [Neufeld, 1995]. Development of missiles provided the means to deliver nuclear weapons and the ability to launch satellites which could be used to improve communications and to spy on the other side [McDougall, 1985]. In addition the peaceful and scientific exploration of space often functioned as a surrogate for the arms race between the Americans and the Soviets. In both the United States and the Soviet Union the space race brought together government, military, academic and industrial research in such a way that it blurred any distinction between science and technology [Bromberg, 1999]. As with nuclear studies, the space program changed the nature of science, particularly planetary science and astronomy. New planetary probes and space telescopes required an interdisciplinary team of astronomers, physicists, aeronautical engineers, mechanical engineers, electrical engineers and computer scientists. Most of the equipment had to be built by private industry and required government funding. As such the actual research had to be managed by national or international laboratories.

6.3 Solid state electronics A third area of research and development that reflected the new concept of technoscience was the field of solid state electronics [Riordan and Hoddeson, 1997]. During the 1920s and 1930s physicists began to apply quantum mechanics to solid state materials and began to develop theories about how electrons behave in a newly discovered class of materials called semiconductors. Just before the beginning of World War II, Bell Labs began fundamental research on the solid state but the research was put aside during the war in favor of work on radar. Wartime work on radar led to a new understanding of the properties of semiconductors, and with the end of the war, Bell returned to research on semiconductors in hopes of finding a solid state version of the vacuum tube triode. Soon after the war Bell Labs established an interdisciplinary team of researchers, led by William Shockley, a theoretical physicist, Walter Brattain, an experimental physicist, and John

The Emergence of the Engineering Sciences: An Historical Analysis

145

Bardeen, a theoretical physicist who also had an M.S. in electrical engineering. Combining theory and experiment Brattain and Bardeen demonstrated the first point-contact transistor in December of 1947, and one month later Shockley had the idea for a junction transistor, which would ultimately replace the point-contact transistor. The U.S. military was a major force behind the development of the transistor. The Army Signal Corps was particularly interested in miniaturizing communication devices and the military became a major consumer of transistors at a time when their high cost limited civilian applications. It was also the military that pushed the electronics industry to switch from germanium to silicon, which was more suitable for use in guided missiles and nuclear powered ships. The military also encouraged the dissemination of knowledge about the transistor to industry and universities. With the military market for transistors declining after the end of the Korean War, new civilian markets began to emerge, such as hearing aids and radios. Many of the new civilian applications of transistors originated in Japan which had been forbidden to have a military and therefore sought other applications for transistors. By the second half of the 1950s Shockley left Bell and created a new company in the Stanford Industrial Park in California which had been created to encourage cooperation between the university and private industry. This would become the beginning of what came to be known as Silicon Valley. The success of new markets for transistors led to improvements in the manufacture of transistor circuits. This culminated in the independent invention of the integrated circuit in 1959 by Jack Kilby at Texas Instruments and by Robert Noyce at Fairchild Semiconductor in California.

6.4

Computers and computer science

One of the most significant applications of transistors and integrated circuits was in the computer. Although the idea of a general purpose computer goes back to the first half of the 19th century when Charles Babbage in England conceived of his Analytical Engine, it was not until World War II that the British and the Americans built actual specialized calculating devices [Goldstine, 1972]. In England a team at Bletchley Park under Alan Turing built a machine called Colossus which was designed to help break the German codes. Also during the war Howard Aitken at Harvard and a team at IBM developed the Mark I, which was used to do ballistic tables and to solve problems arising from the development of the atomic bomb. These early computers were limited by the fact that they were electromechanical devices and designed for specific applications. After the war several breakthroughs led to the creation of the general application electronic computer [Ceruzzi, 2003]. Many of the early improvements were helped by wartime research on radar. The first improvement was the Electronic Numerical Integrator and Computer (ENIAC), the first all electronic computer, designed and built for the U.S. Army by John Eckert and John Mauchy in 1945 at

146

David F. Channell

the University of Pennsylvania. This was followed by the Electronic Delay Storage Automatic Computer (EDSAC), the first full-scale stored program computer that was built in 1949 by Maurice Wilkes at the University of Cambridge, based on a design by John von Neumann. The modern electronic general purpose computer arose with the further improvements brought about by the invention of the random access magnetic core memory by Jay Forrester in 1949 at M.I.T.’s Servomechanisms Laboratory, and Philco’s development of SOLO in 1957, one of the first all transistor computers, and finally by the use of integrated circuits by several firms in the 1960s which resulted in the development of the minicomputer. While the modern computer was the result of a combination of science and technology that could be labeled technoscience, its development also led to the creation of a new technoscience that became known as computer science. At the same time that core memories, transistors and integrated circuits were transforming computer hardware, significant changes were taking place in the development of computer software. A breakthrough came with the realization that computers essentially manipulate symbols and that computer instructions themselves could be “coded” into the machines where they could execute commands [CampbellKelly and Aspray, 1996]. By the late 1950s higher level programming languages, such as FORTRAN and COBOL, began to be developed. In addition researchers began creating operating systems that controlled how computers scheduled tasks when more than one program was running. The combination of developments in hardware and software led to the emergence of what became known in North America as computer science. In 1967 Herbert Simon, Alan Perlis and Allan Newell at Carnegie Institute of Technology in Pittsburgh, argued that computer science was the study of computers in the same way that astronomy was the study of stars, except that computers were humanly constructed artifacts. As such computer science was a “science of the artificial” [Simon, 1969]. By 1968 many in the field of computer science were moving away from focusing on the computer itself and turning toward the notion of computation as the real focus of a science of computing. This led to the study of algorithms by Donald Knuth and others, and in 1968 the Association for Computing Machinery recommended a new curriculum for computer science which replaced courses on computer hardware with courses on logical design, switching theory and algorithms. This new focus on computation further blurred the distinctions between science and technology since computation could be seen as either a human construction, and therefore technological, or as a branch of mathematics, and therefore a foundation of the sciences. By the end of the 20th century the idea of computation was being used to model both physical and biological phenomena, including human intelligence. Work done by Norbert Wiener and Julian Bigelow on anti-aircraft guns during World War II led to the development of mathematical theories of control and feedback in machines, which became the basis of the new field of cybernetics. Beginning in 1950 Alan Turing, influenced by Wiener, proposed the idea that a computer could exhibit intelligent behavior, which helped to create the field of artificial intelligence.

The Emergence of the Engineering Sciences: An Historical Analysis

147

Near the end of the century physicists and biologists began to model natural phenomena using computers which led to the solution of many unsolved problems. The success of computational physics and biology led some researchers to argue that computation was not simply a methodology for solving scientific problems but that the universe itself was essentially computational and could be best understood in terms of information processing. This of course further blurred the distinctions between science and technology with science being seen more as applied technology [Channell, 2004; Forman, 2007].

6.5

Material science: lasers, superconductivity and nanotechnology

An area also closely associated with solid state electronics but which extended beyond it was material science. The application of quantum mechanics to solids during the 1920s and 1930s led to new knowledge of the relationship between atomic structure and the overall properties of materials. This opened the possibility of designing materials with a given set of properties. After Sputnik the U.S. Department of Defense’s Advanced Research Project Agency (DARPA) became interested in developing materials that could function in the extreme environments of the missile and space programs [Leslie, 1993, pp. 218-219]. Through DARPA the U.S. government funded a number of interdisciplinary materials research laboratories at major universities and funded research into new analytical techniques, such as electron microscopy, X-ray diffraction, and nuclear magnetic resonance. One of the most significant achievements of material science was the development of the laser [Bromberg, 1991; Hecht, 2005]. Like the transistor, the laser had roots in both electronics and material science. After World War II the military became interested in generating radio waves in millimeter range which would be difficult to intercept. During the late 1940s the Army, Navy and Air Force created laboratories at Columbia, M.I.T., Stanford, and Harvard to work on the problem. At Columbia Charles Townes discovered that microwave radiation could be amplified by stimulating electrons to a higher state. By 1954 Townes and his graduate student James Gordon produced the maser which produced amplified microwaves by stimulated emission of radiation in ammonia gas. The maser was independently invented by Nikolai Brasov and Alexander Prochorov in the Soviet Union. During the second half of the 1950s a number of researchers, including Townes, used an idea of Joseph Weber to create a solid state maser using ruby crystals, and by 1957 Townes and R. Gordon Gould at Columbia and Arthur Schawlow at Bell Labs independently began work on the laser which would amplify light by stimulated emission of radiation. Although Bell Labs was eventually granted the patent, all three are usually given credit with the invention of the laser, but it was not until 1959-1960 that Theodore Maiman of the Hughes Laboratory produced the first working laser. Another area of research in material science that reflected technoscience was the field of superconductivity, especially high temperature superconductivity [Matricon and Waysand, 2003; Nowotny and Felt, 1997]. The phenomenon of super-

148

David F. Channell

conductivity had been discovered at the beginning of the 20th century by Heike Kamerlingh Onnes at the University of Leyden, but the phenomenon only seemed to occur at temperatures below 23 K and could not be explained by classical physics. It was not until 1957 that John Bardeen, Leon Cooper and J.R. Schrieffer developed the BCS theory which used quantum mechanics to explain one type of superconductivity. Although the theory had limited practical applications it did explain the phenomenon of tunneling which led Gerd Binnig and Heinrich Rohrer at IBM’s Zurich Research Laboratory to invent the scanning tunneling microscope in 1985 [Rasmussen, 1997]. Some of the most significant advances in superconductivity came from experimental research. By 1986 Alex M¨ uller and Georg Bednorz at IBM’s Zurich Research Laboratory discovered a ceramic oxide that was superconducting at 35 K. Within a few months Shoji Tanaka, at the University of Tokyo, M.K. Wu, at the University of Alabama-Huntsville, and Paul Chu, at the University of Houston, discovered other materials that were superconducting at 98 K, which could be attained with inexpensive liquid nitrogen rather than the more expensive liquid helium. While the development of high temperature superconductors represents another blurring of science and technology it may also represent a new phase of technoscience. In contrast to the earlier big science projects, most of this work was the result of small laboratories with teams of one or two individuals. But the work does not represent a return to traditional science. The necessity to empirically consider almost an infinite number of materials led to the establishment of alliances and networks of researchers that Michel Callon has labeled the “extended lab” [Nowotny and Felt, 1997, p. 5]. One of the most recent developments in material science is the emergence of nanotechnology [Regis, 1995]. Some historians trace the origin of the idea of nanotechnology to a talk by physicist Richard Feynman in 1959 in which he argued that miniaturization in technology could be pushed much further and he suggested designing a linked series of machines, each smaller that the other, in which the smallest could be capable of manipulating individual atoms. About the same time Arthur von Hippel, an engineer at M.I.T. suggested that new materials could be custom designed by fabricating them one atom at a time. But it would not be until the 1980s that any significant work in the field would begin. It was then that K. Eric Drexler, a graduate student at M.I.T. suggested using genetic engineering techniques to manipulate proteins in order to produce the equivalent of small machines [Drexler, 1987]. In the mid-1980s Carl Pabo, Jay Ponder and Frederic Richards at Yale “inverted” the usual biological approach and were able to find a sequence of amino acids that would produce a protein with a given shape. At the same time that protein engineering was showing some potential to design nanostructures, Binnig and Rohrer, inventors of the scanning tunneling microscope, discovered that if the tip of the instrument got too close to the surface of a sample an atom would stick to the tip and could be moved [Rasmussen, 1997]. Shortly after, groups at Bell Labs and the IBM Almaden Research Center in San Jose, California created a device that could move atoms to specific locations. Other

The Emergence of the Engineering Sciences: An Historical Analysis

149

researchers at M.I.T. and AT&T used techniques originally developed to manufacture integrated circuits and produced micron-sized machines by etching away silicon from a chip. Although none of these techniques produced any truly practical devices, in 1991 the Japanese Ministry of International Trade Industry (MITI) provided $200 million for the development of nanotechnology. The next year the U.S. Senate Subcommittee on Science, Technology and Space, chaired by Albert Gore, began hearings which led to the National Nanotechnology Initiative (NNI), signed into law in 2000, which would provide billions of dollars for nanotechnology research. Ironically much of the government-funded research followed a different path than that advocated by Feynman or Drexler. Instead of focusing on assembling materials one atom at a time, researchers turned to more traditional chemical techniques. This approach led Richard Smalley, Robert Curl and Harold Kroto at Rice University in Houston to discover a new arrangement of carbon atoms that was similar to the geodesic domes designed by Buckminster Fuller. These new fullerenes, or buckyballs, had a number of interesting applications, such as cagelike structures to deliver medicine to certain parts of the body. By 1991 Sumio Iijima, a researcher at NEC in Japan discovered that fullerenes could exist as long tubes, labeled nanotubes. Such tubes held the promise of use in high voltage electric lines or as textile-like materials that would be stronger than Kevlar. Although still just emerging at the end of the 20th century, nanotechnology reflects many of the characteristics of technoscience. Research into nanotechnology cannot be easily divided into science versus technology or basic versus applied. Work aimed at developing practical nanotechnology has often resulted in new basic scientific discoveries, such as the first artificial proteins or fullerenes. At the same time, basic research, such as work on protein folding or the search for new fullerenes is always done with one eye on practical applications. But, like superconductivity, nanotechnology may represent a new phase of technoscience. While it has some military applications, most of the funding has come from nonmilitary government agencies, such as MITI or NNI. Also, while billions of dollars have been allocated for nanotechnology research, that funding has been divided among a number of projects which has led to alliances and networks that may also reflect Callon’s concept of the “extended lab” [Nowotny and Felt, 1997, p. 5].

6.6

Biotechnology

An area connected to computer science and to material science that reflects the notion of technoscience is biotechnology [Judson, 1979; Cherfas, 1982]. Robert Bud has argued that the modern conception of biotechnology emerged out of the industrialization of fermentation during the late 19th and early 20th centuries, which was used to supply Germany with animal fodder and ingredients for explosives during World War I [Bud, 1993]. But the real advancement of biotechnology took place after World War II when in 1953 Francis Crick and James Watson determined the structure of DNA [Watson, 2003]. The discovery of the structure of DNA led

150

David F. Channell

to research into the way DNA carried genetic information [Kay, 2000]. Using an idea put forward by Erwin Schr¨ odinger in 1943 that genes store information as a code-script, researchers began to apply World War II code-breaking techniques and computers to the problem of deciphering the genetic code. Although the crypto-analytic technique came close to breaking the code, the problem of the way in which DNA functioned was ultimately solved by more traditional techniques in 1961 by Marshall Nirenberg and Heinrich Matthaei at the National Institutes of Health in Washington, D.C., but researchers still often thought of DNA in terms of codes and information. The breaking of the genetic code opened up a new range of possibilities for biotechnology. With the combination of Stanley Cohen’s discovery in 1971 of the use of plasmids to “infect” E. coli with new genetic properties, and the discovery in 1967 of enzymes, labeled ligase, which could glue strands of DNA, and the discovery in 1970 of restriction enzymes, which cut strands of DNA, the elements to manipulate DNA were in place. By 1973 Cohen and Herbert Boyer, working in the San Francisco Bay area, combined two plasmids, each conferring resistance to a different antibiotic, and created a plasmid with recombinant DNA that conferred resistance to both antibiotics. Boyer and Cohen’s success led to the founding of Genentech in 1976 which became the first biotechnology company based on genetic engineering. This was soon followed by a large number of new companies such as Biogen and Amgen. Since many of these companies were founded by university researchers a number of universities soon began to create laboratories and institutes whose specific purpose was to create new commercial biotech products and processes. This further blurred the lines between industrial and academic research and between pure and applied science. The potential of biotechnology led to a proposal in 1985 by Robert Sinsheimer of the University of California at Santa Cruz to determine the genetic code of the human genome. The plan was made possible by the development during the late 1970s of gene sequencing techniques by Walter Gilbert at Harvard and Fred Sanger at Cambridge. By 1988, with funding from the U.S. Department of Energy and the National Institutes of Health, the Human Genome Project was established. Although the project was conceived in the United States, a significant amount of work was done in Great Britain, France, Germany and Japan. Debates over sequencing technologies led to the establishment of a competing private commercial program to sequence the human genome at Celera Genomics headed by Craig Venter. By 2000 both the government program which primarily funded academic science and the commercially driven program had led to a “rough draft” of the human genome. As a result of this success not only have the distinctions between pure and applied research and between science and technology again become blurred, but so have the distinctions between the organic and the technological [Channell, 1991].

The Emergence of the Engineering Sciences: An Historical Analysis

151

BIBLIOGRAPHY [Alder, 1997] K. Alder. Engineering the Revolution: Arms and Enlightenment in France, 17631815. Princeton University Press, 1997. [Anderson, 1997] J. Anderson, Jr. A History of Aerodynamics. Cambridge University Press, 1997. [Artz, 1966] F. Artz. The Development of Technical Education in France, 1500-1850. MIT Press, 1966. [Beer, 1959] J. Beer. The Emergence of the German Dye Industry. University of Illinois Press, 1959. [de B´ elidor, 1729] B. de B´ elidor. La science des ing´ enieurs. C. Jombert, 1729. [Billington, 1983] D. Billington. The Tower and the Bridge: The New Art of Structural Engineering. Basic Books, 1983. [B¨ ohme et al., 1978] G. B¨ ohme, W. van den Daele, and W. Krohn. The Scientification of Technology. In: The Dynamics of Science and Technology, W. Krohn, E. Layton, Jr. and P. Weingart, eds., pp. 219-250. D. Reidel, 1978. [Bromberg, 1991] J. Bromberg. The Laser in America, 1950-1970. MIT Press, 1991. [Bromberg, 1999] J. Bromberg. NASA and the Space Industry. Johns Hopkins University Press, 1999. [Brown, 1999] L. Brown. A Radar History of World War II. Institute of Physics Publishing, 1999. [Brunot and Coquand, 1982] A. Brunot and R. Coquand. Le Corps des Ponts et Chauss´ ees. Centre National de la Recherche Scientifique, 1982. [Bud, 1993] R. Bud. The Uses of Life: A History of Biotechnology. Cambridge University Press, 1993. [Buderi, 1996] R. Buderi. The Invention that Changed the World. Simon & Schuster, 1996. [Calvert, 1967] M. Calvert. The Mechanical Engineer in America, 1830-1910. Johns Hopkins University Press, 1967. [Campbell-Kelly and Aspray, 1996] M. Campbell-Kelly and W. Aspray. Computer: A History of the Information Machine. Basic Books, 1996. [Capshew and Rader, 1992] J. Capshew and K. Rader. Big Science: Price to Present. In: Science After ’40, A. Thackray, ed. Osiris, 7, 3-25, 1992. [Cardwell, 1971] D. Cardwell. From Watt to Clausius: The Rise of Thermodynamics in the Early Industrial Age. Cornell University Press, 1971. [Cardwell, 1995] D. Cardwell. Technology: The Norton History of Technology. W.W. Norton, 1995. [Ceruzzi, 2003] P. Ceruzzi. A History of Modern Computing, 2nd ed. MIT Press, 2003. [Channell, 1982] D. Channell. The Harmony of Theory and Practice: The Engineering Science of W.J.M. Rankine. Technology and Culture, 23, 39-52, 1982. [Channell, 1989] D. Channell. The History of Engineering Science: An Annotated Bibliography. Garland Publishing, 1989. [Channell, 1991] D. Channell. The Vital Machine: A Study of Technology and Organic Life. Oxford University Press, 1991. [Channell, 2004] D. Channell. The Computer at Nature’s Core. Wired Magazine, 2, 79-80, 2004. [Cherfas, 1982] J. Cherfas. Man-Made Life. Pantheon Books, 1982. [Clagett, 1959] M. Clagett. The Science of Mechanics in the Middle Ages. University of Wisconsin Press, 1959. [Clagett, 1963] M. Clagett. Greek Science in Antiquity. Collier, 1963. [Clow and Clow, 1952] A. Clow and N. Clow. The Chemical Revolution: A Contribution to Social Technology. Batchworth Press, 1952. [Constant, 1980] E. Constant, II. The Origins of the Turbojet Revolution. Johns Hopkins University Press, 1980. [de Vries, 2005] M. de Vries. 80 Years of Research at Philips: The Role of the Philips Natuurkundig Laboratorium (1914-1994). University of Amsterdam Press, 2005. [Dee, 1975] J. Dee. The Mathematical Preface to Euclid 1570. Science History Publications, 1975. [Dijksterhuis, 1961] E. Dijksterhuis. Mechanization of the World Picture. Oxford University Press, 1961.

152

David F. Channell

[Drachmann, 1948] A. Drachmann. Ktesibios, Philon and Heron: A Study of Ancient Pneumatics. Ejnar Munksgaard, 1948. [Drachmann, 1962] A. Drachmann. The Mechanical Technology of Greek and Roman Antiquity. University of Wisconsin Press, 1962. [Drake and Drabkin, 1969] S. Drake and I. Drabkin, eds. Mechanics in Sixteenth Century Italy. University of Wisconsin Press, 1969. [Drexler, 1987] K. E. Drexler. Engines of Creation. Anchor Books, 1987. [Durand, 1929] W. Durand. Robert Henry Thurston. American Society of Mechanical Engineers, 1929. [Easton, 1952] S. Easton. Roger Bacon and His Search for a Universal Science. Greenwood Press, 1952. [Emmerson, 1977] G. Emmerson. John Scott Russell: A Great Victorian Engineer and Naval Architect. John Murray, 1977. [Farrington, 1951] B. Farrington. Francis Bacon: Philosopher of Industrial Science. Lawrence and Wishart, 1951. [Ferguson, 1962] E. Ferguson. Kinematics of Mechanisms from the Time of Watt. United States National Museum Bulletin, no. 228. Smithsonian Institution, 1962. [Ferguson, 1977] E. Ferguson. The Mind’s Eye: Non-Verbal Thought in Technology. Science, 197, 827-836, 1977. [Forman, 2007] P. Forman. The Primacy of Science in Modernity, of Technology in Postmodernity, and of Ideology in the History of Technology. History and Technology, 23, 1-152, 2007. ´ [Fourcy, 1987] A. Fourcy. Histoire de L’Ecole Polytechnique. Belin, 1987. [Fox and Guagnini, 1998–1999] R. Fox and A. Guagnini. Laboratories, Workshops, and Sites. Concepts and Practices of Research in Industrial Europe, 1800-1914. Historical Studies in the Physical and Biological Sciences, 29, 55-140 and 191-294, 1998-1999. [Galilei, 1954] G. Galilei. Discorsi e Dimostrazioni Matematiche, Intorno a due Nuovue Scienze. Elzevir, 1638. Translation. Two New Sciences. Dover, 1954. [Galison, 1997] P. Galison. Image and Logic: A Material Culture of Microphysics. University of Chicago Press, 1997. [Geiger, 1992] R. Geiger. Science, Universities, and National Defense, 1945-1970. In: Science After ’40, A. Thakray, ed. Osiris, 7, 26-48, 1992. [Gillispie, 1971] C. Gillispie. Lazare Carnot: Savant. Princeton University Press, 1971. [Gillmor, 1971] C. Gillmor. Coulomb and the Evolution of Physics and Engineering in Eighteenth-Century France. Princeton University Press, 1971. [Gispen, 1989] K. Gispen. New Profession, Old Order: Engineers and German Society, 18151914. Cambridge University Press, 1989. [Goldstine, 1972] H. Goldstine. The Computer from Pascal to von Neumann. Princeton University Press, 1972. [Hacker, 1997] B. Hacker. Greek Catapults and Catapult Technology: Science, Technology and War in the Ancient World. In Technology and the West, T. Reynolds and S. Cutcliffe, eds., pp. 49-66. University of Chicago Press, 1997. [Hall, 1973] M. B. Hall, ed. The “Pneumatics” of Hero of Alexandria: A Facsimile of the 1851 Woodward Edition. American Elsevier, 1973. [Hartcup, 1988] G. Hartcup. The War of Invention: Scientific Developments, 1914-18. Brassey’s Defense Publishers, 1988. [Hartcup, 2000] G. Hartcup. The Effects of Science on the Second World War. Macmillan, 2000. [Hartenberg and Denavit, 1964] R. Hartenberg and J. Denavit. Kinematic Synthesis of Linkages. McGraw-Hill, 1964. [Hecht, 2005] J. Hecht. Beam: The Race to Make the Laser. Oxford University Press, 2005. [Heyman, 1972] J. Heyman. Coulomb’s Memoir on Statics: An Essay in the History of Civil Engineering. Cambridge University Press, 1972. [Hughes, 1983] T. Hughes. Networks of Power: Electrification in Western Society, 1880-1930. Johns Hopkins University Press, 1983. [Israel, 1992] P. Israel. From Machine Shop to Industrial Laboratory. Johns Hopkins University Press, 1992. [Israel, 1998] P. Israel. Edison: A Life of Invention. John Wiley, 1998. [Jacob, 1988] M. Jacob. The Cultural Meaning of the Scientific Revolution. Temple University Press, 1988.

The Emergence of the Engineering Sciences: An Historical Analysis

153

[Judson, 1979] H. Judson. The Eighth Day of Creation: Makers of the Revolution in Biology. Simon & Schuster, 1979. [Kay, 2000] L. Kay. Who Wrote the Book of Life: A History of the Genetic Code. Stanford University Press, 2000. [Kline, 1995] R. Kline. Construing ‘Technology’ as ‘Applied Science’: Public Rhetoric of Scientists and Engineers in the United States, 1880-1945. Isis, 86, 194-221, 1995. [K¨ onig, 1996] W. K¨ onig. Science-Based Industry or Industry-Based Science? Electrical Engineering in Germany before World War I. Technology and Culture, 37, 70-101, 1996. [Kranakis, 1997] E. Kranakis. Constructing a Bridge: An Exploration of Engineering Culture, Design and Research in Nineteenth Century America. MIT Press, 1997. [Krohn et al., 1978] W. Krohn, E. Layton Jr., and P. Weingart, eds. The Dynamics of Science and Technology. D. Reidel, 1978. [Kuhn, 1977] T. Kuhn. The Essential Tension: Selected Studies in Scientific Tradition and Change. University of Chicago Press, 1977. [Landels, 1978] J. Landels. Engineering in the Ancient World. University of California Press, 1978. [Latour, 1987] B. Latour. Science in Action. Harvard University Press, 1987. [Laudan, 1984] R. Laudan, ed. The Nature of Technological Knowledge: Are Models of Scientific Change Relevant? D. Reidel, 1984. [Layton, 1971] E. Layton, Jr. Mirror-Image Twins: The Communities of Science and Technology in 19th Century America. Technology and Culture, 12, 562-580, 1971. [Layton, 1974] E. Layton, Jr. Technology as Knowledge. Technology and Culture, 15, 31-41, 1974. [Layton, 1979] E. Layton, Jr. Scientific Technology, 1845-1900: The Hydraulic Turbine and the Origins of American Industrial Research. Technology and Culture, 20, 64-89, 1979. [Leslie, 1993] S. W. Leslie. The Cold War and American Science: The Military-IndustrialAcademic Complex at MIT and Stanford. Columbia University Press, 1993. [Lindberg, 1983] D. Lindberg. Roger Bacon’s Philosophy of Nature. Clarendon Press, 1983. [Lindberg, 1992] D. Lindberg. The Beginnings of Western Science. University of Chicago Press, 1992. [Lindberg and Westman, 1990] D. Lindberg and R. Westman. Reappraisals of the Scientific Revolution. Cambridge University Press, 1990. [McDougall, 1985] W. McDougall. The Heavens and the Earth: A Political History of the Space Age. Basic Books, 1985. [McEvoy, 1982] J. McEvoy. The Philosophy of Robert Grosseteste. Clarendon Press, 1982. [Manegold, 1992] K. Manegold. Technology Academised: Education and Training of the Engineer in the Nineteenth Century. In Dynamics of Science and Technology, W. Krohn, E. Layton, Jr. and P. Weingart, eds., pp. 137-158. D. Reidel, 1992. [Marsden, 1992] B. Marsden. Engineering Science in Glasgow: Economy, Efficiency and Measurement as Prime Movers in the Differentiation of an Academic Discipline. British Journal for the History of Science, 25, 319-346, 1992. [Martin, 1992] J. Martin. Francis Bacon, the State, and the Reform of Natural Philosophy. Cambridge University Press, 1992. [Matricon and Waysand, 2003] J. Matricon and G. Waysand. The Cold Wars: A History of Superconductivity. Rutgers University Press, 2003. [Mayr, 1970] O. Mayr. The Origins of Feedback Control. MIT Press, 1970. [Meyer-Thurow, 1982] G. Meyer-Thurow. The Industrialization of Invention: A Case Study from the German Dye Industry. Isis, 73, 363-381, 1982. [Neufeld, 1995] M. Neufeld. The Rocket and the Reich. Free Press, 1995. [Newman, 2004] W. Newman. Promethean Ambitions: Alchemy and the Quest to Perfect Nature. University of Chicago Press, 2004. [Nowotny and Felt, 1997] H. Nowotny and U. Felt. After the Breakthrough: The Emergence of High-Temperature Superconductivity as a Research Field. Cambridge University Press, 1997. [Oldenziel, 2006] R. Oldenziel. Signifying Semantics for the History of Technology. Technology and Culture, 47, 477-485, 2006. [Prager and Scaglia, 1970] F. Prager and G. Scaglia. Brunelleschi: Studies of His Technology and Inventions. MIT Press, 1970. [Pritchard, 1987] J. Pritchard. The Professionalization of French Naval Shipbuilders. Technology and Culture, 28, 1-25, 1987.

154

David F. Channell

[Rasmussen, 1997] N. Rasmussen. Picture Control: The Electron Microscope and the Transformation of Biology in America, 1940-1960. Stanford University Press, 1997. [Regis, 1995] E. Regis. Nano: The Emerging Science of Nanotechnology. Little, Brown and Company, 1995. [Reich, 1985] L. Reich. The Making of American Industrial Research: Science and Business at GE and Bell, 1876-1926. Cambridge University Press, 1985. [Reynolds, 1983] T. Reynolds. Stronger than a Hundred Men: A History of the Vertical Water Wheel. Johns Hopkins University Press, 1983. [Reti, 1974] L. Reti, ed. The Unknown Leonardo. McGraw-Hill, 1974. [Riordan and Hoddeson, 1997] M. Riordan and L. Hoddeson. Crystal Fire: The Birth of the Information Age. W.W. Norton, 1997. [Rhodes, 1986] R. Rhodes. The Making of the Atomic Bomb. Simon and Schuster, 1986. [Rossi, 1970] P. Rossi. Philosophy, Technology and the Arts in the Early Modern Era. Harper and Row, 1970. [Rouse and Ince, 1957] H. Rouse and S. Ince. History of Hydraulics. Iowa Institute of Hydraulic Research, 1957. [Sassower, 1995] R. Sassower. Cultural Collisions: Postmodern Technoscience. Routledge, 1995. [Schofield, 1963] R. Schofield. The Lunar Society of Birmingham. Oxford University Press, 1963. [Seely, 1993] B. Seely. “Research, Engineering, and Science in American Engineering Colleges: 1900-1960.” Technology and Culture, 34, 344-386, 1993. ´ [Shinn, 1980] T. Shinn. Savoir Scientifique et Pouvoir Social: L’Ecole Polytechnique, 17941914. Foundation National des Sciences Politiques, 1980. [Simon, 1969] H. Simon. The Sciences of the Artificial. MIT Press, 1969. [Smith, 1998] C. Smith. The Science of Energy: A Cultural History of Energy Physics in Victorian Britain. University of Chicago Press, 1998. [Steele, 1997] B. Steele. Muskets, and Pendulums: Benjamin Robins, Leonhard Euler, and the Ballistics Revolution. In Technology and the West, T. Reynolds and S. Cutcliffe, eds., pp. 145-180. University of Chicago Press, 1997. [Stewart, 1992] L. Stewart. The Rise of Public Science: Rhetoric, Technology, and Natural Philosophy in Newtonian Britain, 1660-1750. Cambridge University Press, 1992. [Taylor, 1991] J. Taylor. The “Didascalicon” of Hugh of St. Victor. Columbia University Press, 1991. [Timoshenko, 1953] S. Timoshenko. History of Strength of Materials. McGraw-Hill, 1953. [Todhunter, 1886–1893] I. Todhunter. A History of the Theory of Elasticity and the Strength of Materials from Galileo to the Present, 2 vols. Cambridge University Press, 1886-1893. [Vincenti, 1990] W. Vincenti. What Engineers Know and How They Know It. Johns Hopkins University Press, 1990. [Vincenti and Rosenberg, 1978] W. Vincenti and N. Rosenberg. The Britannia Bridge: The Generation and Diffusion of Technological Knowledge. MIT Press, 1978. [Von Braun and Ordway, 1976] W. Von Braun and F. Ordway. The Rockets’ Red Glare. Anchor Press, 1976. [Watson, 2003] J. Watson. DNA: The Secret of Life. Alfred A. Knopf, 2003. [Weiss, 1982] J. Weiss. The Making of Technological Man: The Social Origins of French Engineering Education. MIT Press, 1982. [White, 1978] L. White, Jr. Medieval Religion and Technology. University of California Press, 1978. [Wise, 1985] G. Wise. Willis R. Whitney and the Origins of U.S. Industrial Research. Columbia University Press, 1985. [Yates, 1979] F. Yates. Giordano Bruno and the Hermetic Tradition. University of Chicago Press, 1979.

COHERENCE AND DIVERSITY IN THE ENGINEERING SCIENCES Gerhard Banse and Armin Grunwald

1

INTRODUCTION

The philosophy of technology has devoted relatively little attention to the engineering sciences. The central focus has been primarily upon the concept of technology, the facets of technical action, the relationship between humans and technology or between technology and society [Ferr´e, 1995; Mitcham, 1994], and the relationship between technology and art or between technology and nature (see for example most contributions in [Lenk and Maring, 2001] and in [Mitcham and Mackey, 1983]). By contrast, the theory of science has concerned itself much more predominantly with the classical scientific disciplines — particularly theoretical physics — than with the engineering sciences, despite the fascinating and even specific relationship demonstrated by the latter between theory and practice. Although technical practices are an inextricable aspect of the cultural history of mankind, the engineering sciences emerged at a relatively late stage as a consequence of the industrial revolution and became institutionally anchored in response to the societal need for technical knowledge to be made systematic, for engineers to be well-educated and for technical knowledge and know-how to continue to expand. The coherence and diversity of the engineering sciences cannot therefore be considered in isolation of this history. It is precisely in this relationship between theory and practice that these sciences manifest themselves. Further clarification of this relationship is a task shared by the philosophy of technology and the theory of science. At first sight, the engineering sciences appear to be a conglomeration of a wide variety of scientific disciplines the only apparent common ground being their affiliation to engineering and technology, to technical systems and the way in which they are developed and applied. The systems themselves do, indeed, represent enormous diversity. Upon closer examination uniform features and unifying elements do, however, crystallise within this diversity. It is the aim of this chapter to make these common features clear. To this end, we shall first define the specific nature of engineering sciences and examine their development. This will enable us to lend substance to the coherence and multiplicity of engineering sciences whilst simultaneously taking their historical development into account. This introduction (Section 2) thus provides Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

156

Gerhard Banse and Armin Grunwald

information on both the perspective from which technological sciences is viewed in this chapter and the options open to us for doing justice to their diversity. Having chosen to view the coherence and diversity of engineering sciences in relation to the specific relationship between theory and practice, it seems only logical to follow through this perspective both methodologically (Section 3) and substantively (Section 4). Not surprisingly, in terms of methodology, we will primarily be forced to deal with the coherence of engineering sciences, for instance, in engineering design, while in terms of substance the focus will be on classification approaches and thus on diversity. In addition to these considerations, we will finally present our “general technology” considerations. That is where we consider at a conceptual level the relationships of coherence and diversity discussed so far (Section 5). 2

COHERENCE AND DIVERSITY — THE SPECIFIC NATURE AND DEVELOPMENT OF THE ENGINEERING SCIENCES

2.1 The specific nature of engineering sciences The collective term “engineering sciences” is used to group together all the scientific disciplines that pertain to technology, mainly in the form of technological systems (“artefacts”), materials, work procedures and technological processes (see also Mitcham and Schatzberg’s chapter in this Volume, Part I). These sciences have to fulfil the following dual function: • First, it is particularly the technical characteristics of existing technical systems that are recorded by observation or measurement and are then analysed empirically and theoretically. These results are presented in terms of both the natural sciences and the engineering sciences and, where possible, are justified mathematically and generalised. This forms the basis for potentially improving technical systems, for example in terms of functionality, effectiveness, user-friendliness, potential applications, reliability and safety, or for better controlling their application. • Secondly, new technical objects and technological procedures (or changes to existing procedures) are anticipated using approaches guided by methodology that is based on theoretical knowledge and inferences as well as existing practical experience. They are evaluated according to external requirements and are drafted (design, engineering design). Two things thus become apparent: (a) technology cannot be reduced to its concrete realizations because it also encompasses the human actions behind technical construction as well as the technological production processes which are based on the use of technical systems and serve to create things of value. (b) The engineering sciences domain does not only incorporate already existing and functioning technical systems but also those not yet created, those which are still to be intellectually anticipated, designed and projected.

Coherence and Diversity in the Engineering Sciences

157

In this context, the engineering sciences deal with specific combinations of possible technical options dictated by the laws of nature and by social objectives, requirements, constraints and needs. These specific combinations are subject to the influence of what is possible in terms of laws of nature, what is technically and technologically realisable, economically feasible, ecologically sensible, socially desirable and enforceable, and humanly acceptable. They can be recognised by the characteristic structures of technical systems and the corresponding external1 functions which, although ultimately enabled by laws of nature, cannot be sufficiently explained within the framework of natural scientific knowledge. Technical characteristics or technical properties have to be characterised both with respect to nature and the field of human activity. “They reflect the objects’ ability to carry out this or that technical function” [Tscheschew and Wolossewitsch, 1980, p. 163]. They express the dialectics of nature and society, of the natural and social components that are characteristic of engineering and technology. These technical characteristics are first and foremost the object of engineering sciences which also set out to investigate the relations between components of the laws of nature (or combinations of these components) and the technical conditions of their realisation or effect. At the forefront of scientific activity in engineering sciences is the systematic anticipation and complex evaluation of the structure and function of technical systems, of the pathways to their realisation and of the procedures underlying their application. It is the process of the intellectual anticipation of the new, that is, of what does not yet exist, or not in a given form. The results include possible solutions to technical tasks and problems, technical ideas, innovations, developments and inventions. Engineering sciences as a whole is equally concerned with cognition and practice. The cognitive goals are to create new knowledge, for instance, in the form of functional and structural rules, technological laws or ecological-sociotechnological systems. The practical goals are to anticipate technology, for example in the form of new or improved technical systems, human-technological interaction or socio-technical structures. This not only defines the practical objective of work in engineering sciences, but it also makes clear that technical systems are created by “thinking, planning and constructing” and that the prerequisite of the “intended purpose is practical effectiveness” [Spur, 1998b, pp. 1ff.; present authors’ emphasis]. By being linked to practical realisation, engineering sciences become what the German physicist Hans-Peter D¨ urr termed “machinations” or “sciences urr, 1988, p. 172, 179].2 In this way more emphasis comes to be of making” [D¨ placed on both the “madeness” of technical systems and the work of humans and thus on human-technical interaction in the production and use of these systems (sociotechnical systems within a cultural “frame”). Engineering sciences are thus 1 As distinct from the “internal” functions which can be explained purely in terms of the natural sciences. 2 In contrast to our own interpretations, D¨ urr saw the creation of technology by engineers and technical scientists as primarily a matter of craftsmanship without any scientific basis.

158

Gerhard Banse and Armin Grunwald

sciences of action. They are concerned with technical artefacts, processes, methods of action and procedures and they are designed to support human action in many ways by means of technology or by providing the knowledge and know-how for technology. The specific scientific concern with technical action serves the systematic investigation of the conditions of successful action and the extension of the possibilities for action. Technology and engineering sciences pursue three kinds of targets on this abstract level (see [K¨ onig, 2006]): 1. Practical goals: Engineering sciences should support and want to support and improve technical practice with regard to the usability, economic viability, efficiency, safety, functionality, and so on, of technical procedures, products or systems. The central criterion of success for engineering sciences is their usefulness to society through the supporting of technical practice and the educating of engineers. With regard to their practical aims, the engineering sciences demonstrate a high degree of social relevance. 2. Cognitive Goals: Technical knowledge, in other words, knowledge of technical contexts and procedures, of material processing and exploitable physical or chemical processes should be produced, stabilised, systematised and improved. The central criterion for success here is the truth of the relevant knowledge in conjunction with the achieving of recognition within and outside the sciences. The central concept is therefore scientific excellence. 3. Professional Goals: The scientific community affiliated to these technical disciplines (e.g. of mechanical or civil engineering) should be consolidated through scientific journals, societies, congresses and scientific meetings, through the stabilisation and expansion of the knowledge system and through further specialisation and differentiation. The duality of practical and cognitive goals as found in the first two kinds of targets is crucial to our further discussion while the third kind is of less relevance to the present study. The duality of practical goals (with all the promises of relevance, usefulness and feasibility) and cognitive goals (with all the requirements of excellence and truth) represents a significant, specific element of engineering sciences which contrasts with the natural sciences. For example, physics — at least as it is traditionally understood — lacks the aspect of following “external” goals of societal practice but focuses rather on “internal” goals. Here practice and research are methodically ordered. The duality of practiceorientation and knowledge-creation, of relevance and excellence, of usefulness and truth, of feasibility and reproducibility is organised according to the primacy of practice. Social practice, for which engineering sciences provide technical knowledge, is the ultimate objective of research in the engineering sciences. The guarantee of technical feasibility under “real” world conditions is thus a crucial objective

Coherence and Diversity in the Engineering Sciences

159

of technical action. In contrast, excellence and the claim to truth do not become ends in themselves in engineering science research. They should rather be viewed as instruments, for instance in the hope that excellent and universal knowledge, by means of being transferred to new contexts, will display signs of greater utility and innovative force. Alternatively it may be supposed that the claim to excellence (which is strengthened, e.g. through evaluations and the increasing pressure of competition) will indirectly raise the innovative competence of engineering sciences. The primacy of practice also corresponds to historical development in the engineering sciences. The requirements of practice often lead to the establishment of new directions of research and new professorships (see, e.g., [K¨onig, 1995] with regard to electrotechnology). The termination of the requirements from outside science leads, in turn, to the termination or reduction of corresponding activities (e.g. in nuclear technology following the decision made in Germany to phase out atomic energy). This primacy of practice even leaves room for innovation that is driven by research, and it by no means implies that it is “demand-pull” alone that dictates the direction of engineering sciences. New ideas for products, processes and systems often derive from engineering practice thus constituting a technology push. Such ideas are, however, also based on assumptions about the social need for them or for some other type of “utility”. In such cases, primacy of practice implies that a demand or need for new technical solutions is initially accepted or assumed to be valid for reasons that may or may not be good. This occasionally leads to a dead end if social requirements are misinterpreted (such as in the case of the Transrapid monorail in the German transportation network), but it sometimes also results in socially and economically successful lines of technical development (such as in the case of MP3 players). The primacy of practice has certain implications. With regard to social practice, the concrete facet is essential to engineering sciences, as technical applications always take place in very concrete contexts: spatially, temporally and with regard to groups of people. Of course, technical knowledge must satisfy the criteria of universalisation and abstraction and be transferable to new contexts, but as a rule such transferability encounters practical and concrete limits. One general aspect of engineering sciences is, therefore, to reflect on the limits of technical knowledge transferability where concrete solutions are concerned. In the engineering sciences even theoretical work of this kind must ensure that there is always a path from theoretical, that is, abstract knowledge back to concrete practice. The coherence of engineering sciences thus lies in the specific relationship between theory and practice. The objective of engineering sciences is ultimately practical involving the social utilisation of technology and thus concrete realisations. Abstraction and theory are not ends in themselves but function rather as means, for instance, to systematise knowledge and facilitate the transferability of knowledge to other contexts. The natural sciences, by contrast, abstract their knowledge in the form of predicates that are kept as general as possible and

160

Gerhard Banse and Armin Grunwald

make use of concretisations (i.e. experiments) only in an instrumental capacity, to test hypotheses, for example.3 Whereas knowledge from the natural sciences thus moves out of the “real world” and into more or less abstract model worlds, the engineering sciences aim directly at practical goals in the real world. This structural coherence within the engineering sciences accommodates two phenomena. One pertains to the diversity of the engineering sciences themselves, ranging from technical mechanics and materials sciences to civil engineering, mechanical engineering and chemical process engineering, and from microelectronics and nanotechnology to technical informatics and all the various biotechnologies. The other phenomenon lies in the great diversity of the objects of engineering sciences. They include micromechanical objects and simple machine components as well as complex chemical syntheses and even worldwide information and communication networks, all of which have to cover an enormous range of application conditions (e.g. with regard to temperature, pressure, speed or radiation). The specific nature of engineering sciences can be summarised according to the following three characteristics: 1. Engineering sciences are target-oriented sciences. A central concept is the application of scientific knowledge and practical experience to satisfy the “technical needs” of society. 2. Engineering sciences have a constructive character. At the forefront are the intellectual anticipation and evaluation of the structure, function and realisation of new technical systems. 3. Engineering sciences are integrating sciences. Since technical systems represent a specific coherence of natural and social components, knowledge is incorporated from the natural sciences, from other technical sciences and from the social sciences.

2.2 The development of the engineering sciences Technical action (in terms of production and/or the use of technical systems) is as old as humanity itself. Engineering sciences, by contrast, arose — via various early forms — much later (see also Channell’s chapter in this Volume, Part I). In the eighteenth and nineteenth centuries, many mathematicians and natural scientists endeavoured to apply mechanical principles to technical problems. This had varying degrees of success, leading to positive results in simple technical cases and failing predominantly in the case of more complex technical problems which taxed the resources of the time when it came to finding solutions on the basis of theory and calculation. However, this was just the beginning of a cross-fertilisation between natural scientific and technical knowledge. 3 Philosophers of experiment, however, have criticised this theory-dominant view of experimentation, putting emphasis on the value of the experimental practices in themselves (see, e.g., [Hacking, 1983; Radder, 2003]).

Coherence and Diversity in the Engineering Sciences

161

Thanks less to technical questions than to organisational and practical administrative matters in the form of the guidance, management and state “supervision” of manufacture and trade, eighteenth century Germany saw technology develop as a science. It was particularly Johann Beckmann (1739–1811), from the University of G¨ ottingen, who systematised the technical knowledge underlying the various work being done in craft areas and factories to make it all more teachable. He did that especially in his book “Anleitung zur Technologie . . . ” (“Instructions on Technology . . . ”), written in 1777, in which the equality or similarity of (technical) procedures was perceived as the basis of organisation [Beckmann, 1780]. He was concerned with the connection between “true basic principles” and “reliable experience”. By “true basic principles” he meant the basic scientific principles for the theoretical elucidation of technological processes, while “reliable experience” was what he saw as the stable element, as what is gained from living in the world, from technical, formal know-how based on handed down experience. In that way, Beckmann strove to rationally explain working procedures. Beckmann’s approach received much acclaim, but it soon became clear that it was not possible to develop a technological science at that time. Nevertheless many attempts were made to apply this concept. To name but two such attempts there was first Karl Karmarsch (1803-1879) “Grundriß der mechanischen Technologie” (“Outline of Mechanical Technology”, 1837/41; later editions were entitled “Handbuch der mechanischen Technologie” (“Textbook of Mechanical Technology”) and then there was Friedrich L. Knapp (1814-1904) and his “Lehrbuch der chemischen Technologie” (“Textbook of Chemical Technology”, 1847). A detailed overview of further approaches to and institutions of technology can be found elsewhere, for example in Spur [Spur, 1998a]. Beckmann’s approach was not taken up again until the mid-twentieth century (for further details, see Section 5). Engineering sciences, as institutionalised sciences, evolved then in the eighteenth century but particularly in the nineteenth century which saw the rise of industrial production, the move from hand to machine operation and the step from manufacture to “big industry” (Marx), which accompanied the invention and further dissemination of labour machines (“Arbeitsmaschinen”), the use of steam power and the tempestuous development of the natural sciences. This process is still ongoing today: new disciplines in the engineering sciences are still emerging (like, for instance, nanotechnology). The accumulation of the technical experience of many generations made it possible to systematise and generalise our empirical knowledge of technical systems, processes and methods for the first time. Knowledge was based on how to organise and manage industrial production on a large scale and on how to exploit natural forces and employ technical systems (more) effectively. Inevitably the developing technical basis in advanced countries (Great Britain, France, USA, Germany) required the conscious use of science. In those days, the laws of mechanics and thermodynamics were therefore applied to machines, and mathematical theorems were used to successfully solve static construction problems and to control processes. In that way constructive-technical knowledge was created. In particular, the development of widely applicable machines (initially

162

Gerhard Banse and Armin Grunwald

the steam engine, later the electromotor), a branching system of transfer mechanisms and the tool machine were accompanied by the increasing use of scientific knowledge and methods. Changes in the means of work (steered, for instance, by the various available energy forms), in the basic materials used (metals, non-metals, natural products, “synthetics”, composite materials) and in the processing conditions to be technically controlled (such as in micro and macro areas, temperatures, pressures, precision, handling and processing times) have led and still do lead to continual further developments in the engineering sciences. With the emancipation and consolidation of the engineering sciences came institutionalisation in the form of polytechnics and technical colleges (including the right to obtain a PhD). The same era saw the rise of engineers as a professional group with its own organisational forms and societies, including technical and “polytechnical” journals. This “scientification of technology” unfolded very differently in each country (compare, for instance, France, Germany, Great Britain and the USA) in the areas of scientific organisation and practice (for instance with regard to links between theory and practice, the significance of the natural sciences and mathematics as a conceptual basis and integration into existing universities or colleges), all of which is too much to be presented in any detail here (see, e.g., [Buchheim and Sonnemann, 1990]).

2.3 Debates on the specific nature of the engineering sciences From the very early days, the specific nature of engineering sciences and its status within the system of sciences has remained a subject of discussion, above all else with regard to its relationship to the natural sciences. From the point of view of their substantive nature, intensity and consequences these theoretical reflections have followed very different paths. One might, for example, cite the methodological controversies of the second half of the nineteenth century and the debate on technology as an applied science aired in Technology & Culture in the 1960s. In some form or another, the subject always revolved around people’s image of the engineering sciences, the actors involved (engineers, technical scientists), and the relevant teaching and research institutions (polytechnics, technical colleges, and technical universities). In the first half of the nineteenth century, Ferdinand Jakob Redtenbacher (1809– 1863) attempted to provide students of mechanical engineering with the practical tools they needed to learn about construction. His goal was to create his own approach in order to transcend both the English empirical orientation and the French school of engineering training approach with its method that was strongly rooted in mathematical theory. His aim was to unite people’s knowledge of physicalmechanical nature with their experience, capabilities and skills, none of which could be entirely gained through theoretical education. He appealed to engineers to link scientific knowledge to technical ability, “calculation” and “feeling” [Schneider, 1987, p. 174], and cognition (knowledge) to creation (making). However, he

Coherence and Diversity in the Engineering Sciences

163

did not consider their relations (cognitive, methodological and so on) to each other, or their — indisputably historically variable — relative importance to, for example, engineering training courses of the day. He thus provided a starting point for opposing lines of discussion on systematic ways of finding solutions to technical problems. One strand of thought emphasised the machine-theoretically oriented direction whilst the other dwelt on the creative character of construction engineering. Ultimately it was the development of science, technology and industry that was to provide the solution to this methodological controversy that had been so vehement. Both approaches proved to be unconditionally necessary, complementary and elucidatory. This led to a synthesis within engineering training; to the combining of theoretical lectures, drawing board draughtsmanship, laboratory experiments and experimentation with life-size machines under real conditions. In the 1960s discussion launched in Technology & Culture that was initiated by the article “Toward a Philosophy of Technology” [Technology, 1966] the emphasis was somewhat different. There the primary focus was on a philosophy of technology or discussion in terms of scientific theory and it was particularly conducted by non-experts in the engineering sciences. Proceeding from the contributions made to a symposium, it was also concerned with the specific nature of engineering sciences in terms of cognition and method, the relationships to the (natural) sciences (as independent sciences or as “applied natural science”) and the concept of “technology” (e.g. as knowledge or know-how). The opinions emphasise “technology as knowledge” on the one hand and “technology as activity” on the other hand (for more details, see [Mitcham, 1994, pp. 192ff.]). Subsequent considerations pertaining to the philosophy of scientific and technical theory illustrated the fact that these two sides are related to each other and are inseparable (e.g. as units of knowledge and as design — see [Banse et al., 2006]). Since the end of the 1960s, the engineering sciences and their specific nature have once again been at the centre of intellectual debate, particularly regarding the philosophy of technology and scientific theory, but new interest also arose in the sociology of technology, in economics and in the history of technology. At the moment, from the perspective of a philosophy of technology alone — and that is the focus here — the current themes embrace the relationships between the natural sciences and the engineering sciences, forms of knowledge in engineering sciences and the establishment of a general technology (see, e.g., [Lenk and Maring, 2001]).

2.4

Perspectives on coherence and diversity

The relationship between the engineering sciences and practice is the point of departure when considering the above-mentioned coherence of engineering sciences (see Section 2.1). At this level, mechanical engineering, process engineering, civil engineering and production engineering, to give but a few examples, do indeed display a common structure. However, even at first sight, clear differences can be recognised between the disciplines of the engineering sciences with regard to the relevant subject areas, materials and tasks. When it comes to the coherence

164

Gerhard Banse and Armin Grunwald

and diversity of the engineering sciences, one must thus seek further levels of differentiation in order to pursue assumptions about coherence and, in a more concrete sense, detect the facets of diversity in the engineering sciences. This can be done in various ways: • Methods in the engineering sciences presented as approaches and procedures in order to gain a knowledge of the subject areas and the technical problem solving methods (practice and design, cf. Section 3): the examples that can be cited here include methods of analysis, design and construction, heuristic models, experimental methods, and modelling and simulation. The activities of the engineers in the various fields are also considered, such as in research, development, production, distribution, marketing, maintenance and quality control. • Substance in the engineering sciences: this can be carried out according to the classical tripartite division into matter, energy and information. Differentiation may be made according to areas of application (e.g. energy provision, material research, space technology or robotics), whereby in each case the conditions valid for different contexts are considered. Alternatively the basis could be to adopt approaches to systematising or abstracting technical knowledge (cf. Section 4 for such divisions according to content). The substance issue also comprises a look at the products of engineering sciences in the form of the available technical knowledge and know-how as well as its materialisation in products, processes and systems. • Generalities in the different methodical or substantial directions which can be interpreted within the framework of a general technology (cf. Section 5). The identification of aspects of coherence or diversity between the sub-disciplines of the engineering sciences has its implications. The topics and questions that can generally be attributed to engineering sciences can be addressed both at research and at theoretical level and in an interdisciplinary manner. Courses offered at universities, for instance, for engineers in training can be conceived in such a way that they cross faculty and disciplinary borders. However, where diversity is dominant, these activities will always be the responsibility of sub-disciplines such as process engineering, construction engineering or materials science. 3 COHERENCE AND DIVERSITY IN TERMS OF METHOD The central theme of this section is the fact that the engineering sciences display strong common features in terms of method and these common features correlate with the specific theory-practice relationship mentioned above. The diversity which is indeed present in terms of method subordinates itself structurally to coherence in the characterisation of technical design as a problem-solving strategy with the use of technical knowledge (see, e.g., [Hubka and Eder, 1996]).

Coherence and Diversity in the Engineering Sciences

3.1

165

The methodology of the engineering sciences

Methods are target-orientated procedures aimed at the intersubjectivisation (objectivisation) of information and knowledge. Using methods, the production and examination of knowledge can be conducted systematically and in such a fashion that it becomes learnable as well as teachable. Methods are also the central element of scientific quality assurance, of education and of scientific progress. Engineering sciences have developed a plethora of methods which are in part intrinsic and have, in part, been taken over or adapted from other disciplines. These can be classified as follows — albeit with much overlap — (see [Banse et al., 2006]): • Methods of design: If technological design is thought to solve problems by means of technical knowledge and know-how it is possible, depending on the phase division of the processes of technological design chosen in each case, to distinguish between various classes of methods. Intuitive-heuristic methods proceed from the plurality of available knowledge forms and attempt to deploy creativity techniques to find new solutions. By contrast, rationalsystematic methods start from an analysis of the system under consideration in terms of problem-solving and infer appropriate solutions top-down. In both directions, evaluations and selections from the possible options must ultimately be made (both according to the criteria of technical functionality and criteria external to technology such as economic viability, customer acceptance or ecological compatibility). For this reason, evaluation and selection methods are also included in the spectrum of design-orientated methods included in the engineering sciences (for more details, see Section 4). • Methods of research: The provision and examination of the technical knowledge necessary for the technical design goals aimed at societal practice (see Section 3.2) require their own methods. Here too, heuristic models play a role and serve above all else to structure the further search for knowledge. Of special significance is the distinction — familiar from the classical natural sciences — between theoretical-deductive and empirical-inductive procedures. The theoretical-deductive procedures chiefly emphasise the mathematisation of functional connections and the use of mathematical processes. Empiricalinductive procedures, by contrast, are laboratory orientated. They revolve around appropriate measuring techniques and experimental procedures such as materials processing or process regulation. • Methods of implementation: The technological designing of products, processes or systems is something that must be implemented by engineering as well. Implementation methods comprise various approaches to creating prototypes, to demonstrating technical feasibility and finally to creating the physical hardware that can provide the required capabilities. In all areas, the introduction of computer-aided methods has led to many changes. To a certain extent mathematical modelling and simulation have now replaced

166

Gerhard Banse and Armin Grunwald

laboratory experimentation which, in turn, often permits considerable cost-saving (see for more details Part IV of this Handbook). At the same time, however, new challenges arise when it comes to validating model-simulated knowledge in the real world.

3.2

The methodological characterisation of technical knowledge

The products of the engineering sciences are inherent in the systematised and quality-assured knowledge of technical processes and systems and they are, to some extent, anchored in technical artefacts. This knowledge, which has been gathered, extended and refined over the course of decades and centuries — and which in each case includes knowledge of the necessary know-how — constitutes the starting point for technical design plans. Technical knowledge is knowledge for action. If such knowledge is to be teachable and learnable it must be or become explicit. In technical practice, however, implicit knowledge (tacit knowledge) plays an important role. 3.2.1

Technical rules as a form of technical knowledge

The origin of the engineering sciences in the nineteenth century can be traced back to an increasing demand for explicit, systematic technical knowledge, for the systematic production of new technical knowledge and for well-educated engineers. If technical knowledge is to be acknowledged as true knowledge it has to satisfy certain demands of invariance in different situations and transcend the subjective. This implies that this type of knowledge can be taught and learned or, in other words, systematically passed on in education. Together with other systematisations that are thus made possible, for example in the form of text books for sub-disciplines within the engineering sciences, a process of systematic accumulation of technical knowledge is therefore initiated. The basis of this process, which has proved exceptionally successful in the last 200 years, is the validity of technical knowledge. This validity extends beyond individual persons and historically coincidental constellations and can be understood methodologically since technical knowledge is characterised as a set of technical rules. These technical rules provide statements about means–ends relationships. A technical rule describes knowledge about how something can be caused. It describes the relationship between something that is to be achieved and the means required to achieve it (see [Bunge, 1974; 1983; Kornwachs, 2001]). This obviously represents a cause-and-effect relationship which is constructive in nature — in contrast to the natural sciences where the nature of things tends to be explanatory. In other words, it aims to achieve something, and is used in a practical context. Technical rules are valid with regard to the observance of certain conditions of applicability, such as those made by the requirements of a technical process on temperature, pressure or humidity. The justification of technical rules thus demands the designation of the corresponding attribute of the area of applicability.

Coherence and Diversity in the Engineering Sciences

167

The character of technical rules consists precisely in the fact that an area of applicability can be given, within which the technical rule is valid, independent of individuals and situations (as described in greater detail in [Grunwald, 2002a]). It is precisely this fact that satisfies the requirements of being teachable and learnable: first, the rule itself is taught as a means-end relationship and then its area of applicability is taught. The totality of such explicit technical knowledge can be understood in this respect as a set of technical rules. The reproducibility of technical rules, regardless of the situation, is the precondition for being able to teach them and learn them in practice. It must be possible to prove and validate technical rules over and over again, and thus to recreate the initial situation, at least with regard to the relevant issues laid down by the rules’ area of applicability. This makes it possible to confirm any rule experimentally by repeating tests, like in a laboratory. To a high degree the reliability of technology is based on the reproducibility of technical rules that emanate from the regularity of technical processes, like in the engine of a car. Technology no longer functions if the conditions of validity are no longer satisfied, such as if diesel is put into a petrol engine or a non-waterproof watch is immersed in water. Technical knowledge does not consist solely of technical rules that can be explicitly described. A foreman knows how to handle “his” machine so that it functions properly although this knowledge may not satisfy the clear, logical criteria for technical rules (see [Vincenti, 1990]). Implicit knowledge is a part of technical practice and it also plays a role in the engineering sciences (see [Baumard, 1999; Collins, 1974; Polanyi, 1966]). Implicit knowledge is, however, personal knowledge [Polanyi, 1962] and, as such, can only be taught or learned to a very limited extent. Thus, the task of all the engineering sciences includes making the implicit explicit, because only then does technical knowledge become teachable, systematic and open to direct improvement. In that way the engineering sciences may also be said to be involved in demystifying implicit and personal knowledge and transforming tacit knowledge as far as possible into teachable knowledge about technical actions. The way in which this knowledge is teachable becomes the formulation of this knowledge as a set of technical rules. This feature is common to all disciplines in engineering sciences.

3.2.2

The properties of technical rules

Although technical rules do not represent universal laws of nature, with regard to generalisability, they do cover a wide spectrum ranging from a high degree of precision to rather heuristic rules of experience. The crucial factor in evaluating the degree of generalisability is knowledge of the area of a rule’s validity both with regard to the direct area of validity and the quality and reliability of such knowledge. The generalisability of technical rules is often the object of research in itself in that the areas of validity are, for instance, systematically tested in laboratory experiments.

168

Gerhard Banse and Armin Grunwald

The characterisation of knowledge in engineering sciences given here is abstract. At this level of abstraction, technical rules are characteristic of the coherence of engineering sciences. The diversity of the various technological sciences can be seen when their concrete forms of knowledge are considered more closely, depending on the nature of the tasks, the type of objects, the materials used, etc. Beyond the level at which they can be formulated as technical rules, classifications can be introduced that are content-motivated and empirically strong (for more details, see Section 4). A relationship can be described between technical rules and the fundamentals of technical progress since such rules show a specific ambivalence. On the one hand, they express knowledge of what is technically feasible. For situations inside the rule’s area of applicability they provide knowledge that can be used for good reason in technical applications or handed down in teaching. On the other hand, however, they also contain knowledge of what is not yet feasible: the assumption that a rule will not function outside the area of a rule’s applicability is well-founded. Thus technical rules display a facet that points to certain knowledge and another facet in which technical knowledge is obviously insufficient. The latter also functions as an impetus to transform the boundaries of knowledge and know-how through research that extends the areas of a rule’s applicability. Technical rules do not generally permit specific measures to be identified as essential to the achieving of desired effects as there are frequently different technical rules that will lead to the same effect (e.g., there is often a choice between various materials when components are first conceived). Technical design thus precedes the need to choose between various options. This opens up the market to optimisation according to specific criteria (minimisation of development time, optimisation of the cost-benefit balance etc.). Technical rules are options for achieving a certain effect but usually there is no requirement that a particular technical rule must be abided by to achieve a particular end. 3.2.3

The origin and validation of technical rules

The production of new technical knowledge or know-how is partly deliberate. One might think, for instance, of systematically varying parameters or material properties in a laboratory so that the scope of a given rule’s applicability can be experimentally determined. Such good experimental practice procedures are a part of the methodological inventory of engineering sciences (see [Banse et al., 2006]). The sources of new ideas for technical solutions (typically in relation to inventions), include trying something out in a tentative and creative fashion. One might term this an intuitively driven heuristic procedure based on many years of experience and even rather random and spontaneous impulses (see [Banse and M¨ uller, 2001; Ferguson, 1992]). The “tacit” knowledge (see Nightingale’s chapter in this Volume, Part II) and intuition of experienced engineers that reflects rich experience but is hard to explain plays an important part in the practical progress of engineering sciences, see [Vincenti, 1990], namely with regard to the process according to which technical rules are created.

Coherence and Diversity in the Engineering Sciences

169

Such knowledge that is intuitive or gained by unconventional means must then be validated according to the established procedures and scientific testing criteria. Only then can the knowledge and know-how gained be intuitively regarded as certified and reliable according to the standards of engineering sciences (and, for example, used or taught for the purposes of technical application). One task of the engineering sciences is thus to make intuitively gained knowledge and know-how systematic to ultimately formulate it as technical rules, see [Grunwald, 2002a]. Technical rules arise during the systematic process of making repeatable an action that has succeeded once but was perhaps intuitively conducted (e.g. the realisation of a technical process in a laboratory) and then going on to examine the conditions under which it will remain successful. In contrast to the practice of craftsmen, the scientific character of the engineering sciences is based on the systematically tested characteristic rules of technical activity. The methodological coherence of the engineering sciences includes the role of tacit knowledge and extends to efforts to achieve systematisation and explanation in terms of “objective” technical rules. 3.2.4

Methodological problems

The methodological coherence of the engineering sciences is also evident from the fact that they share certain fundamental problems linked to the production, validation and limitations of technical rules. This particularly includes the following: (a) Temporal inversion of the direction of inference: Technical rules are formulated retrodictively: they describe relations between effects that are to be realised and actions which are regarded as appropriate in the realising of such effects. What is to be achieved lies in the future and that is how inferences are drawn as to what can be done to get there, see [Bunge, 1974]. The corresponding technical knowledge, however, is produced and substantiated in the other time direction. For instance, during an experimental set-up in a laboratory, tests are conducted and the cause/effect-relations which are to be certified in the course of the experiment relate the causes to later effects. Temporal inversions of the direction of inference from cause/effect-relations to effect/action-relations are not, however, trivial either logically or pragmatically, see [Kornwachs, 1995]. If cause C reproducibly causes effect E and a request for achieving effect E is received, knowledge of the cause/effect relation between C and E will not imply that C should be done in order to achieve E for several reasons. One reason is that C may be a cause which does not result from human action. Another reason is that even if C were a human action, there might be other actions that would be more appropriate for achieving E in the specific context, see [Grunwald, 2002b]. (b) The difference between the tested area of applicability and the area of practical application (see Radder’s chapter in this Volume, Part I): The areas of applicability of technical rules are usually determined inductively, by means of trial and error. Certain parameters (e.g., temperature, pressure) are changed

170

Gerhard Banse and Armin Grunwald

repeatedly until the rule no longer works. In technical practice, however, technical rules are often implemented in contexts that differ from those in which the rule was originally tested for correctness. Facilitated especially by the possibilities offered by computer simulation, it has thus become common practice in many areas of technical planning to use a simulation in order to test a technical rule and its area of applicability. A rule of such a kind that has been substantiated under artificial conditions is then realised in practice but under real manufacturing or implementation conditions. Laboratory experiments and simulations reduce the real conditions to the modelled form. The removal of these reductions when transferring the technical rule from the laboratory or simulator to practice thus represents a non-trivial operation. In reality, the “real-case” situation is thus often also the testing situation. This leads to questions regarding the scope of (computer) models, their adequacy in practice with regard to their ends and the transferability and validity of simulation results. (c) The need for pragmatic organisation: It is not just a question of using the right technical rules but also of ensuring their expedient conjunction to a sensible whole. Pragmatic rules regulate the temporal or logical organisation of solutions to complex technical problems on the basis of detailed technical rules (this is particularly relevant in the field of civil engineering; one just needs to think of questions of logistics on a major construction site). As a rule, disregarding the pragmatic organisation usually leads to failure to achieve the desired solution. (d) Dealing with lack of knowledge: Since many if not all problems in the “real world” can be regarded as “wicked problems” [Simon, 1973], the attempt to solve them by technical means is inherently fraught with uncertainty (see [Banse, 2003; Rittel and Webber, 1973]). Strategies for dealing with these uncertainties and for developing “robust” technical solutions form part of the spectrum of activities of all engineering sciences.

3.3 Methodological coherence in technical design In the engineering sciences, design has to do with solving problems with the help of technical products, systems or procedures (see also the chapters on design in Part III of this Handbook). This can only be achieved with the dedicated implementation of knowledge and know-how drawn from the engineering sciences. Technical constructing and drafting form the procedure in which knowledge and design come together. Drafting is thus the methodological location at which the specific relationship of theory and practice in the engineering sciences is most clearly apparent (see, e.g., [Cross, 1989; Gregory, 1966; Hubka et al., 1988; Hubka and Eder, 1996]). Most general descriptions of how technical solutions come about distinguish between the design process, which results in a executable draft of the solution and the execution itself, see [Hubka and Eder, 1996]. In this respect the concept “design

Coherence and Diversity in the Engineering Sciences

171

process” designates the operationalisation — in terms of the engineering sciences — of the drafting or planning of technical solutions. In the following section we describe the procedural characteristics of the design process and the demands placed on results. The methodological coherence of technical planning is evident from these steps. One way of going about this is by establishing the relationship between the requirements for solving specific problems and the available technical knowledge (see above). One might call this: selecting an appropriate combination from the host of available technical rules and presenting that as a solution (in the process, new technical knowledge may also have to be developed). Various ways of structuring the design process have been developed. Although their terminology may differ, in terms of content they have much in common. They emphasise different answers to the question of whether the design process can be characterised as a linear process or whether it primarily consists of feedback loops. The current trend favours feedback loops and iterations because of the opportunities they provide for sequential phases of learning. This is even true of graphic representations (see Fig. 1), in which a linear presentation is frequently preferred for didactic reasons (see [Banse and Friedrich, 2000; Hubka and Eder, 1996]). The design process accordingly consists of a series of steps which are repeated with an increasing degree of detail and greater approximation to technical action (see, e.g., [Eder, 2000, p. 217; Ropohl, 1999, pp. 258ff.]). Problem analysis and planning involves formulating the requirements placed on the technical task in such a way that their fulfillment leads to a solution. This then leads to the formulation of a consistent specification. After the target system has been sufficiently clarified, a lack of knowledge is often encountered in the conception phase with regard to feasibility, either intrinsically or in relation to the given boundaries (e.g. the cost framework). In planning practice it is feasibility or pilot studies that are often used to clarify this problem; a rough preliminary evaluation of a project’s risk or the structure of a possible solution concept is made when there is insufficient information to make a precise evaluation. The main task during the conception phase is to break down, in system-analytic terms, the desired overall function of the technical solution into sub-functions; it then becomes possible to ascertain solution principles and the relevant basic structures. If possible, this should be based on the available scientific and technical knowledge and know-how; otherwise, conception leads to the identification of gaps in knowledge and know-how and thus, in turn, to a need for technical research. The options for solving the sub-functions are integrated into one overall concept or several variants suitable for solving the overall function (see [Ropohl, 1999]). In the draft phase (when a technical system is being developed), this concept forms the basis for the creation of a scale draft (invariably a digital model). Such models are evaluated according to various technical criteria such as feasibility, security and function fulfilment in conjunction with the realisation of necessary preconditions, material, and energy and data flow. At this juncture it is usually already possible to consider criteria from outside the technical field (cf. Section 3.4). At the very least such criteria will include economic aspects and perhaps also

172

Gerhard Banse and Armin Grunwald

Problem analysis Planning

Task selection

Determine development task

Conception Clarify and specify problem definition

Divide overall function into subparts

Ascertain solution principles and structures to fulfil subparts

Combine solution principles and structures to fulfil overall function

Elaborate conceptual variants for principle combination

Draft Create a scaled draft

Technical and economic evaluation of the draft

Create improved draft and fix corrected draft

Elaborate Design and optimisation of individual components

Elaborate explanatory documentation

Check costs

Realisation

Figure 1. Procedure in the design process. Source: modified after [VDI, 1977]

Coherence and Diversity in the Engineering Sciences

173

further social, legal or ethical aspects. It is on this basis that an improved draft or model is drawn up. For temporal sequences the time structures are elaborated and in each case milestones with verifiable requirements are included. These milestones constitute the starting point for projections and project management. The elaboration phase covers further concretisation and detailing while optimising the individual components of the draft and its composition. The systematic connections between the necessary resources are determined and finally the documentation for carrying out the work is produced (e.g. design instructions for a technical system). Rough cost estimates are refined and are made as realistic as possible. The results of any design process are the elaborated drafts of a technical solution. The degree of elaboration can differ depending on the knowledge available and the adequacy and depth of the planning. When drafting a technical solution, one must distinguish between an internal technical side and an external side, each of which is treated differently in terms of method and processed by different disciplines or groups in the further implementation of the solution: • The internal side of the technical drafts serves to guarantee the required technical functionality. What characterises this facet is the requirements placed on materials by energy considerations and by the guidance of data and information flows, security and risk considerations and, subsequently, the provisions made for production, utilisation and disposal. The guaranteeing of these aspects is a task of quality assurance in the strictest sense. • The external side concerns aspects of the technical solution which influence the social context (cf. also Section 3d2.4). These include costs, the resources necessary for research and development, the duration of realisation, the possible inclusion of users, the operational knowledge needed by future users, and the prevention of obvious misuse and predictable abuse. The task of quality assurance in its broadest sense is to oversee these external interfaces to technical models, since success is as significantly dependent on them as it is on the realisation of internal functionalities (see also Radder’s chapter in this Volume, Part V). The significance of the external side of technical solution proposals has increased considerably over the past decades. One can cite examples of technical solutions ultimately becoming a social failure though initially they might have seemed to be exceedingly appropriate in terms of standards of internal rationality (examples are fast-breeder technology in the nuclear energy field and the magnetic high-speed train Transrapid). Once the design process has led to results and a decision has been made about a technical solution, the realisation phase can begin. This is not actually a task of engineering sciences but rather of technical practice. It does, however, remain the task of engineering sciences to support this practice by further developing existing procedures (e.g. conceptions, drafts or quality management), by taking up and

174

Gerhard Banse and Armin Grunwald

managing new challenges (e.g. support for these processes through new computeraided or internet-aided modelling and simulation procedures), and by backing up and evaluating current practice. The process outlined here for creating technical solutions through design and planning on the basis of technical knowledge perfectly illustrates the coherence found in the engineering sciences which emanates from a specific relationship between theory and practice. Insofar as research in the engineering sciences should support technical practice and the level at which technical details should be included, it is all a question of introducing systematisations, improvements and extensions. At the level of overall technical solution processes there is also the matter of considering systems and improving efficiency at procedural level. Engineering sciences operate on various levels. The upper level of drafting and design illustrated here demonstrates a coherent structure, while at the technical level increasing diversity (e.g. in the form of various tasks, application areas and subject areas) is visible and differentiation is necessary (cf. Section 4).

3.4 Technology and engineering sciences within the context of society The successful use of technology is subject to numerous conditions linked to different goals, subject areas and methods in the respective fields (e.g. in automobile technology, power plant technology, refrigeration technology, computer technology or construction). The success of technical endeavours depends both on the satisfactory overcoming of technical challenges and on non-technical factors such as acceptance, the relative advantages over other potential options for solving a problem and user wishes. Examples in Germany such as the breeder reactor in Kalkar, the Wankel motor and the Transrapid train show that technically successful developments by no means automatically become successful applications. Non-technical factors such as success in the market, customer acceptance, and “striking the right note” with regard to lifestyle expectations (such as in the cell phones market) are important factors affecting the success of technical developments. Technical quality is often a necessary but by no means sufficient precondition for social and, particularly, economic success. Technical innovations must also fit into the context of social practice. Since the engineering sciences not only claim to satisfy inner scientific knowledge criteria but also to be socially relevant, the non-technical factors given above as relevant to technology must also constitute an object of technological reflection. The specific relationship between theory and practice in the engineering sciences entails concern not only with technical matters but also with the social context. This places methodological requirements on methods of selection and evaluation whilst also underlining the specific responsibility of the engineering sciences. This is a necessity that is shared by all the disciplines in the engineering sciences [Schaub et al., 1983] and is supported by technology assessment and related activities (see Grunwald’s chapter in this Volume, part V). The need to include non-technical factors finds methodological expression in

Coherence and Diversity in the Engineering Sciences

175

various of the evaluation, selection and decision-making procedures in which both technical criteria and non-technical aspects play a role. Selection and decisionmaking necessities pervade the entire technical problem-solving process and the available room for manoeuvre is utilised to achieve the optimal solution in each context. The selection and decision-making necessities come into play early on with the formulation of the specifications and the first steps towards drafting and the drawing up of concept papers and extend to the phases of elaboration, development and production. They take place at very different levels and range from decisions on details in the laboratory to far-reaching strategic decisions on the general direction to be taken by projects. When preparing a decision for the carrying out of a proposal the drafts, concept papers and plans for solving the technical problems must examine and evaluate quality aspects such as the reasonably assumable technical feasibility, the likelihood of achieving the goal, the possible side effects, costs, efficiency and acceptance. This is often done in a comparative way which means that competing drafts are considered. The following methodological requirements concern the engineering sciences in general: • The Link with Practice: There must be links with the social problems for which engineering sciences aim to provide solutions if their relevance is to be ensured. Even if the majority of the engineering sciences are more oriented to technology push, i.e. to supply rather than to the problem or demand facet, they still have to keep an eye on social responsibilities. • The Establishing of Priorities: From the large number of possible research directions and topics priorities have to be set and choices have to be made, for instance according to the urgency of the problems in hand or the extent of the predicted research. Internal scientific criteria also play a role. One consequence of the demand for technical products, procedures and systems to be economical and competitive is that methods of calculating prospective efficiency attract special attention. • Anticipating User Behaviour : The intended success of the technical scientific results in later societal practice also involves considering social questions such as expected or possible user behaviour and, for example, preventing the obvious misuse or intentional abuse of technical products. • Considering Adverse Consequences: Ultimately in the engineering sciences responsibility also includes considering the possible unintended side effects of developments and participating in technology assessment projects, e.g. in relation to competence and evaluation potential (see, e.g., [Grunwald, 2002b; Rader, 2002; Rip et al., 1995]; see also Grunwald’s chapter in this Volume, Part V). Thus, in the context of a wider society, the methodological coherence of the engineering sciences resides in the fact that — as a result of the specific relationship between theory and practice — they represent coherence in terms of social responsibility which is something that can be realised by means of methods and

176

Gerhard Banse and Armin Grunwald

procedures. Although such responsibility assumes different forms in different contexts, the engineering sciences share the obligation to cautiously observe social practice and to reflect on the way in which their own particular work relates to that practice. When dealing appropriately with this kind of responsibility it is important to critically analyse power relations, to question the predominant technology push approach and to ethically reflect on engineering (see [Pater and Van Gils, 2003; Schaub et al., 1983]; on the special relevance of ethics in engineering design, see [Van de Poel, 2001; Van Gorp, 2005]). In the guidelines laid down by the German Union of Engineers [VDI, 1991] the eight core areas of values identified as criteria for selection and decision-making processes with regard to technical options are these: functionality, safety, economy, prosperity, health, environmental quality, personality development and societal quality. These very general criteria must be made more concrete and operational before they can be implemented to solve technical problems. They must be established in relation to parameters that are concretely measurable or can be evaluated during decision-making processes such as costs, utility, likelihood of failure, use of natural resources, risk of accidents, emissions, time consumption and the demands placed on human time. The solutions to technical problems are usually presented with an abundance of goals that have to be achieved simultaneously: they are supposed to function well, be safe, efficient, economical and environmentally friendly, to name but a few. This complex collection of criteria provides a good idea of the heterogeneous demands placed on finding “good” solutions to technical problems but also of the fact that their quality only partly depends on satisfying technical criteria as it also depends, in part, on the realisation of non-technical expectations. These expectation profiles reveal the complexity of the problems of selection and decision-making and thus of the social factors that the engineering sciences have to take into account and methodologically resolve. Particularly in the technical field, but even more so regarding non-technical criteria, there are often competitive relationships and conflicts between the goals behind different demands. For example, a car body should be as light as possible to reduce emissions but at the same time it also has to satisfy safety requirements in collisions. Similarly, bridges are supposed to be economical whilst also being as safe as possible (see, e.g., [Van Gorp, 2005] for some examples]. Thus, judgements frequently have to be made from evaluations based on very diverse criteria that must be integrated into an overall assessment. In these situations the methods adopted for selecting and decision-making provide help and support decisions that are as transparent and as logical as possible. These methods include quantitative procedures such as risk analysis, material flow analyses, life cycle analyses, environmental balance, cost-benefit analyses and multi-criteria decision-making procedures, as well as qualitative and discourse-orientated procedures (see [Joss and Belucci, 2002]; see also Grunwald’s chapter in this Volume, Part V). All these methodological and conceptual approaches that link developments in the engineering sciences to their social context are common to all engineering sciences and represent the expression of a specific theory-practice relationship. Thus,

Coherence and Diversity in the Engineering Sciences

177

to the extent that these tasks and the methods and procedures employed to solve them (such as cost-benefit analyses, risk assessment, multi-criteria decision-making analyses or technology assessment) are an expression of coherence within the engineering sciences, they can be taught in university courses across and beyond the borders of faculties and disciplines. Their application and the forms they take vary, however, in the different engineering sciences disciplines. 4 COHERENCE AND DIVERSITY IN THE ENGINEERING SCIENCES DISCIPLINES Engineering sciences are subdivided into a large number of disciplines corresponding to the diversity of technical systems and the objectives of technical and scientific-technical activity. There is, however, no recognisable hierarchy in this organisation of the disciplines. We maintain that the reason for this is because the internal structure of the engineering sciences reflects a lengthy historical development process marked by pragmatic classification principles revealing the status of technological developments and their level of scientific inquiry. Scientific discipline pertains here to both a specific system of knowledge (cognitive; statements about a subject area, i.e., about existing and future technical systems) and to a specific community of scientists (social; institutionally organised researchers and teachers affiliated to subject areas with specific technical language, scientific standards, publications, scientific events, etc.). The formation and development of the disciplines in the engineering sciences (genesis of disciplines) depends on the coherence and interaction between these cognitive and social factors. The dynamics which are currently clearly prominent in the development of theories and methods in engineering sciences (differentiation and integration!) have refocused the discussion more sharply on the problems encountered in classifying the engineering sciences. It is understandable that as the tasks confronting the engineering sciences become more complex, so too does their processing and thus also the interdisciplinary interaction of a whole range of internal disciplines. Current knowledge of the structure and classification of the engineering sciences does not give a comprehensive picture but it does allow us to identify certain results and problems which contribute to understanding the nature and development of the field. In addition, it must be borne in mind that any classification — and this includes the engineering sciences — is pragmatic, that is to say, specific and related to a particular end. Such an end could include the organisation of faculties and institutes within a technical college, a library system, the organisation and comparable education of professional groups affiliated to technology (i.e. engineers), the structure of databases, etc. For this reason, the following discussion constitutes only one approach to coherence and diversity in the content side of the engineering sciences (see [Jobst, 1986]). We should first emphasise that probably the “roughest” but at the same time most basic way to classify the engineering sciences might be by making distinctions

178

Gerhard Banse and Armin Grunwald

according to whether the dominant features are technological (manufacturing, processing or procedural techniques), constructive, or materials science-linked tasks. This differentiation, which touches on a large group of disciplines and is justified on the basis of the specific aspects of both the subject and the activity aspects of the engineering sciences, will continue to remain important in the future. In the second place an organisational principle has emerged in the engineering sciences system which in some ways represents a further differentiation of the first point and which is objectively determined by the technical means and processes that are scientifically processed. This principle distinguishes between those disciplines that are usually concerned with the more constructional (structural) and the more technological (procedural) side of groups of products and procedures. Within these disciplines there is often — but not always — a subdivision in line with constructive and technological orientations. This pertains to the quantitative and qualitative status of development of the subject area in question and to the specific nature of the historical development of particular disciplines, to personnel questions, to the size of the scientific institutions, etc. The disciplines which fall under this organisational principle include — from the field of general mechanical engineering — machine tool construction, textile machines and polygraphs. Disciplines from the fields of electrotechnics, electronics and computer technology are also included. In concrete terms these disciplines may, of course, be further subdivided. In the field of machine tool construction one may take, for instance, such areas as production means development, component manufacture, process design, jointing techniques and assembly. This does not, however, contravene the principle that the structure of the disciplines is determined by examining real technical means and processes. Within the engineering sciences system there is, in the third place, the matter of the disciplines that deal with the structural and functional problems occurring in all or many technical means and processes. One could term these cross-sectional disciplines. They include measuring techniques, automation techniques, construction science, standardisation, security engineering and tribo techniques. It is, however, impossible at present to even come close to drawing any definite conclusions on classification in precisely these disciplines for the simple reason that their theoretical and practical significance is still growing in new directions and in terms of their scientific profile (e.g. measuring techniques and quality assurance), they are currently in a crucial phase of theoretical generalisation. The increasing significance of these cross-sectional disciplines is above all an expression of the process-orientated complexity of technical systems and corresponding solutions in the engineering sciences. In their investigations they reflect important moments of integration in the engineering sciences system. This is particularly true of systems that deal with the theoretical control of functions and structures, in other words, with whole classes of technical means and processes or those marked by strong similarity which has to be attributed to different classes. In the fourth place it should be stressed that the engineering sciences do embrace basic disciplines. Their specific nature is particularly present in the gaining

Coherence and Diversity in the Engineering Sciences

179

of mainly theoretically-based knowledge and the developing of models, computational specifications, diagrams, flow charts, etc. These are the basic prerequisites for intellectual anticipation, constructive and technological design, evaluation, monitoring, and the perfection of whole classes of technical means and processes. What is still specific to these disciplines is the fact that in their investigations knowledge of the natural and mathematical sciences is very markedly transformed, combined, modified and applied to arrive at generally applicable theoretical statements on fundamental technical structures and functions. To cite a few typical examples, the following areas can be included in the basic disciplines of engineering sciences: technical mechanics, technical thermodynamics, technical cybernetics, the theory of electrical engineering and information technology theory. It is to be expected that in the future the number and more to the point the significance of such basic disciplines in the engineering sciences will grow as the natural and mathematical sciences increasingly pervade many areas of engineering sciences thus, in the process, expanding their theoretical basis. In short we may conclude that with regard to the coherence and diversity of the content of the engineering sciences there are many different and opposing tendencies associated firstly with content difference (types, functions, dimensions, the input and output of technological systems, the operating conditions such as forms of energy, temperatures, pressures, and so on) and secondly with technological development, research and teaching. 5 GENERAL TECHNOLOGY In addition to the tendencies towards differentiation, opposing tendencies that emphasise common features have always been present in the area of technical knowledge and engineering sciences, both towards a “general heuristics of invention” (see [Nickles, 1980]) and in the direction of a general technology. The term “invention heuristics” refers to all the intellectual efforts which, in a more organisationally systematic way (more or less independent of subject area), are suitable or capable of contributing to the production of something new (i.e. also to innovations in the technical field such as inventions, improvements or optimisations). These intellectual efforts can be traced back a long way to before the establishment of the engineering sciences in the nineteenth century. Examples include the work of Socrates (469–399 BC), Aristotle (384–322 BC), Archimedes of Syracuse (ca. 287–212 BC), Pappus of Alexandria (ca. 320 AD), Ramon Llull (1232/33–1316), Francis Bacon (1561–1626), Ren´e Descartes (1596–1650), Christopher Polhem (1661–1751) and Gottfried Wilhelm Leibniz (1646–1716). What is notable in this regard is also the cinematics of Robert Willis (1800– 1875), the “Theoretical Cinematics” by Franz Reuleaux (1829–1905) from 1875, the combinatorics of Wilhelm Ostwald (1853–1932) and — more recently — the “morphological method” of Fritz Zwicky (1898–1974). However, the ideal of acquiring a logically unambiguous understanding was always hindered by insurmountable difficulties in practice: the number of elementary

180

Gerhard Banse and Armin Grunwald

or at least preliminary elementary concepts has always been and still is much larger than was originally supposed which means that the number of possible combinations is enormous. In 1806, Johann Beckmann (cf. Section 2.2) published the third part of his “Vorrat kleiner Anmerkungen u ¨ber mancherley gelehrte Gegenst¨ ande” (“Collection of Short Notes on Various Learned Subjects”) which included a short text that could be termed the “birth certificate” of a general technology: i.e. the “Entwurf der algemeinen Technologie” (“Draft on General Technology”). This signified a new approach to the understanding, comparing and inventing of technical systems. We have to briefly recapitulate if we are to contextualise the relevant threads of argument. In his work “Anleitung zur Technologie oder zur Kenntnis der Handwerke, Fabriken und Manufakturen” (“Instructions on Technology, or on the Knowledge of Trades, Crafts and Manufacture”), first published in 1777, Beckmann views technology not — as was usual up until then — as the teaching of skills but rather as “the science that teaches the processing of natural things or the knowledge of crafts”, as a science “which explains completely, properly and clearly all works, their results, and their reasons” [Beckmann, 1780, p. 17]. In this first step on the path to a more strongly theoretical basis for engineering, his initial intention was to systematise the various works of crafts and factories by going beyond a mere descriptive list of means and procedures. To that end he called for a stronger contemplation of knowledge from the natural sciences and a unified — and thus also unifying — terminology. While in his “Guide to Technology” Beckmann presents “a justification of the science of technology”, in his 1806 work “Entwurf der algemeinen Technologie” (“Draft on General Technology”), he goes a step further. Initially he divides technology into two parts by distinguishing special technology (the “particular part”) from general technology (the “first or general part of technology”): “Now I desire a directory of all the different purposes which craftsmen and artists have in their various works and together with it a directory of all the means by which they will know how to achieve these ends. I would give a directory of this kind the name of general technology, or the first or general part of technology” [Beckmann, 1806, p. 465]. The “description of the individual crafts” is the function of the special technology in the sense called for in 1777. With general technology, Beckmann then goes beyond a simply comparative systematisation of the exploitable technical means to consider realistic (technical) ends. This general technology should “reveal the common and special intentions of the works and means, explain the principles on which they are based and briefly teach what could serve the understanding and evaluation of the individual means and their selection in transference to other objects”. “For the artists and craftsmen”, he continues “this would facilitate thorough and general concepts of the objects they work on and the customary procedures here, and indeed, it would provide an overview which could lead inventive minds to new, useful improvements [ibid., pp. 465, 480]. As a scientific “reason” for all of this

Coherence and Diversity in the Engineering Sciences

181

— i.e. as the “interest-guiding knowledge”, he emphasises that “[a]nyone who has tried to learn about many crafts and arts and anyone who has tried to gather many of them together in one view must notice that very many crafts, although their materials and goods are different, still use many works to achieve similar ends or that these crafts can achieve the same end in very different ways” [ibid., pp. 464]. Beckmann’s intention — as becomes clear from his considerations — is first to systematise all the technical and technological knowledge collected up until then before going on to give it a secure theoretical basis and justify the methodological programme of a heuristic of invention. In the years that followed, the general technological approach to outlining engineering in its entirety lost its importance. However, efforts were always made to undertake more or less comprehensive generalisations for individual fields of engineering or engineering sciences. An example is the more strongly (constructional) methodological approach adopted by Franz Reuleaux in the second half of the nineteenth century and the studies on “unit operations” and “unit processes” conducted in the 1920s and 1930s, which can be seen as the foundation or grounding of a general chemical technology. In the past 30 years, the general technology discussion has once again been revived: questions on generalisation and approaches to general technology were presented by various scientists from the perspective of different scientific disciplines. Today general technology is still more of a program than a concept that has been elaborated. In these various approaches to general technology, the general nature of technical objects and processes are recorded in terms of technical principles, basic and guiding rules, regularities, statements about effective settings, and so on. For more far-reaching generalisations towards a general technology (above all else as part of the “theoretical justification of the engineering sciences”), one must take into account that there are various, not yet sufficiently debated positions here (see, e.g., [Banse and Reher, 2002; 2004; Koen, 2003; Ropohl, 1999; Simon, 1981; Spur, 1998a; Wolffgramm, 1994/95]). The development of technical knowledge is thus on the one hand marked by embellishment with details and on the other by the promotion of systematising, comparative, integrating and generalising knowledge bases. 6

CONCLUDING REMARKS

The engineering sciences represent an interesting subject of investigation for philosophy and scientific theory. Many cognitive, methodological, normative and — in the strictest sense — science-theoretical questions differ from those in many other sciences. This comes first and foremost from the fact that engineering sciences must be understood as sciences of “doing” (as aim-related and producing activity) and that their practical, final result is real technology. This factor must be borne in mind at all times. In this case “real” technology must be taken to mean different things. First it is used to emphasise the fact that technical systems

182

Gerhard Banse and Armin Grunwald

must function in a real world situation for a particular duration, not in an ideal or laboratory environment in which disturbances can be intellectually or practically isolated. This is also true of the production connection which — though it begins with an idea — only sees completion with the objectivised product. The real “environment” with its manifold influences, interactions and changes must always be remembered — although often it cannot be comprehensively registered in its entirety, nor its effects completely gauged or forecast from the point of view of impact. Ultimately, the application of technical systems relates to real people and to real institutions. Technology must be saleable, purchasable, acceptable, and usable. BIBLIOGRAPHY [Banse, 2003] G. Banse. Technikgestaltung im Spannungsfeld von Plan und Lebenswelt [Design of Technology: The Tension between Plan and Life-World]. In Technikgestaltung zwischen Wunsch und Wirklichkeit [Design of Technology between Desire and Reality], A. Grunwald, ed., pp. 71-87. Springer, 2003. [Banse and Friedrich, 2000] G. Banse and K. Friedrich, eds. Konstruieren zwischen Kunst und Wissenschaft: Idee — Entwurf — Gestaltung [Designing between Art and Science: Idea – Concept — Design]. Edition Sigma, 2000. [Banse et al., 2006] G. Banse, A. Grunwald, W. K¨ onig, and G. Ropohl, eds. Erkennen und Gestalten. Eine Theorie der Technikwissenschaften [Knowing and Designing. A Theory of Engineering Sciences]. Edition Sigma, 2006. [Banse and M¨ uller, 2001] G. Banse and H.-P. M¨ uller, eds. Johann Beckmann und die Folgen. Erfindungen — Versuch der historischen, theoretischen und empirischen Ann¨ aherung an einen vielschichtigen Begriff [Johann Beckmann and his Impact. Inventions — An Attempt at a Historical, Theoretical and Empirical Approximation to a Complex Term]. WaxmannVerlag, 2001. [Banse and Reher, 2002] G. Banse and E.-O. Reher, eds. Allgemeine Technologie — Vergangenheit, Gegenwart, Zukunft [General Technology — Past, Present, Future]. Trafo Verlag, 2002. [Banse and Reher, 2004] G. Banse and E.-O. Reher, eds. Fortschritte bei der Herausbildung der Allgemeinen Technologie [Advances in the Development of General Technology]. Trafo Verlag, 2004. [Baumard, 1999] P. Baumard. Tacit Knowledge in Organisations. Sage Publications, 1999. [Beckmann, 1780] J. Beckmann. Anleitung zur Technologie, oder zur Kenntnis der Handwerke, Fabriken und Manufakturen... [Instructions on Technology, or on the Knowledge of Trades, Crafts and Manufactures . . . ]. 2nd ed. Verlag Vandenhoeck, 1780. [Beckmann, 1806] J. Beckmann. Entwurf der algemeinen Technologie [Draft on General Technology]. In Vorrath kleiner Anmerkungen u ¨ber mancherley gelehrte Gegenst¨ ande. Drittes St¨ uck [Collection of Short Notes on Various Learned Subjects. Third Part], pp. 463-533. Verlag Johann Friedrich R¨ ower, 1806. [Buchheim and Sonnemann, 1990] G. Buchheim and R. Sonnemann, eds. Geschichte der Technikwissenschaften [History of Engineering Sciences]. Verlag Edition Leipzig, 1990. [Bunge, 1974] M. Bunge. Technology as Applied Science. In Contributions to a Philosophy of Technology, F. Rapp, ed., pp. 19-39. Reidel, 1974. [Bunge, 1983] M. Bunge. Towards a Philosophy of Technology. In Philosophy of Technology. Readings in the Philosophical Problems of Technology, C. Mitcham and R. Mackey, eds., pp. 62-76. Free Press, 1983. [Collins, 1974] H. M. Collins. The TEA Set: Tacit Knowledge and Scientific Networks. In Science Studies, 2, 165-186, 1974. [Cross, 1989] N. Cross. Engineering Design Methods. John Wiley & Sons Ltd, 1989. [D¨ urr, 1988] H.-P. D¨ urr. Das Netz des Physikers [The Net of the Physicist]. Carl Hanser Verlag, 1988.

Coherence and Diversity in the Engineering Sciences

183

[Eder, 2000] W. E. Eder. Konstruieren aus der Sicht eines Konstruktionswissenschaftlers [Designing from a Viewpoint of a Designer]. In Konstruieren zwischen Kunst und Wissenschaft: Idee — Entwurf — Gestaltung [Designing between Art and Science: Idea — Concept — Design], G. Banse and K. Friedrich, eds., pp. 193-218. Edition Sigma, 2000. [Ferguson, 1992] E. S. Ferguson. Engineering and the Mind’s Eye. MIT Press, 1992. [Ferr´ e, 1995] F. Ferr´ e. Philosophy of Technology. The University of Georgia Press, 1995. [Gregory, 1966] S. A. Gregory. The Design Method. Butterworths, 1966. [Grunwald, 2000] A. Grunwald. Against Over-estimating the Role of Ethics in Technology. In Science and Engineering Ethics, 6, 181-196, 2000. [Grunwald, 2000a] A. Grunwald. Philosophy and the Concept of Technology. On the Anthropological Significance of Technology. In On Human Nature. Anthropological, Philosophical and Biological Foundations, A. Grunwald, M. Gutmann and E. Neumann-Held, eds., pp. 173-188. Springer-Verlag, 2002. [Grunwald, 2002b] A. Grunwald. Technikfolgenabsch¨ atzung. Eine Einf¨ uhrung [Technology Assessment. An Introduction]. Edition Sigma, 2002. [Hacking, 1983] I. Hacking. Representing and Intervening: Introductory Topics in the Philosophy of Natural Science. Cambridge University Press, 1983. [Hubka et al., 1988] V. Hubka, M. Andraesen, W. E. Eder, and J. P. Hills. Practical Studies in Systematic Design. Butterworths, 1988. [Hubka and Eder, 1996] V. Hubka and W. E. Eder. Design Science. Introduction to Needs, Scope and Organization of Engineering Design Knowledge. Springer-Verlag, 1996. [Jobst, 1986] E. Jobst. Besonderheiten von Klassen von Technikwissenschaften [Specialities of Groups of Engineering Sciences]. In Erkenntnismethoden in den Technikwissenschaften [Methods of Cognition in the Engineering Sciences], G. Banse and H. Wendt, eds., pp. 15-18. Verlag Technik, 1986. [Joss and Belucci, 2002] S. Joss and S. Belucci, eds. Participatory Technology Assessment — European Perspectives. Centre for the Study of Democracy (CSD) at University of Westminster in Association with TA Swiss, 2002. [Koen, 2003] B. V. Koen. Discussion of the Method: Conducting the Engineer’s Approach to Problem Solving. Oxford University Press, 2003. [K¨ onig. Technikwissenschaften [Engineering Sciences]. Verlag Fakultas, 1995. onig, 1995] W. K¨ [K¨ onig, 2006] W. K¨ onig. Geschichte der Technikwissenschaften [History of Engineering Sciences]. In Erkennen und Gestalten. Eine Theorie der Technikwissenschaften [Knowing and Designing. A Theory of Engineering Sciences], G. Banse, A. Grunwald, W. K¨ onig and G. Ropohl, eds., pp. 24-37. Edition Sigma, 2006. [Kornwachs, 1995] K. Kornwachs. Theorie der Technik? [Theory of Technology?]. In Forum der Forschung. Wissenschaftsmagazin der Brandenburgischen Technischen Universit¨ at Cottbus, 1, 11-22, 1995. [Kornwachs, 2001] K. Kornwachs. A Formal Theory of Technology? In Advances in the Philosophy of Technology, H. Lenk and M. Maring, eds., pp. 51-69. Lit-Verlag, 2001. [Lenk and Maring, 2001] H. Lenk and M. Maring, eds. Advances in the Philosophy of Technology. Lit-Verlag, 2001. [Mitcham, 1994] C. Mitcham. Thinking through Technology. The Path between Engineering and Philosophy. University of Chicago Press, 1994. [Mitcham and Mackey, 1983] C. Mitcham and R. Mackey, eds. Philosophy and Technology. Readings in the Philosophical Problems of Technology. Free Press, 1983. [Nickles, 1980] T. Nickles. Scientific Discovery, Logic and Rationality. Reidel, 1980. [Pater and van Gils, 2003] A. Pater and A. van Gils. Stimulating Ethical Decision-Making in a Business Context. Effects of Ethical and Professional Codes. European Management Journal, 21, 762-772, 2003. [Polanyi, 1962] M. Polanyi. Personal Knowledge. University of Chicago Press, 1962. [Polanyi, 1966] M. Polanyi. The Tacit Dimension. Doubleday & Co, 1966. [Radder, 2003] H. Radder. Technology and Theory in Experimental Science. In The Philosophy of Scientific Experimentation, H. Radder, ed., pp. 152-173. University of Pittsburgh Press, 2003. [Rader, 2002] M. Rader. Synthesis of Technology Assessment. In Strategic Policy Intelligence: Current Trends, the State of Play and Perspectives, A. T¨ ubke, K. Ducatel, J. P. Gavigan and P. Moncada-Patern` o-Castello, eds., pp. 27-37. Institute for Prospective Technological Studies (IPTS), 2002.

184

Gerhard Banse and Armin Grunwald

[Rip et al., 1995] A. Rip, T. Misa, and J. Schot, eds. Managing Technology in Society. Cengage Learning, 1995. [Rittel and Webber, 1973] H. W. J. Rittel and M. M. Webber. Dilemmas in a General Theory of Planning. Policy Sciences, 2, 155-169, 1973. [Ropohl, 1999] G. Ropohl. Allgemeine Technologie. Eine Systemtheorie der Technik [General Technology. A Systems Theory of Technology]. 2nd ed. Carl Hanser Verlag, 1999. [Schaub et al., 1983] J. H. Schaub, K. Pavlovic, and M. D. Morris, eds. Engineering Professionalism and Ethics. John Wiley & Sons, 1983. [Schneider, 1987] J. Schneider. Franz Reuleaux und die Theorie der Maschinen [Franz Reuleaux and the Theory of Machines]. In Wissenschaften in Berlin [Sciences in Berlin], vol. 3, T. Buddensieg, K. D¨ uwell and K.-J. Sembach, eds., pp. 173-177. Gebr¨ uder Mann Verlag, 1987. [Simon, 1973] H. A. Simon. The Structure of Ill-structured Problems. Artificial Intelligence, 4, 181-201, 1973. [Simon, 1981] H. A. Simon. The Sciences of the Artificial. MIT Press, 1981. [Spur, 1998a] G. Spur. Deutung der Technologie als Lehre vom Wandel der Technik [The Interpretation of Technology as a Theory of Technological Change]. ZWF — Zeitschrift f¨ ur wirtschaftliche Fertigung, 93, 235-239, 1998a. [Spur, 1998b] G. Spur. Technologie und Management [Technology and Management]. Carl Hanser Verlag, 1998b. [Technology, 1966] Technology. Toward a Philosophy of Technology. Technology & Culture, 7, 301-390, 1966 (with contributions by L. Mumford, J. K. Feibleman, M. Bunge, J. Agassi, J. O. Wisdom, H. Skolimowski and I. C. Jarvie). [Tscheschew and Wolossewitsch, 1980] W. W. Tscheschew and O. M. Wolossewitsch. Gegenstand und Aufgaben der technischen Wissenschaften [Subject and Tasks of Engineering Sciences]. In Spezifik der technischen Wissenschaften [Specifics of Engineering Sciences], pp. 163-192. Fachbuchverlag, 1980. [van de Peol, 2001] I. R. van de Poel. Investigating Ethical Issues in Engineering Design. Science and Engineering Ethics, 7, 429-446, 2001. [van Gorp, 2005] A. C. van Gorp. Ethical Issues in Engineering Design; Safety and Sustainability. Delft University of Technology, 2005. [VDI, 1977] VDI — Verein Deutscher Ingenieure. Richtlinie 2222 “Konstruktionsmethodik: Konzipieren technischer Produkte [Guideline 2222 “Methods of Design: Prepairing of Technical Products”]. VDI, 1977. [VDI, 1991] VDI — Verein Deutscher Ingenieure. Richtlinie 3780 “Technikbewertung — Begriffe und Grundlagen” [Guideline 3780 “Technology Assessment — Terms and Foundations”]. VDI, 1991 (English version available at www.vdi.de). [Vincenti, 1990] W. G. Vincenti. What Engineers Know and How They Know It. Johns Hopkins University Press, 1990. [Wolffgramm, 1994/95] H. Wolffgramm. Allgemeine Technologie [General Technology]. 2 vols. 2nd ed. Verlag Franzbecker, 1994/95.

Part II

Ontology and Epistemology of Artifacts

This page intentionally left blank

INTRODUCTION TO PART II

Wybo Houkes, associate editor Ontology and epistemology are two of the oldest and best-established branches of philosophy. Yet in the long histories of these fields, artifacts and technology have seldom been topics of inquiry. Likewise, engineers have traditionally ignored questions regarding the nature and categorization of the items they produce and regarding the status of the knowledge they produce. Still, it seems obvious that engineering activities change the world by adding objects, such as combustion engines, and also add to our understanding of this changed world, by acquiring knowledge of, for instance, combustion processes. Indeed, technology can be characterized in at least two ways: as a system of artificial objects, and as a system of (practical) knowledge [Mitcham, 1978]. Surprisingly, these intuitions and suppositions have not given rise to more thorough analysis. A short reflection on the reasons for this lack of interest serves to introduce the background of the various contributions to this part. Mutual distrust partly explains the lack of mutual interest. Most philosophers have, ever since Plato and Aristotle, been suspicious of technological advances, and have subordinated the efforts of engineers to those of natural scientists. In the works of Heidegger and Ellul, which shaped much of what is now called the philosophy of technology, these anti-technological sentiments have culminated in deeply pessimistic, essentialist evaluations of technology and of societies that have become irredeemably technological. On their part, engineers largely equate philosophical reflection with empty speculation. The metaphysical engineer, it seems, only exists in poetry.1 More importantly, two standard ideas about artifacts make further philosophical analysis seem useless. The first, ontological idea is that artifacts do not, properly speaking, exist. Engineers cleverly rearrange physical stuff into suspension bridges and combustion engines, but in doing so, they add nothing to the ultimate inventory of the world. The second, epistemological idea is that technology is nothing but applied (natural) science. Both in engineering design and in engineering science, people solve practical problems, but in doing so, they add nothing – or at least nothing fundamental – to the total body of human knowledge. This idea is not exclusive to philosophy. There is a continuous, albeit not constant, tendency among engineers to model their work after the natural sciences. Curricula at engineering schools have become filled with courses in which students are taught to 1 “The metaphysical engineer” is one of the self-descriptions of Alvaro ´ de Campos, a heteronym of the Portugese poet Fernando Pessoa.

Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

188

Wybo Houkes, associate editor

apply scientific theories such as classical mechanics and thermodynamics to practical problems; and in practice, engineers have frequently attempted to “ground” their designs with the experimental method. Alternative (e.g., design-oriented) conceptions of engineering exist, especially since the 1960s, but they are typically prominent countermovements instead of the new orthodoxy. The two standard ideas are probably related, and may root in deeper convictions, e.g., about the reducibility of all objects to physical matter and about the subordination of practical reasoning to theoretical reasoning. What matters here is that, in combination, the ideas entail that, ontologically and epistemologically, engineering is a zero-revenue activity. While this view is clear and provocative, and may be assumed as a null hypothesis, it should only be accepted, if at all, after ample reflection. Such reflection is mostly clearly shown by recent work on artifact ontology. Several philosophers have argued against the ontological null hypothesis, and have concluded that artifacts are a significant class of real objects. These arguments, of which an overview is given in the chapter by Amie Thomasson, pave the way for analyses of the nature of artifacts. A feature of artifacts that has commanded particular attention is their functionality. Many artifacts come in functional kinds, and all appear to be “for” doing something. Analyses of artifact function typically focus on its relation with the intentions of designers and users, but nonintentionalist alternatives are possible. In her chapter, Beth Preston reviews these accounts, with the aid of a set of analysanda. Engineers also show an increasing interest in the nature and categorization of artifacts, albeit from a different perspective. Two areas that illustrate possible convergences between philosophy and engineering are presented in this part — both by, necessarily, explorative papers. Firstly, the chapter by Pieter E. Vermaas and Pawel Garbacz explores functional decomposition and function modelling. These engineering practices revolve around artifact functionality, and its relation to the structural or physical properties of artifacts. The practical value of these practices depends partly on conceptual precision, meaning that philosophy and engineering can strike up a mutually profitable relation. The same goes for the second area, that of formal ontology. For various engineering applications, categorizations of objects and their parts are vital. This has led to a veritable industry of domainspecific ontologies. The chapter by Stefano Borgo and Laure Vieu provides an overview of several ontologies and explores how, for one ontology in particular, characterizations of artifacts and their properties may be included. The last three contributions deal with epistemological issues. For these, the blockade on artifacts and technology was lifted in the 1970s, as the applied-science thesis grew ever more unfashionable. This led to some reflections of the nature of technological knowledge, which often stress its autonomy from scientific knowledge. Indeed, the knowledge produced by engineers and, more generally, knowledge regarding artifacts appears to have some distinctive features, such as its tacit dimension and its relation to practical, rather than purely theoretical interests. An overview of the work on technological knowledge, as provided in the chapter by

Introduction to Part II

189

Wybo Houkes, may lead to doubt whether these features are sufficiently distinctive, but this should not prevent — and has not prevented — philosophers and other researchers from analyzing them. Outside philosophy, the tacit dimension of technology and engineering design has attracted substantial attention. Paul Nightingale reviews the results of this attention and their relevance for a better understanding of engineering. Inside philosophy, such a better understanding may be created by analyzing the role of practical reasoning in engineering. In his chapter, Jesse Hughes explores ways in which this role may be conceptualized. Research on artifact epistemology does not, at the moment, show the vitality of research on artifact ontology, neither in philosophy nor in engineering. However, the contributions make clear that there is room for substantial growth. All papers make clear that philosophers and engineers have only begun to develop an appropriate ontology and epistemology for the realm of artifacts. All papers make specific suggestions for further research, from developing an account of artifact functions that is fully adequate to the complicated phenomenology of use and design to analyzing the role of specificationism in the reasoning of engineers. Indirectly, they shed light on issues that have to remain unexplored in this part, such as the relation between structural part-whole relations and functional decomposition, and the difference between natural and artificial objects. In combination, they also show the need for a richer account of what counts as a real object and of our standards for knowledge and its relation to action – questions that lie at the roots of the disciplines of ontology and epistemology. BIBLIOGRAPHY [Mitcham, 1978] C. Mitcham. Types of technology, Research in Philosophy and Technology 1, 229-274, 1978.

This page intentionally left blank

ARTIFACTS IN METAPHYSICS Amie L. Thomasson Perhaps the most striking feature about discussions of artifacts in recent metaphysics is their paucity. While attention has focused on explicating the basic metaphysical concepts at work in the physical sciences, such as space, time, property, event, cause, and law of nature, artifacts and other common sense objects have been relatively neglected by metaphysicians.1 Where artifacts have been discussed, they are often mentioned only in the contexts of arguments that we should deny that there are any such things. In short, where they are not neglected, artifacts have most often simply been rejected from metaphysical studies of what there is. The failure to study the metaphysics of common sense objects such as artifacts is unfortunate, since understanding the ontological status of artifacts is crucial to understanding the objects of concern in the social, human and technological sciences and to investigating the proper methodologies of these sciences. If that were the only problem, metaphysicians might simply be accused of benignly neglecting an important possible application of their work. I will argue, however, that the problems run much deeper than that: failure to adequately consider the problems presented by artifacts has led to important blindspots within metaphysics itself. Engaging with the problems artifacts present for metaphysics, I will argue, may lead us to rethink some of the most central problems in metaphysics and beyond, including whether various classic metaphysical problems (including problems of colocation, vagueness, and the like) require solution, how we should handle existence questions and conceive of realism, how we should understand the relation between modality and human concepts, and how the social and human sciences differ from their natural science counterparts. While consideration of artifacts is not always the only route to discovering the need to rethink these issues, it is at least a sufficient route — and that is enough to secure the significance of artifacts for metaphysics. 1

ROOTS OF NEGLECT AND REASONS FOR REJECTION

The reasons artifacts have largely been neglected in analytic metaphysics over the past century or so are not difficult to unearth. The small portion of metaphysics 1 An important exception is Randall Dipert’s [1993] detailed study of artifacts. There are also recent signs of an emerging interest in artifacts, e.g., in the work of Lynne Baker [2007] and Crawford Elder [2004], as well as my [2003; 2007a; 2007b].

Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

192

Amie L. Thomasson

that survived the positivist assault was that dedicated to serving as the handmaiden of the natural sciences by explicating their fundamental concepts — a role that left little room for undertaking a metaphysics of artifacts and other objects of the social and human sciences. The idea that metaphysics could provide insight to what exists — not merely to our language and concepts — has, of course, been revived in post-Quinean metaphysics. But here again, metaphysics is conceived of explicitly as of a piece with the natural sciences. Following Quine, moreover, those seeking to revive metaphysics have embraced the idea that the proper methods of determining an ontology involve determining what our best scientific theories (with physics as the paradigm) must quantify over. As long the ‘best scientific theories’ are considered to include only those of the natural sciences, this method provides a more explicit ground for justifying the neglect by holding that we need not accept artifacts and other social and cultural objects in our ontology. (I will have more to say about the Quinean approach to existence questions in §2 below.) Even if we broaden our view to include the theories of the social and technological sciences, as long as one assumes that their claims may ultimately be rewritten in terms that need not quantify over artifacts, one also assumes that artifacts may be safely ignored by metaphysics. (Of course, little has been done to justify this assumption by showing how such claims might be rewritten, and the prospects for doing so across the board are dim.) Recent discussions in metaphysics, however, have done more than tacitly neglect artifacts. A large number of contemporary metaphysicians have argued, on various grounds, that we have positive reason to deny that artifacts exist. These arguments are often based on claims that accepting the existence of artifacts (or other medium-sized composite objects) would violate certain independently plausible general metaphysical principles. Thus, e.g., general prohibitions against colocation: that no two objects may be in the same place at the same time, or may share all of the same parts, are sometimes taken as grounds for denying the existence of artifacts, since an artifact and its constitutive lump of matter apparently do occupy the same place at the same time, and are made of the same parts (at some level of decomposition).2 Others have rejected them for violating Alexander’s Dictum that ‘to be real is to have causal powers’, since all the work commonly supposed to be done by artifacts allegedly may be better attributed to their microscopic parts [Merricks, 2001]. Still others have held that we must deny the existence of artifacts to maintain the metaphysical view that there is no vagueness in the world (since it seems that, if our artifact concepts refer to any objects, they must refer to vague objects);3 or to preserve the idea that there is a uniform principle of composition, determining (in any situation) when various objects compose 2 Indeed Merricks [2001, pp. 40–42] goes further, urging that troubles with colocation give us reason to deny the existence of artifacts and of their constitutive lumps of matter. 3 Horgan [1994], for example, uses this approach to argue that — at least where strict semantic standards are in place — we have reason to deny the existence of artifacts and many other ordinary objects.

Artifacts in Metaphysics

193

some larger object [van Inwagen, 1990; Horgan and Potrc, 2000]. Even where these metaphysical difficulties aren’t taken as direct grounds for denying the existence of artifacts, avoiding such difficulties is commonly seen as an important attraction of eliminativist views. For, as Peter van Inwagen writes, “if there are no artifacts, then there are no philosophical problems about artifacts” [1990, p. 128]. Avoiding these problems simply by denying the existence of artifacts (and their kin) prevents metaphysicians from really engaging with the most central problems that arise for making sense of our common-sense world view, and for making sense of the objects of the social and technological sciences. Moreover, if (as I argue below) we have good reason to accept the existence of artifacts, it seems that we must actually try to confront these classic metaphysical problems, giving us reason to reopen these issues and find ways to solve or dissolve — rather than merely side-step — the problems, making the serious study of artifacts again of broader relevance for metaphysics. Unfortunately, there is not space here to discuss all of these problems directly. Moreover, since these are problems that arise equally for inanimate composite natural objects (such as sticks and stones), they are not unique problems for artifacts. I have discussed these and other alleged metaphysical problems for ordinary objects in depth elsewhere [2007a]. There, I argue that we can diagnose the problems behind all of the various arguments against artifacts and other ordinary objects by accepting certain basic principles about meaning and modality. These principles in turn lead to important conclusions about the proper methods and limits of metaphysics, making a serious study of artifacts (and other ordinary objects) once again centrally relevant to our understanding of metaphysics itself. Rather than reviewing these replies in detail, I refer the reader to the detailed discussions there. Here instead, I will focus on those metaphysical problems that arise more uniquely for artifacts — or at least, for artifacts and other cultural and social objects. On the face of it, artifacts are distinguished from common-sense natural objects in that they are apparently mind-dependent, at least in the sense that (unlike rocks and trees) they would not exist were it not for the beliefs, practices, and/or intentions of the human beings who make and use them. But the apparent mind-dependence of artifacts raises distinctive metaphysical suspicions against them. Some resist the idea that there could be any mind-dependent objects, on grounds that accepting them requires positing ‘magical modes of creation’ that allow that — at least in these ‘special’ cases — human thought or intentions may bring new entities into existence. I will begin in §2 by trying to clarify the claim that artifacts are existentially mind-dependent, and showing why accepting that there are objects that, like artifacts, are existentially mind dependent does not require invoking any ‘magical modes of creation’. The purely descriptive approach to existence questions that I will recommend provides the basis for a direct argument that we should accept the existence of artifacts, and also gives reason for abandoning the standard Quinean approach to existence questions.

194

Amie L. Thomasson

Others have objected not to the idea that artifacts would depend on human intentionality for their existence, but rather to the idea that their natures might be mind-dependent. For, it is often assumed, any object that is a genuine part of our world must have a fully mind-independent nature available for discovery. I will address these arguments in §4, arguing that we have no reason to deny that the natures of artifacts are — in a sense — determined by human intentions in favor of the problematic view that they have mind-independent natures. Accepting that the natures of these kinds are determined by human intentions has important implications for understanding the differences in our epistemic relationship to the objects of the social and technological sciences on the one hand, and the natural sciences on the other, and for understanding differences in the proper methodology for each area of inquiry.

2 DEPENDENCE FOR EXISTENCE It is fairly uncontroversial that artifacts — in contrast with natural objects such as sticks and stones — are mind-dependent objects, in some sense of the term. But appeals to mind-dependence in philosophical discussions often lead more to obfuscation than to clarity, so we must tread carefully here to distinguish the sense(s) of mind-dependence that might rightly be claimed for artifacts. The most obvious sense in which artifacts may be said to be mind-dependent is that artifacts would not exist were it not for the (mental and physical) activities of humans; they are the products of human work. But even this simple point admits of at least two interpretations: there is the causal point, that the intentional activities of humans are causally responsible for the production of tables and chairs, ships and sirens. If this were the only sense of dependence at issue, it might not be of great philosophical interest, since the intentional activities of humans are also causally responsible for the production of a great many natural objects, such as the plants and animals reared in agriculture. The philosophically interesting sense in which artifacts seem to be mind-dependent is not the simple causal sense, but rather the conceptual sense: that is, the very idea of an artifact (as opposed to, say, a cow or a cabbage) is the idea of something produced by intentional human activity. So we can say that artifacts are not just causally but existentially dependent on minds, in the sense that it is metaphysically necessary for something to be an artifact that there be intentional human activities (cf. [Baker, 2007]). This distinction, however, still does not go far enough to distinguish the class of artifacts, for not just anything existentially dependent on intentional human activities counts as an artifact — for something to be garbage or pollution, it seems, it must be produced by humans (in the course of their intentional activities), but these do not — in the strict sense under discussion here — count as artifacts.4 (I leave to one side here the other use of the term ‘artifact’ to refer to unintended 4 In

Dipert’s terms, they are artificial but are not artifacts [1993, p. 33].

Artifacts in Metaphysics

195

byproducts of designs and the like as, e.g., spandrels are said to be artifacts of certain forms of ceiling design.) Unlike garbage and pollution, artifacts proper must be not merely the products of human activities, but the intended products of intentional human activities [Hilpinen, 1992, p. 60].5 Thus, as I have argued elsewhere, we can say that essentially artifactual concepts (as opposed to other concepts that might accidentally include artifacts in their extension), are those for which any member of the kind to be referred to must be the product of an intention to create that very sort of object [Thomasson, 2003; 2007b]. And we can begin to distinguish artifacts from other human products by way of these two features: artifacts are not merely causally but existentially dependent on human intentions; moreover, they are existentially dependent not merely on some intentional human activities or others, but rather on intentions to create that very kind of object. But the very idea that artifacts are existentially mind-dependent leads many metaphysicians to be suspicious of them, or to deny that they really exist or may be real parts of our world. Concerns about accepting the existence of minddependent objects arise from the feeling, as it is often put, that it seems too much like a ‘conjuring trick’ to suppose that our mental activities could bring new entities into existence, as if by thought or proclamation we could (shazam!) add to the inventory of being. This line of thought may be partially behind van Inwagen’s insistence that “Artisans do not create; at least not in the sense of causing things to exist” [1990, p. 127]: that is the thought that, although artisans may intentionally shove the stuff of the world about, surely it takes more than that to add to the ‘ontological inventory’ of the world.6 John Searle similarly notes qualms about accepting the existence of institutional facts, given a pervasive “sense that there is an element of magic, a conjuring trick, a sleight of hand in the creation of institutional facts out of brute facts” [1995, p. 45]. Clearly it would be a problem, and invoke ‘magical modes of creation’ if we held that thought or intentions alone could bring into existence rabbits in hats, or pink elephants on parade. And it is this kind of worry that lies behind the objections to artifacts on grounds of their (alleged) mind-dependence. But this problem does not in the least affect the claim that human intentions may be metaphysically necessary for the existence of artifacts, nor does the mind-dependence of artifacts give us reason to deny their existence.7 Why not? 5 Dipert makes a similar point in noting that artifactuality does not ‘consist in any present physical qualities of a thing’, but must in some sense appeal to its history (as having been intentionally modified) [1993, p. 15]. He also adds an additional condition for being an artifact proper: that the entity be intended to be recognized as an artifact [1993, p. 16]. 6 Van Inwagen’s central justification for this view relies on his argument that any acceptable answer to the special composition question requires us to reject artifacts. (I critically analyze this argument in my [2007, Chapter 7].) Nonetheless, he presents the view that artisans don’t really create as also intuitively plausible, and the contrary view that they do as committed to consequences about when objects are created that are ‘incredible’ [1990, p. 127-9]. 7 Baker [2007] also argues (on somewhat different grounds) that the differences between artifacts and natural objects (including the mind-dependence of the former) do not imply that artifacts are in any sense ‘ontologically inferior’.

196

Amie L. Thomasson

First, to say that artifacts are existentially dependent on human intentionality, of course, is not to say that human intentions, practices, beliefs or desires alone are sufficient to bring artifacts into existence. That would be a crazy view. While some (e.g., [Collingwood, 1938, pp. 134-5]) have defended the idea that works of art may be created merely ‘in the artist’s head’, even this implausible claim was presented to contrast works of art with the products of engineering and other ‘crafts’. For standard artifacts, especially the products of serious works of engineering like bridges, it is clear that their production requires not only an intention to make a bridge, but also some raw materials (steel, concrete), and needs to perform not only mental but also physical activities (of pouring concrete, joining spans, etc.). So, first, there is no risk on this view of considering human intentions and desires alone to be sufficient for the production of artifacts.8 Secondly and still more importantly, the trouble with the clearly problematic purported cases of ‘magical modes of creation’ (of rabbits out of hats, or pink elephants, etc.) is precisely that there are certain substantive criteria for the existence of rabbits or elephants that are not met merely by performing intentional acts, whether of imagination, invocation, desire, etc. Indeed in these cases, what any human believes, thinks, desires, or intends is quite irrelevant to the question of whether there really is a rabbit or elephant in a situation — human intentions are not even plausibly held to be a necessary condition for the existence of such natural entities. So it requires positing some form of ‘magic’ to imagine that human intentionality is in any way relevant to whether or not the application criteria for terms such as ‘rabbit’ or ‘elephant’ are fulfilled [cf. Thomasson 2007a, Chapter 9]. But the same is not the case for artifacts. If the analysis provided above is correct, then the very idea of an artifact is the idea of something that could not exist were it not for certain human intentions and practices. Thus the very idea of an artifact is the idea of something mind-dependent. If the existence of human intentions and practices of certain sorts are at least a necessary condition for a term like ‘paperweight’ (unlike ‘rabbit’) to apply, it requires invoking no magic to think that if a rock is not only physically modified (by some force or other) in certain ways, but also intentionally modified in order to serve as a paperweight, that makes the difference as to whether or not there is a paperweight in that situation. For, according to the standard use of the term ‘paperweight’, that is all it takes for a paperweight (as opposed to just a modified rock) to exist. There is a general lesson to be drawn here about the proper methods for handling existence questions.9 Metaphysicians have often proposed or assumed certain general, across-the-board, criteria for existence, e.g., that anything that exists must contribute novel causal powers [Merricks, 2001], or must be mind-independent in some sense [Lakoff, 1987, p. 184]. In each case, these are criteria that are suitable 8 By contrast, human intentionality may, as I have argued elsewhere [1999], be sufficient to produce things of other sorts, such as imaginary or hallucinatory entities — but that is an issue that need not be pressed here. 9 I argue on independent grounds in favor of this approach to existence questions in my [2007a].

Artifacts in Metaphysics

197

to what it takes to be a realist about things of a certain kind: e.g., it might be a legitimate requirement for there to be a basic physical entity that it contribute (otherwise unaccounted for) causal powers and be mind-independent. By considering too narrow a range of examples, these criteria are illegitimately generalized and adopted across the board as criteria for the existence of ‘anything whatsoever’. But the very idea of such entities as artifacts, institutional entities such as money or marriages, fictional characters, etc., is the idea of entities that — if there are any — are mind-dependent. We cannot use such general substantive criteria for what it takes to exist to argue against the existence of entities that (according to the very idea of such things) are not supposed to be distinctive causal contributors (e.g., beyond the causal contributions of their parts), mind-independent beings, or otherwise characterized by the preferred features. General prohibitions that would deny that anything mind-dependent can ‘really’ exist simply beg the question against the existence of artifacts (and other existentially mind-dependent objects), and do not provide any reasons for their rejection. Careful consideration of a broader range of examples suggests that there are different existence conditions for things of different kinds — with, e.g., mindindependence serving as a relevant criterion for rabbits but not for tables or fictional characters. This does not mean, however, that we cannot formulate a general understanding of how existence questions work. As I have argued elsewhere [2007a], we can gain a more neutral, and non-question-begging approach to existence questions by holding a purely formal criterion for existence: for any term ‘K’, things of kind K exist just in case the application conditions criterially associated with proper use of the term are met. On this model, existence questions (whether specific kind questions or general category questions) are to be answered in two steps: by combining an analysis of the basic application conditions for the term in question with empirical inquiry into whether or not these conditions are (ever) fulfilled. The relevant conditions may vary for different types of entity, so, e.g., the conditions under which there is a rabbit will be very different from those under which there is a paperweight, a dollar bill, or a story. Those conditions appeal to human intentionality in some cases, but not in others: so human intentionality of certain forms may be necessary and sufficient for the existence of a story or an imaginary object, merely necessary for the existence of an artifact, and completely irrelevant to the existence of a rabbit.10 But in each case, provided the relevant criteria are met, we have no reason to deny the existence of the relevant objects. In the case of artifacts, the application conditions for terms like ‘table’ are apparently are satisfied by the circumstances in my dining room and millions of others around the world. So if we combine the basic facts about meaning with obvious empirical facts, we can conclude that there are tables [Thomasson, 2001; 2007a, Chapter 9].

10 In my [1999] I lay out a system of categories, divided according to whether and in what sense(s) purported entities of various kinds would depend on spatio-temporal entities and on mental states.

198

Amie L. Thomasson

But this way of establishing existence claims is again often criticized as invoking a form of magic. Thus, e.g., Stephen Yablo writes of attempts to prove the existence of abstracta from a priori or empirically obvious premises, that such “arguments are put forward with a palpable sense of daring, as though a rabbit were about to be pulled out of a hat” [Yablo, 2000, p. 197]. “Our feeling of hocus-pocus about the ‘easy’ proof of numbers (etc.) is really very strong and has got to be respected” [Yablo, 2000, p. 199]. Such name-calling does not get us very far, however. Why think that there is anything ‘magical’ about the idea that existence questions are to be answered by way of determining what (according to the basic meanings of the terms in question) it would take for there to be an entity of the relevant kind (establishing the a priori premises), and then determining whether those conditions are fulfilled (adding the empirically obvious premises)? Instead of invoking ‘magic’, this seems like a perfectly flat-footed assessment of the truth conditions for existence claims uttered in English. As Michael Beaney notes in another context: “Rabbits can only be pulled out of hats if they are already there” [2007, p. 203]. The conditions it takes, for example, for there to be tables (which are perfectly well satisfied by there being stuff intentionally arranged into a certain firmness and shape, suitable for supporting plates and tea-cups, and for average sized humans to use for dining) are obviously satisfied — and given those conditions, it requires no magic, but only minimal observation and grasp of the application conditions for the English word ‘table’ to conclude that there are tables. This conclusion is of relevance to the study of artifacts, since it gives us reason to hold (against a multitude of recent metaphysical arguments) that there are artifacts. But it also suggests how considering the case of artifacts may be of much broader relevance to metaphysics. For, first, if there are artifacts, we cannot simply sidestep the various metaphysical problems of artifacts (to do with colocation, causal redundancy, vagueness, etc.), but must face these upfront. Second, it involves an approach to existence questions that is far different from the familiar Quinean paradigm, as well as from other approaches that take fulfillment of certain standard substantive conditions — be they mind-independence, causal efficacy, or whatever — as across the board conditions for things of any kind to exist.

3 REAL ESSENCES AND THE NATURES OF ARTIFACTUAL KINDS Quite a different sense of mind-dependence from the existential mind-dependence discussed above has recently come to play a central role in debates about artifacts: the sense in which the natures of artifacts (rather than their mere existence) might turn out to be dependent on human concepts (or thought, or language). But what is meant by this, and what is its relevance for debates about artifacts and broader issues in metaphysics?

Artifacts in Metaphysics

199

The basic idea comes from the direct reference approach to issues about reference and necessity developed by Kripke [1980] and Putnam [1975; 1977]. On that approach, natural kind terms are held to refer directly (via a causal and/or historical relation to a sample) to ‘that kind of thing’, so that we may first refer to a genuine kind in the world, and then go on to investigate its nature. On this view, the essential features of gold, say, are features we may discover only by empirical inquiry; they are not features that our very concept of gold establishes as necessary for anything to count as belonging to that kind. As a result, whatever beliefs competent speakers may have about the nature of gold — about what it takes for there to be gold, or when gold is created or destroyed, etc. — may all turn out to be false; and indeed there may be truths about the nature of gold even if everyone is ignorant of them. At least at first glance, however, artifacts do not seem to have real natures in this sense. Even if we may discover what it takes for something to be gold or to be a tiger, it seems to be our concepts or use of language that determine what counts as a pencil, a coffee table, or a split-level. As Stephen Schwartz put it “What makes something a pencil are superficial characteristics such as a certain form and function. . . They are analytically associated with the term ‘pencil’, not disclosed by scientific investigation” [1978, p. 571]. Considerations such as these led some (e.g., [Schwartz, 1978; 1980]) to hold that, however compelling direct reference theories and their consequences for a conception of ‘real’ essences’ might be for natural kinds, such theories cannot hold across the board as they, e.g., do not apply to artifactual kinds. Artifactual kind terms, Schwartz held, refer via associated descriptions that establish what it takes for there to be a pencil in a certain situation, for a pencil to survive or be destroyed, etc. As a result, such kinds have only a nominal essence established analytically by the criteria we associate with the term and discoverable merely by reflection on speakers’ use of the term ‘pencil’, not a real essence discovered via empirical investigation into the kind in question. Claims such as these generated a great deal of discussion about whether or not artifactual kind terms could be understood on a direct reference model, and correspondingly about whether artifactual kinds themselves have real or only nominal essences. Hilary Kornblith [1980] and James Nelson [1982] argued that, although it is true that many common artifactual kinds cannot be thought to have hidden underlying natures analogous to those of chemical kinds, they should nonetheless be thought of as having real natures of a different sort. As Kornblith puts it “What serves to determine the underlying nature of an object depends, in part, on what kind of object it is” [1980, pp. 111-112]. In the case of artifacts, Kornblith suggests, “At least for the most part, it seems that what makes two artifacts members of the same kind is that they perform the same function” [1980, p. 112]. If we allow that artifacts do have real natures in this sense then, Kornblith argues, we can preserve the idea that direct reference theories apply across the board: in each case, the term refers to whatever has the same nature as members of the relevant sample, where speakers may all be in ignorance or error about what this nature is.

200

Amie L. Thomasson

The insights behind Schwartz’s argument, Kornblith suggests, are simply based on the fact that the sort of nature in question may differ for artifactual kind terms versus (say) chemical kind terms. On closer examination, however, this line of response suggests a much deeper worry about whether or not a thoroughgoing direct reference approach is tenable at all. For suppose we accept Kornblith’s suggestion that artifactual kind terms and chemical kind terms both refer directly to ‘whatever shares the same essence as is instantiated by all or most members of this sample’, merely adding the amendment that artifactual kind terms will pick out whatever shares the same functional essence while chemical kind terms will pick out whatever shares the same chemical essence. Which sort of essence is in question will make all the difference as to which entities are unified as being of a kind, and which are not — according to whether they share a function with other members of the kind, a chemical structure, or both (or neither). But what determines whether or not a term, the reference of which we wish to ground, is to pick out entities that share the same function as these or entities that share the same chemical structure as these (or the same physical structure, shape, legal status, and so on — as presumably there may be kinds whose essences are unified in all these various ways)? It seems that here we must appeal to the intentions of speakers regarding whether the term they intend to ground the reference of is to be a term for an artifactual kind or chemical kind (or physical kind, institutional kind, etc.). In short, as soon as we consider artifactual kind terms alongside terms for (various sorts of) natural kinds, direct reference theories are confronted with a formidable problem in determining whether and if so to what sort of kind our terms refer, given that we may intend our kind terms to refer to a wide range of different kinds (physical, chemical, biological, artifactual, institutional, aesthetic. . . ).11 This is the notorious qua problem faced by direct reference theories. In light of this qua problem, many have argued (e.g., [Devitt and Sterelny, 1999; Thomasson, 2007a, Chapter 2]) that we have reason to give up pure direct reference theories in favor of a hybrid theory of reference. Such hybrid theories allow that our kind terms at least have a basic form of conceptual content specifying the category of kind to be referred to (e.g., artifactual, chemical, biological, institutional. . . ) by establishing what sorts of features are and are not relevant to unifying the kind (e.g., sameness of chemical structure versus sameness of function). If we accept such a hybrid theory of reference, then we must accept that speakers’ intentions establish at least what category of kind each general term refers to (if it refers at all), and what sorts of features (though perhaps not what particular features) will be essential or accidental to members of the kind (cf. [Thomasson, 11 Again, considering artifactual kind terms is not essential to uncovering the qua problem, since considering any range of different types of kind terms would do. Nonetheless, failure to notice the qua problem initially may have resulted from considering too narrow a range of examples, and considering problems of the reference of artifactual kind terms as well as natural kind terms is sufficient for raising the problem, and has in fact played a key role in critical discussion of direct reference theories.

Artifacts in Metaphysics

201

2007b]).12 So, e.g., on this view speakers’ intentions establish that ‘gold’ is to be a chemical kind term, so that gold, if there is any, has its chemical structure essentially. Speakers can nonetheless be entirely in ignorance or error about what the precise chemical structure of gold actually is. On this view, human concepts are relevant to the natures of the kinds they refer to, at least insofar as the intentions of speakers who ground the reference of terms determine what category of kind is to be referred to by establishing what sorts of features are and are not essential to belonging to the kind.13

4

DO ARTIFACTUAL KINDS HAVE MIND-INDEPENDENT NATURES?

The idea that artifactual kinds have some sort of nature, enabling us to refer to whatever shares that nature, leaves open the question: what sort of nature do artifactual kinds have? Given the above approach to reference, we can ask this by asking what sorts of feature those who ground the reference of artifactual kind terms treat as essential to unifying members of artifactual kinds (as opposed to chemical or biological kinds). Many have argued that, although artifacts are brought into existence by humans, they still may have natures that are every bit as mind-independent and open to ignorance, error, and discovery, as the natures of chemical and biological kinds are. As mentioned above, the chief suggestion along these lines has been that artifactual natures are at least largely distinguished by sameness of function rather than by sameness of chemical or genetic structure [Kornblith, 1980, p. 112]. Nelson similarly suggests that the essence of artifactual kinds may be a matter of a shared ‘structure and function’ [1982, p. 363]. The most sophisticated view along these lines has been recently developed by Crawford Elder [2004], who argues that at least a great many artifactual kinds are ‘copied kinds’ possessing discoverable ‘real’ natures, comprised of a cluster of properties centered on three shared features: 1) shape or qualitative makeup 2) proper function (members are produced by a mechanism that copies them from 12 Some have the intuition that speakers may also be wrong about, e.g., the very category of entity referred to (e.g., about whether kangaroos are animals). I discuss this objection in my [2007a, p. 48-53]. 13 This view is sometimes thought to express a form of modal conventionalism — the view that “The essential status of essential properties is mind-dependent” [Elder, 2004, p. 8]. And modal conventionalism, in turn, has been widely rejected since it is generally thought to lead to a form of anti-realism [Elder, 2004; Rea 2002]. As a result, debates about the natures of artifacts have played a central role in rekindling debates about the status of basic modal truths and essences generally, and thus again proven of wider relevance to fundamental issues in metaphysics. But since this debate about modality would take us far afield from the particular issues concerning artifacts, I will not discuss it further here. Instead, I refer the reader to the criticisms of modal conventionalism in [Elder, 2004] and [Rea, 2002], and to my extensive arguments that the view of modality that follows from the above hybrid view of reference does not commit us to any form of anti-realism [2007a, Chapter 3]. I now think that it is best to consider the allied view of modality a form of expressivism about modality — a view I develop and defend elsewhere [forthcoming].

202

Amie L. Thomasson

previous similarly shaped members as a consequence of the previous members’ performance of certain functions) and 3) historically proper placement (being where there are also tokens of other specific artifactual kinds — as, e.g., screwdrivers must be located with slotted screws [2004, p. 137]). According to Elder, we can see that these form a core essential nature by seeing that they cluster together in such a way that whenever two of them are found together, the third is there as well, and whenever one departs, one (or both) of the other two depart [2004, p. 139-140]. In most cases, Elder holds, many other properties cluster around this central core as well, making a richer, more interesting nature of the kind available for discovery. One of the acknowledged limitations of Elder’s ontology is that it does not seem able to include, e.g., merely decorative items for personal ornamentation (such as neckties and nose rings). These are not plausibly construed as members of copied kinds, since they don’t have a characteristic shape the replication of which is due to something previous members could do in virtue of that shape [2004, p. 158], so a great many apparent artifacts of these sorts must be treated as ‘mere projections’. Worse still is the situation for quite ‘general’ kinds of artifacts, such as tables and chairs. For, as Elder again acknowledges, “kinds as broad as chairs and tables can barely be said to have any one ‘shape’ or qualitative character in common at all” [2004, p. 149], and so shape cannot be part of a cluster of essential properties at the core of the kind’s discoverable nature. The problem is not limited to decorative and highly general kinds, however. Practical artifacts like paperclips and corkscrews ought to present something like the best-case scenarios for treating artifacts as copied kinds, but even these lack a consistent shape. A corkscrew may be a delicate t-shaped contraption of wood and metal, a giant handle with internal metal spiral attached by a vice to a bar, a flexible two-pronged puller, an air pump, etc. Paperclips may be elongated ovals, triangles, rectangles, or of any of a great many other shapes, colors, and materials. So again, we seem to lack a copied kind here if those must be characterized by the three clustering essential properties of proper function, shape, and historically proper placement. If there are no tables or chairs, paperclips or corkscrews (certainly no paperweights!), this does not seem to be the route to defending an ontology of artifacts. Elder responds by suggesting that what is preserved is more narrowly construed artifactual kinds (which may have the tightly clustering properties characteristic of copied kinds) such as the 1957-design Eames desk chair, or presumably the gem paperclip, or the ‘jumping jack’ corkscrew. In fact, though, on closer inspection, it seems that Elder cannot even defend the existence of these specific artifactual kinds as being copied kinds in his sense. For there to be a copied kind, recall, there must be three types of property: proper function, historically proper placement, and shape, such that these cluster together in virtue of laws of nature, so that whenever one goes, at least one of the other two goes as well. But the very presence of these widely varying design kinds for chairs, corkscrews, paperclips, and the like (which makes it implausible to defend the existence of very general

Artifacts in Metaphysics

203

artifactual kinds) also seems to demonstrate that shape regularly fails to cluster in this way with function and historically proper placement. The corkscrew, let us imagine, traditionally had the characteristic ‘shape’ of a rigid spiraling piece of metal (of perhaps 1-2 inches) attached perpendicularly to a solid cylinder of about a half-inch diameter and three inches length. Its historically proper placement would have been in contexts where cork-sealed wine bottles were in use, and its proper function obviously would be to open the bottle — with success at that function leading to the production of further members of the kind. But if corkscrews — or even T-corkscrews — form a genuine copied kind, then when we change the shape, one or both of the other core essential properties (historically proper placement or proper function) should go with it [2004, p. 139]. Suppose the shape is changed to a novel design: the two-pronged corkscrew, with two parallel straight flexible pieces of metal about an inch apart, attached to an empty oval on its side, and reproduce these so that they too may be said to have the proper function of opening wine bottles. Clearly this retains the historically proper placement and proper function of the original, despite the drastic change in shape. So contrary to our original supposition, there is not a copied kind characterized by those three essential features — it is simply not the case that “the absence of that third property would require the absence of one or another of the first two properties” [2004, p. 140]. This sort of problem is not unique to cases like corkscrews. A careful study of the history of artifactual kinds shows that function and historically proper placement vastly underdetermine the ‘shape’ of most of our familiar artifacts — function and placement may jointly set up an engineering problem, for which there is a wide variety of solutions that may be selected for all sorts of reasons (aesthetic, economic, social. . . ). Henry Petroski makes precisely this historical argument based on a number of case studies in The Evolution of Useful Things, where he argues extensively against the na¨ıve view that ‘form determines function’ in artifacts [1992, p. 20] — he would clearly also reject the idea that function and historically proper placement determine form. Even the simple paperclip has been made in an enormous variety of shapes and materials, which have changed without change of the function (clipping papers) or placement (where papers are used) of these artifacts. Having run through a history of paperclip designs, Petroski writes: There are still other styles of paper clips offered by other companies. . . and the variety reminds us not only of the nonuniqueness of form for this object but also of the fact that nontechnological (and subjective) factors such as aesthetics can account for the competitive dominance of one particular form over functionally superior forms [1992, p. 74]. Even the Eames desk chair will share proper function and placement with all other desk chairs, though these vary widely in shape, showing that shape may vary independently from the latter two, and so undermining the claim that there is a copied kind here at all — even if we try to make it specific. To say that, where the shape does so vary, we don’t have an Eames (1957) desk chair (since that very

204

Amie L. Thomasson

shape is an essential feature of that kind) is not a response available to Elder, since for him we must discover where clustering uniformities hold to discover what is essential, not (e.g.) infer that shape is essential from the fact that “Eames (1957) desk chair” is a design-specifying term.14 It is of course open to Elder to say that if these (general and specific) artifactual kinds cannot be identified with copied kinds (and thus shown to have real natures), so much the worse for artifactual kinds; we should speak instead of whatever copied kinds we may identify. But that would be to renounce all claim to preserving an ontology of artifacts; it would only give us a replacement ontology of biological and pseudo-biological kinds. Given these difficulties for the cluster view, it would be wise to reconsider the simpler view that the natures of artifactual kinds are just based in a shared function. But what notion of function is at issue here? It seems that it cannot be the actual functional capacities of the objects in question that are relevant. First, that would make every object a member of far too many supposedly artifactual kinds; second, it would rule out broken or malformed can openers (say) from being members of the relevant artifactual kind. So it seems instead that it must be something like ‘proper function’ in Millikan’s [1993] sense that is relevant — that is, the same sense of ‘function’ at issue in Elder’s proposal, though treated on its own, not as one of cluster of properties that jointly make up the core of an artifact’s nature. On Millikan’s account, a proper function may be either ‘direct’ or ‘derived’; a thing’s direct proper function (roughly) is what its ancestors did that led to their reproduction (and thus to this thing’s production); a thing has a derived proper function if it is the product of a prior device that has the relevant proper function and normally performs it by producing something like this. As items created by our own intentional needs, desires, and plans, artifacts would seem to have derived proper functions — so, as Millikan herself notes, artifacts have as derived proper functions ‘the functions intended for them by their makers’ [Millikan, 1999, p. 205]. Beth Preston [1998, p. 245-7] pursues the other account of artifactual proper function, arguing that artifacts may be understood to have direct proper functions in much the same sense as biological entities do. An artifact kind’s proper function, she argues, is whatever those artifacts actually do that leads to their reproduction. But a problem arises here since the reproduction of artifacts (unlike organisms) must be mediated by humans who believe the ancestral artifacts to have performed some useful function and intend the new ones to do the same. So, 14 A

further problem arises with treating the kind Eames (1957) Desk Chair as a copied kind: Clearly, according to ordinary use and application of the term, this is at least in part a historical kind. If something is to be an Eames (1957) desk chair, it must be modeled after the design by Charles Eames, created in his factory or its licensed heirs, etc. If I place an ad on Ebay to sell an Eames (1957) desk chair but am selling a chair of similar design unauthorizedly produced by a factory in China, I may be sued. So Elder’s model of copied kinds also cannot account for the historical element in many artifactual and common sense concepts — indeed Elder explicitly denies that such historical factors as origin may play any role in membership in real kinds [2004, pp. 155-6].

Artifacts in Metaphysics

205

for example, various superstitious and religious artifacts are reproduced because they are thought to increase rain, aid fertility, avert natural catastrophes — even if they do no such thing. Preston argues that in such cases the proper function of members of the artifactual kind in question is whatever they actually do — not what they are intended or (perhaps mistakenly) believed to do — that leads to their reproduction. We could interpret this in two ways: we could either find something the various religious objects actually do apart from altering people’s beliefs about what they do (the usual suggestion is something about aiding social bonding or providing psychological comfort) — but then we would end up classifying all manner of different religious artifacts, social objects, art objects, etc. together, contrary to the practices of classifying artifacts that are central to those social sciences like anthropology and archeology that actually deal with artifacts. Or we could treat their proper function more narrowly, saying that their function is to cause the belief that, e.g., rain will be brought (rather than to cause rain). But if it is the causal relation to our beliefs that is genuinely central to artifacts’ function (since it is these beliefs, not performance, that is relevant to their reproduction), then it seems by the same token we should hold that the proper function of can openers is not to open cans, but to cause beliefs that cans are open, and so on — a very counterintuitive result. In any case, even if we could successfully understand the function of artifacts along these lines, Maarten Franssen [2008] and Peter Kroes [2007] have recently pointed out that no such account of artifactual function can be used as the basis for distinguishing artifactual kinds, as any such attempts would be circular. Theories of (direct) proper function for artifacts aim to identify the proper function of artifacts of a given kind, e.g., of bug zappers or corkscrews. To determine the proper function of an individual (token) artifact, we must ask what its ancestors (or predecessors) did that led to their reproduction. What are an artifact’s ancestors, however? They must be previous members of the same kind (e.g., previous bug zappers or corkscrews) — but this presupposes a categorization of entities into artifactual kinds, a categorization that cannot itself be grounded on sameness of function.15 For all of these reasons, it seems better to hold that the notion of artifactual function that is relevant to our standard ways of delineating at least many artifactual kinds is in fact a notion of intended function (or Millikan’s derived proper function), not (direct) proper function in Preston’s sense. For we can discuss the intended function (unlike the proper function) of an individual (token) artifact without appeal to the kind it belongs to, and can similarly avoid the problem of misclassifying artifacts that fail to perform their intended function, or characterizing all artifacts as having the function merely of producing certain kinds of beliefs in us.

15 This is not in itself a criticism of the way in which proper function theories for artifacts are drawn out (e.g., by Preston) — only of the further idea that proper functions so understood could be used as the basis for unifying and distinguishing artifactual kinds.

206

Amie L. Thomasson

Paul Bloom’s [1996] discussion, however, suggests that even intended function would be too narrow to serve as the feature that unifies artifactual kinds, since there may be members of artifactual kinds that are not intended to fulfill the characteristic function of the kind (say, if they are intended only ‘for show’), and since some artifactual kinds such as art kinds may fail to share any common function (with some intended for decoration, others to provoke political protest, others to make money or add to the glory of the artist, others to express emotion. . . ). So while the appeal to intended (rather than performance) function may have been on the right track, sticking to functions (however understood) may be too narrow, and we may do better to pursue an account of artifactual kinds based more broadly on an appeal to intentions. Along these lines, Bloom suggests that artifact kinds are unified in terms of being products of the same sort of intention: namely, to create something of the same kind as current and previous members of the kind [1996, p. 10].16 I have argued elsewhere [2003; 2007b] that being the product of the same sorts of intentions is indeed relevant to unifying artifactual kinds, but that these intentions must be understood substantively — not just as intentions to make ‘one of these’ (referring transparently to members of a sample), but rather as intentions to make something possessing certain features considered relevant to kind membership. More precisely, I have argued, for any essentially artifactual kind K, something is a K only if it is the product of a largely successful intention to make something of kind K, where such intentions must involve a substantive conception of what sorts of features are K-relevant and this conception largely matches that of some group of prior makers of Ks (if there are any). So, for example, what is crucial to being of the artifactual kind knife is to be the product of intentions to make a knife, where these involve a substantive conception of what properties are relevant to being a knife (say, having a sharp blade and handle, along with capability of cutting standard foods) and where the intention to make something with these features is largely successfully realized. On this model, human concepts and intentions are relevant to the natures of artifactual kinds on two levels. First (as I have argued in §2), speakers’ intentions establish what category of kind is to be referred to by establishing what sorts of features are relevant to unifying artifactual kinds in general (and thus are the sorts of features that will be essential to those kinds). Second, since the sorts of features appealed to involve the intentions of those who make the objects in question, makers’ concepts of what specific features are relevant to belonging to the kind establish which particular intended features are relevant to membership in a given artifactual kind and thus may be said to form the particular nature of the kind [cf. Thomasson 2007b, p. 63]. On this view, then, the particular natures of artifactual kinds are constituted by makers’ intentions regarding what particular features are relevant to kind membership, thus marking an important difference between artifactual kinds and natural kinds. 16 For criticisms of Bloom’s account of how we categorize objects into artifactual kinds, see [Malt and Johnson, 1998].

Artifacts in Metaphysics

207

This view leaves it open that the most important intended features for most artifactual kinds may be functional features — capturing what seemed right about functional accounts of artifactual natures. But it also leaves it open that other intended features may (also or instead) be relevant to membership in an artifactual kind, including intended structure/design, intended use-practices,17 intended reception, etc. In fact, this account leaves open the possibility of distinguishing different sorts of artifacts, according to which sorts of intended features are relevant to kind membership. Thus, e.g., Peter Kroes [2007] suggests that intended function and structure are what is centrally relevant to technical artifact kinds; Jerrold Levinson [2007] has argued that intended reception (i.e. being intended to be regarded or treated as earlier works of art were correctly regarded or treated) is what is centrally relevant to membership in the kind ‘work of art’, etc. Accepting that the natures of artifactual kinds are constituted by human intentions in this way is highly controversial, however. For it is often held that possessing a nature that is entirely independent of human concepts, language, etc., which is open to genuine discovery and about which everyone may turn out to be ignorant or in error, is a central criterion for treating kinds as real or genuine parts of our world [Elder 1989; Lakoff 1987]. So, for example, George Lakoff assumes that any metaphysical dependence on human intentionality vitiates a purported entity’s claim to reality, taking it to be a central feature of objectivism that “No true fact can depend upon people’s believing it, on their knowledge of it, on their conceptualization of it, or on any other aspect of cognition. Existence cannot depend in any way in on human cognition” [1987, p. 164]. Crawford Elder similarly writes, “I shall myself construe realism as a denial of epistemic privilege” [1989, p. 440], namely that: . . . for any component of the world and any set of beliefs about that component, the mere facts that those beliefs are (i) about that component and (ii) are held by the particular believers, by whom they are held, never by themselves entail that that set of beliefs is free from massive error. [1989, p. 441] As a result of this view, he holds that: Realists. . . must either argue that members of a given culture could in fact hold shared beliefs about their own CGKs [culturally generated kinds] that were massively mistaken, or else maintain that CGKs are not genuine components of the world. [1989, p. 427] Indeed it was this line of reasoning that prompted Elder [2004] to defend the existence of (at least many kinds of) artifacts by arguing that we can understand artifactual kinds as possessing mind-independent natures — discoverable in much the same way as the natures of natural kinds are. 17 Pieter Vermaas and Wybo Houkes [2006] have brought out the importance of what they call ‘use plans’ to the nature of artifacts.

208

Amie L. Thomasson

But if we accept the idea that the existence of members of any kind requires that that kind have a completely mind-independent nature regarding which everyone may be totally ignorant or in error, then we would be forced to deny the existence of artifacts if we can’t defend the view that they have mind-independent natures. Given the difficulties that have surfaced above for such views, we seem to be left with the options of giving up an ontology of artifacts and artifactual kinds or giving up the idea that possessing discoverable mind-independent natures is the central criterion for ‘really’ existing. I prefer to take the latter route, holding that although the natures of artifactual kinds turn out to be mind-dependent in ways the natures of natural kinds do not, this should not lead us to deny that artifacts and artifactual kinds are real parts of our world: we should reject this criterion for realism, not the objects that apparently fail to meet it. And indeed the discussion from §2 above gives us principled reason for taking this route. For proposals for ‘across the board’ conditions for ‘really’ existing, whether these include existential mind-independence, or, as in this case, possessing purely mind-independent discoverable natures, merely beg the question against entities like artifacts which, if they exist at all, must both depend existentially and for their natures on human intentionality. Instead of adopting a substantive across the board criterion for ‘real existence’, as I argued above, we should accept a purely formal criterion for existence: for any term ‘K’, things of kind K exist just in case the application conditions criterially associated with proper use of the term are met. If we follow this criterion, we can of course accept that there are corkscrews, tables, chairs, and other artifacts — even if the substantive features relevant to being a member of these artifactual kinds are established by the beliefs and intentions of makers about what substantive features are relevant to kind membership. 5

THE NATURAL SCIENCES VERSUS THE SOCIAL AND HUMAN SCIENCES

Although, on my view, the fact that the natures of artifactual kinds are minddependent in the above sense does not make a difference to whether or not we should accept that these kinds (and their members) really exist, it does make a substantial difference to our epistemic relation to artifactual kinds, and for understanding differences in method between the natural sciences on the one hand, and the social and human sciences on the other hand. Epistemically, it follows from the proposed view of artifactual natures that at least some humans have a much closer epistemic relation to artifactual kinds than anyone has to natural kinds. For as I have argued [2003], for any essentially artifactual kind K, something is a K only if it is the product of a largely successful intention to make something of kind K, where such intentions must involve a substantive conception of what sorts of features are K-relevant and this conception largely matches that of some group of prior makers of Ks (if there are any). Since the substantive features that are relevant to being a member of kind K are

Artifacts in Metaphysics

209

established by makers’ conceptions of what sorts of features are K-relevant, the makers of the kind of artifact in question cannot all be massively mistaken about what it takes for there to be a K; their views on the matter are constitituve of what it takes — at least for Ks of that time and tradition. (Gradual changes over time in what features are K-relevant are possible, given that the requirement is only that the substantive conception largely match that of some prior K-makers, if any there be.) By contrast, no one has any protection from error in their beliefs about what specific chemical or biological structures are relevant to belonging to a natural kind like gold or tiger.18 This in turn has interesting consequences regarding the methodologies suitable for the natural sciences as compared with the social and human sciences. At least where questions about the nature of a kind are concerned, those scientists studying natural kinds can do best at finding out, e.g., the specific chemical nature of gold or biological nature of tigers by directly studying gold or tigers.19 By contrast, an archeologist attempting to determine the nature of a kind of artifact she has apparently discovered must do more than investigate the objects unearthed: she must attempt to determine what they were made for — what the objects’ makers would have considered the relevant features determining whether or not the objects belonged to the relevant kind, and so on. In short, she must also try to unearth — by study not only of the artifacts in their historical setting, but also using background understanding of the culture and peoples in question — what features were considered to be essential to membership in that artifactual kind. Discussion of methodology in the social sciences has often focused on the question of whether or not some form of empathy or understanding (Verstehen) must play a role in the social sciences that is not needed in the natural sciences. Humanists (e.g., [Stein, 1917; Dray, 1957/1994; Geertz, 1983/1994]) have argued that the social sciences do require some way of grasping others’ beliefs, intentions, or way of seeing the world, while naturalists (e.g., [Hempel, 1942/1994]) have denied this. The above result provides at least one sense in which it is true that something like empathy must play a central role in the methodology of the social and human sciences that is unnecessary to the natural sciences, since even identifying the natures of the artifactual kinds that play a role in such sciences as archeology and art history requires understanding the intentions and concepts of others. Similarly, it seems that something like empathy must play a role in the engineering sciences, insofar as designers intend to design something of an extant artifactual kind — or, for that matter, hope to create something of a new kind designed to fulfill certain practical purposes of users other than the maker.

18 For more detailed discussion of the sorts of epistemic privilege that do and do not follow, see my [2003]. 19 Given the above hybrid view of reference, however, the same does not go for discovering what category of kind (e.g., chemical, biological) ‘gold’ and ‘tiger’ are to refer to, should they refer at all. That much must still go by way of a form of conceptual analysis.

210

Amie L. Thomasson

6

CONCLUSION

It is undeniable that issues concerning artifacts have been largely neglected and have played only a peripheral role in mainstream analytic metaphysics over the past century. And it is only recently, especially in discussions of theories of direct reference, modality, and especially in arguments for various forms of eliminativism about ordinary objects, that discussions of artifacts have occasionally come to the forefront. I have been at pains to argue, however, that this disinterest in artifacts, and the ‘one-sided diet of examples’ (to borrow a phrase of Austin’s) that has characterized a metaphysics focused on the natural sciences, has been problematic not just for our understanding of artifacts themselves, but also for much broader issues in philosophy. For ignoring the particular issues that arise with artifacts has led to a great many problems in metaphysics, including thinking that we can sidestep central metaphysical problems to do with causal redundancy, colocation, or vagueness by denying the existence of artifacts, and adopting substantive criteria for ‘real existence’ that (however suitable they may be for the postulates of the natural sciences) are not appropriate as across the board criteria. Even beyond metaphysics itself, careful consideration of artifactual kind terms raises important problems for direct reference theories as well as for the corresponding conceptions of real kinds, and failure to note the differences in the types of nature characterizing things of different kinds has inhibited understanding why we might require different methods in pursuing the natural sciences versus the social and human sciences. More work is needed on all these issues — in fact, given the scanty attention that has been paid to artifacts, debates about most of these topics remain in their infancy. But there is hope, nonetheless, that at least by pointing out the farreaching significance of issues concerning artifacts, their natures, and our terms for them, metaphysicians may once again turn more explicitly to consider the problems presented by artifacts, to the benefit not only of those interested in those objects, but also to the benefit of metaphysics. ACKNOWLEDGEMENTS A prior version of this paper was presented in the session ‘Technical Artefacts as Ordinary Objects’ (American Philosophical Association Central Division Meetings, April 2007). Thanks to the other participants: Lynne Baker, Crawford Elder, Wybo Houkes, Peter Kroes, and Pieter Vermaas, as well as to those in the audience, for a very helpful and interesting discussion.

Artifacts in Metaphysics

211

BIBLIOGRAPHY [Baker, 2007] L. R. Baker. The Metaphysics of Everyday Life: An Essay in Practical Realism. Cambridge University Press, 2007. [Beaney, 2007] M. Beaney. Conceptions of Analysis in the Early Analytic and Phenomenological Traditions: Some Comparisons and Relationships. In The Analytic Turn: Analysis in Early Analytic Philosophy and Phenomenology, M. Beaney, ed. Routledge, 2007. [Bloom, 1996] P. Bloom. Intention, History and Artifact Concepts, Cognition 60, 1-29, 1996. [Collingwood, 1938] R. G. Collingwood. The Principles of Art, Oxford University Press, 1938. [Dipert, 1993] R. Dipert. Artifacts, Artworks, and Agency, Temple University Press, 1993. [Dray, 1994] W. Dray. The Rationale of Action. From Laws and Explanation in History, Oxford University Press, 1957. Reprinted M. Martin and L. C. McIntyre, eds. pp. 173-180, 1994. [Elder, 1989] C. Elder. Realism, Naturalism and Culturally Generated Kinds, Philosophical Quarterly 39, 425-444, 1989. [Elder, 2004] C. Elder. Real Natures and Familiar Objects, MIT Press, 2004. [Franssen, 2008] M. Franssen. The inherent normativity of functions in biology and technology. In Functions in biological and artificial worlds: comparative philosophical perspectives, U. Krohs and P. Kroes, eds. MIT Press, 2008. [Geertz, 1994] C. Geertz. Thick Description: Toward an Interpretive Theory of Culture. From The Interpretation of Cultures. Basic Books, 1983. Reprinted in M. Martin and L.C. McIntyre, eds., 1994. [Heller, 1990] M. Heller. The Ontology of Physical Objects, Cambridge University Press, 1990. [Hempel, 1994] C. G Hempel. The Function of General Laws in History, Journal of Philosophy 39, 35-48, 1942. Reprinted in M. Martin and L. C. McIntyre, eds. pp. 43-54, 1994. [Hilpinen, 1992] R. Hilpinen. On Artifacts and Works of Art, Theoria 58, 58-82, 1992. [Horgan and Potrc, 2000] T. Horgan and M. Potrc. Objectivism and Indirect Correspondence, Facta Philosophica 2, 249-270, 2000. [Kornblith, 1980] H. Kornblith. Referring to Artifacts, The Philosophical Review 89, 109-114, 1980. [Kripke, 1980] S. Kripke. Naming and Necessity, Blackwell, 1980. [Kroes, 2007] P. Kroes. “Theories of Technical Functions and Artefact Kinds”. Paper presented at Central Division APA session: Technical artifacts and ordinary objects: metaphysics meets the philosophy of technology, Chicago, April 20, 2007. [Lakoff, 1987] G. Lakoff. Women, Fire and Dangerous Things, University of Chicago Press, 1987. [Levinson, 2007] J. Levinson. Artworks as Artifacts. In Creations of the Mind: Theories of Artifacts and their Representation, E. Margolis and S. Laurence, eds. Oxford University Press, 2007. [Malt and Johnson, 1988] B. C. Malt and E. C. Johnson. Artifact category membership and the intentional-historical theory, Cognition 66/1, 79-85, 1998. [Martin and McIntyre, 1994] M. Martin and L. C. McIntyre, eds. Readings in the Philosophy of Social Science. MIT Press, 1994. [Merricks, 2001] T. Merricks. Objects and Persons, Oxford University Press, 2001. [Millikan, 1993] R. Millikan. White Queen Psychology and Other Essays for Alice, MIT Press, 1993. [Millikan, 1999] R. Millikan Wings, Spoons, Pills and Quills: A Pluralist Theory of Function, The Journal of Philosophy 96/4, 191-206, 1999. [Nelson, 1982] J. Nelson. Schwartz on Reference, Southern Journal of Philosophy 20. 359-65 1982. [Preston, 1998] B. Preston. Why is a Wing like a Spoon? A Pluralist Theory of Function, The Journal of Philosophy 95/5, 215-254, 1998. [Petroski, 1992] H. Petroski. The Evolution of Useful Things, Vintage Books, 1992. [Rea, 2002] M. Rea. World Without Design: The Ontological Consequences of Naturalism, Oxford University Press, 2002. [Schwartz, 1978] S. Schwartz. Putnam on Artifacts. Philosophical Review 87/4, 566-574, 1978. [Schwartz, 1980] S. Schwartz. Natural Kinds and Nominal Kinds, Mind 89, 182-95, 1980. [Searle, 1995] J. Searle. The Construction of Social Reality, The Free Press, 1995.

212

Amie L. Thomasson

[Stein, 19171989] E. Stein. On the Problem of Empathy. 3rd Revised Edition, ICS Publications, 1917/1989. [Thomasson, 2001a] A. L. Thomasson. Geographic Objects and the Science of Geography, Topoi 20, 149-159, 2001. [Thomasson, 2001b] A. L. Thomasson. Fiction and Metaphysics, Cambridge University Press, 2001. [Thomasson, 2003] A. L. Thomasson. Realism and Human Kinds, Philosophy and Phenomenological Research 67/3, 580-609, 2003. [Thomasson, 2007a] A. L. Thomasson. Ordinary Objects, Oxford University Press, 2007. [Thomasson, 2007b] A. L. Thomasson. Artifacts and Human Concepts. In Creations of the Mind: Theories of Artifacts and their Representation, E. Margolis and S. Laurence, eds. Oxford University Press, 2007. [Thomasson, forthcoming] A. L. Thomasson. Modal Expressivism and the Methods of Metaphysics, Philosophical Topics, Forthcoming. [Van Inwagen, 1990] P. Van Inwagen. Material Beings, Cornell University Press, 1990. [Vermaas and Houkes, 2006] P. E. Vermaas and W. Houkes. Technical functions: a drawbridge between the intentional and structural natures of technical artifacts, Studies in the History and Philosophy of Science 37, 5-18, 2006. [Yablo, 2000] S. Yablo. Apriority and Existence. In New Essays on the A Priori, P. Boghossian and C. Peacocke, eds. Oxford University Press, 2000.

PHILOSOPHICAL THEORIES OF ARTIFACT FUNCTION Beth Preston

1

INTRODUCTION

Function theory is a growth industry in philosophy of biology, but artifact function has received comparatively little attention. Perhaps the most important factor contributing to this neglect is the lack — both historical and contemporary — of a philosophical literature dedicated to the study of artifacts.1 In short, there is no philosophy of material culture in general, and thus no breeding ground for theories of artifact function. Moreover, the only two areas of philosophical research that do study artifacts — aesthetics and philosophy of technology — have been largely preoccupied with questions about the aesthetic, ethical and social value of specific kinds of artifacts (art works and sophisticated, industrial technologies, respectively), and have not addressed the basic ontology and epistemology of artifacts generally. Thus artifact function has not emerged as a major topic of discussion in either of these areas, despite their interest in specific aspects of material culture. A second important factor contributing to the neglect of artifact function is the widespread perception that it is unproblematic in comparison to biological function.2 The background intuition here is that because of the uncontroversial existence of intelligent makers in the artifact case, what the function of an artifact is and how that function is established can be easily traced to the purposes of the human agents who made and/or used it. Thus if there are any theoretical difficulties, they must lie in the analysis of intention and intentional action, not in the analysis of artifacts. On such a view, anything that needs to be said about artifacts can and should be said within the confines of philosophy of mind and action. A separate philosophical account of material culture is neither necessary nor desirable. Furthermore, many authors discount the difficulty of theorizing about artifact function by making the artifact account a mere appendage of the 1 In this same vein, Randall Dipert discusses “the (non-)history of theories of artifacts” [1993, p. 7]. On his view, this (non-)history can be explained by an overwhelming focus on the “passive” aspects of cognition — the environment to mind direction — to the virtual exclusion of the “active” aspects — the mind to environment direction [1993, pp. 7-11]. 2 Paul Griffiths [1993] complains about this perception and makes an effort to show that there are some special difficulties for theories of artifact function. Mark Perlman [2004] acknowledges this perception as well, discusses some of the possible special difficulties for a theory of artifact function, and concludes that none of them is insurmountable.

Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

214

Beth Preston

biological account, usually along the lines of ‘What I have just said about biological function goes for artifact function as well, mutatis mutandis.’ Not surprisingly, then, many philosophical theories of artifact function consist of scattered remarks, or at best a paragraph or two as an afterthought to a thorough discussion of biological function. Because of the sketchy and fragmentary nature of existing accounts, it is at present difficult to see what the central issues would be for a full-fledged theory of artifact function. And taxonomizing existing theories of artifact function is difficult for the same reason. In Section 2 I will begin to deal with the first difficulty by describing the main, characteristic phenomena of material culture for which a full-fledged theory of artifact function should account in the hope that the central issues for such a theory will become manifest by the end of this essay. In Section 3 I will try to deal with the second difficulty by describing existing theories of artifact function in accordance with an organizing theme — the role of human intentions in the establishment, maintenance and change of artifact functions. The result will not be a taxonomy, properly speaking, but it will make manifest one important network of relationships among these theories. Finally, in Section 4 I will assess the theories described in Section 3 with regard to their progress explaining the characteristic phenomena described in Section 2. This will complete the task begun there of ferreting out some of the central issues for full-fledged theories of artifact function, as well as providing some indication of the current state of the discussion with regard to these issues. 2

PHENOMENA OF ARTIFACT FUNCTION

In this section I sketch some characteristic phenomena involved in artifact function. I have not focused particularly on characteristics that are peculiar to material culture, as opposed to biology. Some material culture phenomena pretty clearly have biological analogues; others equally clearly do not. I do not claim that this list exhausts the characteristics of material culture a theory of artifact function could profitably take into account. Nor do I claim that the emphases I have chosen within this list are the only ones possible. What I do claim is that the phenomena described here constitute salient characteristics of material culture; that they make manifest central features of artifact function; and that a fullfledged theory of artifact function ought minimally to account for at least these features. My purpose here is thus to establish an initial benchmark for what would count as a full-fledged, well-integrated theory of artifact function, as opposed to the incipient, scattered fragments of such a theory that are available in the literature at present.3 Artifact functions are multiply realizable. There may be some rare functions that can only be realized by a device with a specific form made out of a specific 3 In a similar vein, Vermaas and Houkes [2003] propose a set of four desiderata they believe an adequate account of artifact function should satisfy. Lewens [2004, pp. 88-89] might be interpreted as proposing something along these lines as well.

Philosophical Theories of Artifact Function

215

material — an object used for a religious ritual with very specific requirements, perhaps.4 But typically artifact functions are realizable in a variety of materials and/or forms, provided some general constraints are satisfied. Take spoons, for instance. They do have to be made out of a relatively rigid material, and have a form that includes a handle attached to a bowl. But other than that, form and material are very variable. Historically, in addition to just about any metal you care to mention, spoons have been made from wood, shell, horn, bone, pottery, porcelain, plastic, silicon, and so on. Similarly, the bowl may be oval, round, pointed or even fluted; and the handle may be long or short, flat, bowed, or looped, decorated or plain. Moreover, the multiple realizability of function means that material cultures often contain several types of things all designed to serve basically the same function. Brooms, mops, vacuum cleaners and “swiffers” are all good for cleaning uncarpeted floors, for instance. And of course the same function is often served by different types of things in different material cultures. The function of conveying solid food from plate to mouth is carried out in Japan with chopsticks and in the United States with a fork, for example. Artifacts are multiply utilizable. Some artifacts are designed to serve only one function, but more typically they are designed to serve several, often simultaneously. Uniforms, for instance, serve the functions of ordinary clothing — keeping the wearer warm, dry and modest — but they also serve identification functions with regard to telling friends apart from enemies, displaying rank, and so on. As Michael Schiffer [1992] points out, and the example of uniforms illustrates, multiple functionality is often a matter of serving social or communicative functions in addition to purely technical ones. A crucial aspect of multiple utilizability is that artifacts are usable — and very often used — for functions they were not designed to serve. For example, an umbrella designed specifically to ward off rain, can also be used as a sunshade, as a weapon, as a lampshade (hung upside down underneath a ceiling light fixture), as a cane, or as a handy extension of the arm for reaching and retrieving things (from the top of the bookcase or the other side of the park railing, for instance). These alternate functions sometimes become standardized or routine, either with regard to a particular artifact or to a type of artifact. A person might set aside a particular spoon to be used to transplant seedlings, for instance; and spoons in general are routinely used to open cocoa tins.5 This aspect of multiple utilizability is, in a sense, the other side of the coin of multiple realizability. Since a given artifact function is realizable in a range of forms and materials, it is no wonder that it can also often be performed by other artifacts originally designed to fulfill different functions. There is an important connection here to recycling, a process in which one type of artifact is used as raw material for the production of another type of artifact with a different function. Like multiple utilizability, recycling is important to an account of artifact function because it bears on the issue of change of function 4 Thanks

to Wybo Houkes for suggesting this example. [Preston, 1998] for a more detailed discussion of routine and standardized alternate functions. 5 See

216

Beth Preston

over time. Most of the examples of multiple utilizability above leave the artifact in a condition to perform its original function(s). But sometimes modifications are required. In order to use an umbrella as a lampshade, for instance, you might need to shorten the handle to the point where using the artifact as an umbrella again would no longer be feasible. A closely related phenomenon is the reuse for a new function of items that can no longer be repaired, as when old clothes are cut up for workshop rags or used as pet bedding. Further along this continuum are cases where the structure is so radically modified that the original item is no longer clearly identifiable, as when old clothes are cut into strips and braided for chair mats or rugs. And finally there is recycling proper — breaking the structure down so thoroughly that it is returned to the raw material state and can be used to produce new items, sometimes of a very different type, as when old clothes are processed to make paper.6 Multiple utilizability and recycling appear to be located on a continuum, so I am not sure whether what we have here is one phenomenon or two. But for clarity and convenience, I will reserve ‘multiple utilizability’ for cases where the artifact is still usable for its original function(s), and ‘recycling’ for cases where there is sufficient modification that the original function(s) can no longer be performed. Another central feature is that the functional structures of material culture are not merely produced but reproduced. That is, they have standardized forms and uses that are relatively stable over years, generations or even millenia. That they are only relatively stable bears emphasizing. As in biology, the process of reproduction in material culture ordinarily involves variation, whether intentional or accidental. And as in biology it is this variation that ultimately accounts for cultural innovation, which is overwhelmingly a matter of incremental changes and extensions of existing cultural forms and functions rather than a matter of radical novelty. For example, electronic computers are only the latest entry in a long historical line of calculating machines, and could have arisen only on the basis of such a line. Similarly, when a material culture does encounter radical novelty, it is usually an import from another material culture rather than a local invention. The point is not that radical novelty does not occur — although I must confess I have been unable to come up with an example that qualifies as absolutely novel — but that it is vanishingly rare. On the other hand, it is important to note that in material culture variation and innovation are sometimes promoted in ways that are not possible in biology. But the bottom line here is that innovation is not separate from reproduction, but rather integral to it. In other words, reproduction with variation is simultaneously the process by which material culture changes over time and the process by which it is preserved relatively unchanged over time — these are two sides of the same coin. In order to indicate the dual nature of this phenomenon we will call it reproduction with variation. It is integral to a theory of artifact function because what are reproduced and varied are not mere physical 6 It is important to remember here that what we call “raw” materials are usually not raw, in the sense that naturally occurring materials like clay or stones are raw, but are already artifacts. Paper and plastic are raw materials only in this latter, already artifactual, sense.

Philosophical Theories of Artifact Function

217

structures to which functions are later attached, but always already functional structures. A rarely remarked group of phenomena closely related to reproduction with variation involve processes like maintenance, repair and rebuilding. These processes are integral to a theory of artifact function because the point of the maintenance, repair or rebuilding is to preserve or restore functionality. Most items of material culture must be maintained in order to continue to perform their functions effectively. Sometimes the need for maintenance is a direct result of the exercise of the function — e.g., spoons become encrusted with food during use and must be washed. In other cases the need for maintenance results from natural deterioration of the material of which the item is made — e.g., silver spoons need to be polished from time to time even if they are not used. Most items of material culture are also subject to damage that calls for repair. Sometimes this is the result of normal functioning, as when socks get holes in the toes or heels; and sometimes it is the result of misuse or accidents, as when a spoon is bent by being used to pry up floor tiles. Items of material culture more complex than socks or spoons sometimes undergo a radical type of repair, usually called rebuilding or remodeling, in which an item is disassembled, its parts refurbished or changed out for new ones, and then reassembled. Maintenance, repair and rebuilding are all intended to return the item to a condition in which it can perform its original function(s), so these processes are the complement of recycling, which has the opposite effect. Another important phenomenon associated with the functional structures of material culture is malfunction. This is something a theory of function needs to account for just as theories of representation need to account for misrepresentation. A malfunction, it should be noted, is not something going wrong with the function itself, but with the structure that subserves the function. This may be due to a variety of causes, among them misuse, manufacturing defects, poor design, accident, or simply wearing out as a result of normal use. Thus an umbrella malfunctions if a rib breaks due to long-term normal use, or if the wrong fabric was used and it is not waterproof, or if its handle is cut off so it can be used as a lampshade, and so on. An important limiting case of malfunction occurs when an artifact is well on the way to losing a particular function altogether — something like becoming vestigial in the biological realm. For example, souvenir spoons are often not at all suitable for use as spoons. They are maintained in the material culture largely because they have acquired a separate, social function as souvenirs. Finally, there is a related but little remarked phenomenon I call phantom function. This occurs when a type of artifact is regularly reproduced to serve a specific function, but no exemplar of it has ever been structurally capable of performing that function, or, in the nature of things, ever will be. Examples of phantom functions are tendentious, because there are often people who believe the artifacts in question do perform the functions they are alleged to perform. But here are a couple that should be relatively untendentious with regard to the audience for this essay. Laetrile is a drug that is produced and used in Mexico for treating cancer. It is not produced or used in the United States (or in Canada, for that matter) be-

218

Beth Preston

cause it has not been approved by the U.S. Federal Drug Administration (FDA), which believes that there is no scientific evidence of its efficacy. Assuming the FDA is right about this, curing cancer is a phantom function of laetrile. Similarly, Europe, North America and the Middle East are rife with amulets for protection against the “evil eye,” the alleged capacity of some persons to cause harm simply by looking at someone or some thing. These amulets come in the shape of “all seeing” eyes, hands with an eye on the palm, horseshoes with eyes on them, and so on.7 On the assumption that there is no evil eye, these amulets do not protect against it. In short, phantom functions appear to be normal functions in terms of the design, reproduction and dissemination of their artifact bearers, but not in terms of the performance of the function, which has never actually occurred.8 These phenomena — multiple realizability, multiple utilizability, recycling, reproduction with variation, malfunction and phantom function — can serve as an initial benchmark for a full-fledged theory of artifact function. First of all, any full-fledged theory ought to have something to say about most or all of them insofar as they all bear in one way or another on recognized issues in function theory. Second, any full-fledged theory of artifact function ought to recognize the complexities of these phenomena and the concomitant difficulties in accounting adequately for the features of artifact function they make manifest. In the next section, I will review current theories of artifact function, and then in Section 4 I will employ these benchmark phenomena to assess the state of the art and recent progress made by theories of artifact function.

3 THEORIES OF ARTIFACT FUNCTION The overwhelmingly most important theme to have emerged to date in the literature concerns the role of human intentions in the establishment, maintenance and change of artifact functions. I will use this as the organizing theme in this section. It is, I think, admitted on all hands that human purposes and intentions have something to do with the functions of artifacts. But a fairly common view is that artifact functions are directly and exhaustively determined by individual and/or collective human intentions. Such views have been termed “intentionalist,” and have been contrasted with “reproduction” views.9 Reproduction views appeal to 7 There is an archive of these amulets, historical and contemporary, at http://www.luckymojo.com/. Oddly enough, the “all seeing” eye shows up on the great seal of the United States masquerading as the “eye of providence.” See http://www.greatseal.com/, or just take a look at the back of a U.S. one dollar bill. 8 It has been suggested to me that phantom function is just a very extreme type of malfunction [Wybo Houkes, private communication]. While this suggestion does make sense on the face of it, I am nervous about accepting it. To treat phantom functions as merely the limiting case of malfunction strikes me as analogous to treating fiction as merely the limiting case of misrepresentation. And that does not seem very plausible, given the important and multitudinous positive roles fiction plays in human life. I suspect phantom function is more like fiction than like misrepresentation. 9 This is the terminology of Houkes and Vermaas [2003].

Philosophical Theories of Artifact Function

219

a history of selective reproduction as the primary determining factor for artifact functions, and they regard human intentions as having a secondary or indirect role, mediated by various non-intentional factors. So at one end of the spectrum are radically intentionalist views, at the other end are views that accord human intention only a secondary role, and then there are some mixed views in between. I will start at the radically intentionalist end. John Searle [1995, especially Chapter 1] takes up the question of artifact function in the context of explaining how social reality is created by human agents. On his view, all functions, biological as well as cultural, are entirely “observer relative” [1995, pp. 9-13, et passim]. They are “assigned” [1995, p. 9, et passim], often collectively, by human agents in accordance with human purposes, goals and values. Searle distinguishes between “agentive” functions that depend on a use to which human agents intentionally put artifacts they have made for this or other purposes, and “non-agentive” functions that are mechanical, causal processes of naturally occurring objects to which human agents have assigned an extrinsic purpose or value [1995, pp. 22-23]. Thus for Searle artifact functions depend on the intentions and intentional states of human agents who impose them through the intentional use of artifacts.10 Searle takes this imposition of agentive functions on artifacts to be paradigmatically a matter of individual action, but he also allows for collective imposition. If it is easy to see how a single person might decide to use some object as a chair or a lever, then I believe it is not difficult to see how two or more people together could decide to use some object as a bench on which they can all sit or to use something as a lever to be operated by several people, rather than just one. Collective intentionality can generate agentive functions as easily as individual intentionality. [Searle, 1995, pp. 38-39] On Searle’s theory, collective intentions are not reducible to individual intentions [1995, pp. 23-25], which implies that for him some artifact functions are irreducibly social. This point connects with his discussion of a special class of agentive functions he calls “status functions,” where a function is assigned to something the physical structure of which is related only in an arbitrary way to the performance of the function, e.g., paper money. On Searle’s view, this phenomenon depends on the prior collective assignment of a special status to the object — the status of medium of exchange, in the case of paper money — which subserves specific assigned functions — in this case, the functions of paying debts, making purchases, and so on. Status functions are important for Searle because their assignment to artifacts creates what he calls an “institutional fact,” a fact which exists only through collective human agency [1995, p. 40 ff.]. Searle’s theory of artifact function is thus not only intentionalist; but in addition it foregrounds collective intentions and the social or institutional results of their operation. 10 See

[Kroes, 2003] for a thorough critique of Searle’s view.

220

Beth Preston

Randall Dipert [1993] also has a radically intentionalist account of artifact function. He begins by defining ‘artifact,’ ‘instrument,’ and ‘tool’ in terms of the intentions of their makers and/or users [1993, Chapter 2]. Not surprisingly, then, he analyzes their functions entirely in terms of agent intentions as well. It should be noted that although Dipert is clearly talking about artifact functions, he uses the term ‘artifact purpose’ instead of the more usual locution. I maintain that artifact purposes are derivative from agent purposes. . . . That is, roughly, anyone who thinks of artifact purposes has a conceptualization that unpacks into agent purposes but is not always fully conscious of this relationship. Talk and thought about artifact purposes, however, acquires a certain conceptual independence through habit of use, the present lack of conscious connection, and, of course, the fact that an artifact purpose is not “just” an agent purpose — that is, the telephone’s purpose is not the telephone’s in the way that my purpose in writing this book is mine. [Dipert, 1993, p. 237] As is clear from this passage, Dipert is very sensitive to the non-conscious nature of many human purposes and intentions, and provides for this in his theoretical framework. For instance, in discussing the intentional activity of the creators of artifacts, he says that to the extent that it is not the result of specific prior deliberation or the subject of conscious attention, it is better described as “halfintentional” [1993, pp. 50-51]. But it is also clear that Dipert, like Searle, does not allow for any other source of artifact function apart from human intentions and the beliefs and desires that contribute to the formation of those intentions. He just allows for these intentional states to operate below the level of conscious awareness. Similarly, Dipert acknowledges that these intentional states may involve “group agency” [1993, p. 126 ff.]. But unlike Searle, he does not make this a feature of his account. It is not clear whether he disagrees outright with Searle about the reducibility of collective intentions/agency to individual intentions/agency. But he argues that group agency is best considered as “virtual” individual agency for the purposes of a theory of artifacts and their features [1993, p. 32 and pp.194-5]. So for Dipert, as for Searle, the paradigm case is that of an individual agent planning and constructing an artifact with a specific purpose in mind and thereby endowing it with a function. But unlike Searle, he recommends that cases of group agency be rationally reconstructed in accordance with the individual model for purposes of analysis and theory construction. Dipert’s views appear to have been an important influence on Peter McLaughlin’s account of artifact function. McLaughlin [2001, especially Chapter 3] holds that since artifacts are not self-reproducing systems they have no functions in their own right; rather their functionality is completely dependent on the intentions of their makers and/or users. The function of an artifact is derivative from the purpose of some agent in making or appropriating the object; it is conferred on the object by

Philosophical Theories of Artifact Function

221

the desires and beliefs of an agent. No agent, no purpose, no function. [McLaughlin, 2001, p. 60] These agent purposes do not have to result in the modification of material or even a change of location — an agent can appropriate an artifact for a particular purpose and thus endow it with a corresponding function simply by noticing the artifact and approving its use for that purpose — an operation McLaughlin refers to as “virtual assembly” [2001, p. 54]. But McLaughlin does distinguish between intentions that precede the actual assembly of the artifact and intentions that follow it, calling functions that result from preceding intentions “design functions” and those that result from following intentions “use functions” [2001, pp. 47-48]. McLaughlin is explicit about the epistemological consequences of this intentionalist analysis of artifact function: it is only the beliefs and desires of agents that explain the features of artifacts. The functions of artifacts, being themselves entirely derivative from agent beliefs and desires, do not explain anything. This echoes Dipert’s view that the purposes of artifacts can be unpacked as the purposes of human agents, and suggests that any discussion of artifact function is properly to be carried out under the aegis of philosophy of mind and/or philosophy of action. Another consequence of McLaughlin’s view — and one he stresses throughout his book — is that artifact functions differ from biological functions in that they benefit the user of the artifact and not the artifact itself. Unlike biological organisms, then, artifacts have no intrinsic “good” or interests of their own, but are only good insofar as they serve the extrinsic interests of human agents [2001, Chapter 9]. Thus McLaughlin definitively divorces the discussion of artifact function from the discussion of biological function by referring artifact functions and other associated features of artifacts to human intentions and other intentional states. Karen Neander [1991] also advances an intentionalist theory that distances artifact functions from biological functions, but for somewhat different reasons. On her view, the functions of biological traits are the effects for which these traits were selected over the course of evolutionary history. Similarly, artifact functions are the effects for which they are intentionally selected by human agents. Despite the quasi-biological terminology retained in the artifact case, Neander takes intentional selection to be toto caelo different from natural selection. Whereas evolved functions must be generalizable over types, artifact functions may be idiosyncratic. Unique inventions, like the additions to James Bond’s brief case, can have proper functions peculiar to them because they can be individually selected for particular effects. Also, because intentional agents do have foresight, there need be no past performances of the functional effect, nor any ‘ancestral’ artifacts to do any performing. It is enough, in the case of intentional selection, if the designer believes or hopes that the artifact will have the desired effect and selects it for that purpose. [Neander, 1991, p. 462] Thus whereas biological selection applies to types and so necessarily involves a history of reproduction, intentional selection can apply to tokens and so is in-

222

Beth Preston

dependent of the history of reproduction, if any, of the object. The intentions of designers and users are on an equal footing here as far as endowing artifacts with proper functions goes. But like McLaughlin, Neander distinguishes several sub-types of proper function related to the designer-user distinction. I suggest that the function of an artifact is the purpose or end for which it was designed, made, or (minimally) put in place or retained by an agent. . . . Since there will often be several agents involved, and each might have a different purpose, we might want to distinguish between ‘design functions’, ‘user functions’, ‘occasion functions’, and so on. The everyday notion of an artifact’s function is context sensitive, and in some contexts one intentional agent can take precedence over another. However, although the particular context might highlight the intentions of the user, rather than the designer, say, the function of an artifact is always whatever it was selected for. [Neander, 1991, p. 462] These distinctions among types of functions are not worked out in any detail by Neander, but they clearly echo McLaughlin’s distinction between design functions and use functions and hint at a somewhat more elaborate scheme of the same sort. Finally, and only in a footnote, Neander mentions that non-intentional trial and error, a process that mimics natural selection more closely than intentional selection, may be responsible for the functions of some artifacts or their components [1991, p. 462, n. 11]. This is the first break in the so far monolithically intentionalist reading of artifact functions.11 This non-intentionalist dimension is elaborated and given a more significant role by Paul Griffiths [1993]. Like Neander, Griffiths has a selectionist account of function in general; and like Neander, he thinks that “human selection does for artifacts what natural selection does for organisms” [1993, p. 419]. But unlike Neander, he emphasizes that human selection is often not intentional. Many features of artifacts make no intended contribution and yet have proper functions. In societies with low-level technologies, artifacts are often designed by trial and error over periods of many generations. The contribution that a feature makes to the performance of overall function may never be appreciated. Such features do not have an intended use but they do have functions, and they can be explained by their functions. [Griffiths, 1993, pp. 418-419] The prevalence of an artifact, or an artifact trait, can be explained by selective processes in which people meet their needs, sometimes by 11 The possibility of non-intentional selection of artifacts by trial and error is also pointed out by Bigelow and Pargetter [1987]. More recently, Mark Perlman has sounded a note of caution for intentionalist accounts of artifact function, warning that there is much evidence from psychology, sociology, and anthropology that artifacts often have functions unknown to, and unintended by, their users — and perhaps in some cases even their designers [2004, pp. 33-34]. However, he fails to note that Karen Neander and (as we shall see) Paul Griffiths both provide for this.

Philosophical Theories of Artifact Function

223

conscious design, sometimes by trial and error, and sometimes by an amalgam of the two. [Griffiths, 1993, p. 419] Griffiths’s view is thus much more clearly of a mixed, intentionalist/nonintention alist type, because it is possible for cultural selection to be either intentional or non-intentional or a bit of both. Griffiths is also more at pains to spell out the analogies and disanalogies between natural and human selection. For example, natural selection requires selection between competing variations. Griffiths points out that non-intentional trial and error parallels natural selection in this respect, with competition between actual variations under actual, real world conditions. But in the case of intentional selection there is often intentional design, and here the competition may be “hypothetical,” taking place between imagined variations under imagined, virtual world conditions [Griffiths, 1993, p. 419]. Griffiths notes that in these cases fallible human designers may select something that never has and never can perform the function intended — for example, the tapered tails of early racing cars, which their designers theorized would reduce drag, but which did not in fact do that. Such cases are phantom functions, in the terminology introduced in the previous section. Griffiths points out that this can never happen in biology, since natural selection can only act on an actual performance, whereas intentional selection can act on an imagined performance. And he analyzes this phantom function phenomenon, too, in terms of hypothetical alternatives. It occurs when such alternatives are evaluated and selected in light of false beliefs or theories about the world. Ruth Millikan [1984; 1999] also appears to mix intentionalist and non-intentionalist components in her theory of artifact function, but in a different way.12 She is mainly interested in what she calls proper function — what a thing is supposed to do, and is malfunctioning if it cannot do.13 Most biological proper functions are established by a (non-intentional) history of selection and reproduction for this performance. But on Millikan’s view, some proper functions are established derivatively, through some other, already proper functional, trait. Her favorite example is chameleon skin color — when the skin color changing mechanism of a chameleon sitting on my deck turns it brown, that brown color has the proper function of camouflaging that chameleon from predators (my wretched cats, say, who are altogether too fond of lizard for dessert) even if it is a unique shade of brown that has not been specifically selected for in the course of evolution. As applied to artifacts, Millikan argues, this means that they typically have two, simultaneous sources of proper function. They have direct proper functions in virtue of their history of reproduction, just as biological traits do; and they have derived 12 Millikan is not explicit enough about certain points for this interpretation to be on absolutely sure ground. But it seems to be at least a plausible reading of what she does say, for reasons I will explain in a moment. 13 She also acknowledges another type of function she usually calls Cummins functions, in reference to a well-known article by Robert Cummins [1975]. But she does not make much use of Cummins functions for theoretical purposes in what she has to say about artifacts, although she does sometimes use artifacts as examples of things with Cummins functions (e.g., in [Millikan, 2002]).

224

Beth Preston

proper functions in virtue of the intentions of their producers [1999, p. 205]. Usually these artifactual proper functions coincide. For example, the loaves of bread you bake today have both the direct proper function of providing nourishment in virtue of the long history of similar loaves providing such nourishment, and the derived proper function of providing nourishment in virtue of your current desire for nourishment. But sometimes these proper functions diverge. For instance, if you bake bread in order to sell it rather than eat it, the loaves have the direct proper function of providing nourishment, but the derived proper function of providing income — derived from your desire on this occasion for income rather than nourishment from these particular loaves of bread. This divergence is not problematic in principle, because things often have more than one proper function at the same time. Moreover, Millikan points out, some artifacts — novel prototypes, in particular — do not have direct proper functions because they do not (yet) have a history of reproduction. So if they are to have proper functions at all, they must have derived proper functions grounded in the intentions of their designers [1984, especially Chapter 1; 1999, pp. 204-205].14 For Millikan, then, some artifacts have only an intentional source of proper function; some have only a non-intentional source in a history of reproduction (e.g., the trial and error scenario suggested by Griffiths and Neander); but most of them have both sources of proper function simultaneously.15 Pieter Vermaas and Wybo Houkes [2003] also appear to have a mixed theory, although as with Millikan, the case is not entirely clear. They begin by setting forth four desiderata they think any adequate theory of artifact function should satisfy. It should 1) distinguish proper (“standardly ascribed”) functions from accidental (“ascribed only occasionally”) functions, 2) be able to ascribe proper functions even to malfunctioning tokens, 3) entail that the physical structure of the artifact be sufficient to perform its ascribed function, and 4) provide for an ascription of proper functions to novel prototypes [2003, pp. 265-266]. They argue that only an intentionalist theory that ignores the reproductive history of artifacts can succeed in meeting all of these desiderata, and in particular only an intentionalist theory can meet the novel prototype desideratum. On the other hand, Vermaas and Houkes point out, due attention to the design process shows that designers do not form intentions about the purposes of artifacts in a vacuum — they deliberate and form beliefs about proposed physical structures of artifacts with regard to their capacity to carry out proposed functions. They criticize both Neander and Millikan for ignoring this physical structure desideratum, and propose a more satisfactory alternative.

14 For more details and a critical discussion of Millikan’s distinction between direct and derived proper functions, see [Preston, 1998]. See [Millikan, 1999] for a reply. 15 Tim Lewens [2004, especially Chapters 5 and 7] seems to have a similar idea in mind. He distinguishes between intended artifact functions and evolutionary artifact functions, and stresses that they represent different sources of function. Artifacts, on his view, may get their functions from either one of these sources, depending on the circumstances [2004, p. 166].

Philosophical Theories of Artifact Function

225

Consider a theory that ascribes a function to an artefact on three conditions: (1) the designers intended to design the artefact to have the function, (2) these designers determined the phyiscal structure of that artefact on the basis of their technological and scientific knowledge, and (3) by means of this knowledge, they can provide explanations as to why the artefact with the thus-determined physical structure can perform the function. [Vermaas and Houkes, 2003, p. 287] This might well be interpreted as a pure intentionalist view, since it takes artifact functions to depend entirely on the intentional states of human agents. But the physical structure desideratum is fundamentally a non-intentionalist consideration, since it appeals to the actual physical makeup of the artifact. This is captured in the passage above by the requirement that the designers be applying knowledge, and not just engaging in wishful thinking, irrational fantasies or blatantly false beliefs. Beth Preston [2003] argues that Vermaas and Houkes fail to satisfy their first desideratum, namely, to maintain a theoretical distinction between proper functions and accidental functions of artifacts. Perhaps in response to this criticism, the authors address this distinction directly in a later paper [Houkes and Vermaas, 2004]. Their approach is to derive a theory of artifact function from a theory of artifact use and design. On our theory, an artifact function is a capacity, supposed or actual, which has a preferential status in the context of certain actions and beliefs. It is therefore a highly relational property, which supervenes on both the actual physical makeup of an artifact and on the beliefs and actions of human agents, designers as well as users. [Houkes and Vermaas, 2004, p. 67] Proper functions can be distinguished from accidental functions on this approach via a prior distinction between standard use plans and various kinds of alternative (non-standard) use plans. It is also clear from this passage that Houkes and Vermaas wish to maintain the non-intentionalist emphasis on actual physical structure they stressed in their earlier article, and this supports the interpretation of their view as a mixed intentionalist/non-intentionalist one.16 Marcel Scheele [2005; 2006] follows Houkes and Vermaas in emphasizing use as essential to artifact function, as well as in analyzing use in terms of the plans of individual or collective agents. But he emphasizes the social aspects of use to a much greater extent than they do. Indeed, he argues [2006, p. 32 ff.] that not only their theory, but also the non-intentionalist theory of Beth Preston [1998; 2003], can be improved through greater attention to the social aspects of use. Scheele [2005, especially Chapters 2 and 3] begins by accepting the now familiar distinction between the proper function and the system function of artifacts, i.e., 16 Houkes and Vermaas continue to refine and extend their theory, which they refer to as the ICE (intention/cause/evolution) theory of artifact function. See [Vermaas and Houkes, 2006; Vermaas, 2006; Houkes, 2006], for instance.

226

Beth Preston

between their standard uses and their occasional, non-standard uses. He then argues that standard uses, and thus proper functions, must be understood in terms of social institutions. Social institutions, on his view, are collective, relatively stable patterns of action that are socially enforced. Scheele [2006, p. 28 ff.] also stresses the changes that the functions of an artifact often undergo, and analyzes these changes in terms of the social roles of the designers vs. users, or of original users vs. current users. Scheele does not claim to have a distinct theory of artifact function. But he does have a distinctive theory of the social dimensions of artifact function that can be used to amend or supplement existing theories, many of which are alarmingly thin when it comes to an analysis — or even an acknowledgement — of the social character of artifacts. Beth Preston [1998; 2003; unpublished] is the only theorist to espouse a nonintentionalist account of artifact function. Her initial concern [1998] is to put forward a pluralist account.17 Artifacts, on her view, have two different kinds of functions neither of which is reducible to the other. On the one hand, they have the functions they have historically been reproduced to serve, which Preston calls ‘proper’ functions in reference to Ruth Millikan’s [1984] theory of proper function which Preston adopts and adapts. On the other hand, artifacts often are used for functions they were not reproduced to serve but are capable of serving — e.g., spoons can serve as musical instruments. Preston calls these ‘system’ functions in reference to Robert Cummins’ [1975] theory of biological function as the role a component plays in the overall context of a system. Since both Millikan’s theory and Cummins’ theory are focused on giving an account of biological function, neither is intrinsically intentionalist; and Preston resists reformulating them in intentionalist terms, as Vermaas and Houkes [2003] in effect do, for instance. In general, then, non-intentionalist theories of function maintain the distinction between proper function and accidental [system] function by reference to the distinction between the embedding of a performance in a history of selection and reproduction, and the embedding of a performance in a currently operating system. But intentionalist theories of function eschew reference to either the ancestry or the system context of an artifact, and rest everything on the intentional states of agents. [Preston, 2003, pp. 603-604] [I]ndividual intentional states do have a role to play in a non-intentionalist theory, because they are necessary for the implementation of histories of reproduction and social systems. . . . But regarding individuals’ intentional states collectively as a necessary condition for the establishment of artifact function in this way is quite a different matter from regarding the intentional states of individual agents as a sufficient condition, as the intentionalist theory. . . does. [Preston, 2003, p. 611] 17 Millikan [1989; 1999] endorses a pluralist account of function in general, although not specifically with reference to artifacts.

Philosophical Theories of Artifact Function

227

Preston [2003; unpublished] argues that only non-intentionalist theories can adequately distinguish proper functions from system functions. Intentionalist theories are forced to elide the distinction because they have no adequate way of distinguishing between designers’ intentions and users’ intentions. She concedes, on the other hand, that non-intentionalist theories like her own have no way of ascribing proper functions to novel prototypes. But she argues that this is neither necessary nor desirable. Preston [unpublished] also gives a non-intentionalist account of phantom functions. It appeals to the history and the predominant patterns of use of an artifact type rather than to the designers’ intentions in a context of false beliefs, as Griffiths does in his account of phantom functions. This completes the review of current philosophical theories of artifact function.18 In the next section I will assess the state of the art as described here with regard to progress in accounting for the features of artifact function made manifest by the material culture phenomena described in Section 2. This exercise will also help bring into better focus the most important issues full-fledged theories of artifact function have to face. 4

ASSESSING THEORIES OF ARTIFACT FUNCTION

In Section 2 I described six central and characteristic phenomena of material culture that make manifest important features of artifact function — multiple realizability, multiple utilizability, recycling, reproduction with variation, malfunction and phantom function. I noted there that multiple realizability of function and multiple utilizability of artifacts are two sides of the same coin. Because any given artifact function can be realized in a number of different physical structures, not only are there different types of artifacts with the same standard function, but artifacts for which that function is not standard may be coopted to serve it on occasion. In other words, multiple realizability leads directly to multiple utilizability in the form of alternative uses of artifacts meant for some other purpose. And multiple utilizability in turn implies multiple realizability given that the motive for alternative uses of existing artifacts is typically unavailability of the type of artifact for which the use is standard — e.g., you tighten a loose screw on a doorknob with a tableknife because you can’t find your screwdriver, or don’t want to go all the way out to the garage to get it, or don’t have a small enough one, or whatever. This very close connection between multiple realizability and multiple utilizability makes it reasonable to take them together for the purpose of describing how these two characteristics show up in theories of artifact function. Most of the theories outlined in the previous section make some kind of a distinction between the standard functions of artifacts and alternative, occasional func18 There are a number of other function theorists who claim, or sometimes only imply, that their preferred theory of biological function also applies to artifacts. But they say so little about artifacts, or their theories as applied to artifacts overlap so much with one or more of the theories already discussed, that a separate review does not seem warranted. Examples are [Bigelow and Pargetter, 1987; Kitcher, 1993], and [Cummins, 1975].

228

Beth Preston

tions, although they employ different terminology. Preston and Scheele distinguish between proper function and system function; Houkes and Vermaas distinguish between proper function and accidental function; McLaughlin distinguishes design function from use function; Millikan accepts a distinction between proper function and Cummins functions; and Neander suggests a tripartite distinction between design function, user function, and occasion function. All of these schemes are clearly aimed at analyzing the phenomena associated with multiple utilizability and multiple realizability. Searle emphasizes the relatively unconstrained nature of physical structure with respect to function, especially for his status functions. This is at least a comment upon multiple realizability, if not multiple utilizability. Thus it seems that most theorists of artifact function do recognize these twin phenomena as important, and distinguish them terminologically. However, how best to account for these phenomena theoretically is a very large question that remains to be answered. The predominantly intentionalist cast of theories of artifact function predisposes theorists to try to account for the relevant distinctions in terms of intentional action. But the exchanges between Preston [1998] and Millikan [1999], as well as between Vermaas and Houkes [2003] and Preston [2003] show that the intentionalist approach is at least problematic in this regard. In addition, multiple realizability and multiple utilizability taken together indicate that the relationship between function and physical structure of artifacts is in principle — and very often in fact — a many-many relationship. Thus the more general, underlying issue here is the nature of the relationship between function and structure. This issue has not really been addressed head on, although Searle and Vermaas and Houkes have taken some steps in that direction. The third characteristic I described above is recycling, which I noted may well be best understood as a special case of multiple utilizability where the utilized artifact loses its capacity to perform its original function in the process of being modified for the performance of an alternative function. This phenomenon really has not been discussed as such, although Scheele’s emphasis on the various processes involved in change of function seems like a promising beginning. But there are two points to be made, nonetheless. First, recycling involves a concomitant change of function and change of physical form, so the nature of the relationship between function and structure underlies this phenomenon just as it does the twin phenomena of multiple realizability and multiple utilizability. Second, the neglect of recycling in the philosophical literature on function is symptomatic of the general neglect of material culture in philosophy I noted at the very beginning of this essay. It shows up here as a failure to examine in any detail the full range of human interactions with material culture that are relevant to questions of function. An examination of recycling is especially important in this regard because it would provide a counterweight to the overwhelming emphasis on design that is detectable in the intentionalist theories of artifact function outlined above. I would be willing to bet that a much greater amount of human time, energy, and ingenuity is spent on reuse and recycling of existing artifacts than on designing

Philosophical Theories of Artifact Function

229

new artifacts for specific purposes. Theorists of artifact function should find out the facts and take them into account when constructing their theories.19 The fourth characteristic I listed was reproduction with variation, but I will save that for last because it is perhaps the most important and interesting of the lot. Let us go on to the fifth characteristic, then: malfunction. The centrality of this phenomenon hinges on its connection to the widely accepted idea that artifacts have standard functions (or proper functions, as most theorists now call them). Proper functions are normative; they are what an artifact is supposed to do. If, as all too frequently happens, an artifact is broken or defective and cannot perform its proper function, it is malfunctioning. So theorists of artifact function must account for malfunction, just as theorists of linguistic or mental representation must account for misrepresentation. Searle [1995, p.15], mentions the normative valence of function assignments and the consequent possibility of identifying malfunctions. Dipert [1993, p.143 ff.] discusses malfunction under the heading of failures of function, and gives quite an extensive description of the common sources of such failures. Millikan’s [1984] original theory of proper function contains perhaps the best known account of malfunction. Other theorists who have adopted some version of Millikan’s theory of proper function (e.g., Preston [1998]; Scheele [2005]) can, of course, help themselves to it. Neander [1995] gives an account of malfunction that explicitly relates it to misreprentation. Vermaas and Houkes [2003] list accounting for malfunction as one of their four desiderata for an adequate theory of artifact function, and provide a provisional intentionalist explanation of this phenomenon. Thus it seems that malfunction is not only recognized as important, but is usually, if not universally, addressed by theorists of artifact function.20 This is not the case with the sixth characteristic on the list — phantom function. Artifacts have phantom functions when they are designed and reproduced to perform a specific function, but are in principle unable to do so, e.g., amulets to ward off evil or bring good luck, patent medicines, and so on. Griffiths [1993] gives an intentionalist account of this phenomenon. Preston [1998] gives one non-intentionalist account; and then later [unpublished] a different, but equally non-intentionalist account. Otherwise, phantom function seems to have gone unnoticed. This may be partly due to a tendency to focus on instrumental functions, and to ignore social or communicative ones. The failure of an artifact type to have the bare capacity to perform its instrumental function is likely to be recognized at some point, and the likely consequence is modification of the artifact so that it does have the required capacity, or being dropped from the production schedule if modification is fruitless. Consider Griffiths’s example of the tapered tails of early racing cars. Once the false theory of drag that spawned those tails was rejected, 19 Thinkers in other disciplines have a good headstart on this, from which philosophers have much to learn. See, for instance, the work on reuse and recycling by Schiffer, Downing and McCarthy [1981]. 20 Malfunction is also discussed by Maarten Franssen [2006] in the context of an article on the normativity of artifacts.

230

Beth Preston

the tails disappeared as well. But the most important phantom functions are ones where this sorting out process is unlikely to occur for social reasons or for reasons having to do with the psychological and cognitive makeup of human beings. Good luck charms, artifacts with religious significance, and popular remedies, for example, continue to be reproduced without let or hindrance regardless of their capacity to perform their functions. Phantom functions are thus both common and persistent. Moreover, the sharp disagreement between Griffiths and Preston is evidence that accounting for them is not going to be easy or straightforward. So here again, facts need to be gathered and more attention needs to be paid to them during theory construction. Finally, let us consider reproduction with variation. On the face of it, the theories referred to above as reproduction theories — Preston’s theory and the non-intentionalist component of Millikan’s theory — do recognize reproduction with variation as central to theories of artifact function. But neither Preston nor Millikan has really theorized reproduction in material culture in its own right, as opposed to appealing to it for the purposes of analyzing other phenomena of artifact function, such as the distinction between proper function and system function. Millikan [1984, p.19 ff.] does have an account of copying. But it is couched in the most generic of terms so as to cover both biology and material culture, and so avoids saying anything specific about the sources of persistence and variation. Since these sources are widely held to be very different in the two realms, philosophical issues and problems concerning reproduction in material culture are simply bypassed — as are issues and problems concerning reproduction in biology, of course. Preston [unpublished] does make an attempt to rectify this situation in her last chapter, where she discusses the reproduction of structure, function, and intention in material culture. But this is only a start.21 With regard to the rest of the theories outlined above, the focus on human intentions predisposes theorists to focus on production to the neglect of reproduction. McLaughlin, for instance, has a great deal to say about self-reproducing systems — which, on his view, artifacts are not — and then talks only about intentional “assembly” with regard to artifacts. Dipert, too, focuses exclusively on production and refers this activity back to the intentions of the producing agent. A correct description of an artifact as an artifact describes the artifact in the way that its creator conceived of it — at least as much as is now possible. Specifically, an object is contemplated by an agent, and some of its properties are intentionally modified (or perhaps, intentionally left alone); the production of an artifact is the goal of some intentional activity. [Dipert, 1993, pp. 15-16] Vermaas and Houkes [2003] do not even mention reproduction in the four desiderata they set forth for theories of artifact function. On the other hand, their fourth 21 One philosopher who has a lot to say about reproduction in material culture is Karl Marx, e.g., Chapter 23 of Capital. Similarly, there is relevant work in other disciplines. See, for example, [Schiffer and Skibo, 1997] and [Miller, 1985] on types and sources of artifact variability in the reproduction process.

Philosophical Theories of Artifact Function

231

desideratum concerns the ascription of proper functions to novel prototypes, echoing Millikan’s insistence that theoretical provision must be made for things that are “new under the sun” to have proper functions [1984, p. 19; 1999, p. 204]. Dipert falls into this camp as well, since the ultimate aim of his book is an account of works of art — which are, of course, widely held to be unique and idiosyncratic productions. This emphasis not just on production, but on novel productions, is, it seems to me, a red herring. The salient characteristic of material culture is not production, but reproduction; not novelty, but standardized forms and functions. Indeed, as indicated above, novelty in any radical sense is vanishingly rare in material culture. What is common is minor or incremental variations on existing artifacts; and that is what primarily needs to be examined and explained. Another crucial issue that arises in connection with reproduction concerns the obviously social and collaborative nature of the processes involved. Scheele does not focus on reproduction in material culture, but his salutary emphasis on the social aspects of artifact function certainly provides a starting point for doing so. This is also true of Searle, who discusses artifact function in the context of a work devoted to the nature of social institutions in general. For the intentionalist wing, the social aspects of material culture are most naturally understood as Searle understands them — in terms of collective intentions.22 But philosophical theories of collective intention are in their infancy, since action theory has traditionally concentrated exclusively on individual intentional action. Moreover, the theories of collective intention that have been advanced so far, including Searle’s, only apply to small, close-knit, egalitarian groups. As Christopher Kutz [2000] complains, this leaves unanalyzed the larger, more diffuse, non-egalitarian groups that are paradigmatically the subjects of social processes — tribes, trade unions, the Vatican, military units, the faculty of [your institution’s name here], extended families, and so on. And with the possible exception of those small and isolated human cultures of which there are fewer and fewer these days, the reproduction of material culture is carried out mostly by these larger and more diffuse groups. In short, an intentionalist approach to the social aspects of reproduction with variation is barely off the ground and faces some rather large questions. Moreover, it will undoubtedly face non-intentionalist alternatives. Indeed, a non-intentionalist approach is most plausible precisely with regard to large, diffuse, non-egalitarian groups, where the overall nature and ultimate result of the activity undertaken may be explicitly intended only by some or even by noone involved. So what is the current state of the art in theories of artifact function? There is some discussion of the twin phenomena of multiple realizability and multiple utilizability, although many significant issues in this area have hardly been touched. And current theories do seem to have some leverage on the issue of malfunction. But recycling, phantom function, and reproduction with variation present some 22 There is one other option, represented here. Dipert [1993] prefers to reconstruct group or collective activity as “virtual” individual activity for the purposes of theorizing about artifacts. However, this is more a way of bypassing the social aspects of material culture than of analyzing them.

232

Beth Preston

very difficult problems and have hardly been broached. Overall, then, almost everything remains to be done. BIBLIOGRAPHY [Bigelow and Pargetter, 1987] J. Bigelow and R. Pargetter. Functions, The Journal of Philosophy 84, 181-196, 1987. [Cummins, 1975] R. C. Cummins. Functional analysis, The Journal of Philosophy 72, 741-765, 1975. [Dipert, 1993] R. R. Dipert. Artifacts, Art Works, and Agency. Temple University Press, 1993. [Franssen, 2006] M. Franssen. The normativity of artifacts, Studies in History and Philosophy of Science 37, 42-57, 2006. [Griffiths, 1993] P. E. Griffiths. Functional analysis and proper functions, British Journal for the Philosophy of Science 44, 409-422, 1993. [Houkes, 2006] W. Houkes. Knowledge of artefact functions, Studies in History and Philosophy of Science 37, 102-113, 2006. [Houkes and Vermaas, 2004] W. Houkes and P. E. Vermaas. Actions versus functions: A plea for an alternative metaphysics of artifacts, The Monist 87, 52-71, 2004. [Kitcher, 1993] P. Kitcher. Function and design, Midwest Studies in Philosophy 18, 379-397, 1993. [Kroes, 2003] P. A. Kroes. Screwdriver philosophy: Searle’s analysis of technical functions, Techn´ e 6, 22-35, 2003. [Kutz, 2000] C. Kutz. Acting together, Philosophy and Phenomenological Research 61, 1-31, 2000. [Lewens, 2004] T. Lewens. Organisms and Artifacts: Design in Nature and Elsewhere. The MIT Press, 2004. [McLaughlin, 2001] P. McLaughlin. What Functions Explain: Functional Explanation and SelfReproducing Systems. Cambridge University Press, 2001. [Miller, 1985] D. Miller. Artefacts as Categories: A Study of Ceramic Variability in Central India. Cambridge University Press, 1985. [Millikan, 2002] R. G. Millikan. Biofunctions: Two paradigms. In Functions: New Essays in the Philosophy of Psychology and Biology, A. Ariew, R.C. Cummins, and M. Perlman, eds., pp. 33-59. Oxford University Press, 2002. [Millikan, 1999] R. G. Millikan. Wings, spoons, pills, and quills: A pluralist theory of function, The Journal of Philosophy 96, 191-206, 1999. [Millikan, 1989] R. G. Millikan. An ambiguity in the notion ‘function’, Biology and Philosophy 4, 172-76, 1989. [Millikan, 1984] R. G. Millikan. Language, Thought, and Other Biological Categories: New Foundations for Realism. The MIT Press, 1984. [Neander, 1995] K. Neander. Misrepresenting and malfunctioning, Philosophical Studies 79, 109141, 1995. [Neander, 1991] K. Neander. The teleological notion of ‘function’, Australasian Journal of Philosophy 69, 454-468, 1991. [Perlman, 2004] M. Perlman. The modern philosophical resurrection of teleology, The Monist 87, 3-51, 2004. [Preston, unpublished] B. Preston. The Stuff of Life: Towards a Philosophy of Material Culture. Unpublished. [Preston, 2003] B. Preston. Of marigold beer — a reply to Vermaas and Houkes, British Journal for the Philosophy of Science 54, 601-612, 2003. [Preston, 1998] B. Preston. Why is a wing like a spoon? A pluralist theory of function, The Journal of Philosophy 95, 215-254, 1998. [Searle, 1995] J. Searle. The Construction of Social Reality. The Free Press, 1995. [Scheele, 2006] M. Scheele. Function and use of technical artefacts: Social conditions of function ascription, Studies in History and Philosophy of Science 37, 23-36, 2006. [Scheele, 2005] M. Scheele. The Proper Use of Artefacts: A Philosophical Theory of the Social Constitution of Artifact Functions. Simon Stevin Series in the Philosophy of Technology, 2005.

Philosophical Theories of Artifact Function

233

[Schiffer and Skibo, 1997] M. B. Schiffer and J. M. Skibo. The explanation of artifact variability, American Antiquity 62, 27-50, 1997. [Schiffer et al., 1981] M. B. Schiffer, T. E. Downing, and M. McCarthy. ‘Waste not, want not’: An ethnoarchaeological study of reuse in Tucson, Arizona. In The Archaeology of Us: Modern Material Culture, R.A. Gould and M.B. Schiffer, eds., pp. 67-86. Academic Press, Inc., 1981. [Vermaas, 2006] P. E. Vermaas. The physical connection: Engineering function ascriptions to technical artefacts and their components, Studies in History and Philosophy of Science 37, 62-75, 2006. [Vermaas and Houkes, 2003] P. E. Vermaas and W. Houkes. Ascribing functions to technical artefacts: A challenge to etiological accounts of functions, British Journal for the Philosophy of Science 54, 261-289, 2003. [Vermaas and Houkes, 2006] P. E. Vermaas and W. Houkes. Technical functions: A drawbridge between the intentional and structural natures of technical artefacts, Studies in History and Philosophy of Science 37, 5-18, 2006.

This page intentionally left blank

FUNCTIONAL DECOMPOSITION AND MEREOLOGY IN ENGINEERING Pieter Vermaas and Pawel Garbacz

1

INTRODUCTION

A key characteristic of the engineering sciences is that their descriptions of physical objects and processes are cast not only in structural terms but in functional terms as well. The engineering sciences share this characteristic with biology and with some of the humanities. If, however, it is added that these functional descriptions express the aims for which objects and processes can be employed, the characteristic becomes more discriminative to the engineering sciences. Functional descriptions in the engineering sciences have been analysed in philosophy. The focus of these analyses has largely been on determining what is meant when an individual technical object or process is ascribed a function.1 Functional descriptions in the engineering sciences are, however, much richer than individual ascriptions of functions. Design methodologists, for instance, often characterise the initial conceptual phase of engineering designing as one in which engineers reason from a required overall function of some product-to-be to a number of subfunctions that can make up this overall function.2 In reverse engineering and other explanatory tasks the reasoning may be the other way round, deriving an overall function from a series of subfunctions. And in engineering knowledge bases, functional descriptions of technical objects, processes and their parts are related to one another.3 This functional reasoning in the engineering sciences leads to descriptions of technical systems in which different functions are related to one another. Functional reasoning leads in particular to functional decompositions, which are descriptions in which one overall function is related to a series of subfunctions that, together, make up the overall function. And functional reasoning can be taken to define what can be called a functional part-whole relationship by identifying the subfunctions in a functional decomposition as parts of the overall function that they make up. These interrelated functional descriptions have 1 See

Beth Preston’s chapter “Philosophical Theories of Artifact Function” in this Handbook. [Pahl and Beitz, 1996, Section 2.1; Umeda and Tomiyama, 1997; Chittaro and Kumar, 1998; Stone and Wood, 2000; Chakrabarti and Bligh, 2001; Kitamura et al., 2005/2006; Bell et al., 2007]. See also Peter Kroes’ chapter “Foundational Issues of Engineering Design” in this Handbook. 3 See William H. Wood’s chapter “Computational Representations of Function in Engineering Design” in this Handbook, in which reverse engineering also plays a role. 2 E.g.,

Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

236

Pieter Vermaas and Pawel Garbacz

received little to no philosophical attention, which probably becomes more clear when it is noted that they are about functions only. In functional decompositions overall functions are related to subfunctions and not to, say, the structural parts (the components) of the technical systems that have the overall functions. The functional part-whole relationship described above is similarly a part-whole relationship directly between functions and not a functionally-defined structural part-whole relationship between technical systems and their structural parts. A wall, for instance, may be a functionally-defined structural part of a house but the wall is not a subfunction part of the function of the house; rather the function to support is a functional part of the function to provide shelter. In this chapter we consider interrelated functional descriptions of technical systems, specifically functional decompositions and the functional part-whole relationship that such decompositions define.4 Yet, given the embryonic stage of research, we cannot do more than announcing them as a topic for philosophical analysis; in this chapter we give a first explorative analysis and an incomprehensive sketch of what this analysis may have in store.

1.1 Functional descriptions in engineering To pin down functional descriptions, functional decompositions and the functional part-whole relationship to which they lead, we introduce some initial notation, anticipating a more thorough exposition in Section 2. Let Φ be the function that is decomposed and let φ1 , φ2 , . . ., φn be the subfunctions into which Φ is decomposed. We write this down as Decomp(Φ, Org(φ1 , φ2 , . . ., φn )), where Org(φ1 , φ2 , . . ., φn ) refers to a functional organisation, that is, a set of functional orderings of the subfunctions φ1 , φ2 , . . ., φn . This organisation is introduced, and in Section 2 we do this more systematically, in order to capture that the ordering of subfunctions matters in functional decompositions. If, for instance, the functions φ1 to heat with 150 degrees centigrade and φ2 to cool with 150 degrees centigrade are temporally ordered, one has two possibilities, which lead to different overall functions: in the area of food processing, the ordering “φ1 and then φ2 ” may make up the function to bake; and the ordering “φ2 and then φ1 ” may make up the function to refrigerate. Hence, an initial reading of Decomp(Φ, Org(φ1 , φ2 , . . ., φn )) is that the subfunctions φ1 , φ2 , . . ., φn ordered by the organisation Org(φ1 , φ2 , . . ., φn ) provide a decomposition of the function Φ. In general there exist in engineering more than one decompositions of a given function Φ, hence a reading of Decomp(Φ, Org(φ1 , φ2 , . . ., φn )) in which the subfunctions φ1 , φ2 , . . ., φn ordered by Org(φ1 , φ2 , . . ., φn ) are presented as the unique decomposition of Φ, is to be avoided. A decomposition Decomp(Φ, Org(φ1 , φ2 , . . ., φn )) is a description in which the subfunctions φ1 , φ2 , . . ., φn are related to one another by the organisation Org(φ1 , φ2 , . . ., φn ), and are in that organisation making up the function Φ. Moreover, the subfunctions φ1 , φ2 , . . ., φn can be taken as the functional parts of the overall function Φ. 4 See [Simons and Dement, 1996] for an analysis of the structural part-whole relations between technical systems.

Functional Decomposition and Mereology in Engineering

237

With this notation Decomp(Φ, Org(φ1 , φ2 , . . ., φn )), functional decomposition is analysed in primarily functional terms. The analysis is made broader by also including the objects and processes that are described in functional descriptions. Let S be a technical system that is described functionally by Φ, and let s1 , s2 , . . ., sn be the systems that are described by φ1 , φ2 , . . ., φn , respectively (the systems s1 , s2 , . . ., sn are functionally-defined structural parts of S; in Section 2.1 we discuss the relation between the systems s1 , s2 , . . ., sn and S in detail). With references to these systems we can, for instance, characterise more precisely the engineering activities in which functional descriptions play a role. Conceptual designing, for instance, can be analysed as follows. The starting point of conceptual designing is an overall function Φ and the aim is to determine the physical description of a system S, that is, the product-to-be, that can perform this function Φ. Assuming that this aim cannot be realised immediately by the engineers, i.e., assuming that they cannot derive the physical description of S directly from the description of the overall function, engineers reason by the following intermediate steps. First, they determine a series of subfunctions φ1 , φ2 , . . ., φn and an organisation Org(φ1 , φ2 , . . ., φn ) that defines a decomposition Decomp(Φ, Org(φ1 , φ2 , . . ., φn )) of the overall function. Second, engineers determine objects and processes s1 , s2 , . . ., sn that can perform these subfunctions φ1 , φ2 , . . ., φn , respectively. And, finally, they arrive at a physical description of S from the organisation Org(φ1 , φ2 , . . ., φn ) of the subfunctions5 and the physical descriptions of s1 , s2 , . . ., sn .6 If the description of the overall function Φ is very detailed, one may assume that the entity S consists of only the entities s1 , s2 , . . ., sn , but if this description is a coarse-grained one, say when Φ is only the primary function of the product-to-be, then S may also contain other entities. Think, for instance, of an aeroplane. If Φ is merely the primary function to fly, then the subfunctions in a functional decomposition Decomp(Φ, Org(φ1 , φ2 , . . ., φn )) of this overall function do not single out the systems that enable emergency evacuations of the passengers and crew. But if Φ refers to the more detailed function of to fly safely, then a functional decomposition should identify more of these latter systems. In reverse engineering the overall function Φ and the physical description of the system S are initially known, and the aim is to derive the subfunctions φ1 , φ2 , . . ., φn , their organisation Org(φ1 , φ2 , . . ., φn ) and the systems s1 , s2 , . . ., sn that perform these subfunctions.7 5 De Ridder [2007, Chapter 4] has argued that functional decompositions help engineers in reasoning from a purely functional description Φ of a product-to-be S to its physical description since especially the organisation of the subfunctions φ1 , φ2 , . . ., φn gives engineers early in the design process information about the spatiotemporal structure of the product: even if the systems s1 , s2 , . . ., sn are still only functionally characterised by φ1 , φ2 , . . ., φn , the organisation Org(φ1 , φ2 , . . ., φn ) fixes how these systems s1 , s2 , . . ., sn are spatiotemporally related. 6 Conceptual designing is not a process in which these three steps are taken one after the other; in engineering design literature it is emphasised that designing is an iterative process. 7 In reverse engineering some of the systems s , s , . . ., s may also be known initially. The n 1 2 objects in the set {s1 , s2 , . . ., sn } are technical components of S, and it seems reasonable to assume that engineers are able to recognise some of those components. The processes in {s1 , s2 , . . ., sn } may, however, be more difficult to identify.

238

Pieter Vermaas and Pawel Garbacz

In engineering knowledge bases, functional descriptions of technical systems contain all functions Φ, φ1 , φ2 , . . ., φn related to one another through organisations Org(φ1 , φ2 , . . ., φn ) and decompositions Decomp(Φ, Org(φ1 , φ2 , . . ., φn )).8 These functional description may also contain the systems S, s1 , s2 , . . ., sn , and, conversely and more typically, descriptions in engineering knowledge bases are descriptions of the systems S, s1 , s2 , . . ., sn to which the functions Φ, φ1 , φ2 , . . ., φn are added. With these characterisations of engineering activities, one can also to some extent formulate criteria that functional descriptions and functional reasoning should meet in order to be useful. In conceptual designing the subfunctions φ1 , φ2 , . . ., φn should be ones for which, given the technological state of the art, one has available or can find easily the systems s1 , s2 , . . ., sn that can perform them (see also Section 5). Decomposing, for instance, a function to emit light into the subfunctions to collect acoustic energy and to convert acoustic energy in light does not help in finding via the entities s1 and s2 a physical description of a light source S: currently systems s2 that can perform the function to convert acoustic energy in light are technologically not available. In reverse engineering, assuming that the description of the overall function Φ is sufficiently detailed, the systems s1 , s2 , . . ., sn should make up together a substantial part of S, in order to avoid the conclusion that the original designers of S added all kinds of spurious systems to their design of S. Finally, one of the goals for developing knowledge bases is to enhance communication about functional descriptions among engineers of different disciplinary backgrounds, between engineers and computer tools like CAD/CAM systems, and among computer systems. For achieving this goal, at least the subfunctions φ1 , φ2 , . . ., φn can be chosen from a standardised set (cf. [Hirtz et al., 2002]; see also Section 5).

1.2 Relevance The analysis of functional descriptions will be of relevance to a number of existing topics in philosophy, thus providing new and renewed links between the philosophy of the engineering sciences and other more classical domains in philosophy. We see four of such domains: philosophy of technology and philosophy of biology, both specifically with respect to accounts of functions, and epistemology and mereology. Existing philosophical accounts of the concept of technical functions provide input to the analysis of functional descriptions, and this analysis can in turn be seen as a next step in developing these accounts. The analysis of functional descriptions may in this way yield criteria for judging the accounts, for it is not yet clear if all these accounts can provide for a basis sufficient for taking this next step (we come back to this point in Section 3). 8 Alternatively one can analyse functional descriptions in engineering knowledge bases as descriptions containing functions Φ, φ1 , φ2 , . . ., φn related to one another through organisations Org(φ1 , φ2 , . . ., φn ) and a composition function (in the mathematical sense) Comp(Org(φ1 , φ2 , . . ., φn )) = Φ expressing that the subfunctions φ1 , φ2 , . . ., φn ordered by Org(φ1 , φ2 , . . ., φn ) make up Φ. We introduce this composition relation in Section 2.

Functional Decomposition and Mereology in Engineering

239

Since the analysis of technical functions is typically related and contrasted to the analysis of biological functions, it is to be expected that the analysis of functional descriptions establishes a similar interaction with philosophy of biology.9 In that domain there already is attention for the biological counterpart of functional decompositions (e.g., [Wimsatt, 2002]), which is especially manifested in the philosophy of science and epistemology literature on mechanistic explanations. In that literature scientific discoveries in biology and other sciences are interpreted as discoveries and explanation of mechanisms [Machamer et al., 2000; Craver, 2001; Bechtel and Abrahamsen, 2005], where the concept of mechanisms is sometimes explicitly introduced by means of technical artefacts [Craver and Bechtel, 2006]. In such explanations activities of mechanisms are analysed in terms of the organised10 objects and activities that make up mechanisms. Mechanistic explanations and functional descriptions in engineering are clearly related (Bechtel and Abrahamsen [2005, pp. 432—433] even speak about functional decompositions in the context of mechanistic explanations). A mechanism can arguably be interpreted in engineering as the entity S, the mechanism’s activity as the overall functions Φ, and the entities and activities that make up the mechanism as the objects and processes s1 , s2 , . . ., sn , revealing the epistemic dimension of functional decompositions as explanations. Functional reasoning in engineering is thus a form of explanatory reasoning, albeit one that provides more types of explanations than the one referred to in mechanistic explanations. In conceptual designing a functional decomposition of the overall function Φ of the product-to-be S implies also an explanation of the function Φ in terms of the organised subfunctions φ1 , φ2 , . . ., φn of the systems s1 , s2 , . . ., sn (see [de Ridder, 2006]). In reverse engineering and in knowledge bases, functional descriptions may, however, represent reasoning in which the subfunctions φ1 , φ2 , . . ., φn and their organisation are, in reverse order, explained in terms of overall functions Φ. We will not further develop these links between engineering functional descriptions, functional descriptions in biology and functional reasoning in epistemology,11 apart from a few isolated remarks in our concluding Section 6. Functional decompositions define in two ways part-whole relationships and their analysis may contribute to logic and specifically mereology. First, functional decompositions define a part-whole relation between the systems S and s1 , s2 , . . ., sn , with S taking the role of whole and s1 , s2 , . . ., sn taking the role of parts. These parts are called functional components by Peter Simons and Charles Dement [1996, p. 264] and we called them functionally-defined structural parts. Simons and Dement have argued that this part-whole relationship is different to a number of other part-whole relationships that are in use in engineering, and that it is not 9 Through this interaction with biological theories of functions, it may be expected that the analysis of functional decompositions in engineering will eventually also interact with philosophy of mind and the philosophy of the cognitive sciences, since accounts of biological functions are applied in these two domains. 10 Our notion of the organisation of functions Org(φ , φ , . . ., φ ) is adapted from the notion n 1 2 of organisation as used in the literature on mechanistic explanations. 11 See [Houkes, 2006] for a broader discussion of functional reasoning in engineering.

240

Pieter Vermaas and Pawel Garbacz

coinciding with the standard notion of the part-whole relationship as defined in mereology. Second, functional decompositions define a part-whole relation directly on the level of functions, that is, between the functions Φ and φ1 , φ2 , . . ., φn , with Φ taking the role of whole and φ1 , φ2 , . . ., φn taking the role of parts. We called this the functional part-whole relationship and this second relationship is to our knowledge not yet considered in mereology even though it may be taken as inducing the first structural part-whole relationship between the systems S and s1 , s2 , . . ., sn , since these systems are singled out by the functions Φ and φ1 , φ2 , . . ., φn (we come back to the functional part-whole relationship in Section 4). In addition to being relevant to philosophy, the analysis of functional descriptions can also be of benefit to engineering. Clearly engineering is initially primarily a source to this analysis. But when it will advance, philosophical analysis may contribute to the different engineering uses of functional descriptions. Conceptual clarity provided by the different accounts of technical functions may be of benefit to engineering as a whole; research on the epistemology of functional decompositions will be more of use to design methodology and to functional reasoning, whereas research on mereology may prove to be of value for developing engineering ontologies for knowledge bases (we come back to these points in Section 5).

1.3 Our plan In the next section we develop our characterisation of functional descriptions by defining the organisation Org(φ1 , φ2 , . . ., φn ) of sets of functions and the composition Comp(Org(φ1 , φ2 , . . ., φn )) of such sets. Then we discuss in Section 3 the extent to which existing philosophical accounts of technical functions provide already means for carrying out the analysis and how these accounts may in turn be affected by the analysis. In Section 4 we focus on mereology and in Section 5 we consider engineering work on functional decompositions and illustrate how this may interact with philosophical research on the topic. 2 FUNCTIONAL ORGANISATION AND FUNCTIONAL COMPOSITION In the previous section we introduced functional decomposition as a relation Decomp(Φ, Org(φ1 , φ2 , . . . , φn )) that is to be read as that the subfunctions φ1 , φ2 , . . . , φn ordered by the organisation Org(φ1 , φ2 , . . . , φn ) provide a decomposition of the function Φ. For defining this relation more formally, we here describe functional descriptions in general and more systematically. We start by introducing organisations Org(φ1 , φ2 , . . . , φn ) of functions, then we introduce compositions Comp(Org(φ1 , φ2 , . . . , φn )) of such organisations of functions, and finally we define decompositions Decomp(Φ, Org(φ1 , φ2 , . . . , φn )) in terms of such compositions. Let, firstly, F be the set of all functions of technical systems. This set contains thus both the overall function Φ and the subfunctions φ1 , φ2 , . . . , φn in the case of a decomposition Decomp(Φ, Org(φ1 , φ2 , . . . , φn )).

Functional Decomposition and Mereology in Engineering

241

Define, secondly, “→” as a relation of functional ordering between two functions φ and φ in F that expresses that functional output of φ is functional input to φ . Functional input and output of a function are for now primitive terms; in Section 5 we discuss an engineering approach to functional decomposition in which this input and output consist of flows of materials, signals and energies, yet other choices — the functional input and output may consist of forces and fields — are not ruled out. But, having this approach in mind, we assume that a functional ordering φ → φ implies the temporal ordering that φ is not later than φ . The ordering is in general neither symmetric nor reflexive, but for specific functions φ and φ it may hold that φ → φ and φ → φ, or that φ → φ: a force can be functional output of φ and functional input to φ , while the reaction force is output of φ and input to φ; heat may be functional output of φ and functional input to other functions but also input to φ itself. A functional ordering φ → φ is represented by an ordered pair φ, φ  that belongs to the Cartesian product F ×F . Define, thirdly, a functional organisation Org(φ1 , φ2 , . . ., φn ) of a set {φ1 , φ2 , . . ., φn } of functions as a set {φi → φj } of functional orderings of φ1 , φ2 , . . ., φn . A functional organisation Org(φ1 , φ2 , . . ., φn ) is thus a, not necessarily connected, network of orderings between the functions φ1 , φ2 , . . ., φn , as depicted in the figures 1 and 2. A functional organisation Org(φ1 , φ2 , . . ., φn ) is represented by a set {φi , φj } of ordered pairs from the Cartesian product F ×F .

Figure 1. A linear functional organisation Networks of organised sets {φ1 , φ2 , . . ., φn } of functions in F , like those depicted in figures 1 and 2, are in engineering taken as making up other functions Φ in F . We capture this by defining, fourthly, functional composition Comp(Org(φ1 , φ2 , . . ., φn )) which maps the functions φ1 , φ2 , . . ., φn in their organisation Org(φ1 , φ2 , . . ., φn ) to this function Φ, that is, Comp(Org(φ1 , φ2 , . . ., φn )) = Φ. Formally, the general notion of functional composition is represented by a set of ordered pairs {φi , φj }, Φ that contains a set {φi , φj } of ordered functions from F ×F and a function Φ defined on F . Yet, functional composition Comp does not map every set {φi , φj } in F ×F onto another function Φ in F . Engineering constraints rule out some ordered pairs φk , φl  as representing possible functional orderings φk → φl , and if a set

242

Pieter Vermaas and Pawel Garbacz

Figure 2. A more complex functional organisation

{φi , φj } is containing one or more of such impossible orderings, then this set {φi , φj } neither represents a functional organisation Org(φ1 , φ2 , . . ., φn ), nor is mapped by Comp to another function in F . A general example of an engineering constraint on functional orderings φk → φl is that φk is not later than φl since then functional output of φk cannot possibly be functional input to φl . And in the approach in which functional inputs and outputs are flows of materials, signals and energies, one can derive more specific constraints: if, for instance, φk has electricity as its functional output, it cannot provide input to a function φl that has only water as its functional input. With functional composition defined, we are now able to define the general notion of functional decomposition as a relation Decomp(Φ, Org(φ1 , φ2 , . . ., φn )) represented by a set of ordered pairs Φ, {φi , φj } such that a function Φ from F occupies their first argument place and a set {φi , φj } of ordered functions from F ×F occupies the second argument place. For this relation it holds that Decomp(Φ, Org(φ1 , φ2 , . . ., φn )) if and only if Comp(Org(φ1 , φ2 , . . ., φn )) = Φ. In other words, the relation Decomp is the inverse relation to Comp. A more verbose reading of this decomposition relation Decomp(Φ, Org(φ1 , φ2 , . . ., φn )) is thus that the subfunctions φ1 , φ2 , . . ., φn ordered by the organisation Org(φ1 , φ2 , . . ., φn ) provide a decomposition of the function Φ, since the composition Comp(Org(φ1 , φ2 , . . ., φn )) of φ1 , φ2 , . . ., φn ordered by Org(φ1 , φ2 , . . ., φn ) is equal to Φ. We assume that functional composition Comp(Org(φ1 , φ2 , . . ., φn )) is unique in the sense that a set of functions {φ1 , φ2 , . . ., φn } organised by Org(φ1 , φ2 , . . ., φn ) composes one function Φ. Yet we assume also that other compositions Comp(Org (φ1 , φ2 , . . ., φn )) may compose that same function Φ as well; engineering practices

Functional Decomposition and Mereology in Engineering

243

provide evidence for this latter assumption. Hence, a given function Φ may be decomposable in more than one way, meaning that Decomp(Φ, Org(φ1 , φ2 , . . ., φn )) should indeed not be read as that the set of functions {φ1 , φ2 , . . ., φn } ordered by Org(φ1 , φ2 , . . ., φn ) is the unique decomposition of Φ. In sum, both functional composition Comp and functional decomposition Decomp are relations, but only the former is in the mathematical sense a function.

2.1

Technical systems

With the above concepts and definitions, the analysis of functional descriptions is given in primarily functional terms; only in the functional ordering relation between two functions φ and φ there is a reference to the temporal ordering: φ is not later than φ . The functions in functional descriptions are, however, functions of technical systems, and we now broaden our analysis to those systems to make the concepts and definitions more tangible. Let s be the system that is described by the function φ, and adopt the convention that s1 , s2 , . . ., sn and S are the systems described by the functions φ1 , φ2 , . . ., φn and Φ, respectively.12 In general a function φ does not single out uniquely one system s due to the underdetermination that exists between functional and structural descriptions of systems; a function φ rather fixes a set of entities {s} that have that function: the function to conduct an electric current can be performed by a copper wire, but also by other technical systems. Conversely, for a system s there exists in general a set {φ} of functions it can perform: a copper pipe can conduct electricity but also guide a fluid or gas. A technical system s can, moreover, be an endurant or a perdurant;13 up to now we sloppily alluded to this distinction by distinguishing between technical systems that are objects and those that are processes, respectively. A consequence of these observations is that a functional organisation of a set of functions {φ1 , φ2 , . . ., φn } does not in general translate straightforwardly into an associated spatiotemporal organisation of a set of systems {s1 , s2 , . . ., sn }, and vice versa. There exist, for instance, functions φ that can describe both endurants and perdurants: the function to prevent depletion of soil, for instance, describes fertilizers (endurants) and rotary crop systems (perdurants).14 A functional organisation of a set of functions {φ1 , φ2 , . . ., φn } containing 12 In the examples discussed in this chapter the entities s , s , . . ., s are all different systems: n 1 2 system si is thus by our convention straightforwardly described by the function φi . More complex cases are possible as well. Two functions φj and φk , may, for instance, describe the same system, which then implies that the associated systems sj and sk are one and the same. 13 We use the terms “endurant” and “perdurant” in the standard philosophical sense, i.e., an endurant is an entity that persists through time by being wholly present at more than one time and a perdurant is entity that persists through time by having different temporal parts or stages at different times (cf., for instance, [Lewis, 1986]). Nonetheless, we do not want to engage in the debate between threedimensionalism and fourdimensionalim, in which these terms are applied. Without accepting or rejecting the actual existence of endurants and perdurants, we just provide a theoretically broad framework, which makes room for both types of entities. 14 One can distinguish in general three types of functions: functions φ that single out only endurants {s} (the function to support the back and bottoms of humans, for instance, seems to

244

Pieter Vermaas and Pawel Garbacz

such a “hybrid” function, clearly defines spatiotemporal organisations of the set of systems s1 , s2 , . . ., sn , that may be mutually quite different. And two ordered or unordered functions φ1 and φ2 may be performed by one system (the two functions to conduct an electrical current and to guide water may be performed by one single copper pipe), meaning that the functional organisation may even be lost all together. For analysing relationships between functional organisations of sets of functions and the associated organisations of the sets of technical systems, we continue by considering four special cases defined by the ontological opposition between endurants and perdurants. We sketch how in these examples the functional organisation of the functions φ1 , φ2 , . . ., φn is related to the spatiotemporal organisation of the systems s1 , s2 , . . ., sn . In the first two cases the systems in the set {s1 , s2 , . . ., sn } are all endurants and each si is functionally described by one separate function φi from the set of functions {φ1 , φ2 , . . ., φn }. In the first case the function Φ that composes φ1 , φ2 , . . ., φn in their functional organisation, i.e., Φ = Comp(Org(φ1 , φ2 , . . ., φn )), describes a system S that is an endurant as well; in the second case this function describes a system S that is a perdurant. In the third and fourth case the systems {s1 , s2 , . . ., sn } are all perdurants, is each si functionally described by one separate function φi , and is the composite function Φ describing an endurant S or a perdurant S, respectively. Table 1. The four special cases

Case Case Case Case

1 2 3 4

s 1, s 2, . . . , s n endurants endurants perdurants perdurants

S endurant perdurant endurant perdurant

So, let us start with considering the first two cases in which the functions φ1 , φ2 , . . ., φn are all functions of different endurants s1 , s2 , . . ., sn . These endurants are spatially positioned relative to one another and this positioning determines which of the endurants s1 , s2 , . . ., sn can physically interact with one another in a technological relevant way.15 There may, for instance, be a technologically relevant single out different chairs but not processes); functions φ that single out only perdurants (say, the function to provide health care); and functions for which the set {s} contains both endurants and perdurants (to prevent depletion of soil, as argued in the main text). Further examples of the third type reveal also that a function φ of an endurant s may sometimes be reinterpretable as a function φ of a perdurant s that takes place in that endurant s, and vice versa: the function to tear down city walls of the endurant “cannon”, for instance, can be taken alternatively as a function of the perdurant “shooting cannon balls” with that cannon, and vice versa. 15 Engineering determines what technologically relevant physical interactions are. In electrical engineering, electromagnetic interactions are relevant, in mechanical engineering, action and reaction forces are relevant, and so on.

Functional Decomposition and Mereology in Engineering

245

interaction between the endurants s1 and s2 , but not between s1 and s3 , and so on. Let us now capture this spatial positioning by the set of unordered pairs {{si , sj }} of endurants that do interact and call this the spatial organisation of the endurants {s1 , s2 , . . ., sn }. Assuming that the functional output of one function φ can only be the functional input to another function φ through physical interactions between the systems s and s that these functions are describing, this spatial organisation of {s1 , s2 , . . ., sn } puts a direct constraint on the functional ordering of the functions φ1 , φ2 , . . ., φn and thus on their functional organisation: Org(φ1 , φ2 , . . ., φn ) can consist of only a set {φi → φj } of functional orderings for which holds that for each element φi → φj the pair {si , sj } is an element of the spatial ordering of the endurants {s1 , s2 , . . ., sn }. The system S that is functionally described by the function Φ=Comp(Org(φ1 , φ2 , . . ., φn )), is now in its turn either an endurant or a perdurant. If S is an endurant, it is to be taken as an endurant that contains the physical composite of the endurants s1 , s2 , . . ., sn in their spatial organisation.16 If S is a perdurant, it is to be taken as a perdurant in which an endurant that contains that physical composite participates. An example of the first case is the composition of the three functions to attach to the seabed (φ1 ), to fix distance (φ2 ) and to attach to the vessel (φ3 ) of the three endurants “anchor”, “rope” and “knot”, yielding the function to fix the location of a vessel in open sea of an endurant “anchoring system”. The right spatial positioning of the anchor, rope and knot, allows action and reaction forces between the anchor and rope, and between the rope and the knot. Hence the spatial organisation of these endurants is captured by the set {{anchor, rope}, {rope, knot}}. The functional organisation Org(φ1 , φ2 , φ3 ) of the three functions is {φ1 → φ2 , φ2 → φ1 , φ2 → φ3 , φ3 → φ2 } with the functional inputs and outputs all forces, and this organisation satisfies the constraint that the spatial organisation puts on it. The endurant S consisting of anchor, rope and knot in their spatial organisation can perform the function Comp(Org(φ1 , φ2 , φ3 )) = Φ of fixing a ship. An example of the second case is the composition of the functions to remove solid particles (φ1 ), to absorb particles in solution (φ2 ) and to sieve bacteria (φ3 ) of the different reservoirs s1 , s2 and s3 (which are assumed to be endurants) part of a wastewater plant, yielding the function to purify water of the perdurant “ water treatment” (i.e., the process) that is performed by the plant. Assuming a linear spatial positioning of the reservoirs, fluids can flow from s1 to s2 , and from s2 to s3 letting the spatial organisation be equal to {{s1 , s2 }, {s2 , s3 }}. The functional organisation of the three functions is {φ1 → φ2 , φ2 → φ3 } (see also Figure 1) with the functional inputs and outputs all fluids in different phases of cleansing, and this organisation satisfies the constraint that the spatial organisation puts on it. The perdurant S is the water treatment process that takes place in the reservoirs 16 One may assume that the endurant S is just the physical composite of the endurants {s1 , s2 , . . ., sn }. This assumption is however challenged by the occurrence of all types of back-up systems and other safety systems in technical systems. It thus seems more tenable to hold that the physical composite of {s1 , s2 , . . ., sn } is part of S.

246

Pieter Vermaas and Pawel Garbacz

in their linear spatial organisation and this process can perform the composite function Φ to purify water. In the third and fourth case that we consider, the functions φ1 , φ2 , . . ., φn are all functions of different perdurants s1 , s2 , . . ., sn . These perdurants are spatiotemporally positioned relative to one another and this positioning again determines which pairs of perdurants can physically interact with one another in a technological relevant way. If two perdurants si and sj take place simultaneously and in one another’s vicinity, such an interaction can take place from si to sj and fromsj to si . Such bidirectional interactions are represented in the spatiotemporal organisation of the perdurants s1 , s2 , . . ., sn by unordered pairs {si , sj }, similar to the interactions between the endurants discussed in the first two cases. If a perdurant sk takes place before another sl , interaction is possible only from sk to sl . These unidirectional interactions are particular to sets of perdurants and we represent them by ordered pairs sk , sl . The spatiotemporal organisation of the perdurants s1 , s2 , . . ., sn thus has the form {{si , sj }, sk , sl }. Assuming again that the functional output of one function can only be the functional input to another function through a physical interaction, this spatial organisation of {s1 , s2 , . . ., sn } puts again a constraint on the functional organisation of these functions: Org(φ1 , φ2 , . . ., φn ) can consist of only a set {φi → φj } of functional orderings for which holds that for each element φi → φj the pair {si , sj } or si , sj  is an element of the spatiotemporal ordering of the perdurants {s1 , s2 , . . ., sn }. The system S that is functional described by the function Φ=Comp(Org(φ1 , φ2 , . . ., φn )) is an endurant or a perdurant. If S is an endurant, it is to be taken as an endurant which participates (possibly only partially) in the perdurants s1 , s2 , . . ., sn in their spatiotemporal organisation. If S is a perdurant, it is to be taken as a perdurant that consists of the perdurants {s1 , s2 , . . ., sn } in their spatiotemporal organisation. An example of the third case is the composition of the functional organisation of the functions to spin (φ1 ), to collect water (φ2 ), to produce hot dry air (φ3 ) and to vent humid air (φ4 ) of the four processes s1 , s2 , s3 and s4 that perform them in a drying machine (the endurant S), yielding the function Φ to dry clothes of that machine. The four processes can take place simultaneously, giving a spatiotemporal organisation containing every possible combination {si , sj }. The functional organisation Org(φ1 , φ2 , φ3 , φ4 ) is {φ1 → φ2 , φ3 → φ4 } with the functional inputs and outputs all consisting of water, and this organisation satisfies clearly the constraint that the spatiotemporal organisation of the processes {s1 , s2 , s3 , s4 } puts on it. The endurant S “drying machine” can perform the function Comp(Org(φ1 , φ2 , φ3 , φ4 )) = Φ of drying clothes. indent An example of the fourth and final case is the composition of the functions to emit radio waves with a specific frequency (φ1 ), to detect radio waves with the same frequency (φ2 ) and to display the direction and the delay of the reflected waves (φ3 ) of processes s1 , s2 and s3 (perdurants) that take place in the radar equipment, yielding the function to detect plane positions of a process S (also a perdurant) that includes these three (sub)processes {s1 , s2 , s3 }. If the

Functional Decomposition and Mereology in Engineering

247

processes s1 , s2 and s3 take place one after the other, their spatiotemporal organisation is given by {s1 , s2 , s2 , s3 }. The functional organisation Org(φ1 , φ2 , φ3 ) is {φ1 → φ2 , φ2 → φ3 } with the functional inputs and outputs all signals, and this organisation satisfies the constraint that the spatiotemporal organisation of the perdurants {s1 , s2 , s3 } puts on it. The perdurant S containing the three processes s1 , s2 and s3 can perform the function Comp(Org(φ1 , φ2 , φ3 )) = Φ of detecting the positions of planes. These four special cases suggest the following generalisation. For a functional description of the technical systems {s1 , s2 , . . ., sn , S} by means of the functions {φ1 , φ2 , . . ., φn , Φ}, where Comp(Org(φ1 , φ2 , . . ., φn )) = Φ and where the technologically relevant physical interactions between the systems {s1 , s2 , . . ., sn } are given by the spatiotemporal organisation {{si , sj }, sk , sl }, the following constraint on the functional organisation holds: Org(φ1 , φ2 , . . ., φn ) can consist of only a set {φi → φj } of functional orderings for which holds that for each element φi → φj the pair {si , sj } or the pair si , sj  is an element of the spatiotemporal ordering of the perdurants {s1 , s2 , . . ., sn }. This constraint is most probably not the only one that is possible or reasonable. One can envisage also that in engineering one wants to limit the number of technological relevant (and irrelevant) physical interactions between the systems {s1 , s2 , . . ., sn } that are not required by the functional organisation Org(φ1 , φ2 , . . ., φn ); such ‘spurious’ interactions may, for instance, lead to unintended effects. Moreover, the systems {s1 , s2 , . . ., sn } are all systems that have by definition functions, meaning that the endurants and perdurants part of S that do not have functions are ignored; including such nonfunctional systems in the description will most probably again amount to all kinds of constraints.

3

FUNCTIONAL DESCRIPTIONS AND PHILOSOPHICAL ACCOUNTS OF TECHNICAL FUNCTIONS

Philosophy has, as we have noted, produced a number of accounts that spell out what it means to describe individual technical objects or processes functionally.17 These accounts may be taken as a starting point for the analysis of functional descriptions, and this analysis can in turn be seen as a next step in the development of the accounts. We present here three archetypical approaches towards technical functions18 and assess them for their ability to be developed to also describe more complex functional descriptions. The analysis of functional descriptions becomes as such also a criterion for judging the versatility of the existing accounts of technical functions to incorporating engineering activities such as functional decomposition. In the first approach functions of technical systems are analysed in terms of the intentions of their designers or of their users: the function of a system is taken as 17 See

Preston’s chapter Philosophical Theories of Artifact Function in this Handbook. and Vermaas, 2009].

18 [Houkes

248

Pieter Vermaas and Pawel Garbacz

the capacity or goal for which it is designed or for which it is used. This approach can be called the intentionalist approach. An example is the account of technical functions by Karen Neander in which “the function of an artifact is the purpose or end for which it was designed, made, or (minimally) put in place or retained by an agent” [1991a; 1991b, p. 462]. The second approach emphasises the physical structure of technical systems and identifies the functions of parts of a system as the parts’ actual physical capacities that, together, causally contribute to a physical capacity of that system. This approach can be called the causal-role approach (the systems functions approach may be an alternative) and it was Robert Cummins [1975] who put it forward in detail. The third approach towards technical functions is one that defines functions relative to long-term developmental histories of technical systems. It takes distance from individual design processes of technical systems and individual uses, and focuses instead on their cultural dissemination. It identifies the function of a system with the capacity for which the system is reproduced for a longer period of time. This final approach can be called the evolutionist approach.19 The example is now Ruth Garrett Millikan’s notion of proper function [1984; 1993]. Of these three approaches especially the causal-role approach provides more than just an analysis of what it means to describe individual systems functionally. In this approach a set of actual capacities c1 , c2 , . . ., cn of a set of parts s1 , s2 , . . ., sn of a technical system that, together, causally contribute to a physical capacity C of that system, are all simultaneously taken as the functions φ1 , φ2 , . . ., φn of the parts s1 , s2 , . . ., sn . Hence, in this approach one has at once a functional description that consists of multiple functions {φ1 , φ2 , . . ., φn } and of a spatiotemporal organisation of the parts s1 , s2 , . . ., sn that puts constraints on the functional organisation Org(φ1 , φ2 , . . ., φn ) of these functions. Moreover, the interactions between the parts s1 , s2 , . . ., sn required for letting the capacities c1 , c2 , . . ., cn causally contribute to the physical capacity C, provides more definite information about the functional organisation Org(φ1 , φ2 , . . ., φn ).20 Conversely, functional descriptions containing the functions φ1 , φ2 , . . . , φn and Φ with Comp(Org(φ1 , φ2 , . . . , φn ))=Φ, and describing sets of technical systems {s1 , s2 , . . . , sn , S}, plausibly fit the causal-role account. In this account the functions φ1 , φ2 , . . . , φn , Φ single out capacities c1 , c2 , . . . , cn , C of the systems s1 , s2 , . . . , sn , S for which has to hold that, firstly, the systems s1 , s2 , . . . , sn are parts of S and, secondly, the capacities c1 , c2 , . . . , cn contribute causally to the capacity 19 The evolutionist approach has its origin in the analysis of biological functions and theories that fall under this approach are called etiological theories in that domain. In Preston’s chapter in this handbook, the evolutionist approach falls under the heading of non-intentionalist reproduction views. 20 In the causal-role approach a system s can have more than one function: s may have a capacity c as its function φ since c causally contributes to a capacity C of a system S, s may have a capacity c , different to c, as its function φ since c causally contributes to a capacity C  of a system S  , and so on. We here ignore questions about the relations between these multiple functions and consider only the organisation of the functions that parts s1 , s2 , . . ., sn have on the basis of their causal contributions to one capacity C of one system S.

Functional Decomposition and Mereology in Engineering

249

C. On the basis of the discussion of the four cases of functional composition given in section 2.1, the first condition plausibly holds and the second also although it introduces an explicit commitment that functions compose overall functions due to causal contributions. Consider, for instance, the example of the composition of the three functions to attach to the seabed, to fix distance and to attach to the vessel of anchor, rope and knot, yielding the function to fix the location of a vessel in open sea of an anchoring system. This example fits the causal-role approach, since the anchor, rope and knot are parts of the anchoring system and the capacities corresponding to their functions are causally contributing to the capacity “fixing the location of a vessel in open sea” of the anchoring system as a whole. Intentionalist approaches, in which users determine by their intentions the functions of technical systems, provide less means to analyse functional descriptions; intentionalist approaches that put designers at centre stage in determining functions, fare better. A technical system that is used for a specific capacity can in a user-intentionalist approach be ascribed that capacity as a function. Yet, for also ascribing functions to parts of that system, one has to assume that these parts are also intentionally used for specific capacities. This latter assumption seems, however, in general less tenable. In the anchoring-system example, for instance, it can be maintained that the system as a whole is intentionally used to fix the location of a vessel in open sea, but it is less tenable to maintain that a sailor who is throwing out an anchor line, uses the anchor for its capacity to attach to the seabed, the rope for its capacity to fix distance, and the knot for its capacity to attach to the vessel. Hence, user-intentionalist approaches provide for descriptions consisting of single functions ascribed to single systems, but may fail to give more complex functional descriptions. A designer-intentionalist approach can create such complex functional descriptions, since it can be maintained that designers in addition to designing technical systems as wholes, also design their parts intentionally. In the anchoring-system example, for instance, it can be said that the system as a whole, and the anchor, rope and knot were designed for the capacities to fix the location of a vessel in open sea, to attach to the seabed, to fix distance and to attach to the vessel, respectively. A first conclusion about the intentionalist approach seems therefore that it should de-emphasise the relevance of user intentions in the determination of functions of technical systems, in favour of designer intentions: designer-intentionalist approaches seem better equipped to incorporate more complex engineering functional descriptions. The account of technical functions as put forward by Wybo Houkes and Pieter Vermaas,21 is an example of a designer-intentionalist approach (it incorporates also elements of the causal-role and evolutionist approaches). In this account engineers can ascribe a capacity C as a function Φ to a technical system S as a whole, ascribe capacities c1 , c2 , . . . , cn as functions φ1 , φ2 , . . . , φn to parts s1 , s2 , . . . , sn of S, and relate these functions [Vermaas, 2006]. The difference with the causal-role approach is that in the Houkes-Vermaas account the entities S and s1 , s2 , . . . , sn need not actually have the capacities C and c1 , c2 , . . . , cn , respectively; the engineers need 21 [Houkes

and Vermaas, 2004; Vermaas and Houkes, 2006].

250

Pieter Vermaas and Pawel Garbacz

only to have justified beliefs about S and s1 , s2 , . . . , sn having these capacities and about the capacities c1 , c2 , . . . , cn contributing to the capacity c. Conversely, functional descriptions containing the functions {φ1 , φ2 , . . . , φn , Φ} with Comp(Org(φ1 , φ2 , . . . , φn ))=Φ, and describing sets of technical systems {s1 , s2 , . . . , sn , S}, plausibly fit designer-intentionalist approaches, since, for instance, such functional descriptions can be taken as supported by the beliefs of designers. For evaluating the construction of functional descriptions in evolutionist approaches, again a distinction is to be made between approaches that put users at centre stage and those that favour designers. Let a designer-evolutionist approach be one in which a function φ of a technical system s is the capacity c for which designers reproduce s’s, say by including s’s repeatedly in their designs as systems that can perform the capacity c. Such evolutionist approaches can provide for functional descriptions like functional compositions. Anchor systems are repeatedly included in designs of ships for the capacity to fix the location of a vessel in open sea, so this capacity becomes the anchor system’s function Φ in a designer-evolutionist approach. And also anchors, ropes and knots are repeatedly included in designs for the capacities to attach to the seabed, to fix distance and to attach to the vessel, respectively, so also these capacities become the functions φ1 , φ2 and φ3 of the anchor, rope and knot part of the anchoring system. In a user-evolutionist approach, in which a function φ of a technical system s is the capacity c for which users (let) reproduce s’s by using s’s repeatedly for the capacity c, more complex functional descriptions may in general be harder to create. The reason for this is that it is less tenable to maintain that users (let) reproduce parts s of larger systems S for a specific capacity c. Consider again the anchoring system S and the anchor s1 , rope s2 and knot s3 . The reproduction of the whole system S due to user demands does now not necessarily imply that the parts s1 , s2 and s3 are also reproduced due to user demands. Due to the underdetermination that exists between functions and systems, a technological innovation may make that at some point in time some of the parts of the anchoring system are changed. The rope s2 may, for instance, be changed into a chain s2 and the knot s3 by a welded joint s3 . In an extreme case there may exist for a specific system S with a fixed function Φ a number of different sets {s1 , s2 , . . . , sn }, {s1 , s2 , . . . , sn }, . . . of parts by means of which S can be constructed, showing that reproduction of S need not imply reproduction of its parts. This conclusion may be rejected by taking systems consisting of different parts also as different systems: one could take the position that the anchoring system consisting of the anchor s1 , chain s2 and welded joint s3 is a different system S  as compared to the anchoring system S consisting of the anchor s1 , rope s2 and knot s3 . By that position reproduction of the old anchoring system S still implies reproduction of the anchor s1 , rope s2 and knot s3 as well, and reproduction of the new anchoring system S  implies reproductions of the anchor s1 , chain s2 and welded joint s3 . For anchoring systems this position may be tenable, but from an engineering perspective this position is less plausible for more complex technical systems. In

Functional Decomposition and Mereology in Engineering

251

technical systems like cars and industrial plants, simple parts like wires, pipes and switches can typically be replaced by alternative parts, and this replacement does not immediately turn those systems from an engineering point of view into new systems. The conclusion that this reproduction of technical systems S due to user demands does not in general imply that the parts of S are reproduced by those user demands as well, may also be rejected by noting that in the case of parts the users should be identified with the designers of the system S: user demands make that whole systems are reproduced and make that designers reproduce the parts of those systems. This position, however, turns a user-evolutionist approach towards functions into a designer-evolutionist approach. Hence, our first conclusion about evolutionist approaches is that they should de-emphasise the relevance of the reproduction of systems due to user demands in the determination of their functions, in favour of the reproduction of those systems by designers.22 Conversely, functional descriptions containing the functions {φ1 , φ2 , . . . , φn , Φ} with Comp(Org(φ1 , φ2 , . . . , φn ))=Φ, and describing sets of technical systems {s1 , s2 , . . . , sn , S}, plausibly fit designer-evolutionist approaches. In those approaches the systems {s1 , s2 , . . . , sn , S} have the their functions {φ1 , φ2 , . . . , φn , Φ} only if designers have reproduced for a longer period those systems for the capacities corresponding to their functions {φ1 , φ2 , . . . , φn , Φ}, and the fact that designers did so, provides support to the conclusion that the functions φ1 , φ2 , . . . , φn compose in their functional organisation to Φ. Our assessment of the three philosophical approaches towards technical functions is clearly a preliminary one that needs to be developed. This development is bound to correct our conclusions that the causal-role approach is suitable to an analysis of functional decomposition, and that the intentionalist and evolutionist approaches are so only if they de-emphasise the role of user intentions and of user demands in their respective analyses of functions of technical systems.

4

THE FUNCTIONAL PART-WHOLE RELATIONSHIP AND MEREOLOGY

Functional descriptions, and specifically functional decompositions Decomp(Φ, Org(φ1 , φ2 , . . . , φn )) and compositions Comp(Org(φ1 , φ2 , . . . , φn ))=Φ, amount to a part-whole relation directly on the level of functions, where the overall function Φ takes the role of whole and the subfunctions φ1 , φ2 , . . . , φn take the role 22 Another attempt to save user-evolutionist approaches may consist of an argument that draws on the distinction between “selection for” and “selection of” as made in [Sober, 1993] in the context of biological evolutionary theory. One may assume that the evolution of technical systems can be described by a technological version of this theory and then argue that user demands amounts to a selection of systems S “for” the capacities corresponding to their functions Φ, which in turn amounts to a selection “of” the parts s1 , s2 , . . . , sn with the capacities corresponding to their functions {φ1 , φ2 , . . . , φn }. This argument seems, however, again blocked by the underdetermination phenomenon that a particular system S can in principle be constructed from different sets {s1 , s2 , . . . , sn }, {s1 , s2 , . . . , sn }, . . . of parts: a selection of S “for” the capacity corresponding to its function Φ does not unambiguously amount to a selection “of” a specific set of parts s1 , s2 , . . . , sn .

252

Pieter Vermaas and Pawel Garbacz

of parts. This analysis of functional descriptions faces, however, a serious problem with the understanding of the term “part”. If we understand the relation of parthood in the sense establish by standard mereology, then one can derive a number of consequences for subfunctions and overall functions that at least at first sight are incompatible with the engineering understanding of functional descriptions. In this section we present this problem and discuss possible solutions. First, we briefly sketch one of the possible axiomatisations of mereology. Then we expose the consequences involved in expressing the relation of being a subfunction in terms of the mereological relation of parthood. Finally, we try to investigate whether the current state of the art in philosophy and logic provides with some feasible means to work around the problem. So let us start with mereology itself. We cannot provide here a comprehensive exposition of the formal theory of mereology.23 The following axiomatisation, which is one among a number of equivalent systems, is given just for the sake of reference. Alfred Tarski formalised the standard mereology originally formulated by Stanislaw Le´sniewski (cf. the English translation of his works in [Srzednicki and Rickey, 1984]) by means of one primitive term: the relation of (improper) parthood denoted here by “≤”, where “x ≤ y” is to be read as “x is an improper part of y”. Tarski’s axiomatisation contains two axioms: (4.1)

If x ≤ y and y ≤ z, then x ≤ z.

(4.2)

If X = ∅, then there exists exactly one x such that x SUM X.

The expression “x SUM X” means that x is the mereological sum of the set X. The relation SUM is defined as follows:24 (4.3)

x SUM X ≡ ∀y ∈ X (y ≤ x) ∧ ¬∃y (y ≤ x ∧ ∀z ∈ Xy ∫ z).

(4.4)

x ∫ y ≡ ¬∃z(z ≤ x ∧ z ≤ y).

The expression “x ∫ y” is to be read as: x is disjoint from y. The complement of the relation of disjointness is the relation of overlap, which is usually denoted by “”. The problem now arises when we assume that axioms 4.1 and 4.2 apply to the domain of technical functions. If mereology is to be considered as a useful tool in analysing the part-whole relationship between subfunctions and overall functions, we must assume some kind of correspondence between this relationship and mereological terms. A suitable candidate mereological term for establishing this correspondence seems to be the relation SUM. (4.5) 23 An 24 Cf.

The functions {φ1 , φ2 , . . ., φn } are subfunctions of the function Φ iff Φ SUM {φ1 , φ2 , . . ., φn }. interested reader may consult [Simons, 1987] and [Casati and Varzi, 1999]. [Tarski, 1956].

Functional Decomposition and Mereology in Engineering

253

There are, however, a number of reasons to reject this choice. First, one can argue that SUM is not the proper relation to connect subfunctions and overall functions. Axiom 4.2 guarantees that for any non-empty set of functions, there exists the mereological sum of these functions. By means of 4.5 one thus has for any set of functions {φ1 , φ2 , . . . , φn } an overall function Φ that has them as subfunctions, meaning that for any set of functions {φ1 , φ2 , . . . , φn } there is a function Φ that can count as the composition of these functions. Yet, one can envisage sets of functions for which this will not be the case. Consider, for instance, the set {to maintain a pressure of 1 atmosphere in vessel x at time t, to maintain a pressure of 2 atmosphere in vessel x at time t}; by their intentional meaning it seems impossible to compose these two functions.25 Using the terminology defined in Section 2, we can explicate this fact by saying that for some sets of functions there does not exist any organisation by which they can be composed: an organisation of the function to maintain a pressure of 1 atmosphere in vessel x at time t and the functions to maintain a pressure of 2 atmosphere in vessel x at time t does not exist since these two functions cannot be temporally ordered one after the other, nor ordered simultaneously. Second, also if sets of functions {φ1 , φ2 , . . . , φn } properly compose overall functions Φ by SUM, there may be scientific and engineering reasons to nevertheless deny that they define reasonable compositions. The functions in the set {to cool, to allow a magnetic degree of freedom, to process a login request signal, to sand } may be taken as composing the overall function to smooth surfaces. But that composition seems also technologically nonsensical by the spurious first three elements. The reason for this embarrass de richesse is the fact that mereology lacks the conceptual tools needed to express any kind of scientific or engineering constraint or standard. Again, using the terminology defined in Section 2, we can formulate this second fact by saying that for some sets of functions specific organisations are from a scientific or engineering point of view (better) ruled out. Third, Axiom 4.3 guarantees that any non-empty set of functions composes exactly one other function. Consequently, if there are two engineering models in which the same set of subfunctions composes different functions due to different orderings between the subfunctions, the embedding of the functional part-whole relationship into mereology by 4.5 turns out to be inadequate. That such sets exist was briefly indicated at the beginning of section 1.1. Let φ1 be the function to heat with 150 degrees centigrade, let φ2 be to keep the temperature fixed, and let φ3 be to cool with 150 degrees centigrade. If these functions are performed sequentially in the order given, they may be taken as composing the function Φ1 to bake, but if they are performed in reverse order, they compose the overall function Φ2 to refrigerate. Again, the source of this incongruity seems to be the 25 [Simons, 2006] is one of the recent attempts at restricting the general principle of composition. Simons proposes to restrict this principle to the equivalence classes of mereologically disjoint objects. That is to say, if X is an equivalence class of this kind, then there exists such x that x SUM X. For instance, if we define in the set of protons and neutrons the relation: x exchanges gluons with y, then this relation will yield the equivalence classes of protons and neutrons such that each such class compose a single nucleus.

254

Pieter Vermaas and Pawel Garbacz

fact that we cannot express in mereology any kind of order among subfunctions, whereas ordering seems vital to functional composition and was therefore part to our characterisation of it. Using the terminology defined in Section 2, we can formulate this consequence by saying that for some sets of functions there can exist more than one organisation by which the functions in these set compose mutually different overall functions: for the functions {φ1 , φ2 , φ3 } introduced above one has, for instance, Org 1 (φ1 , φ2 , φ3 ) given by the orderings {φ1 → φ2 , φ2 → φ3 }, which defines the composition Comp(Org 1 (φ1 , φ2 , φ3 ))=Φ1 , and Org 2 (φ1 , φ2 , φ3 ) given by {φ3 → φ2 , φ2 → φ1 }, which defines Comp(Org 2 (φ1 , φ2 , φ3 )) = Φ2 = Φ1 . In sum, 4.5 is not a plausible candidate for a conceptual bridge between mereology and the functional part-whole relationship. Now the problem with applying mereology to functional descriptions is that there are not many alternatives to 4.5 available.26 The above consequences derived from 4.5 suggest replacing the SUM relation with a more flexible expressive composition relation since some sets of functions do not (reasonably) compose27 whereas other sets of functions have more than one composition. In a recent proposal to describe components of technical systems in mereological terms as given by Peter Simons and Charles Dement [1996] this flexibility may seem to be present.28 Their proposal presupposes one of the strategies of reconciling mereology with the real world, consisting of proclaiming that besides the notion of parthood defined in standard mereology, there are a number of more specific relations of parthood, e.g., being a functional part or being a component (cf. [Casati and Varzi, 1999, pp. 33—36]). These more specific relations need not satisfy all the requirements imposed by Le´sniewski on the general relation of parthood. First, we need to emphasise that Simons and Dement focus on functional parts of technical systems and not on subfunctions part of overall functions. Thus, their theory applies to what we have called the functionally-defined structural partwhole relationship between physical systems. Being in a seminal way sensitive to the gap between standard mereology and actual engineering, Simons and Dement suggest substituting the general notion of part with a more specific notion that would be applicable to technical systems. The latter notion is tensed: x is part of y at time t. The following axioms establish the formal properties of the relation that captures this specific notion: (4.6)

If x exists at time t, then x is part of x at t.

26 The problems with applying mereology outside the domain of mathematics are well investigated (cf. Rescher [1955], Casati and Varzi [1999], Pribbenow [2002], Johansson [2004]). But, due to the fact that functional descriptions have not received much philosophical attention, there are not yet solutions available that deal specifically with the problems associated with applying mereology to functional descriptions. 27 [Simons, 2006] takes this approach in mereology, as is described in an above footnote. 28 Other attempts at defining the notion of parthood suitable for technical systems may be found in [Tzouvaras, 1993], [Salustri and Lockledge, 1999], [Johansson, 2004] and [Vieu and Aurnague, 2005]. Nonetheless, none of these accounts concerns functional descriptions as discussed of this chapter.

Functional Decomposition and Mereology in Engineering

255

(4.7)

If x is part of y at time t, then x exists at t.

(4.8)

If x is part of y at time t and y is part of z at t, then x is part of z at t.

(4.9)

If x is a proper part of y at time t, then there exists z such that z is a proper part of y at t and x is disjoint from (i.e., does not overlap) z at t.

Simons and Dement define the relation of proper parthood and the relation of overlap in the following way. (4.10) x is a proper part of y at time t iff x is part of y at t and y is not part of x at t. (4.11) x overlaps y at time t iff there exists some z such that z is part of x at t and z is part of y at t. If one ignores for a moment the tensed character of the relation parthood in question, the theory developed in [Simons and Dement, 1996] might be seen as a weaker version of standard mereology. In particular, we do not find here the counterpart of Axiom 4.2, so none of the aforementioned mereological paradoxes occurs here. Simons and Dement claim further that the relation they define provides the most general framework for speaking about the mereology of technical systems. From a philosophical point of view, we may distinguish within this framework the following kinds of parts: • assembly components, which are those parts that are manipulated as units during the processes of assembly or manufacturing, • functional components, which are those parts that perform certain functions, • maintenance components, which are those parts that are manipulated as units during the process of repairing, and a number of other kinds. The actual engineering practice involves however more specific notions. Simons and Dement draw our attention to three kinds of mereological structures related to three different engineering specifications of parts of technical systems. The engineering bill of materials represents the mereological components of the abstract physical architecture of a given system. The manufacturing bill of materials represents the mereological structure determined by a manufacturing schema for constructing the technical system in question. Finally, the logistic bill of materials specifies those components of the system that are salient for maintaining it in a state of readiness (cf. [Simons and Dement, 1996, pp. 268—271]). Despite its conceptual complexity and despite the fact that it avoids the aforementioned paradoxes of standard mereology, Simons and Dement’s theory of the mereological structure of technical systems is not meant to be applied to the modelling of functional part-whole relations, and seems also not to be applicable to

256

Pieter Vermaas and Pawel Garbacz

it. Their theory does not make room for our notion of organisation or for some similar notion, which we believe is an indispensable aspect of any adequate conception of the functional part-whole relation. As a result, they can say neither that two parts of a technical system are sometimes not meaningfully or consistently combinable, nor that two subfunctions are sometimes not meaningfully or consistently combinable. And as a result, Simons and Dement cannot say that two parts of a technical system, or two subfunctions, are sometimes in multiple ways meaningfully and consistently combinable. As we can see again, mereological language as available in the literature is not expressive enough to capture functional part-whole relationship. In our characterisation of functional descriptions we introduced the concept of organisation to create this expressiveness. A way to improve on the conceptual bridge 4.5 between the functional part-whole relationship and mereology, may now seem be one in which this concept is explicitly introduced into the bridge. Using the notation defined in Section 2, we can rewrite 4.5 as 4.5*. (4.5*) Comp(Org(φ1 , φ2 , . . . , φn ))=Φ iff Φ SUM {φ1 , φ2 , . . . , φn }. 4.5* clearly implies 4.12.29 (4.12) For any set of functions φ1 , φ2 , . . . , φn , (i)

there is an organisation Org of these functions such that there exists a function Φ, for which it holds that Comp(Org(φ1 , φ2 , . . . , φn ))=Φ, and (ii) for any two organisations of those functions Org 1 and Org 2 , if Comp (Org 1 (φ1 , φ2 , . . . , φn ))=Φ1 , and Comp(Org 2 (φ1 , φ2 , . . . , φn ))=Φ2 , then Φ1 =Φ2 . In general, 4.12 is false. As we argued above, for some sets of functions there do not exist organisations or reasonable organisations of the functions and for other sets of functions there exist more than one (meaningful) organisation relative to which they compose mutually different functions. Still, in some restricted domains of engineering the specific organisation of subfunctions might not be that relevant, i.e., it may not affect the overall functions to which these subfunctions compose. If one describes a domain of this sort, then 4.5 can be seen as a definition of the purely mereological type of functional part-whole relation (within this domain). The example of the purely temporally organised functions to heat with 150 degrees centigrade, to keep the temperature fixed and to cool with 150 degrees centigrade, shows that temporal functional part-whole relations are in general not of such a mereological type. But more special cases of such temporal functional part-whole relations, or of specific spatial functional part-whole relations may be. Obviously, all these types are borderline cases of the more general spatiotemporal part-whole relationship defined in Section 2 and this general relationship is not mereological in the standard sense. 29 The implication from the left-hand side to the right-hand side in 4.5* is innocent; it is the reverse implication that should be blamed here.

Functional Decomposition and Mereology in Engineering

5

257

AN ENGINEERING APPROACH TO FUNCTIONAL DECOMPOSITIONS

Engineering designing is, as we mentioned in the introduction, one of the engineering domains in which functional descriptions are in use. In the first conceptual phase of designing initial requirements about the product-to-be — user needs and additional specification about, for instance, safety — are translated into overall functions of the product and these functions are then by functional decompositions analysed in terms of series of subfunctions. Yet, the description of conceptual designing is far from being unambiguous: the initial design requirements are rarely standardised and often acknowledged to change during the unfolding of the design process, and the resulting overall functions and their decompositions are usually expressed in informal terms, not meeting rigorous constraints. As a result design methodologists interested in analysing and improving conceptual designing, are facing the problem how to define and represent functions and their decompositions more rigorously, a problem that has become increasingly important by the growing use of computers systems, such as CAD/CAM tools, to support engineering design. Among the different conceptual models that are devised to solve this problem,30 we report here about what has become known as the Functional Modelling approach, since it provides relatively well-defined descriptions of functional decompositions, by which we can illustrate how engineering and philosophical research on functional decompositions can benefit from one another. The origin of this approach can be located in the fundamental work of Gerhard Pahl and Wolfgang Beitz [1996]; current research centres on a framework proposed by Robert Stone and Kristin Wood [2000]. We start by discussing Pahl and Beitz’ original ideas of associating functions with flows, and then move to current research on what has become known as the Reconciled Functional Basis.31 Pahl and Beitz define a function as a relation between an input and an output of a technical system (under a specific goal) and claim that technical functions are derived from flows [1996, p. 31]. A flow is either a conversion of material (e.g., a chunk of clay being converted into a vase), a conversion of energy (e.g., electrical energy being converted into heat), or a conversion of signal (e.g., a safety buzz indicating the high pressure of a vapour). Pahl and Beitz do not spell out what it means that functions are derived from flows. But in their definitions and examples they presuppose that any function boils down to a flow, for instance, when they refer to a function denoted by the expression “transfer torque”, which clearly is a flow of torque.

30 Research in design methodology is not converging to a single approach to functional descriptions or to functional decomposition. Authors use, and acknowledged to use, different notions of functions of technical systems and different ways of representing systems (see, e.g., [Umeda and Tomiyama, 1997; Chittaro and Kumar, 1998; Hubka and Eder, 2001]) and arrive at different models of, for instance, functional modelling (see, e.g., [Chandrasekaran, 2005; Far and Elamy, 2005; Van Eck et al., 2007] for surveys). 31 The functional modelling approach is also discussed in William H. Wood’s chapter “Computational Representations of Function in Engineering Design” in this Handbook.

258

Pieter Vermaas and Pawel Garbacz

Pahl and Beitz then continue with observing that an overall function of a technical system can be complex or less complex in three senses: the relationships between the input and the output of the technical system may be relatively opaque or rather transparent; the underlying physical processes might be intricate or simple; and, finally, the number of assemblies and component might be higher or lower. They suggest that in the case of functions of high complexity it is advisable to decompose such functions into subfunctions for three reasons: • to facilitate the subsequent search for design solutions, • to obtain simple and unambiguous function structures, and • to modularise the process of developing technical systems and their subsystems. These subfunctions are in turn relations between inputs and outputs defined in terms of the three flows, and are linked to one another by a causal net of such flows. This structure of subfunctions and flows as a whole should now establish the same relation between an input and an output as the overall function it decomposes. Any two subfunctions that are linked together by a certain flow in such a causal net need to be “compatible” and all subfunctions of a given overall function need to be combined in a “meaningful” way. Although it is unclear what Pahl and Beitz mean with these terms, they observe that the relationship between subfunctions and overall function is very often governed by certain constraints, inasmuch as some subfunctions have to be satisfied before others. Moreover, the Pahl and Beitz methodology leads to the question of whether there is a level at which the decomposition stops or effectively stops. Pahl and Beitz consider five types of conversions that they take as reasonably not sub-dividable, being to channel, to connect, to vary, to change and to store. Together with their distinction between flows of material, energy and signal, one arrives at a taxonomy of 15 basic functions occurring at the lower level of functional decompositions. In later developments, topic of the next part of this section, especially this taxonomy has been changed. Captured in our terminology, Pahl and Beitz arrive at functional decompositions Decomp(Φ, Org(φ1 , φ2 , . . . , φn )), where Pahl and Beitz’ overall function is Φ, the subfunctions are φ1 , φ2 , . . . , φn , and the causal net of flows defines the organisation Org(φ1 , φ2 , . . . , φn ) of these subfunctions (each flow in the net from one subfunction φi to another φj , defines a functional ordering φi → φj , and vice versa, meaning that Org(φ1 , φ2 , . . . , φn ) represents all connections in the net). Compared to our analysis of functional descriptions as given in Section 2, Pahl and Beitz introduce a number of additional assumptions. It seems at first sight that in their methodology the overall function Φ and the subfunctions φ1 , φ2 , . . . , φn in a functional decomposition are all typically describing systems S, s1 , s2 , . . . , sn that are perdurants, since conversions of flows refer to processes. However, the examples mentioned by Pahl and Beitz are functions of endurants — objects — and not of perdurants. That is, even if a function is a flow, i.e., a subtype

Functional Decomposition and Mereology in Engineering

259

of perdurant, this function is not a function of a perdurant; functions are special cases of perdurants that are ascribed to endurants. Second, by taking functions as conversions of flows of materials, energies and signals, functions seem to have to comply with physical conservation laws for such flows. The conversion of a signal flow representing a small amount of energy, into a much larger electromagnetic energy flow, seems not to be a possible function in Pahl and Beitz’ methodology.32 Finally, in a functional decomposition Decomp(Φ, Org(φ1 , φ2 , . . . , φn )) the subfunctions φ1 , φ2 , . . . , φn , are according to Pahl and Beitz ultimately not just any functions from the general set F of functions, but to be taken from the set of 15 conversions that are reasonably not sub-dividable. These subfunctions can thus only be to channel, to connect, to vary, to change and to store for materials, energies and signals. When evaluating our analysis of functional descriptions with the methodology of Pahl and Beitz, a first remark can be that our functional ordering relation φi → φj may be (too) coarse-grained. In our analysis this relation holds already if there is “something” that can count as functional output of φi that is functional input to φj . In the Pahl and Beitz methodology this something is categorised as (types of) materials, energies and signals. This opens the possibility to develop our analysis by distinguishing between (associated) different types of functional ordering relations between functions. Other ways of developing our analysis can be drawn from the second and third additional assumptions sketched above. These assumptions are also made in more recent work in the Functional Modelling approach to functional decompositions, suggesting that our analysis of functional descriptions as given in Section 2 is too liberal: in order to let it cohere more with engineering work on functional descriptions, we should incorporate a requirement that functional descriptions comply with physical conservation laws and a requirement that all functions can be decomposed in terms of what can be called basic functions. Our analysis does not provide the means to formulate the first requirement; for incorporating the second requirement we can define a set BF of basic functions and the condition that for all functions Φ in F there exists a decomposition Decomp(Φ, Org(φ1 , φ2 , . . ., φn )) with {φ1 , φ2 , . . ., φn } ⊆ BF . Taking some distance from the work of Pahl and Beitz, one can, however, doubt that especially the second additional assumption that functional descriptions have to comply with physical conservation laws holds for all engineering work on functional decomposition. Bell et al. [2007], for instance, accept functions that have a signal representing a small amount of energy as their input and a much larger 32 That functions in the Functional Modelling approach have to comply with conservation laws is not explicitly said by Pahl and Beitz or in the key publications on the Reconciled Functional Basis discussed later in this section. Yet, examples of conversions of material and energy flows that are clearly violating conservation laws are hard to find, and the tracking of (conserved) flows seems to be an important device in developing functional decomposition in Functional Modelling. Modarres and Cheon [1999], however, make an explicit link between functions and conservation laws in (their work on) Functional Modelling.

260

Pieter Vermaas and Pawel Garbacz

electromagnetic energy flow as their output.33 The physics that underlies such a function will clearly have to comply with conservation laws, but the input-output description of the function itself does not. The analysis of functional descriptions as given in Section 2, may therefore also be taken as a more general analysis, compared to which the additional assumptions in engineering methodologies for functional decomposition like the ones made by Pahl and Beitz can be analysed. This brings us to the benefits of philosophical research on functional descriptions to engineering: a philosophical analysis of such descriptions can help design methodologists with making their assumption explicit and with developing their work on functional decomposition. According to Pahl and Beitz the functions in decompositions are functions of endurants, but the examples given in Section 2.1 show that one can generalise functional decompositions to apply to also functions of perdurants. According to Pahl and Beitz functions comply with physical conservation laws, but this requirement may be dropped. And also the requirement that functions always have to be decomposable into functions from a set of basic functions may be questioned. One can, for instance, argue that this requirement has a context-dependent meaning. If functional decomposition is considered in the context of conceptual designing, this requirement may be that functions have to be decomposable into functions φ1 , φ2 , . . . , φn from a set of easily solvable functions, that is, from a set of functions for which, given the technological state of the art, one has available or can find easily the systems s1 , s2 , . . . , sn that can perform them (see also Section 1.1). Such a set of easily solvable functions varies with that technological state of the art. If, however, functional decomposition is considered in the context of engineering knowledge bases aimed at enhancing communication about functional descriptions among engineers, this requirement may have the form that functions are to be decomposable into functions φ1 , φ2 , . . ., φn from a standardised set of functions, irrespectively of whether the functions in this set are easily solvable. Such a standardised set clearly should not vary (too much) over time.

5.1 The reconciled functional basis A more recent research project that originates with the foundational work of Pahl and Beitz is the Reconciled Functional Basis project. This Reconciled Functional Basis (RFB, from now on) is the result of an effort towards establishing a standard taxonomy of basic technical functions (see, e.g., [Hirtz et al., 2002]) by reconciling two previous taxonomies: the NIST taxonomy (cf. [Szykman, et al., 1999]) and the older versions of the Functional Basis (developed in [Little et al., 1997; Stone et al., 1998; McAdams et al., 1999; Stone et al., 1999; Stone and Wood, 2000]). Each of these taxonomies is a result of empirical generalisation of engineering specifications. RFB analyses the notion of a functional decomposition against the background of its taxonomy of functions, which is based on a taxonomy of flows. RFB modifies 33 See,

for instance, their model of a torch [Bell et al., 2007, p. 401].

Functional Decomposition and Mereology in Engineering

261

the meaning of the term “flow” since here “flow” does not mean “a process of flowing” (e.g., removing debris), but “a thing that flows” (e.g., debris).34 More precisely speaking, in some papers, e.g., in [Stone and Wood, 2000] this term is used in both meanings, but the RFB taxonomy of flows is based on the latter sense. This shift in meaning is, to be sure, justifiable since it is hard to see how one might differentiate between a process of flowing and a function given the conception of Pahl and Beitz. The RFB whole taxonomy of flows is depicted in Table 2. RFB also contains a three-layer classification of what are called basic functions. Each type of function is accompanied by a definition (in natural language), example, and a set of synonymous names. The basic functions are divided in a first layer into eight primary types. Then, some primary basic functions are divided into types of secondary basic functions, and some of these secondary basic functions are in turn divided into types of tertiary basic functions. The whole taxonomy is depicted in Table 3. Of course, the RFB taxonomy of basic functions is not a model of functional decomposition. For instance, the fact that Divide and Extract are subtypes of Separate does not mean that the former are subfunctions of the latter. Moreover the basic functions are not functions in the sense the overall functions are, since the overall functions are (complex) modifications of specific input flows into specific output flows, whereas the basic functions are modifications generalised for the flows subjected. Hence, the basic subfunctions are in the RFB to be identified with basic functions operating on specific primary, secondary and tertiary flows. In RFB a functional decomposition is a conceptual structure that consists of an overall function that is decomposed, its subfunctions into which the overall function is decomposed, the flows which are modified by the subfunctions, and a net that links these modifications in an ordered way. The overall function to be decomposed is defined in terms of the flows it modifies, which are taken from the RFB taxonomy of flows. Each of its subfunctions is defined both in terms of the flows the respective subfunction modifies and in terms of its type of modification, which is taken from RFB taxonomy of basic functions. For instance, the overall function of a screwdriver, which is to tighten/loose screws, is defined by means of the following ten input flows and nine output flows (see also Figure 3). • input flows for the function tighten/loose screws: – energy flows: electricity, human force, relative rotation and weight; – material flows: hand, bit and screw ; – signal flows: direction, on/off signal and manual use signal ; • output flows for the function tighten/loose screws: – energy flows: torque, human force, heat, noise and weight; 34 In engineering design the term “flow” is used in the specific sense in which it is roughly equivalent to the term “process”.

262

Pieter Vermaas and Pawel Garbacz

Table 2. The RFB taxonomy of flows [Hirtz et al., 2002] Primary flow

Secondary flow Human Gas Liquid Solid

Tertiary flow

Object Particulate Composite

Plasma Material

Mixture

Status Signal Control

Gas-gas Liquid-Liquid Solid-solid Solid-liquid Liquid-gas Solid-gas Solid-liquid-gas Colloidal Auditory Olfactory Tactile Taste Visual Analog Discrete

Human Acoustic Biological Chemical Electrical Electromagnetic Energy

Optical Solar

Hydraulic Magnetic Mechanical Pneumatic Radioactive/Nuclear Thermal

Rotational Translational

Functional Decomposition and Mereology in Engineering

263

Table 3. The RFB taxonomy of functions [Hirtz et al., 2002] Primary functions Branch

Secondary functions Separate

Tertiary functions Divide Extract Remove

Distribute Import Export Channel

Transfer Guide

Connect

Couple

Transport Transmit Translate Rotate Allow degree(s) of freedom Join Link

Mix Actuate Regulate Control magnitude

Change

Stop Convert Provision

Increase Decrease Increment Decrement Shape Condition Prevent Inhibit

Convert Store

Contain Collect

Supply Sense Signal

Support

Indicate Process Stabilize Secure Position

Detect Measure Track Display

264

Pieter Vermaas and Pawel Garbacz

Figure 3. The RFB modelling of the overall function of a screwdriver [Stone and Wood, 2000, Fig. 2] – material flows: hand, bit and screw ; – signal flows: looseness/tightness. On the other hand, one of the subfunctions in the functional decomposition of this overall function tighten/loose screws is called convert electricity to torque (see Figure 4), which means that it is a function of the convert-type (cf. Table 3), and modifies one input flow to three output flows: • input flows for the subfunction convert electricity to torque: – energy flows: electricity; – material flows: none; – signal flows: none. • output flows for the subfunction convert electricity to torque: – energy flows: heat, noise and torque; – material flows: none; – signal flows: none.

Functional Decomposition and Mereology in Engineering

265

Figure 4. The RFB functional decomposition of a screwdriver [Stone and Wood, 2000, Fig. 4]

266

Pieter Vermaas and Pawel Garbacz

The task of a designer who performs a functional decomposition is to link any input flow of the overall function to be decomposed with some of the output flows. Any such link that starts with an input flow of the overall function and ends with one of its output flows is called a function chain. In RFB one distinguishes between two types of function chains: sequential and parallel. A function chain is sequential if it is ordered with respect to time, i.e., if any temporal permutation of its subfunction may in principle result in failing to perform the overall function. A parallel function chain is a fusion of sequential function chains that share one or more flows. In RFB one assumes that each subfunction of an overall function to be performed by a technical system S is realised by a component of S; however, the relation between subfunctions and components is many-to-many, i.e., one subfunction may be realised by several components and one component may realise more than one subfunction. The notion of functional decomposition developed within RFB plays an important role in what is called the concept generator, which is a web-based computational tool for enhancing conceptual design.35 The concept generator is to present a designer with a number of different solutions to his or her design problem on the basis of previously developed (and stored) high-quality designs. One of the input data to be provided for this tool is a function chain for a product to be newly developed. The output solutions describe the design solution in terms of the technical systems whose descriptions are loaded into the knowledge base of the concept generator. The functional decomposition links the overall function established by the generator with the conceptual components that compose a general description of the product that is construed here as a solution of the initial design problem [Strawbridge et al., 2002; Bryant et al., 2004]. The RFB proposal adds precision and a wealth of empirical details to the methodology of Pahl and Beitz. Its explicit aim to contribute to the standardisation of conceptual models in engineering makes it even more valuable for specifically mereological analysis of functional modelling. In our terminology, the overall function of an RFB functional decomposition Decomp(Φ, Org(φ1 , φ2 , . . ., φn )) may be any function Φ but the subfunctions φ1 , φ2 , . . ., φn are to be identified with RFB basic functions from Table 3 operating on specific RFB primary, secondary and tertiary flows from Table 2. The net of flows between the subfunctions φ1 , φ2 , . . ., φn defines their organisation Org(φ1 , φ2 , . . ., φn ). In RFB the overall functions Φ and the subfunctions φ1 , φ2 , . . ., φn in functional decompositions Decomp(Φ, Org(φ1 , φ2 , . . ., φn )) may be describing systems S and s1 , s2 , . . ., sn that are endurants and perdurants, but like in the methodology of Pahl and Beitz, again the additional assumptions are made that functions comply with physical conservation laws for flows, and that the subfunctions φ1 , φ2 , . . ., φn , are to be taken from a set of basic functions. A further additional assumption seems to be that the functional orderings φi → φj making up the organisations 35 See

http://function.basiceng.umr.edu/delabsite/repository.html.

Functional Decomposition and Mereology in Engineering

267

Org(φ1 , φ2 , . . ., φn ) of the subfunctions, are always asymmetric: flows between two subfunctions in functional decompositions like depicted in Figure 4, always go in one direction. The benefit of philosophical research on functional descriptions to engineering can again lie in making these assumptions explicit and in challenging them. The requirement that functions always have to be decomposable into RFB basic functions operating on specific RFB flows introduces again a tension between the goal of functional decomposition to facilitate designing and to facilitate communication. Consider, for instance, the basic function convert acoustic energy in electrical energy. The identification of this basic function in a decomposition of an overall function may be useful to a shared understanding of this overall function but will not help designers to easily find a corresponding design solution. A requirement that subfunctions are only ordered in one direction may in turn be helpful in engineering for managing the flow of materials, energies and signals, but may also be revealed to be an unnecessary constraint to the decomposition of functions. 6 PROVISIONAL CONCLUSIONS In this chapter we have introduced interrelated functional descriptions and specifically functional decompositions as a topic for philosophical analysis. We identified conceptual designing, reverse engineering and engineering knowledge bases as the engineering domains in which such functional descriptions are given. Special cases of interrelated functional descriptions that are important to engineering are functional decompositions: descriptions in which an overall function Φ of a technical system S is analysed in terms of a series of mutually ordered subfunctions φ1 , φ2 , . . . , φn describing systems s1 , s2 , . . . , sn part of S. Features of interrelated functional descriptions that are of interest to philosophy are that they define two different part-whole relationships: a functionally-defined structural part-whole relationship by which the systems s1 , s2 , . . . , sn are part of the system S, and a functional part-whole relationship by which the functions φ1 , φ2 , . . . , φn are part of the function Φ. We introduced in Section 2 a series of formally defined relations for capturing interrelated functional decompositions. A functional ordering φ → φ exists between two functions if functional output of φ is functional input to φ . The functional organisation Org(φ1 , φ2 , . . . , φn ) of a set of functions φ1 , φ2 , . . . , φn is defined as the set of (pair-wise) functional orderings that exists between these functions. Functional composition Comp(Org(φ1 , φ2 , . . . , φn )) maps the functions φ1 , φ2 , . . . , φn in their organisation Org(φ1 , φ2 , . . . , φn ) to another function Φ. Functional decomposition was finally taken as the inverse of composition and captured as a relation Decomp(Φ, Org(φ1 , φ2 , . . . , φn )) for which holds that Comp(Org(φ1 , φ2 , . . . , φn ))=Φ. These relations were, moreover, illustrated with a series of cases and examples in which also the systems s1 , s2 , . . . , sn and S described by the functions φ1 , φ2 , . . . , φn and Φ, respectively, were considered. The discussion of these cases/examples suggest that functional organisations in functional descrip-

268

Pieter Vermaas and Pawel Garbacz

tions are constrained by the spatiotemporal ordering of the systems described: φ → φ can hold only if there is a physical interaction from s to s . In this chapter we have, moreover, identified four areas in philosophy that may benefit from the analysis of functional descriptions: research on functions in philosophy of technology and in philosophy of biology, epistemology and mereology. Finally, we presented engineering work on functional decompositions, and considered how it can contribute to and benefit from our first exploration of functional descriptions. Here we sum up our findings, with the necessary provisos; since the analysis of engineering functional descriptions is a new topic in philosophy, this chapter can merely be an appetizer, shaped as well as limited by the directions we have chosen in our exploration, but hopefully rewarding nevertheless. The first philosophical area that can benefit from the analysis of functional descriptions is research on the concept of technical functions in philosophy of technology. Philosophical accounts of technical functions may initially be taken as mere means to developing the analysis of functional descriptions. We now argued in Section 3 that of the three archetypical approaches towards such accounts, only Cummins’ causal-role approach seems fully equipped to provide for such means; the other two approaches, that is, the intentionalist and the evolutionist ones, seem to be able to do so only if designers and not users are determining technical functions by their intentions or by reproduction, respectively. Hence, as its precondition, the analysis of functional descriptions requires that the role of users in the determination of technical functions in these latter two approaches is de-emphasised in favour of the role of designers. Analyses of technical functions in philosophy of technology and of biological functions in philosophy of biology are often in interaction, for instance, by using them as contrasts to one another, or by attempting to unify them. By this interaction, it may be assumed that the analysis of functional descriptions in engineering may also have an impact on philosophy of biology. We have not elaborated on the possible results of this interaction. In Section 4 we considered the area of mereology. We considered in particular the functional part-whole relationship between the functions φ1 , φ2 , . . . , φn and the overall function Φ given by their composition Comp(Org(φ1 , φ2 , . . . , φn )). We have argued that this functional part-whole relationship cannot be understood with standard mereology, but requires a modelling in mereology that can accommodate organisations of the subfunctions. In Section 5 we presented an engineering approach to functional decompositions, called functional modelling and described it in terms of the concepts part of our analysis of functional descriptions. Relative to this analysis functional modelling introduced a number of additional requirements. The more important ones were that functional descriptions have to comply with physical conservation laws, that there exists a set of basic functions into which other functional can be decomposed, and that functional ordering are always asymmetric functional orderings. We argued that these additional requirements could be incorporated into our analysis but could also be taken as assumptions in engineering approaches to functional

Functional Decomposition and Mereology in Engineering

269

descriptions that can be questioned, thus establishing the worth of philosophical analyses to engineering. The remaining philosophical area to which the analysis of functional descriptions can contribute is epistemology. We have described in Section 1 the relations between functional descriptions, functional reasoning and explanation, and linked especially functional decomposition to the literature on mechanistic explanation, a link that establishes another relation between analysing functional descriptions and philosophy of biology. Yet, we did not delve deeper into these relations. One reason for not considering epistemology in detail is that we still have to acknowledge that the analysis of functional descriptions is currently in a first phase. We hope to have shown with our analysis and with the presentation of the engineering approach to functional decomposition that a philosophical analysis of engineering functional descriptions is feasible and beneficiary. Yet, our analysis is still a first step aimed primarily at clarity on conceptual and mereological aspects of functional descriptions, and — hopefully — opening the way to a versatile and more broader analysis of functional decomposition in the engineering sciences. ACKNOWLEDGMENTS Research by Pieter Vermaas was supported by the Netherlands Organization for Scientific Research (NWO). BIBLIOGRAPHY [Bechtel and Abrahamsen, 2005] W. Bechtel and A. Abrahamsen. Explanation: A mechanist alternative. Studies in History and Philosophy of Biological and Biomedical Sciences, 36, 421—441, 2005. [Bell et al., 2007] J. Bell, N. Snooke, and C. Price. A language for functional interpretation of model based simulation. Advanced Engineering Informatics, 21, 398—409, 2007. [Bryant et al., 2004] C. Bryant, D. McAdams, R. Stone, T. Kurtoglu, and M. A. Campbell. A computational technique for concept generation. In Proceedings of IDETC/CIE 2005, Long Beach, CA, paper no. DETC2005-85323, ASME, 2004. [Casati and Varzi, 1999] R. Casati and A. Varzi. Parts and Places. MIT, 1999. [Chakrabarti and Bligh, 2001] A. Chakrabarti and T. P. Bligh. A scheme for functional reasoning in conceptual design. Design Studies, 22, pp. 493—517, 2001. [Chandrasekaran, 2005] B. Chandrasekaran. Representing function: Relating Functional Representation and Functional Modeling research streams. Artificial Intelligence for Engineering Design, Analysis and Manufacturing, 19, 65—74, 2005. [Chittaro and Kumar, 1998] L. Chittaro and A. N. Kumar. Reasoning about function and its application to engineering, Artificial Intelligence in Engineering, 12, 331—336, 1998. [Craver, 2001] C. F. Craver. Role functions, mechanisms, and hierarchy. Philosophy of Science, 68, 53—74, 2001. [Craver and Bechtel, 2006] C. Craver and W. Bechtel. Mechanisms. In The Philosophy of Science: An Encyclopedia, S. Sarkar and J. Pfeifer, eds., pp. 469—478. Routledge, 2006. [Cummins, 1975] R. Cummins. Functional analysis. Journal of Philosophy, 72, 741—765, 1975. [de Ridder, 2006] J. de Ridder. Mechanistic artefact explanations. Studies in History and Philosophy of Science, 37, 81—96, 2006. [de Ridder, 2007] J. de Ridder. Reconstructing Design, Explaining Artifacts: Philosophical Reflections on the Design and Explanation of Technical Artifacts. Simon Stevin Series in the Philosophy of Technology, Vol 4 (Delft University of Technology Ph.D. Thesis). 2007.

270

Pieter Vermaas and Pawel Garbacz

[Far and Elamy, 2005] B. H. Far and A. H. Elamy. Functional reasoning theories: Problems and perspectives. Artificial Intelligence for Engineering Design, Analysis, and Manufacturing, 19, 75—88, 2005. [Hirtz et al., 2002] J. Hirtz, R. Stone, D. McAdams, S. Szykman, and K. Wood. A functional basis for engineering design: Reconciling and evolving previous efforts. Research in Engineering Design, 13, 65—82, 2002. [Houkes, 2006] W. Houkes. Knowledge of artefact functions. Studies in History and Philosophy of Science, 37, 102—113, 2006. [Houkes and Vermaas, 2004] W. Houkes and P. E. Vermaas. Actions versus functions: A plea for an alternative metaphysics of artefacts. Monist, 87, 52—71, 2004. [Houkes and Vermaas, 2009] W. Houkes and P. E. Vermaas. Useful Material: An Action-Theory of Artefacts and Their Functions, Delft/Eindhoven Universities of Technology manuscript. 2009. [Hubka and Eder, 2001] V. Hubka and W. E. Eder. Functions revisited. In International Conference on Engineering Design, ICED 01-C586/102, Glasgow, Scotland, August 21-23, pp. 69—76. 2001. [Johansson, 2004] L. Johansson. On the transitivity of the parthood relations. In Relations and Predicates. H. Hochberg and K. Mulligan, eds., pp. 161—181. Ontos Verlag, 2004. [Kitamura et al., 2005/2006] Y. Kitamura, Y. Koji, and R. Mizoguchi. An ontological model of device function: Industrial deployment and lessons learned. Applied Ontology, 1, 237—262, 2005/2006. [Lewis, 1986] D. Lewis. On the Plurality of Worlds. Blackwell, 1986. [Little et al., 1997] A. Little, K. Wood, and D. McAdams. Functional analysis: A fundamental empirical study for reverse engineering, benchmarking and redesign. In Proceedings of the ASME Design Theory and Methodology Conference, Sacramento, CA, paper no. 97DETC/DTM-3879, ASME, 1997. [Machamer et al., 2000] P. Machamer, L. Darden, and C. F. Craver. Thinking about Mechanisms. Philosophy of Science, 67, 1—25, 2000. [McAdams et al., 1999] D. McAdams, R. Stone, and K. Wood. Functional interdependence and product similarity based on customer needs. The Journal of Research in Engineering Design, 11, 1—19, 1999. [Millikan, 1984] R. G. Millikan. Language, Thought, and Other Biological Categories: New Foundations for Realism. MIT, 1984. [Millikan, 1993] R. G. Millikan. White Queen Psychology and Other Essays for Alice. MIT, 1993. [Modarres and Cheon, 1999] M. Modarres and S. W. Cheon. Function-centered modeling of engineering systems using the goal tree–success tree technique and functional primitives. Reliability Engineering and System Safety, 64, 181–200, 1999. [Neander, 1991a] K. Neander. Functions as selected effects: The conceptual analyst’s defense. Philosophy of Science, 58, 168–184, 1991. [Neander, 1991b] K. Neander. The teleological notion of “function”. Australasian Journal of Philosophy, 69, 454—468, 1991. [Pahl and Beitz, 1996] G. Pahl and W. Beitz. Engineering Design: A Systematic Approach. Springer, 1996. [Pribbenow, 2002] S. Pribbenow. Meronymic relationships: From classical mereology to complex part-whole relations. In The Semantics of Relationships: An Interdisciplinary Perspective, R. Green, C. Bean and S. Myaeng, eds., pp. 35—50, Kluwer, 2002. [Rescher, 1955] N. Rescher. Axioms for the part relation. Philosophical Studies, 6, 8–11, 1955. [Salustri and Lockledge, 1999] F. Salustri and J. Lockledge. Towards a formal theory of products including mereology. In Proceedings of the 12 th International Conference on Engineering Design, Birkhofer, Meerkamm, and Vajna, eds., pp. 1125–1130. Technische Universit¨ at M¨ unchen, 1999. [Simons, 1987] P. Simons. Parts: A Study in Ontology. Clarendon, 1987. [Simons, 2006] P. Simons. Real Wholes, Real Parts: Mereology Without Algebra. Journal of Philosophy, 103, 597–613, 2006. [Simons and Dement, 1996] P. Simons and C. Dement. Aspects of the mereology of artifacts. In Formal Ontology, Volume 53 of Nijhoff International Philosophy Series, R. Poli and P. Simons, eds., pp. 255–276. Kluwer, 1996.

Functional Decomposition and Mereology in Engineering

271

[Sober, 1993] E. Sober. The Nature of Selection: Evolutionary Theory in Philosophical Focus. University of Chicago Press, 1993. [Srzedncki and Rickey, 1984] J. Srzednicki and V. Rickey, eds. Lesniewski’s Systems: Ontology and Mereology. Kluwer, 1984. [Stone and Wood, 2000] R. Stone and K. Wood. Development of a functional basis for design. Journal of Mechanical Design, 122, 359–370, 2000. [Stone et al., 1998] R. Stone, K. Wood, and R. Crawford. A heuristic method to identify modules from a functional description of a product. In Proceedings of DETC98, Atlanta, GA, paper no. DETC98/DTM-5642, ASME, 1998. [Stone et al., 1999] R. Stone, K. Wood, and R. Crawford. Product architecture development with quantitative functional models. In Proceedings of DETC99, Las Vegas, NV, paper no. DETC99/DTM-8764, ASME, 1999. [Strawbridge et al., 2002] Z. Strawbridge, D. McAdams, and R. Stone. A computational approach to conceptual design. In Proceedings of DETC 02, Montreal, paper no. DETC2002/DTM-34001, ASME, 2002. [Szykman et al., 1999] S. Szykman, J. Racz, and R. Sriram. The representation of function in computer-based design. In Proceedings of the 1999 ASME Design Engineering Technical Conferences, Las Vegas, NV, paper no. DETC99/DTM-8742, ASME, 1999. [Tarski, 1956] A. Tarski. Foundations of geometry of solids. In Logic, Semantics, Metamathematics, A. Tarski, pp. 24–29, Oxford University Press, 1956. [Tzouvaras, 1993] A. Tzouvaras. Significant parts and identity of artifacts. Notre Dame Journal of Philosophy, 34, 445–452, 1993. [Umeda and Tomiyama, 1997] Y. Umeda and T. Tomiyama. Functional reasoning in design. IEEE Expert, 12(2), 42–48, 1997. [van Eck et al., 2007] D. van Eck, D. McAdams, and P. E. Vermaas. Functional decomposition in engineering: A survey. In 2007 ASME International Design Engineering Technical Conferences and Computers and Information in Engineering Conference DETC2007, September 4-7, 2007, Las Vegas, NV, paper no. DETC2007-34232, ASME, 2007. [Vermaas, 2006] P. E. Vermaas. The physical connection: Engineering function ascriptions to technical artefacts and their components. Studies in History and Philosophy of Science, 37, 62–75, 2006. [Vermaas and Houkes, 2006] P. E. Vermaas and W. Houkes. Technical functions: A drawbridge between the intentional and structural natures of technical artefacts. Studies in History and Philosophy of Science, 37, 5–18, 2006. [Vieu and Aurnague, 2005] L. Vieu. and M. Aurnague. Part-of relations, functionality and dependence. In Categorization of Spatial Entities in Language and Cognition. M. Aurnague, M. Hickmann and L. Vieu, eds., pp. 483–509. John Benjamins, 2005. [Wimsatt, 2002] W. Wimsatt. Functional organization, analogy, and inference. In Functions: New Essays in the Philosophy of Psychology and Biology. A. Ariew, R. Cummins and M. Perlman, eds., pp. 173–221. Oxford University Press, 2002.

This page intentionally left blank

ARTEFACTS IN FORMAL ONTOLOGY Stefano Borgo and Laure Vieu

1

INTRODUCTION

Since the early 1990s, there has been an increasing interest, in the knowledge representation area, in formal systems which aim at describing general notions. Indeed, as the research community became aware of the limits of ad hoc approaches such as expert systems (that focus on very specialized domains and pay no attention to flexibility and reusability) and conceptual modeling frameworks (which are limited to capturing the distinctions explicitly needed in the target application or domain), the need to provide clear and unambiguous meaning to notions across knowledge domains became self-evident. Simultaneously, researchers started to look to the philosophical tradition for the characterization of general concepts and relations and to build formal systems based on ontological distinctions. All this led to a new area of research now widely, but perhaps naively, known as applied ontology or even simply ontology. Research in this area is both theoretical and application driven since it aims to reconstruct and to organize philosophical views into sophisticated formal systems whilst achieving the semantic integration of various information systems. Although the term ontology has been endowed with several meanings in the knowledge representation field, it does not directly refer to the discipline that goes back at least to Aristotle and has influenced most of western philosophy. The two disciplines are clearly distinct since the general goal of applied ontology is to construct and apply knowledge structures in order to reliably and automatically manipulate information content, and is motivated by research in areas like information retrieval, data management and conceptual modeling. Nonetheless, applied ontology is strongly linked to the philosophical approach because it relies on general philosophical principles and considerations to justify the various adopted knowledge structures. In this chapter we are interested in the ontological systems, hereafter referred to as ontologies, which satisfy the two main requirements of being formal and foundational. Roughly speaking, an ontology is formal if it is expressed in a logic language endowed with clear semantics (for instance in model-theoretic terms as first-order predicate logic [Hodges, 1983]). This choice is not determined by application concerns (at least not primarily), it emphasizes the relevance that semantic transparency has in this domain. By foundational ontologies we mean those knowledge Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

274

Stefano Borgo and Laure Vieu

systems that focus on very general and basic concepts (like object, event, state, quality) and relations (such as constitution, participation, dependence, parthood). Often the term formal ontology is used to cover both the above requirements, thus reminding us of Husserl’s distinction between formal logic and formal ontology. In this specific meaning, formal ontology is the study of the interconnections between entities, properties, parts, wholes and collectives. These are considered to be “formal” because they can be exemplified by objects in all domains of reality [Smith, 1998]. To take yet another perspective, one can say that formal ontology is the study of formal (logical) systems which are: general, since they include the most usable and widely applicable concepts; reliable, as they are logical theories with clear semantics, a rich axiomatization and carefully analysed formal consequences (theorems); and well organized, because they are based on philosophical principles the choice of which is explicitly motivated and remains independent from particular domains. In this work, when using the expression “formal ontology” we will be referring to this latter general characterization. Among the ontological systems that have been used in applications, there are just a few that more or less satisfactorily present these properties. They are bfo, dolce, gfo, opencyc, and sumo. They will be discussed in the following sections. We anticipate that unfortunately only a couple of these ontologies will include an explicit characterization of artefacts, the topic of this chapter, and that these existing characterizations will be quite shallow. An ontological system that properly models artefacts would have large potentialities in applications where artefacts are central. For instance, information systems that control processes in the manufacturing industry must rely on a rich but coherent notion of artefact. The integration of artefact data is crucial to concurrent engineering scenarios and to product lifecycle management. Our goal is to show that a satisfactory characterization of the artefact category can be made within a system as complex as a foundational ontology and within the constraints of classical formal logic, in other words, via an axiomatization. In particular, this means formalizing the notion in such a way that it does justice to several of the crucial properties (derived from philosophical considerations as well as from practical usage) that we usually ascribe to artefacts. At the same time, the definition has to remain independent from particular application domains. We achieve this goal by working within a specific foundational ontology (dolce) which, in its present form, lacks a characterization of the notion of artefact. However, our work goes beyond the extension of this system since it provides a general analysis of the category of artefacts which is helpful when formalizing this and related notions in other formal systems as well. It is important to note that in this enterprise we place ourselves within the framework of an ontology of social reality. Social reality [Smith, 1995; Searle, 1983] has to do with the part of reality that covers groups of agents and the social relationships therein, actions that are either collective or directed towards a social group, and the whole range of relevant resulting “social entities”, such as contracts or companies. Such entities are often dependent on mental attitudes, either

Artefacts in Formal Ontology

275

individual or collective. A formal ontology dedicated to social reality takes into account all such entities in its domain and attempts to characterize them by modeling general properties and facts. The ontology of social reality and the ontology of mind need to be separated from epistemological studies that would account for the ways in which an agent constructs his or her beliefs about reality, for instance by categorizing entities. Formal ontology takes for granted an objective1 point of view on reality, that is, a point of view that is external to any particular agent. This stance justifies the introduction of notions like “intentional selection” or “social artefact”, which are important to our approach as well as to the philosophical debates that inspired us.

2

FORMAL ONTOLOGY, SYSTEMS AND CHOICES

In this section, we shall begin by giving a brief description of the relevant foundational ontologies. This presentation, though admittedly brief and limited, provides information on the development and quality of the available ontological systems.2 Later we shall go on to discuss some of the ontological issues which help in the characterizing and comparing of these systems. Although the literature on foundational ontologies and their comparison remains scarce, the ontological topics we will review have largely been analyzed within the context of the philosophical tradition [Rea, 1997].

2.1

Existing formal ontologies

Basic Formal Ontology, BFO. The development of bfo3 was initiated in 2002 by the Institute for Formal Ontology and Medical Information Science (IFOMIS, first at the University of Leipzig and later at Saarland University). What characterizes this ontology is the careful description of both the general philosophical viewpoint and the organization of its structure. The ontology is only partly axiomatized and is not aligned with other knowledge systems, e.g. lexical resources. bfo is actually a framework of sub-ontologies linked together by formal relations. Every sub-ontology must be conceived of as a particular perspective on reality: the user selects the sub-ontology that she finds most appropriate to capture the aspects of the world she is interested in. The most important ontologies in bfo are: SNAP (a series of time-indexed snapshot ontologies, these are ontologies of endurants which, roughly speaking, are objects) and SPAN (a single ontology of perdurants which are, in rough terms, events). SNAP-bfo provides a list of all the entities existing in time such as cars, animals and mountains. They can be seen as 1 Certain ontologies adopt a cognitive approach: the categories of entities and the relations used to represent reality are chosen for their compatibility with those arguably used by humans in their language structures and/or their conceptual notions. If a cognitive approach is adopted this does not necessarily mean that the represented facts have to be subjective. 2 Data in this section have been collected in July 2007. 3 http://www.ifomis.org/bfo

276

Stefano Borgo and Laure Vieu

“a snapshot of reality” with no temporal extension. By contrast SPAN-bfo is a catalogue of events which necessarily occur over the course of time such as races, deaths and avalanches. SNAP and SPAN are intertwined via transontological relationships since SNAP entities participate in SPAN entities. bfo (version 1.1) consists of about 40 classes (categories) and is formalized in the weak language known as OWL (Web Ontology Language [Antoniou and van Harmelen, 2004]). It is partially available in first-order logic as well [Masolo et al., 2003]. bfo, which is freely available, has so far been mainly applied in the biomedical domain. Descriptive Ontology for Linguistic and Cognitive Engineering, DOLCE. dolce4 has been developed at the Laboratory for Applied Ontology (LOA), which is part of the Italian ISTC-CNR, as a reference module for a library of ontologies (mainly within the context of the WonderWeb Project). dolce has a definite cognitive bias since it aims at capturing the ontological categories underlying natural language and human common sense. The modelling choices and the various dolce terms and expressions are influenced by philosophy and linguistics. It has a tree-structure that is obtained by applying a “top-down” type of methodology. dolce provides a rich axiomatization of the different main categories and their relationships and it has been aligned to WordNet [Fellbaum, 1998; Pr´evot et al., 2005]. It is publicly distributed (see licence on the web site) and available in first-order logic (including KIF) and weaker languages like OWL, DAML+OIL and RDFS. It is also distributed as a software running in CASL, the Common Algebraic Specification Language,5 which makes available certain theorem provers and graphical devices. It is actively used in several projects in a variety of domains such as manufacturing, linguistics and the Semantic Web. Further information will be provided on this ontology in Section 4. General Formal Ontology, GFO. gfo6 was developed at the Onto-Med Research Group of the University of Leipzig. It consists of a taxonomy of entities, a taxonomy of relations, and an axiomatization in first-order logic (an axiomatization that is currently still under development). gfo is philosophically well motivated and its ontological choices clearly stated. A crucial guideline for gfo ontology resides in the notion of “levels of reality”. These are the material level (biological, chemical and physical), the mental level (that of psychological phenomena) and the social level (where we find agents, organizations and societies). This structure of levels forms the very basis of the ontology from the point of view that every element in gfo is required to participate in at least one of these levels. Note that gfo is a component of a larger perspective since it forms part 4 http://www.loa-cnr.it/DOLCE.html 5 http://www.brics.dk/Projects/CoFI/CASL.html 6 http://www.onto-med.de/en/theories/gfo/. At the time when this paper was being written a new version of gfo had just been presented. We therefore regret it has not been possible to fully analyze this ontology and compare it with all the others.

Artefacts in Formal Ontology

277

of the Integrated Framework for the Development and Application of Ontologies (IFDAO), an evolution of the General Ontological Language (GOL) project which dates back to 1999. The OWL axiomatization of the previous gfo version consisted of about 80 classes, some 100 subclass-relations and around 70 properties. We expect the new version to be of a comparable size. The Onto-Med group is active in the biomedical domain which explains why gfo has been predominantly applied in that area but it has also been implemented in conceptual modeling. As far as we know, no link to WordNet [Fellbaum, 1998] has yet been developed. The ontology is available through a modified BSD Licence. OpenCyc and Cyc. Cyc, owned by Cycorp Inc., is not a proper foundational ontology but rather a very large, multi-contextual knowledge base enriched with inference engines. It started being constructed in 1984 the aim being to develop a computer program “equipped” with a large amount of commonsense knowledge. The name “Cyc” derives from encyclopedia. The purpose of the project is not, however, to build an electronic encyclopedia, but rather to “complement” such information resources [Guha and Lenat, 1990]. The idea was to create a knowledge base with millions of everyday terms, concepts and rules that would capture the na¨ıeve human knowledge bank of reality. To ensure the consistency of the system, the information stored in Cyc is segregated into hundreds of contexts or “microtheories” which are ontologically primitive entities. Essentially, a microtheory is a bundle of assertions that share common assumptions about the world (microtheories are isolated by topics, granularity, culture, etc.). One basic assumption in Cyc is that everything is a member of some microtheory. opencyc7 was first presented as the “semantic heart” of Cyc but in reality it is just the open source version of the whole Cyc. Ontologically it appears to be deeply affected by cognitive assumptions since its categories try to capture na¨ıeve conceptions of the real world or, more simply, common sense knowledge. opencyc adopts a dedicated language (CycL), an extension of first-order logic. It adopts a set-theoretical approach and makes available hundreds of thousands of terms (categories and relations) and millions of assertions (facts and rules). Clearly, only a mere fragment of more general categories may be qualified as foundational. This fragment is not explicitly singled out by the developers. If one looks at the inheritance relation opencyc has a graph-like structure revealing a “bottom-up” approach, that is to say, the organization of the ontology was developed at a later stage to systematize the data present in the knowledge base. opencyc provides connections with other formalisms and domains: (partial) translators to OWL and Lisp, a connection to WordNet’s synset structure [Fellbaum, 1998; Pr´evot et al., 2005], and a Cyc-to-English generator. It should be emphasized that the overall system suffers from its commercial targets. For example, the semantics of Cyc partly depends on the implemented 7 http://www.opencyc.org

278

Stefano Borgo and Laure Vieu

inference engines which shows the relevance of performance at run-time and blurs the semantic transparency of categories and relations. In addition, the characterization of the ontological choices on which Cyc and opencyc are based seems to be a secondary task: the documentation is still incomplete and references to the established literature are lacking. opencyc does not claim to be a foundational ontology, nor is it a proper top-level ontology because of its size. Nevertheless, it formalizes top categories and their mutual relationships, provides fairly extended descriptions for most of the categories and the examples provided are helpful in conveying (at least in part) the intended meaning of the terms. Finally, the success of this long-standing project (it has been running for almost 25 years) is definitely questioned today.8 Suggested Upper Merged Ontology, SUMO. sumo9 was created by a private company in 2000 when publicly available specialized ontologies were merged into a single structure in an attempt to obtain a top-level ontology that would be suitable for interoperability, communication and search in the Semantic Web area. The project includes Sowa’s upper level ontology [Sowa, 2000] and the work of Guarino and colleagues on theories of space, time and objects [Borgo et al., 1996]. sumo is not directly influenced by a specific theoretical approach instead it selects from various ontological proposals the categories that seem to be of general use and are broadly accepted by the knowledge representation community. Thus, like OpenCyc, sumo is not a proper foundational ontology. Nonetheless, it is often included among the others because it is a large ontology used in several applications and one that provides an axiomatization of its terms in a rich language. sumo is divided into “sections” or kinds of complementary ontologies that cluster about 1000 terms and relations, 4000 axioms, and 750 rules (but the volume increases considerably if we include all the related domain ontologies). The sections isolate relevant topics: the Mereotopology ontology, for example, contains concepts that deal with the formalization of a general part/whole relation, while the Unit of Measure ontology provides definitions for unit systems. The inheritance structure of this ontology forms a tree obtained through a top-down methodology. It provides an axiomatization of the categories and their relationships in a version of first-order logic known as SUO-KIF as well as in OWL and can be exploited via several theorem provers. It is available in different natural languages and linked to WordNet [Fellbaum, 1998; Pr´evot et al., 2005]. sumo has been implemented in several projects. The distribution of the ontology is regulated by a licence (see the web site).

8 In specialized mailing lists like SUO (http://suo.ieee.org/) and ontolog-forum (http://ontolog.cim3.net/) a number of discussions have been conducted on this issue. 9 http://www.ontologyportal.org/

Artefacts in Formal Ontology

2.2

279

Ontological choices

So, ontologies are knowledge systems. They provide a framework in which every entity and relation we want to talk about can be classified. The construction of such a general framework is not simple and it relies on various basic principles, principles that are studied in depth in metaphysics. The best ontologies do indeed refer to the philosophical literature. Universals, particulars and tropes. The ontological distinction between universals and particulars can be characterized by taking the primitive relation of instantiation: particulars are entities that cannot have instances; universals are entities that can have instances. Linguistically, proper nouns are normally considered to refer to particulars, while common nouns refer to universals. For example, Pavarotti, the Italian tenor, is an instance of “person”, but he cannot himself be instantiated. (This characterization of the concept of universal is admittedly imprecise since it does not, for instance, clarify whether sets, predicates and abstract entities should be considered as universals or not. A complete presentation of the different notions demands an analysis of these other entities. Fortunately, we do not need to go into further details to understand the overall position of the ontologies described in Section 2.1.) By adopting a different ontological perspective, one can reject universals and rely on other entities in the way done in the trope theory [Campbell, 1990]. There one claims that the “whiteness” of the specific piece of paper one is holding is a trope (a located property or individual quality) while the universal “white” does not in fact exist. Roughly speaking, tropes are properties of specific material entities upon which they ontologically depend because if the entity ceases to exist, so too does the trope. Tropes do not have instances and cannot be confused with universals. The crucial ontological choice is the decision to include universals in the domain of the ontology, a necessary step if we want to refer to and classify them within the formalism. Some ontologies, such as dolce and sumo, are examples of foundational ontologies of particulars that do not refer directly to universals. opencyc and bfo admit both particulars and universals into the domain. dolce and bfo include some forms of tropes as well. Abstract and concrete entities. Abstract entities (or abstracts) are entities that do not exist in space or time which means to say that they are not located. In contrast, concrete entities (or concretes) are defined as entities that do exist, at least in time. Mathematical objects (like numbers and sets) are examples of abstracts, while ordinary objects (like cars and planets) and events (such as the 2008 Olympiad and the Second World War) are examples of concretes. The ontological formalization of abstracts seems to depend on negative properties (i.e. a lack of location) but that is not quite correct: one can take a different tack by claiming that abstracts are eternal and immutable in that they exist at all times

280

Stefano Borgo and Laure Vieu

and are unchangeable. A third alternative definition is based on the “causal criterion”: abstracts possess no causal power while concretes do. Note that in this way we have already switched to a different notion of abstract entity: if abstracts are “timeless”, as in the first definition, then it seems awkward to include them in causal relations; conversely it is possible to individuate entities located in time and space that lack any causal power, like the center of mass of the solar system [Lowe, 1998]. Existing ontologies tend to focus on the first kind of characterization. In dolce, temporal and/or spatial locations are not defined for categories like Abstract Quality and Abstract, so it is roughly the “negative” perspective that is adopted.10 dolce also distinguishes between “direct” and “indirect” location. Some entities do not have a direct location but they inherit their locations from entities on which they depend: tables inherit their temporal locations from the temporal locations of events of which they are participants. In opencyc, instances of the class SetOrCollection do not have spatial or temporal locations, thus opencyc adopts the same “negative” perspective on abstracts. Elements in TemporalThing, a subclass of Individual, are at least located in time while SpatialThing, also a subclass of Individual, are at least located in space. It is not clear if there are instances of SpatialThing that are not also instances of TemporalThing, that is to say, individuals that are located in space but not in time. If not, then all SpatialThing (like all TemporalThing) are concretes. In sumo, the distinction between Physical and Abstract is very similar to the distinction between concretes and abstracts: elements in Physical are said to be entities “that have a location in space-time” and in Abstract they are entities that “cannot exist at any particular place and time without some physical encoding or embodiment”. As far as we can see, the bfo ontology only takes into account entities existing in space and/or in time, that is to say, only concrete entities. Endurants and perdurants. Classically, endurants (also sometimes called continuants or objects) are characterized as entities that “are” in time; they are wholly present (all their proper parts are present) at any given time of their existence. On the other hand, perdurants (also called occurrents or events) are entities that “happen” in time, they extend in time by accumulating different “temporal parts”, so that, at any time t when they exist, only their temporal parts at t will be present. For example, the car you now own can be viewed as an endurant because it is now entirely present, while “your driving to the office” is a perdurant because “your driving out of the garage” is not present when “your driving through the city centre” happens (assuming that these are events that actually occur when you drive to the office). Sometimes only perdurants are admitted in an ontology. It then becomes possible to distinguish between ordinary objects (like “a person”) and events or processes (like “a person’s life”), relying on properties that lie outside 10 The courier font is used to denote the names of categories or classes of entities in the ontologies described. It is a notational system that is also adhered to in the quotations regardless of the authors’ chosen system of notation.

Artefacts in Formal Ontology

281

spatio-temporal aspects. (It should be noted that other ways of characterizing endurants and perdurants have also been proposed.) dolce assumes a classical view which accepts both the concept of endurant and that of perdurant. opencyc has a similar view: the class SomethingExisting (i.e. entities that remain relatively stable throughout their lifetimes) corresponds fairly closely to the classical concept of endurant. Analogously, SituationTemporal (or the union of Event and StaticSituation) corresponds to the classical notion of perdurant. In the case of sumo the distinction is between Object and Process. Here processes are characterized as “the class of things that happen and have temporal parts or stages”, while for objects a less standard interpretation is accepted: “an Object is something whose spatio-temporal extent is thought to divide into spatial parts roughly parallel to the time-axis”. In any case, note that in sumo objects and processes are considered to be necessarily located in the space-time. In bfo the distinction endurant-perdurant forms the basis to the development of two separate sub-ontologies (in other words, the two types of entities do not coexist in the same ontology): SNAP-bfo contains only endurants, while SPAN-bfo contains only perdurants. Co-located entities. No matter what one decides about the ontological status of space and time, one can include spatially and/or temporally co-located objects. It is natural to accept objects that are temporally co-located (at least in part), like the moon and the earth or oneself and one’s clothes but the embodiment of spatially (or rather spatio-temporally) co-located distinct objects can sometimes be questioned. This issue is addressed by posing questions like: Is a hole different from the region of space it occupies? Is a statue different from the matter which constitutes it? Is a person different from his or her body? The subject is complex and includes rather difficult relations like identity across time, material constitution, essentiality and modality. dolce, which takes a multiplicative approach, uses spatial co-location and the relations of dependence and constitution in order to “stratify” co-located entities. For example, persons (elements of Agentive Physical Object) are constituted by their bodies (Non-agentive Physical Object), and elements of Physical Object are constituted by elements of Amount of Matter. opencyc has a weaker position. It sometimes adopts a genuine multiplication of co-located entities (e.g. it considers a statue and the matter that constitutes it as distinct co-located entities). On the other hand, it takes persons to be entities in the class CompositeTangibleAndIntangibleObject, so that Marilyn Monroe, for instance, has two components: a “body” and a “mind”. sumo, as far as we can see, does not have an explicit position on this issue. It might be that this ontology suffers from the heterogeneity of the basic theories on which it is founded (see page 278). Finally, bfo approaches this issue by distinguishing between different SNAP ontologies: a statue would be an element of an ontology of art (or of social reality) while the material it is made of would fall into an ontology of physical reality.

282

Stefano Borgo and Laure Vieu

3

ARTEFACTS IN EXISTING ONTOLOGIES

This section examines the present situation regarding the formalization of the notion of artefact in formal ontology. In this respect, it constitutes a preliminary step to our study that begins in Section 5. Generally speaking, the study of artefacts has attracted the attention of researchers from different domains ranging from engineering to philosophy and psychology to linguistics. Despite this wide interest, all attempts to either formally or informally characterize a shared notion of artefact have come up against serious problems. Existing formal ontologies indirectly register this fact. Indeed, of the five systems listed above, only opencyc and sumo include a category of artefacts. This might be partly attributable to contingent aspects: certain formal ontologies are still strengthening their top-level concepts, while others focus primarily on domains where the role of artefacts is marginal, such as in the biomedical study of living organisms.11 In other cases, it is all the result of specific choice: the notion of artefact may not be considered by some to be so general and basic that it warrants inclusion in foundational ontology. Nonetheless, we suspect that were a clear and shared characterization of artefact available in the literature, all formal ontologies would happily make it part of their system. After all, it is indisputable that artefacts are omnipresent components of our social life. These considerations highlight the need to extend and enrich the debate on the properties that distinguish artefacts from other entities, a topic that will be resumed later in the paper. For now we shall consider how the category of artefact is introduced into the two formal ontologies that deal with it. Since we are not interested in the particular formalization of these systems, we shall concentrate on the overall notion by looking at the inheritance structure for this category and at the explanations accompanying the relevant categories but we will not take into account the formal issues.

3.1 Artefacts in opencyc In opencyc,12 the class Artifact is part of UniversalVocabulary, one of the most general microtheories of opencyc. From the given description, Cyc’s assertions on this concept are “intrinsic to the [artefact] concept’s nature and cannot be violated in any context”. That is, no exceptions are possible, not even within other microtheories. The top class is actually called Artifact-Generic and is described as “a collection of things created by agents” where an Agent-Generic is a “being that has desires or intentions, and the ability to act on those desires or intentions” (it includes social organizations like legal corporations and animals). Elements 11 See,

for example, the Gene Ontology: http://www.geneontology.org/index.shtml. and citations are from opencyc 1.0.2: http://www.opencyc.com.

12 Data

Artefacts in Formal Ontology

283

of Artifact-Generic like a hammer or a bird nest can be tangible or alternatively intangible like a set of laws. The category Artifact-Generic breaks down into Artifact and Artifact-Intangible. If we ignore the latter (which collects entities like computer languages and legal agreements), an element of Artifact may be said to be an inanimate thing which is “at least partially tangible” and “intentionally created by an agent (or group of agents working together) to serve some purpose or perform some function”. The result of an assembling operation or of a modification of existing matter may not be an artefact unless the creating agent performs it intentionally and with a purpose. In opencyc some amounts of matter are classified as artefacts under the class ArtificialMaterial, a subclass of Artifact. According to the informal description, this class contains “portion[s] of artificial stuff that was intentionally made by some agent(s), such as Plastic...” but excludes the byproducts of such activities. Note that the top category Artifact-Generic has a second (orthogonal) partition. The subcategories here are Artifact-NonAgentive and Artifact-Agentive. The first class collects artefacts which are not agents, like bicycles whilst the latter collects agents which are themselves created by agents, such as organizations. Unfortunately the opencyc documentation gives no information on the underlying view and no link with the specialized literature is provided. Some rationalisations and underlying intuitions are to be inferred from the category descriptions, when provided, and the formalization itself (for which the licence is needed). Finally, the hierarchy of concepts below the Artifact category seems to be more driven by application goals (like the need to have a detailed and broad coverage of concepts of specific domains) than by ontological factors. Otherwise, it seems difficult to justify the presence (at the same level in the hierarchy) of Artifact subcategories like InstrumentalArtifact (“A sub-collection of Artifact. Each instance is an artifact (or system of artifacts) that is instrumental in accomplishing some end.”), ItalianCuisine (“the collection of instances of what many Americans tend to think of as Italian food”), and StuffedToy (no description provided).

3.2

Artefacts in sumo

In sumo13 , an artefact is described as a “CorpuscularObject that is the product of a Making” where an element of the CorpuscularObject is a “SelfConnectedObject whose parts have properties that are not shared by the whole”. These descriptions do not provide a clear view because sumo adopts a very general notion of property which means that if one has two entities (e.g. an object and one of its proper parts) it will seem possible to find properties with which to distinguish them (for instance, properties relative to size). From the above definitions, it must follow that sumo artefacts are located in space-time and are self-connected. In other words, sumo artefacts are non-scattered and physical. This confinement to physical entities is adopted in several discussions even in the philosophical litera13 Data and citations derive from the sumo webpage: http://www.ontologyportal.org/ (Sept. 2007).

284

Stefano Borgo and Laure Vieu

ture but the constraint on self-connected objects is new and not explicitly justified: a jigsaw puzzle would count as an artefact in sumo only when assembled. Another peculiarity resides in the description of the category Making seen above and said to characterize artefacts. Making is described as follows: “The subclass of Creation in which an individual Artifact or a type of Artifact is made”. In turn, Creation is said to be “the subclass of Process in which something is created”. These descriptions are hopelessly circular, they do not help us to understand the intended meaning of the categories. When tracing the links between categories, we did not find a direct relationship between the part of the hierarchy containing Making and the category Agent. This is surprising because in the relevant literature the direct and strong connection which exists between artefacts and agents is crucial to the ontological status of artefacts. From the information collected, we can conclude that the class Artifact in sumo captures the notion of physical, self-connected, and made (created) objects. Since the terms and relations used in sumo are poorly characterized and documented, the true extent of this class of sumo remains obscure.

3.3 Artefacts and the other ontologies The foundational ontologies bfo, dolce and gfo do not introduce artefacts in their hierarchy. However, two of them, dolce and gfo, seem to be in a position to provide a definition for the category. From our previous description, it can be concluded that bfo has a limited number of categories and thus few expressive tools to introduce artefacts. If it is true that the ontology has a minimal notion of function, the lack of axiomatization and the limited number of classes makes the formalization of a category of artefacts problematic: one should introduce several preliminary notions simultaneously. We are not aware of any attempt to extend bfo with a category of artefacts. dolce does not provide a notion of function but it has a very expressive framework to deal with qualities. However, some categories in the hierarchy are not fully formalized while some of these (e.g. Social Object) are crucial to model artefacts. To our knowledge, there has been no attempt to extend the ontology in this direction. gfo has carried out an interesting study on the notion of function and it has a fairly rich hierarchy that may provide the tools to define the category of artefacts or, at least, a generalization of it in terms of functionalities. The developers of gfo have some idea of how to tackle the artefact category14 but the ontology is still being developed and no extension of this is expected in the near future. Since artefact is a notion that has direct consequences for applications one might assume that expanding our analysis to include non-foundational ontologies would lead to an interesting characterization of this notion. However, ontologies developed within certain application domains only rarely introduce categories of 14 H.

Herre, personal communication.

Artefacts in Formal Ontology

285

artefacts (but one often finds a subcategory for products). Even then their characterization of artefact (or of product, for that matter) is of little or no help. These ontologies are poorly characterized and the descriptions of their categories show that they are based on practical considerations and rely on the implicit knowledge shared in the community they address. Their proposals are therefore only valid when applied to their particular domain but remain, even in these cases, quite minimal. 4

THE DOLCE FOUNDATIONAL ONTOLOGY

Our next step is to elaborate a formal notion of artefact that is philosophically motivated. To surmount the problems shown above we will make explicit our modeling choices while also relating our approach to major philosophical positions in this area. This analysis will, we hope, be widely applicable. We want it to go beyond a philosophical discussion though. We also want to establish a formal characterization based on a specific formal framework. In the following pages we will therefore try to establish a balance between achieving a general analysis of the notion of artefact and recognizing the constraints imposed by the language and ontological choices behind the specific ontology used. If we leave aside opencyc and sumo which, as pointed out in their respective descriptions, are not proper foundational ontologies, we can choose between bfo, dolce and gfo. These systems all seem quite promising but each has its own particular drawbacks. If we bear in mind that bfo is only partly axiomatized and provides only a few categories, and that gfo is still working out the formalization of its new version, then it would seem that we would be better off working with the dolce ontology. A positive feature of this latter ontology is its rich and flexible framework for modeling qualities which provides an interesting theoretical tool for the capturing of formal and practical distinctions. It will therefore be exploited extensively in our work. On the negative side, the ontology only focuses on particulars (individuals), as it will become clear below.

4.1

An introduction to dolce

The Descriptive Ontology for Linguistic and Cognitive Engineering, dolce [Masolo et al., 2003] (www.loa-cnr.it/DOLCE), concentrates on particulars, that is, endurants, perdurants, qualities and abstract entities. It does not attempt to provide a taxonomy of properties and relations which are only included in the system if deemed crucial for characterizing particulars. We mentioned above that dolce adopts a multiplicative approach: it assumes that different entities can be co-located in the same space-time. For example, a car and its matter are captured in dolce as two distinct entities (as opposed to being different aspects of the same entity). The reason for this lies in the different sets of properties that these entities enjoy: the car ceases to exist if a radical change of shape occurs (e.g. when it is crushed and cannot be repaired) while the amount of

286

Stefano Borgo and Laure Vieu

matter is not affected by this type of change (changes in spatial properties, like in shape or connectedness, is irrelevant for the identity of an amount of matter; only mereological properties can affect it). Going back to the classical example of the statue made of clay, for example, dolce might be said to model the statue and the amount of clay as different entities which share the same spatial (and possibly even temporal) location; the amount of clay used constitutes the statue. This allows the user to capture the strong intuition that a scratched statue is different (since it is scratched) while still remaining the same statue that it was before. In dolce this is possible because the identity of the statue itself might not be affected by minor scratches, but the identity of the clay is because scratches are the result of parts of the clay breaking off. The category Endurant collects objects like cars and bits of matter like steel blocks, while events like the making of this car and the moving of that steel block fall into the category of Perdurant. The term “object” itself is used in the ontology to capture a notion of unity or wholeness as suggested by the partition of the category Physical Endurant into Amount of Matter whose elements are (amounts of) gold, air, etc.; Feature (a hole, a boundary); and Physical Object (a hammer, a human body). See Figure 1. Some of the categories are informally described in Section 4.2.

PT Particular

ED Endurant

PED Physical Endurant

M Amount of Matter

F Feature

POB Physical Object

PD Perdurant

NPED Non-physical Endurant

…

EV Event

AS Arbitrary Sum

NPOB Non-physical Object

Q Quality

TQ Temporal Quality

STV Stative

ACH Achievement

ACC Accomplishment

ST State

PRO Process

…

…

…

…

… TL Temporal Location

PQ Physical Quality

… SL Spatial Location

AB Abstract

AQ Abstract Quality

…

…

TR Temporal Region

… APO Agentive Physical Object

NAPO Non-agentive Physical Object

MOB Mental Object

SOB Social Object

ASO Agentive Social Object

SAG Social Agent

T Time Interval

Fact

Set

PR Physical Region

… S Space Region

R Region

AR Abstract Region

…

NASO Non-agentive Social Object

SC Society

Figure 1. Taxonomy of dolce basic categories. (From [Masolo et al., 2003])

Artefacts in Formal Ontology

287

Every subcategory of Endurant and Perdurant is associated with a group of qualities. Qualities and their values (qualia) form distinct categories of entities in dolce and the distinction between individual quality, quale, and quality space has been established in order to capture several common sense intuitions in a coherent and consistent way. Individual qualities, like the colour of this pen, inhere in specific individuals meaning that the colour of this pen is different from the colour of that pen no matter how similar the two pens may be. These qualities can change over the course of time since the colour of this pen can match the colour red today and the colour pink tomorrow. In contrast to individual qualities, qualia are not entity dependent. An example of a quale is a specific colour, like, for instance, red. Intuitively, these entities are obtained by abstracting individual qualities from time and from their hosts (see the discussion on tropes in Section 2.2). If the colour of this pen and the colour of that pen match the same shade of red, then they have the same (colour-)quale. In this respect, qualia represent perfect and objective similarities between (aspects of) objects. Quality spaces correspond to different ways of arranging qualia. They are motivated by similarities between objects or aspects of objects. By allowing different spaces for the same quality, different structures can be imposed on qualia (for example, a geometry, a metric, or just a topology) and this makes it possible to differentiate several quantitative and qualitative degrees of similarity (consider, for instance, the different ways of classifying and measuring colours). The actual list of qualities associated with an entity depends on the user. Standard examples of qualities are shape and weight (usually taken to be qualities of endurants) and duration and direction (which are usually qualities of perdurants). However, these examples are not enforced by the ontology itself which is indeed neutral on the topic.

4.2

Some categories and relations in dolce

Several of the categories given in Figure 1 will be used to characterize artefacts. Here we shall just consider a few of them (including their relations) by way of informal introduction to the dolce terminology. Their formal names as used in the next sections are given in italicised parenthesis. The interested reader can find in [Masolo et al., 2003] the formal system together with a more detailed discussion. Let us first recall the general category Endurant (ED) which collects entities that are wholly present at any time when they are present like, for instance, Bush, the first car built by Ferrari and the steel of the Eiffel tower. The elements of Physical Endurant (PED) are the endurants located in space-time, e.g. Gandhi’s glasses as opposed to La Divina Commedia poem. This latter entity is classified as a Non-physical Endurant (NPED). Amount of Matter (M ), e.g. some oxygen, Feature (F ), e.g. a curve, and Physical Object (POB ), e.g. a car have already been mentioned. Regarding agency, Non-agentive Physical Object (NAPO) pertains to the physical objects to which one cannot ascribe intentions, beliefs or

288

Stefano Borgo and Laure Vieu

desires (like a product or a ticket). A person falls into the category Agentive Physical Object (APO) which is different from the category of social entities, called Social Object (SOB ), where we find things like organizations, companies, and their institutional artefacts such as constitutions and cheques. Social systems, such as a linguistic community, the people of a village or western society, form a subcategory denoted as Society (SC ). Entities that happen in time fall into the Perdurant (PD) category. Recall that, in informal terms, a perdurant is an entity that is only partially present whenever it is present. In this category we find happenings like football games and productions. These entities have temporal parts (like the first half of the game) as well as spatial parts, i.e. parts that are spatially identified (like the event being restricted to half of the football field during the game). Note that endurants are not parts of perdurants but they participate in them instead (this relation is labelled PC ). Some perdurants (like finishing a race or reaching the top of a mountain) are further classified in the subcategory Achievement (ACH ). They are distinguished according to two properties: they have no temporal parts (e.g. instantaneous events) and their type is not preserved by sum: if we add together two consecutive events consisting in, say, finishing a book we get a new complex event which does not add up to the finishing of a book. Contrast this with events like drilling or walking: if we add together two consecutive walking events we still have a (possibly complex) walking event. Perdurants that have temporal parts but behave in the same way regarding their sum, like football games, fall into the Accomplishment (ACC ) category. Note the distinction between finishing a book (an achievement) and reading a book (an accomplishment). Achievements and accomplishments form the category of eventive perdurants (EV ). Entities of a different type are found in the Quality category (Q), which covers all individual qualities. As discussed above, individual qualities can be seen as instantiations of basic properties of endurants or perdurants (shape, weight, duration, electric charge; usually qualities can be perceived or measured). The term “individual” is used to mark the fundamental role of the inherence relationship between an entity and its own qualities. Every endurant (or perdurant) comes with its physical (or temporal) qualities. Note that qualities are particulars in dolce that are not to be confused with properties (universals). Expression qt(q, x) stands for “q is an individual quality of x”. Qualities are associated with quality spaces and the position an individual quality has in a space is called a quale. We write ql(r, q, t) to indicate that “r is the quale of the endurant’s quality q during time t” while qlT (t, x) stands for “t is the quale of the temporal quality of x”. (Note the temporal parameter in ql(r, q, t). If we want to evaluate “John is 5 feet tall”, we have to be explicit when this sentence is stated as John’s height changes over the course of time. Instead, relation qlT (t, x) describes temporal location and it is used to formalize, e.g. “the party last Sunday lasted from sunrise to sunset”.) Each quale informally identifies a class of equivalence with respect to some individual quality, that is, with respect to an aspect of the entities. For instance, the same weight quale is associated with all the weight

Artefacts in Formal Ontology

289

individual qualities that are ontologically indistinguishable (i.e. independently of any measuring instrument we have). From the remaining relations we will make use of parthood as in “x is part of y”, written as P (x, y). The relationship being present states when an entity exists in the world, thus one writes PRE (x, t) to mean that “x is present in the world at time t”. Earlier, we mentioned participation: expression PC (x, y, t) stands for “endurant x participates in perdurant y during time t”. Constitution, another crucial relation in dolce denotes a strong form of dependence: K(x, y, t) stands for “x constitutes y during t”. That is the relationship that holds between an amount of matter and a statue so that the statue cannot be present unless the material it is made of is also present. Finally, we will use an extension of dolce proposed in [Masolo et al., 2004], and consider the category of concepts (CN ) together with the relationship classification, written CF , that relates concepts and their “instances” at any one time. One writes CF (x, y, t) to state that “at time t, x satisfies the concept y”. Concepts are not standard universals because concepts are individuals that depend on agents who create them or possibly on societies that adopt them. To account for their dependence, concepts are classified as particulars in this extension of dolce. Above all else, they are endurants, not abstracts, since they exist in time. Creating a concept means among other things providing a definition for it; the satisfaction of a concept is characterized by the constraints stated in the description defining a concept. For instance, the concept of Italian President has been created and defined by the Italian constitution which has been adopted by the Italian people. 5

EXTENDING DOLCE TO ARTEFACTS

We now turn to examining how the formal ontology dolce can be extended to include a category of artefacts. As explained above, this endeavour aims at showing that philosophical findings in this little explored domain can be successfully incorporated into an axiomatic first-order theory. This is not to say that we provide a general definition of artefact tout court. Indeed, in the literature the term “artefact” has been associated with a variety of meanings depending on the research domain and on the specific viewpoint of the authors. Here, we formally develop a coherent view formalized in a way that is compatible with the basic dolce choices.

5.1

Artefacts in the taxonomy

Most authors acknowledge that the notion of artefact seems to cover entities in a large variety of basic categories. Following the dolce taxonomy of basic categories depicted in Figure 1, it can be easily argued that artefacts may be either endurants (bottles and laws) or perdurants (judgements, performances and wars).15 Endurants can be physical (bottles, glass and robots) or non-physical 15 There are no abstract artefacts because according to dolce, all abstract entities are out of time (cf. p. 279). All artefacts are created, even non-physical ones like logical theories or novels.

290

Stefano Borgo and Laure Vieu

(amounts of music, laws and social institutions). For some authors, the whole Social Object category appears to fall under the larger class of non-physical artefacts [Searle, 1995]. Included in physical endurants are amounts of matter (pieces of glass or plastic), physical objects, which can be non-agentive (bottles, pens and paperweights) as well as agentive (robots, and arguably, bred animals and perhaps intended babies), and features (folds in a skirt, tunnels in mountains). So it would seem that artefacts are not a separate category in the ontology, but rather a class of entities overlapping a variety of categories. As we will see below though, the identity criteria for artefacts, that is to say, their intentional nature, force us to regard artefacts as entities which actually are in a separate category. As a starting point for providing a more general notion, we will only focus here on a subclass of the larger category of artefacts. This paper considers artefacts that correspond to physical endurants only, and among them, amounts of matter and non-agentive physical objects only.16 For this first step, we will simply add to the dolce categories the category of Physical Artefact which falls under the category of Physical Endurant as a new sibling of Amount of Matter, Physical Object, and Feature. This category contains the most prototypical artefacts (e.g. tools like knives and pens) or, in other words, the least controversial ones so that we can be confident that it is covered by any specific view on artefacts. It arguably is the most studied category in the literature [Baker, 2004; Kroes and Meijers, 2006; Elder, 2007; Thomasson, 2007]. It is also the easiest to grasp in an ontology that is particularly well developed in the domain of material entities and the related fundamental relations, as is dolce. The category of artefacts considered is quite big and presents a variety of interesting subclasses. We will not go into more specialized notions though; that is, we are not going to provide definitions to distinguish, for instance, “technical artefacts” or “works of art” within this class. These are crucial subclasses but a justification and presentation of their specific distinctions would be too detailed for the purposes of the present chapter. Before proceeding further, we should however make it clear that we are dealing primarily with specific tokens, e.g. with the telephone that sits on Mary’s desk, and not with artefact types like the telephone. This implies that we are ignoring here the important process of designing (possible or impossible) artefacts, a process that often precedes the actual creation of any technical artefact token. The focus on tokens is natural within the dolce framework since, as pointed out in Section 4, this ontology is about particulars. Nonetheless, one sees that from the formalization, a notion of artefact type does emerge. We shall introduce and discuss this notion in Section 7.1. Focusing on token artefacts in an ontology theory means being concerned with the nature of these objects, or in other words with the essential properties that make the difference between artefacts and non-artefacts, They all have a creation time before which they don’t exist and after which they do, i.e. they are in time, and thus non abstract. 16 We actually even exclude living entities from Non-agentive Physical Objects (e.g. plants, viruses), but the remaining subclass is not identified as a category in dolce at present.

Artefacts in Formal Ontology

291

and with the relationships between artefacts and other entities. As stated earlier, we are not concerned with the epistemological processes of recognition and categorization by an agent in the presence of a new object which may possibly be an artefact, nor by the process of designing a new artefact for a given purpose.

5.2

The approach

In accordance with the quite limited, although recently significantly increasing, literature on the ontology of artefacts [Dipert, 1993; Baker, 2004; Houkes and Meijers, 2006; Kroes and Meijers, 2006; Elder, 2007; Thomasson, 2007], our approach seeks to do justice to the complex nature of artefacts, which blends a physical substrate (regarding the category of physical artefacts we are concerned with here) with intentional aspects as well as social aspects. We adopt the view that artefacts have an ontological status, in other words that they are full citizens of our ontology, and that artefacts are essentially the result of an intentional act of their creator. As we shall see below, two entities are therefore distinguished, the purely physical object or amount of matter that constitutes the artefact and the artefact itself, which emerges when it is created with both its physical and intention-based properties. Among these artefacts, generated by the private intention of their creators, we can further distinguish social artefacts (or artefacts proper in Dipert’s terminology [Dipert, 1993]), whose intention-based properties take on a social dimension, more specifically, they are artefacts which are recognized as such by members of a specific society, e.g. cars.17 In accounting for these complex aspects, we will insist on developing the minimal formal apparatus required to characterize artefacts. For instance, although we need to model intentional aspects, we will avoid as much as possible the direct reference to theories of mental attitudes, which are not per se the subject of this paper and are by no means consensual. Similarly, we will not deal directly with the extremely elusive and much debated notion of purpose or function but will refer instead to a space of capacities, by making use of a general formal tool for describing the qualities of entities. We will not analyze in this paper the space of capacities itself: the structure of such space and the relationship between capacities and functions are issues that warrant further analysis. Here we shall merely introduce the foundations of the formal machinery.

17 The social character of artefacts is intended here in the broader sense. Social artefacts are not confined to the more restricted class of the physical artefacts that have a marked social purpose and use identity (like money and schools) nor to non-physical artefacts which fall in the category of social objects (e.g. laws and organizations).

292

Stefano Borgo and Laure Vieu

6

MODELING ARTEFACTS

6.1 Intentional selection Let us consider first the matter of agent intention which underlies the creation of an artefact. Two aspects need to be distinguished: the intention of obtaining an entity with some desired property (that perhaps makes it suitable for a particular purpose) and the intention of physically modifying or processing some pre-existing entity or entities in order to produce the artefact. We focus here on the first one only because we do not see the action of physical modification as an essential aspect of the creation of an artefact. In other words, artefacts do not need to be artificial entities: a pebble can make a paperweight, and a fallen tree a bench.18 In addition, residues, such as sawdust, are intentionally processed but not intentionally selected for having certain properties and use. Artefacts are, in nuce, created entities in which “created” refers to a mental event, not to a physical modification. One might remark that we do not base artefacts on actual use. The paperweight exists as soon as the agent selects it on the beach, and not just at the moment when he places it on a pile of loose papers. Similarly, knives coming out of a factory already exist as such; they are sold as knives, even though they have never at that stage cut anything. This approach therefore departs somewhat from Dipert’s views [Dipert, 1993]. Our artefacts are what Dipert terms “contemplated instruments”, in that they may still lack a use to be called “instruments”. In addition, as just explained, we do not restrict artefacts to physically modified entities, so our artefacts may not be termed “tools” (intentionally modified instruments) in Dipert’s terminology. Contemplated instruments, and even only once-used tools, may be far too many for some people’s tastes, but we believe this is not really an issue as one could define and focus on a relevant subclass of artefacts, “stable artefacts”, based on the creator’s repeated use according to her original purpose and/or her maintaining of some specific mental attitudes (memory of the creation, intention to use in the future), regardless of one of the remaining subclasses, the multitude of one-time-used or even one-time-contemplated artefacts. The class of social artefacts is another relevant subclass which will be described in this paper, in Section 7.2. This class arguably overlaps the class of stable artefacts, though one cannot ignore the matter of the many manufactured tools lying, yet unused, in stores and warehouses. So the essence of any artefact lies in the creator’s intention. It is certainly possible to explicitly represent the intentions or goals of the creator in an ontology of mental attitudes [Ferrario and Oltramari, 2004] and to reason about them adopting some dedicated logical formalism, for instance a so-called “Belief, Desire, Intention” logic [Rao and Georgeff, 1991]. As indicated above, for reasons 18 Even though one could argue that transporting the pebble to one’s desk forms some kind of modification, the tree trunk can come to serve as a bench in the very place where it fell. It may furthermore be argued that physical modification has to be restricted to change in intrinsic physical properties, thus disregarding spatial location [Geach, 1969].

Artefacts in Formal Ontology

293

of brevity we prefer to focus on the intentional creation event only and on the product of such an event, the artefact itself. The paperweight is the result of some agent intentionally selecting a pebble and attributing to it certain capacities. The artefact itself is the new entity whose physical realization is the selected object and which has attributed capacities. In particular, the paperweight is a selected pebble together with the attributed capacity to stand firm and hold down paper without damaging it. Ultimately, the artefact might prove not to have the capacities the agent attributed to it, as it could be flawed or malfunctioning. More will be stated on this point later.

6.2

Constitution

We then suppose that the paperweight is not the pebble. The paperweight only starts to exist when it is created, usually well after the pebble has come into existence; the two objects, although co-located when both present, may have different lifetimes and are therefore different. The alternative solution which would be to consider artefactuality as a property that physical endurants may or may not have would avoid the multiplication of entities. But, by not granting artefacts an ontological status, it would not do justice to the notion of creation, intended as a notion distinct from physical modification, nor to the common view that artefacts are sortals [Elder, 2007]. We are thus confronted with two co-located entities, the artefact and the underlying physical object. In addition, the former is dependent on the latter, as the paperweight cannot exist without the pebble. In short, the pebble constitutes the paperweight [Rea, 1997]. The same physical object can constitute two different artefacts, for example the same pebble can constitute both a paperweight and a pestle. Only physical objects or amounts of matter may constitute an artefact, as it is only material artefacts that we are considering here. So, when an artefact is apparently selected from another artefact, e.g. when a coffee-grinder is chosen to be used as a spice-grinder [Scheele, 2005], it is in fact the physical object constituting the first artefact which is selected again. Although we do not dwell here on the special cases of artefacts constituted by aggregates and those which are copies of previously existing models, we agree with Baker and Elder [Baker, 2004; Elder, 2007] that constitution is, in this instance, a powerful tool. As pointed out above, dolce already adopts the corresponding multiplicative approach, in particular to distinguish the statue from the amount of matter that constitutes it. However, in this extension, it is important to note that what directly constitutes the paperweight here is the physical object pebble, and not simply the amount of (rock) matter that in turn constitutes the pebble.19 The pebble is not an amount of rock because it is shape-dependent: the amount of rock persists after crushing, but the pebble does not —we obtain small stones or sand grains. Artefacts therefore bring yet another layer, an intentional level, to the constitution hierarchy. As a result, since the statue is an artefact, we actually need 19 The

amount of matter also constitutes the artefact, as constitution is transitive in dolce.

294

Stefano Borgo and Laure Vieu

to distinguish three co-located entities, and not simply two as argued in dolce and more generally in the literature on material constitution: the intentionally created statue, the specifically-shaped and structured physical object, and the mereologically determined amount of matter.20 We will see below that this further stratification is useful in understanding what happens when an artefact is repaired.

6.3 Capacity What, then, are capacities? Our notion of capacity builds on Cummins’s work on functions [Cummins, 1975]. His behavior-based approach avoids both the etiological account of function often given in philosophy of biology theories and the intentional approach adequate only for artefacts. We do take into account the intention of the agent in the creation event but we characterize, as does Cummins, the function of the artefact in agent-independent ways. To this end, we use the notion of quality in dolce, and assume that all physical endurants, restricted here to the categories Amount of Matter (M ), and Non-agentive Physical Object (NAPO), have a single individual quality named capacity that characterizes all the capacities the physical endurant has. The capacity of an entity is an individual, just as is its colour. This quality maps into a quale that is a region (possibly a sum of atomic qualia) in the capacity space, which can be seen as some sort of functional conceptual space [G¨ ardenfors, 2000]. The quale corresponding to the capacity of an entity at a given time collects all the various dispositions [McLaughlin, 1995; Mumford, 1998] or behaviours the entity is able to express at that time. For instance, the capacity of this pen now has the quale of writing finely in black when drawn over paper, fitting in one’s hand when grasped, and making a certain noise when it contacts the table. The capacity space is certainly complex, possibly founded on more elementary spaces of quality that provide the “bases” to such dispositions. We also assume that this space extends beyond standard (conditional) dispositions to also include structural properties, that is, the internal arrangement of parts. Our purpose here is not to analyse and describe this space in detail but to give the overall architecture of a possible formal ontological view of artefacts. Further study will certainly be required if we are to understand the structure of capacity space. It may in particular assess the need to use several distinct such spaces instead of a single one, and accordingly, several capacity qualities instead of a single one.

6.4 Attributed capacity In addition to the capacity possessed by any physical endurant,21 artefacts also have an attributed capacity, another quality associated with qualia in the same space. The fact that actual dispositions and intended functions are elements of a 20 When the artefact is not selected from a physical object but from an amount of matter, as with an amount of glass, there are of course two layers only. 21 We assume that an artefact’s capacity is inherited from its constituting entity.

Artefacts in Formal Ontology

295

same space has a number of advantages. It first of all permits the definition of malfunctioning (see Section 7.3 below). It furthermore demonstrates how the dual nature of artefacts that are physical endurants can be reconciled as it forms the interface for the mental and the physical aspects of artefacts [Kroes and Meijers, 2006]. This unique space also makes possible a future account of the design process. As the capacity space also covers structural properties, the very design plans of technical artefacts could be conceived as part of the attributed capacity. Although capacity and attributed capacity map into the same space of qualia, the former remains a physical quality whereas the latter is an intentional quality as it is dependent on the intentions of the creator at the time of the creation event.22 Capacity and attributed capacity also differ in the following way: the quale associated with the attributed capacity does not change in time as it is fixed by the creation event; moreover, this quale is a set of regions of the capacity space because the intended behaviour of the artefact need not be specified in a precise way, and may present vagueness. For instance, when one is looking for something to write on a board and one selects a piece of coal the value of the attributed capacity is only to write on the board and fit in the hand. Therefore, the attributed capacity maps into the qualia space in a variety of alternative regions corresponding to different possible writing behaviours, e.g. writing finely in black on the board and fitting in one’s hand, writing thickly in black on the board and fitting in one’s hand, writing finely in brown on the board and fitting in one’s hand, etc. On the other hand, the creator of a commercial pen has probably designed it precisely and has therefore chosen an attributed capacity which maps to a reduced number of regions or even to a unique one. This region, though, is a priori smaller than the capacity’s quale region, e.g. the creator of the pen did not design it for the noise it produces when crushed with a rock. So typically the quale of the artefact’s capacity strictly includes one member of the quale of its attributed capacity. This obviously does not hold in the case of malfunctioning or faulty artefacts: one can create an artefact with an attributed capacity’s quale that the selected physical endurant’s capacity (and the artefact’s capacity) will not always have or may not have from the start and perhaps ever (see Section 7.3).

6.5

Identity criteria

If we are to grant an ontological status to artefacts, a delicate point now needs to be addressed. We need to examine their identity criteria. We have seen that artefacts are distinct from the physical objects (or amounts of matter, in the case of artefactual matter) that constitute them. They should therefore have distinct identity criteria. Indeed, artefacts can be repaired and have some parts substituted, thus changing the entity that constitutes them for another without losing their identity. Such change comes at the cost of the former constituting entity disappearing simultaneously with the newer constituting entity coming into exis22 This dependence will not, however, be formally expressed here, as we deliberately refrained from introducing intentions.

296

Stefano Borgo and Laure Vieu

tence, though maintaining a certain degree of spatio-temporal continuity between the two. In fact, no artefact can “jump” from one material entity to a separate preexisting one at will. If Theseus’s ship [Rea, 1997, Introduction], an artefact, does not disappear when a plank is substituted, the physical object that constitutes it, the planks-and-nails assembly, changes so that the former assembly ceases to exist and a new assembly comes into existence.23 By pointing out the property that an artefact cannot jump from one physical object to another, we can shed some light on the important distinction between artefacts and artefact roles. Roles, in general, can be played by different entities (e.g. different persons at different times can play the role of president of the US) [Masolo et al., 2004] and the change between players can be seen as a “jump”, as the previous player usually survives the change and the successor often already exists. Physical artefacts are more stable. They are not roles. This distinction is evident, for instance, in the house/home contrast. A house is an artefact which can play the role of being someone’s home. One’s home changes, there is a jump from a house to another when one moves house, so “home” is not a type of artefact subsumed by “house”, but rather a role. The gradual change in the constituting material entity may only occur with artefacts selected from physical objects and not with those selected from amounts of matter. It is reasonable to assume that amounts of plastic or of glass cannot switch over just as quantities of matter cannot interchange. Indeed, amounts of matter in dolce have purely mereological identity criteria.24 Non-agentive physical objects have more complex identity criteria, which vary from sortal to sortal. It is not the purpose of this paper to establish those criteria, but as a general guideline, we will take shape and internal structure to be part of these criteria. We assume though that minor changes in shape and in the constituting amount of matter, like those induced by a scratch, are allowed. Granularity is certainly an issue here. With artefacts, an obvious characteristic for determining their identity criteria is their intentional aspects, that is, their attributed capacity. The identity criteria should among other things determine when an artefact disappears all together. Ordinary malfunctioning does not make an artefact disappear, so its identity criteria cannot be simply based on a match between attributed capacity and capacity. Nor is the artefact’s disappearance simply based on its constituting entity’s disappearance, since that can be substituted, as we have just seen. So, the loss of much of the attributed capacity must be involved. We do not intend to solve here the infamous ship-of-Theseus puzzle [Rea, 1997, Introduction], but we believe that we can nevertheless safely assume that the identity criteria of artefacts are based on a combination of significant degree of spatio-temporal continuity of the constituting entities, the existence of all specific essential parts if any (e.g. for a car, its frame), 23 The

term “assembly” denotes here an aggregate in a specific arrangement. does not take into account the nature of the substance of which the amount of matter is made. As a result, it does not consider homogeneity conditions. A different choice would not not affect the present discussion. 24 dolce

Artefacts in Formal Ontology

297

and the actuality of a significant amount of attributed capacity, i.e. a significant overlap between one region member of the quale of the attributed capacity and the region quale of the capacity. Note that since the attributed capacity is not restricted to the overall or main function of the artefact and since it covers structural specifications, a malfunctioning artefact does possess most of its attributed capacity. Even a badly designed artefact, like a medieval flying machine, possesses most of its attributed capacity.

6.6

Axiomatics

We now turn to the formal theory that corresponds to the above choices. A physical artefact or artefact for short is an element of Physical Artefact category. It is dependent on a constituting entity of category Amount of Matter (M ) ∪ Non-agentive Physical Object (NAPO), called for short “material entity” in the remainder. For instance, an amount of glass is an artefact constituted by an entity of M category, while a paperweight is an artefact constituted by an entity of NAPO category. An artefact x is created by an intentional association of a material entity y and a quality q which is of the type AttributedCap, a new primitive predicate denoting attributed capacities.25 The intentional association that generates the artefact is a special event of type CreationEv (creation event). To characterize this we use the IntentionalSel (intentional selection) primitive relation which takes as its arguments an event (EV ) e, an agent p, physical (APO) or not (ASO, e.g. a company), a physical artefact (PhysArt) x, a material entity y, and a quality (Q) q. IntentionalSel(e, p, x, y, q) should be read as “e is the event of p obtaining the artefact x by intentionally selecting y and attributing to it capacity q.” Our first axiom states that artefacts, that is, the elements of the category dubbed Physical Artefact and represented by the primitive predicate PhysArt, are the product of some intentional selection event: (A1) PhysArt(x) ↔ ∃e, p, y, q IntentionalSel(e, p, x, y, q). Next we constrain the primitive IntentionalSel as indicated above: (A2) IntentionalSel(e, p, x, y, q) → EV (e) ∧ (APO(p) ∨ ASO(p)) ∧ PhysArt(x) ∧ (M (y)∨NAPO(y)) ∧ AttributedCap(q) ∧ qt(q, x) ∧ ∃t (ql T (t, e)∧ PC (y, e, t) ∧ PC (x, e, t) ∧ PC (p, e, t) ∧ K (y, x, t))). Axiom (A2), in addition to restricting the arguments of IntentionalSel, specifies a number of assumptions. The quality q is a quality (qt) of the artefact x. The agent, the artefact and the material entity all participate (PC ) in the selection event for the time of the event. One consequence is that these three entities are present (PRE ), that is, exist, during the event. For non-instantaneous events, 25 The new —primitive or defined— predicates introduced to characterize artefacts are given in sans-serif font to distinguish them from the predicates denoting dolce categories and relations.

298

Stefano Borgo and Laure Vieu

in other words accomplishments (ACC ) as opposed to achievements (ACH ), it is nevertheless somehow arbitrary to say that the artefact x exists during the event; one could assume it exists only immediately after the event instead. This decision is not crucial to our approach. Lastly, the artefact x is constituted (K ) by the material entity y during the selection event. Axiom (A5) below will guarantee that this constitution relation lasts until the artefact disappears or until its constituting entity is replaced. The constituting entity y does not need to exist before e, since e could co-occur with a physical creation event, in which case both x and y would be created simultaneously. As far as the existence persistence of the artefact after the creation event is concerned, this can be at best a default rule because nothing prevents its destruction. We assume though that while the original constituting entity is present, the artefact will also be present. This partially underlines the fact that for an artefact to disappear it needs to undergo a major change which cannot happen without altering the identity of the constituting entity. So, as long as the original constituting entity is present, the artefact will also be present. The opposite is not true, though, because the artefact can change its constituting entity as in the case of repairing by substituting a component. While the artefact exists, there is a unique entity of the same category of the material entity originally selected, Amount of Matter or Non-agentive Physical Object, that constitutes it.26 As explained above, if the selected material entity is an amount of matter, this cannot change: (A3) (IntentionalSel(e, p, x, y, q) ∧ M (y) ∧ PRE (x, t)) → (K(y, x, t) ∧ ∀z(¬z = y → ¬K(z, x, t))). (A4) (IntentionalSel(e, p, x, y, q) ∧ NAPO(y) ∧ PRE (x, t)) → ∃!z (K(z, x, t) ∧ NAPO(z)). While the constituting material entity exists (which may or may not be the original entity), it continually constitutes the artefact and so, as a consequence, the artefact still exists: (A5) (PhysArt(x) ∧ K(y, x, t) ∧ (M (y) → ∃e, p, q IntentionalSel(e, p, x, y, q)) ∧ PRE (y, t ) ∧ t < t ) → K(y, x, t ).27 A consequence of the previous axioms is that if the artefact is constituted by different physical objects at different times, these physical objects will not exist simultaneously (thus ruling out “jumps”): 26 It is impossible to simply assert that the entity that constitutes an artefact is unique since an artefact constituted by a physical object is also constituted by the amount of matter that constitutes it. All physical objects are constituted by some amount of matter, and constitution is taken to be transitive in dolce. dolce assumes that Amount of Matter is the lowest substrate, in other words that nothing constitutes an amount of matter. 27 This very partial account of continuity ignores the case of assemblies having an intermittent existence, as in the Theseus’s ship puzzle, in which the original constituting physical object is reassembled. We leave this issue for further developments.

Artefacts in Formal Ontology

299

(PhysArt(x) ∧ K(y, x, t) ∧ NAPO(y) ∧ K(y  , x, t ) ∧ NAPO(y  ) ∧ ¬y = y  ) → (¬PRE (y  , t) ∧ ¬PRE (y, t )). Since a material entity may constitute several different artefacts, the formula (IntentionalSel(e, p, x, y, q) ∧ IntentionalSel(e , p , x , y, q  )) → x = x is neither a theorem nor an axiom. Further constraints are needed. First, for a given artefact, the attributed capacity is unique and the quale of the attributed capacity does not change in time: (A6) (IntentionalSel(e, p, x, y, q) ∧ IntentionalSel(e , p , x, y  , q  )) → q = q  . (A7) (IntentionalSel(e, p, x, y, q) ∧ ql(v, q, t) ∧ ql(v  , q, t )) → v = v  . As asserted above, we assume that the same artefact can be selected several times, by possibly different agents (or societies) like, for example, the same “tree trunk bench” in the woods, so the event and the selector are not necessarily unique. This choice is not essential to the approach. However, for a given intentional selection event, the artefact and the selector must be unique (as well as the attributed capacity quality because of the axiom above): (A8) (IntentionalSel(e, p, x, y, q) ∧ IntentionalSel(e, p , x , y  , q  )) → (x = x ∧ p = p ). We also make sure that attributed capacities are only qualities of artefacts:28 (A9) (AttributedCap(q) ∧ qt(q, x)) → PhysArt(x). On the other hand, all physical endurants, including artefacts, have a capacity (A10). The capacity of physical endurants is inherited through constitution, in the sense that the quale of the capacity of the constituted entity includes that of the constituting entity (A11). Note that the two qualia need not be identical. The capacity of a physical object may include shape-based dispositions, while the capacity of the amount of matter constituting it cannot. Similarly, when creating a socially relevant artefact, like a cheque, new capacities are created that the constituent itself, i.e. the rectangular piece of paper, does not possess. (A10) PED(x) → ∃q (Capacity(q) ∧ qt(q, x)). (A11) (PhysArt(x) ∧ K(y, x, t) ∧ Capacity(q) ∧ qt(q, x) ∧ ql(v, q, t) ∧ Capacity(q  ) ∧ qt(q  , y) ∧ ql(v  , q  , t)) → P (v  , v). To ensure that capacities and attributed capacities map to qualia in the same space of capacities, we use a unary predicate CR, for capacity region, to characterize those qualia. However, we need to allow the attributed capacity of an artefact to have a set — or some sort of collection — of such regions for quale, as mentioned above. Sets, collections and aggregates are not yet formalized in dolce. We will 28 In

dolce a quality inheres in a unique entity so given q there is a unique x such that qt(q, x).

300

Stefano Borgo and Laure Vieu

nevertheless use a fairly intuitive primitive predicate IN to denote “membership” (its characterization does not concern us here): (A12) (Capacity(q) ∧ ql(v, q, t)) → CR(v). (A13) (AttributedCap(q) ∧ ql(v, q, t) ∧ IN(x, v)) → CR(x). Finally, an intentional selection is a creation event unless the same artefact has already been selected. CreationEv is a defined predicate: (D1) CreationEv(e, x) =def ∃p, y, q IntentionalSel(e, p, x, y, q) ∧ ∃t (qtT (t, e) ∧ ∀t (t < t → ¬PRE (x, t ))). We ensure that there is a creation event for each artefact and, as a consequence, that it does not already exist before the first intentional selection event: (A14) PhysArt(x) → ∃e CreationEv(e, x). The creator of an artefact is the first selector; it is unique as long as there are no simultaneous first intentional selections of the same artefact: (D2) Creator(p, x) =def ∃e, y, q(CreationEv(e, x) ∧ IntentionalSel(e, p, x, y, q)). No additional axiom is introduced to model the conditions in which a given artefact disappears, essentially through lack of means to adequately model the difficult vagueness issues involved in the identity criteria of artefacts as described above. 7

BEYOND THE BASICS

7.1 Artefact types The approach developed so far allows us to characterize a notion of artefact type within dolce. Since agents and societies develop concepts to discriminate between types of entities, it is natural to view concepts about artefacts as providing definitions of artefact types. Artefacts are endurants with a particular quality known as attributed capacity. It suffices for the concept to discriminate between the attributed capacities of the artefacts to coherently collect artefacts “of the same type”. We thus assume that the definition of a concept classifying artefacts, i.e. an artefact type, is based on the comparison of the qualia of these attributed capacities with the attributed capacity of what we would call a prototype. So, we suppose that the definition of an artefact type, say Hammer, isolates the collection of the artefacts whose attributed capacities are such that all the regions in their quale include one of those of a generic or prototypical hammer. The prototypical hammer does not need to exist, but there must be a specific attributed capacity value v, i.e. a set of capacity regions, that characterizes what would count as a prototypical hammer. Evidently, the existence of an artefact type, a concept, is independent of the creation of any artefact token of this type.

Artefacts in Formal Ontology

301

Formally, we use the classification (CF ) relation. In Section 4.2 we asserted that CF (x, y, t) stands for “at t, x satisfies concept y”, which we sometimes write “x is classified by y at t”. (D3) ArtefactType(a) =def CN (a) ∧ ∃v (∃u IN(u, v) ∧ ∀ w(IN(w, v) → CR(w)) ∧ ∀x, t (CF (x, a, t) ↔ ∃qx , vx (PhysArt(x) ∧ AttributedCap(qx ) ∧ qt(qx , x) ∧ ql (vx , qx , t) ∧ PRE (x, t) ∧ ∀w(IN(w, vx ) → ∃w (IN(w , v) ∧ P (w , w)))))). This definition states that all and only the artefacts classified by concept a have an attributed capacity’s quale whose member regions all include a region of the (non-empty) set of regions v characterizing a. Note that the artefacts of a given type do not need to be present at the same time, since, as we have seen, the quale of an attributed capacity cannot change in terms of time. For the same reason, we obtain a theorem which claims that an artefact is classified by the same type or types throughout its existence: ∀x, a, t, t ((ArtefactType(a) ∧ CF (x, a, t) ∧ PRE (x, t )) → CF (x, a, t )). Clearly, using mereological relations between the elements of the sets of capacity regions used to characterize artefact types generates a network of types so that we can relate subtypes to types, e.g. Carpentry hammer or Physician’s hammer to Hammer. As for all other concepts, artefact types are dependent on the agents or societies that define them [Masolo et al., 2004]. In fact, artefact types often apply to social artefacts, a subject to which we can now turn.

7.2

Social artefacts

Informally, a social artefact is an artefact whose type is recognizable by the members of a certain society. A once-used object such as the bench-and-table that you selected from a fallen tree for your last picnic in the woods is not what we commonly call an artefact, nor is it a stable artefact repeatedly but privately used, such as the spice-grinder that was selected from the physical object constituting a coffee-grinder [Scheele, 2005]. Often, artefacts are recognized as such by agents other than their creators: we buy knives assuming that someone has made them suitable for cutting when used in a certain way. Societies share the knowledge of recognizing many different artefacts, that is of recognizing part of the attributed capacity of a given entity through its type: pens and knives, glass and flour, and so on so that most of the time there is no need for the creator to explain their purpose. As Dipert puts it, a proper artefact is in effect an entity for which the attributed capacity (Dipert calls it the creator’s intention) is recognizable [Dipert, 1993]. To emphasize their dependence on a given society we call these items social artefacts. We depart somewhat from Dipert’s proposal by requiring that only the part of the quale of the attributed capacity defining an artefact type, in other words, the attributed capacity value characterizing a prototype, be recognized.

302

Stefano Borgo and Laure Vieu

This makes it possible that certain non-central or idiosyncratic aspects of the attributed capacity to be ignored, avoiding the assumption that members of society have unrealistic mind-reading abilities. A recognition event has to be distinguished from an intentional selection, as the intention (recognized in the event) is attributed to someone else, even if that someone is unknown. Of course, one may assume that someone attributed a capacity to this entity thereby believing that one recognizes an artefact, but one might just be wrong. This is an epistemological issue and as such it is not a subject for ontological analysis. Our aim is to account for the ontological import of the fact that a given society shares knowledge about some kinds of artefacts. As mentioned above, the formal ontology of social reality does not seek to analyze how and on what grounds an agent of society is able to fulfil the recognition of a given artefact. It has to deal with the fact that some artefacts acquire a social status. Note, though, that the existence of the artefact is not affected by having or not having the property of being recognizable: social artefacts are essentially artefacts. So general consensus among a community of archaeologists on the artefactual nature of a collection of entities that are actually only residues makes them neither simple nor social artefacts.29 The recognition of a social artefact relies on at least one of two distinct elements: the object structure and the context. In the first case, the act of recognition is intrinsically related to the material entity constituting the artefact — its structure, its physical qualities, its actual capacity, etc. — and to the structure and properties of artefacts previously encountered. The recognition of knives and cars falls into this category. In the second case, recognition is based on the broader context in which the entity is observed. For instance, if we see a pebble (of a certain size) on someone’s desk on top of a heap of papers, we will assume that the pebble constitutes a paperweight, while generally we do not identify paperweights on the seashore (though we may intentionally select some). Similarly, if we see in a shop a pile of pebbles labelled “paperweight”, we all assume the shop is actually selling paperweights. There are obvious limits to this: a label “paperweight” on a heap of sand will not be sufficient to make us recognize there artefacts because certain constraints on the capacity of the material entity need to be satisfied to convince us that someone did select that material entity and attributed it a certain capacity. We thus distinguish between (intrinsic) social artefacts and contextual social artefacts. To ontologically represent such notions, we shall introduce the new primitive predicate Recognizable(a, x, s, t), that reads as “the type a of artefact x is recognizable by society s at time t”. The basic constraints on this relation are: (A15) Recognizable(a, x, s, t) → (ArtefactType(a) ∧ PhysArt(x) ∧ CF (x, a, t) ∧ SC (s) ∧ PRE (s, t)).

29 The gap between ontology and epistemology is a particularly difficult one to bridge when there is societal discontinuity with knowledge loss, viz. the “Nineveh lens” cf. http://www.badarchaeology.net/data/ooparts/nineveh.php.

Artefacts in Formal Ontology

303

To constrain Recognizable further is not an easy matter if one wants to avoid referring explicitly to the mental attitudes of agents. To gain a better grasp of what this predicate is supposed to mean, we can sketch what could be a definition, if we were using a theory that allows for doxastic modalities (the modal belief operator [Hintikka, 1964] Belx,t φ standing for “x believes proposition φ at t”), as well as the arguably simpler primitives Member (between an agent (APO) and a society (SC ) at some time) and Perceives (describing the event of an agent (APO) perceiving a physical endurant (PED)). The type a of artefact x is recognizable by society s at time t if the artefact and the society are present at t and any member30 of the society s believes the artefact x is an artefact of type a whenever during t the agent perceives31 it: Recognizable(a, x, s, t) =def ArtefactType(a) ∧ PhysArt(x) ∧ CF (x, a, t) ∧ SC (s) ∧ PRE (s, t) ∧ ∀e, y, t ((Member(y, s, t ) ∧ Perceives(e, y, x) ∧ ql T (t , e) ∧ P (t , t)) → Bely,t (PhysArt(x) ∧ CF (x, a, t ))). A(n) (intrinsic) social artefact for society s is then an artefact whose type is recognizable by society s at all times when the artefact and the society are present: (D4) SocialArt(x, s) =def PhysArt(x) ∧ ∃a ∀t ((PRE (x, t) ∧ PRE (s, t)) → Recognizable(a, x, s, t)). Let us now turn to contextual social artefacts. For the sake of presentation, contexts are taken here as a category of entities32 and we introduce a new predicate InContext to relate an endurant (ED) to a context at a time, assuming that InContext(x, c, t) entails PRE (x, t). With these tools, we can define a contextual social artefact (ContextualSocialArt) for context c and society s to be an artefact whose type is recognizable by s whenever the artefact is in the context c: (D5) ContextualSocialArt(x, s, c) =def PhysArt(x) ∧ ∃a ∀t ((InContext(x, c, t) ∧ PRE (s, t)) → Recognizable(a, x, s, t)).

7.3

Malfunction

As mentioned above, our notion of artefact includes malfunctioning or even failed artefacts. With our approach, it is rather straightforward to state that an artefact is malfunctioning at t. It simply does not possess all the capacities attributed to it: 30 There is surely a need here to restrict the conditional antecedent to qualified members of the society and to thus disregard babies, drunk people and so on. 31 The nature of the perceptive events involved (seeing, hearing, touching, etc.) may depend on both the artefact and the society; we can safely assume that perceiving an artefact is equivalent to perceiving the material entity that constitutes it. 32 The real nature of contexts is by no means obvious and the very issue of their reification raises some criticism. Contexts are not currently included in dolce and we could manage without them by using “descriptions” as introduced in [Masolo et al., 2004]. However, as it would be impossible to introduce here the notion of description, we shall rely for the present purposes on the intuitive notion of context.

304

Stefano Borgo and Laure Vieu

(A16) MalFunctioningArt(x, t) =def PhysArt(x) ∧ PRE (x, t) ∧ ∀q, q  , v, v  , w ((AttributedCap(q) ∧ qt(q, x) ∧ Capacity(q  ) ∧ qt(q  , x) ∧ ql (v, q, t) ∧ ql (v  , q  , t) ∧ IN(w, v)) → ¬P (w, v  )). This definition is based on the fact that “possessing the attributed capacities” means that at least one of the alternative regions in the quale of the artefact’s attributed capacity is part of the region corresponding to the current quale of the artefact’s capacity. Note that we talk of the capacity of the artefact itself and not of the capacity of its constituent. As posited above in (A11), the capacity of the constituting entity is inherited by the artefact.

7.4 Limitations of the framework We conclude this section by highlighting a couple of open-ended problems that we think should be addressed if we are to understand the advantages and the limitations of this framework. First, our formalization is not compatible (as it stands here) with the intuition that an artefact may gain or lose attributed capacities (more precisely, that the regions in the quale of its attributed capacity might grow or shrink) while it remains the same artefact. We have instead exploited the multiplicative approach of the dolce ontology, assuming that the same material entity can simultaneously constitute different artefacts, e.g. the paperweight and the pestle (both private or contextual social), the (intrinsic social) coffee grinder and the (private) spice grinder, or the (intrinsic social) anvil and the (contextual social) doorstop. Still, it might be possible to adopt a notion of attributed capacity that depends on time and so render, even in this approach, the intuition of artefact evolution. We have not studied that option here. Similarly, concepts like artefact types may evolve (a nice example relates to the evolution of Aspirin from painkiller to painkiller and blood-thinner as detailed in [Houkes and Meijers, 2006]). The evolution of concepts has not been fully addressed in the theory developed in [Masolo et al., 2004] which is exploited here. It is implicitly assumed, however, that we need to distinguish each change in concept as a creation of a new concept historically dependent on the previous one. This appears to be a reasonable solution here too. We have just noticed that the theory developed makes extensive use of the multiplicative approach of dolce, the formal ontology chosen here. As explained on p.281, this feature is rather a specificity of dolce though it is not incompatible with other foundational ontologies. On the other hand, the multiplicative approach developed here is related to the “constitution view” developed by Baker [Baker, 2004] which has been criticized by Houkes and Meijers in [Houkes and Meijers, 2006]. Let us then examine the reasons for rejecting such an approach. Houkes and Meijers’s first criticism of the constitution view is that this approach leads to unnecessary “ontological stacking”. We note that the multiplicative approach has many other applications, giving for instance a straightforward answer to the puzzle of multiple event descriptions which have different causal explanation

Artefacts in Formal Ontology

305

power [Pianesi and Varzi, 2000]. The multiplicative approach does not convince all philosophers, though. There are arguments in favour of and against the reductionist and multiplicative points of view, in all domains. We believe that the artefact domain is not essentially different in this respect. Their second criticism concerns the impossibility of Baker’s proposal to account for the Realizability Constraints (RC), the idea that an ontology of artefacts must provide grounds for constraining the possible entities that may constitute an artefact — given its function — as well as providing grounds to constrain the artefacts an entity could constitute, given its structure. In our opinion, the multiplicative approach does not in itself provide any explanation for this, nor does it prevent a further formal account of RC. Constitution is a generic dependence relation which constrains the existence of the related entity, but that does not explain why such a constitution may or may not hold. As mentioned above, although we have left the designing process out of the picture, we believe that RC could be accounted for by comparing the qualia of the actual and attributed capacities of an artefact, something which is facilitated by the use of a single capacity space. This is not at all straightforward, though. As argued above, the conditions of existence of an artefact are indeed based on a match between its actual and its attributed capacities but such a match is of necessity vague to allow for the existence of malfunctioning artefacts. And vagueness is a notoriously difficult issue, especially when ontological matters are at stake. Finally, we point out once more that our notion of artefact relies on one important quality space, namely the capacity space. This space has not been analyzed here and it is not yet well understood. It includes both functional and structural aspects in an interesting setting that certainly does deserve more attention. Furthermore, it seems necessary to study the dimensions of this space, the relationship to the other quality spaces such as weight, shape and colour, and the overall structure if we are to properly formalize other crucial notions such as that of technical artefact. 8

CONCLUSION

In this chapter we analyzed ontology research on the notion of artefact. After looking at existing ontologies and highlighting several shortcomings, we presented and discussed a new formalization that defines artefacts to be endurants with a special quality known as attributed capacity, which justifies their special status with respect to other endurants. This new quality allowed us to formalize a series of notions which were justified on the basis of philosophical distinctions as well as commonsense intuitions. The theory proposed, although not self-contained and still requiring further development, shows the feasibility of extending a foundational ontology, namely dolce, to grasp the non-trivial notion of artefact. The theory of course reflects certain philosophical choices which will not be palatable for all researchers in the field. Similarly, some of its technical aspects strongly rely on the multiplicative

306

Stefano Borgo and Laure Vieu

structure of dolce, a feature which might not be easily transposed in all other formal frameworks. Nevertheless, we believe that this work has much potential for concrete applications where artefacts are central and semantic integration is an issue. It also illustrates the fecundity of applying philosophical studies to knowledge representation in computer science. BIBLIOGRAPHY [Antoniou and van Harmelen, 2004] G. Antoniou and F. van Harmelen. Web Ontology Language: OWL. In S. Staab and R. Studer, eds., Handbook on Ontologies, pp. 67–92. Springer, 2004. [Baker, 2004] L. R. Baker. The ontology of artifacts. Philosophical Explorations, 7(2):99–111, 2004. [Borgo et al., 1996] S. Borgo, N. Guarino, and C. Masolo. A pointless theory of space based on strong connection and congruence. In L. Carlucci Aiello, J. Doyle, and S. C. Shapiro, eds., Proceedings of the International Conference on Principles of Knowledge Representation and Reasoning (KR’96), pp. 220–229. Morgan Kaufmann, 1996. [Campbell, 1990] K. Campbell. Abstract Particulars. Basil Blackwell, Oxford, 1990. [Cummins, 1975] R. Cummins. Functional analysis. Journal of Philosophy, 72:741–765, 1975. [Dipert, 1993] R. Dipert. Artifacts, Art Works, and Agency. Temple University Press, Philadelphia, 1993. [Elder, 2007] C. Elder. On the place of artifacts in ontology. In E. Margolis and S. Laurence, eds., Creations of the Mind: Theories of Artifacts and Their Representation, pp. 33–51. Oxford University Press, Oxford, 2007. [Fellbaum, 1998] C. Fellbaum, ed. WordNet. An Electronic Lexical Database. MIT Press, Cambridge (MA), 1998. [Ferrario and Oltramari, 2004] R. Ferrario and A. Oltramari. Towards a computational ontology of mind. In A. C. Varzi and L. Vieu, eds., Formal Ontology in Information Systems, Proceedings of the Intl. Conf. FOIS 2004, pp. 287–297. IOS Press, 2004. [G¨ ardenfors, 2000] P. G¨ ardenfors. Conceptual Spaces: the Geometry of Thought. MIT Press, 2000. [Geach, 1969] P. Geach. God and the Soul. Routledge and Kegan Paul, London, 1969. [Guha and Lenat, 1990] R. V. Guha and D. Lenat. Cyc: A mid-term report. AI Magazine, 11(3):32–59, 1990. [Hintikka, 1964] J. Hintikka. Knowledge and Belief. Cornell Univ. Press, 1964. [Hodges, 1983] W. Hodges. Elementary predicate logic. In D. Gabbay and F. Guenthner, eds., Handbook of Philosophical Logic, volume I, pp. 1–131. Dordrecht: Reidel, 1983. [Houkes and Meijers, 2006] W. Houkes and A. Meijers. The ontology of artefacts: The hard problem. Studies in History and Philosophy of Science, 37(1):118–131, 2006. [Kroes and Meijers, 2006] P. Kroes and A. Meijers. The dual nature of technical artefacts. Studies in History and Philosophy of Science, 37(1):1–4, 2006. [Lowe, 1998] E. Lowe. The Possibility of Metaphysics. Clarendon Press, Oxford, 1998. [Lowe, 2002] E. Lowe. A Survey of Metaphysics. Oxford University Press, Oxford, 2002. [Masolo et al., 2003] C. Masolo, S. Borgo, A. Gangemi, N. Guarino, and A. Oltramari. Ontology Library (Wonder-Web Deliverable D18). Available at http://wonderweb.semanticweb.org/deliverables/documents/D18.pdf, 2003. [Masolo et al., 2004] C. Masolo, L. Vieu, E. Bottazzi, C. Catenacci, R. Ferrario, A. Gangemi, and N. Guarino. Social roles and their descriptions. In D. Dubois, C. Welty, and M. Williams, eds., Proceedings of the 9th International Conference on the Principles of Knowledge Representation and Reasoning (KR), pp. 267–277, 2004. [McLaughlin, 1995] B. P. McLaughlin. Dispositions. In J. Kim and E. Sosa, eds. A Companion to Metaphysics, pp. 121–124. Oxford: Blackwell Publishers, 1995. [Mumford, 1998] S. Mumford. Dispositions. Oxford University Press, Oxford, 1998. [Pianesi and Varzi, 2000] F. Pianesi and A. Varzi. Events and Event Talk: An introduction. In J. Higginbotham, F. Pianesi and A. Varzi, eds. Speaking of Events, pp. 3–47. New York: Oxford University Press, 2000. [Pr´ evot et al., 2005] L. Pr´ evot, S. Borgo, and A. Oltramari. Interfacing ontologies and lexical resources. In Ontologies and Lexical Resources: IJCNLP-05 Workshop, pp. 1–12, 2005.

Artefacts in Formal Ontology

307

[Rao and Georgeff, 1991] A. Rao and M. Georgeff. Modeling rational agents within a bdiarchitecture. In J. F. Allen, R. Fikes, and E. Sandewall, eds., Principles of Knowledge Representation and Reasoning (KR’91), pp. 473–484. Morgan Kaufmann, 1991. [Rea, 1997] M. Rea, ed. Material Constitution: A Reader. Rowman and Littlefield Publishers, Lanham, MD, 1997. [Scheele, 2005] M. Scheele. The Proper Use of Artefacts: A Philosophical Theory of the Social Constitution of Artefact Functions. PhD dissertation, Technical University of Delft, 2005. [Searle, 1983] J. R. Searle. Intentionality. Cambridge University Press, 1983. [Searle, 1995] J. R. Searle. The Construction of Social Reality. The Free Press, New York, 1995. [Smith, 1995] B. Smith. Formal ontology, common sense and cognitive science. International Journal of Human Computer Studies, 43(5/6):626–640, 1995. [Smith, 1998] B. Smith. Basic concepts of formal ontology. In N. Guarino, ed., Proceedings of the First International Conference FOIS 1998, pp. 19–28. IOS Press, 1998. [Sowa, 2000] J. Sowa. Knowledge Representation: Logical, Philosophical, and Computational Foundations. Brooks/Cole, Pacific Grove, CA, 2000. [Thomasson, 2007] A. Thomasson. Artifacts and human concepts. In E. Margolis and S. Laurence, eds., Creations of the Mind: Theories of Artifacts and Their Representation, pp. 52–73. Oxford University Press, 2007.

This page intentionally left blank

THE NATURE OF TECHNOLOGICAL KNOWLEDGE Wybo Houkes

1 FROM APPLIED SCIENCE TO EPISTEMIC EMANCIPATION Two decades ago, John Staudenmaier took stock of twenty-five years of research published in Technology & Culture, a leading journal for historians of technology. He identified three key debates, one of which is the relation between science and technology. This debate was largely shaped by the “technology-is-applied-science” thesis, often attributed to Mario Bunge [1966], and the objections to this thesis. Staudenmeier ends his overview of the debate with an intriguing conjecture, worth quoting in full: Thus, it would appear that a substantial number of [historians who published in Technology & Culture] interpret technological praxis as a form of knowledge rather than as an application of knowledge. By their discussions of scientific concepts, problematic data, engineering theory, and technical skill, the authors have begun to develop a complex and provocative model. If these discussions are, in fact, the beginning of a new theme in [Technology & Culture], we may find that the more limited science-technology question will take its place as a subtheme within the more inclusive model. [Staudenmeier, 1986, p. 120] Twenty years later, the antecedent of the last statement has proved false. Staudenmeier’s conjecture about the start of a new theme, around say 1980, can be supplemented with a statement about the development of this theme after 1986: neither in Technology & Culture, nor elsewhere has this “more inclusive model” been developed. What is worse, after the publication of Walter Vincenti’s What Engineers Know and How They Know It [1990], research concerning the nature of technological knowledge seems to have come to a standstill. Historians of technology have lost interest in the topic. One illustration is Samuel Florman’s [1992] review of Vincenti’s book in Technology & Culture. In the review, Florman complains about Vincenti’s excessive interest in epistemological details at the price of attention to people and organizational issues. Philosophers have not rushed in to fill the gap left by historians. Technological knowledge is Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

310

Wybo Houkes

not even a minor theme in journals that cover epistemological and methodological issues. The handful of papers that have been published on the topic reverse Staudenmaier’s prediction: they typically address the science-technology relation and treat the nature of technological knowledge as a side issue. Furthermore, all papers are isolated efforts, and often repeat points that have made in the technology-asapplied-science debate before the 1980s. There is no philosophical research tradition regarding technological knowledge, in which authors build upon, or even respond to each other’s work. Even individuals have seldom published more than a few papers on the topic, meaning that there are not even personal research traditions. It cannot be concluded from twenty years of disinterest in technological knowledge that the subject is not interesting. It may, perhaps temporarily, have gone out of fashion among historians and philosophers. The situation does, however, pose an interesting challenge for an overview of the work on technological knowledge. Therefore, in this paper, I review several different, occasionally quite slender bodies of literature to find out whether there are possibilities to revive the interest in technological knowledge. In particular, I consider to what extent the scattered, divergent research on technological knowledge — organized into several themes in this overview — supports a strong, and therefore interesting epistemological claim made at the start of the Staudenmaier quote: that technological praxis may be interpreted as a form of knowledge. This claim is not unique to Staudenmaier. Take, for instance, George Wise’s summary of historical findings as: Treating science and technology as separate spheres of knowledge, both man-made, appears to fit the historical record better than treating science as revealed knowledge and technology as a collection of artifacts once constructed by trial and error but now constructed by applying science. [Wise, 1985, p. 244; emphasis added] Vincenti approvingly quotes Wise and several other researchers, including Barnes and Layton, as concluding that “technology appears, not as derivative from science, but as an autonomous body of knowledge, identifiably different from the scientific knowledge with which it interacts.” [1990, pp.1-2; emphasis added]. Layton in turn seems to derive this view from the work of Alexandre Koyr´e, writing that [Koyr´e] held that technology constituted a system of thought essentially different from that of science. Technology generated its own independent rules which came ultimately to constitute a body of technological theory. [Layton, 1974, p. 40] These quotes show two things that are useful for an overview. Firstly, they express an aim that shapes several existing studies of technological knowledge. This aim may be called the epistemic emancipation of technology, i.e., to establish that technology is epistemically distinct from science. This emancipation aim makes sense against the background of the technology-as-applied-science debate in the 1960s and 1970s. In the last half of the 1980s, denying that technology merely

The Nature of Technological Knowledge

311

involves the application of scientific knowledge was no longer in need of further argumentation. Thus, a next step could be considered: that technology involves its own form of knowledge. However — this is the second useful aspect of the quotes — the epistemicemancipation claim can be interpreted in various ways. In one sense, which I call “weak emancipation”, it says that scientific and technological practice result in bodies of knowledge that are as distinct as our knowledge of plants and animals, or perhaps more strongly, as distinct as physics and chemistry. “Distinct” here means no more than that there is, as yet, no way of incorporating one body of knowledge into the other. One might think that this is a defensible, and sufficiently emancipatory claim about the relation between scientific and technological knowledge. Yet many authors might also be interpreted as making a stronger claim. Calling our knowledge of plants and animals “separate spheres”, “autonomous bodies”, or “of different forms” sounds exaggerated: irreducibility does not entail separation, let alone autonomy. If autonomy is taken in its standard sense of “self-government”, or the ability to set one’s own rules, it leads to a far stronger emancipatory claim than the minimal one considered above. Technological praxis results in an autonomous body of knowledge if this knowledge answers to its own epistemic rules, not those of science. Physics and chemistry are not mutually autonomous in this sense, since they answer to approximately the same rules; justifying a claim in physics is not qualitatively different from justifying a claim in chemistry, although specific methods may of course differ. Thus, calling technological knowledge “autonomous” more strongly emancipates it from scientific knowledge. Given the applied-science debate, this strong emancipation seems attractive. Having denied the thesis that technological praxis is epistemically dependent on science, one might be eager to prove that technology is epistemologically self-supporting, and not necessarily related to science. Reviewing the existing literature on technological knowledge shows that the strong autonomy thesis plays an important role in it. But this does not mean that authors have tried to validate the thesis in exactly the same way, nor that they have successfully established it. In this chapter, I give a critical overview of the literature by distinguishing four emancipation strategies that have been developed — without claiming that every author pursues only a single strategy, or that authors have appreciated the differences between the various strategies. These strategies are: • to contrast directly scientific and technological knowledge (Section 2). • to construct a taxonomy of technological knowledge (Section 4). • to appeal to the “tacit” nature of technological knowledge (Section 6). • to appeal to the prescriptive nature of technological knowledge (Section 8).

312

Wybo Houkes

After showing how the strategy has been endorsed, expressed and developed, I critically analyze its current success in establishing strong emancipationism.1 Furthermore, most review sections are succeeded by a section that contains a more general argument against the effectiveness of the strategy. To anticipate my conclusion: although the literature on technological knowledge is significantly shaped by the strong-emancipation ideal, efforts to realize it have not only been scattered and idiosyncratic, but also significantly underestimate the difficulties in establishing the ideal. My arguments do not show that strong emancipation is impossible to defend, but they do show that current arguments are ineffective. The critical review is complemented by a short look at one reason why historians and philosophers of technology might have abandoned Staudenmaier’s epistemic theme and the emancipation quest (Section 9). Still, I end the chapter on a more constructive note. In the course of my critical analysis I identify less ambitious and more detailed issues that might be addressed to improve our understanding of technological knowledge. I conclude with offering some suggestions for reviving the study into the nature of technological knowledge (Section 10).

2

CONTRASTING NATURAL AND ENGINEERING SCIENCE

A popular strategy for studying technological knowledge and arguing for epistemic emancipation is to contrast science and technology — more specifically: to look at differences between natural and engineering science. The latter is certainly not equivalent to technology, but I shall show that the narrowing of scope is understandable. Here, I review existing developments of the contrastive strategy. A first thing to note is that most authors who develop this strategy also share a key intuition. This intuition — which is not exclusive to the contrastive strategy — is that technology is, in all its aspects, aimed at practical usefulness. Thus, whether technological knowledge concerns artefacts, processes or other items, whether it is produced by engineers, less socially distinguished designers, or by consumers, the prima facie reason to call such knowledge ‘technological’ lies in its relation to human goals and actions. And just as scientific knowledge is aimed at, or more tenuously related to, the truth, so technological knowledge is shaped by its relation to practical usefulness. This ‘truth vs. usefulness’ intuition — TU-intuition for short — is repeated, in slightly different wordings, in many works, especially those in which a rough-andready characterisation of technology or technological knowledge is sought. Take, for instance: 1 Neither the weak nor the strong emancipation ideal is made explicit in the literature on technological knowledge. Some of the work reviewed in this essay might be interpreted as arguing for weak, rather than strong emancipation. Given my critical analysis, this interpretation might be more charitable. It is also less interesting, since establishing weak emancipation is a rather trivial aim. Thus, I have taken the liberty of reviewing/reconstructing the literature with regard to its effectiveness in achieving a more difficult, perhaps even unattainable, goal.

The Nature of Technological Knowledge

313

Technology. . . aims to be effective rather than true [Jarvie, 1972, p. 55] Science seeks basic understanding (. . . ). Technology seeks means for making and doing things. [Hindle, 1966, p. 4] Science concerns itself with what is, technology with what is to be. [Skolimowski, 1972, p. 44] The TU-intuition also shapes Walter Vincenti’s work. From virtually all his individual case studies, Vincenti draws the conclusion that technological knowledge is distinct from science because it is related to practical purposes. One telling quote is the following:2 [T]he criterion for retaining a variation in engineering must be, in the end, Does it help in designing something that works in solution of some practical problem? The criterion for scientific knowledge, however we put it, must certainly be different . . . Borrowing a phrase used by Alexander Keller . . . I would venture it more or less as follows: Does it help in understanding ‘some peculiar features of the universe’ ? [Vincenti, 1990, p. 254]; (original emphasis) Few authors go beyond expressing the TU-intuition and arguing why it shows that technology involves more than applying scientific knowledge. This is unfortunate, because the intuition alone does not at all establish epistemic emancipation of any variety. For the intuition does not only present a questionable image of science (an objection that shall be considered in Section 3), but it is also unclear on the technology side: does it address engineering practice, engineering science, engineering design and/or technology in one of the possible senses of the term? These meanings can be distinguished more or less clearly (see e.g., [Mitcham, 1978]), and it is often useful to do so. However, a focus on epistemological issues might make the distinctions less relevant. The reason is that not all technological practices are, on the face of it, equally relevant to technological knowledge. Engineering practice, design, and also the use of technical artefacts typically involve knowledge, and might often lead to acquiring knowledge, but they are not primarily knowledge-producing activities. Engineering science is. Therefore, it is a natural starting point for inquiries into the nature of technological knowledge.3 Those who go beyond expressing the TU-intuition frequently focus on the engineering sciences and the role of theories and models in these disciplines. The common supposition is that this role is instrumental. More than natural scientists, 2 Passages

in which Vincenti expresses the TU-intuition in slightly different, artefact-oriented words, are: “In scientific knowledge the purpose is understanding of nature; in engineering science the ultimate goal . . . is the creation of artefacts”[Vincenti, 1990, p. 135] and “Engineers use knowledge primarily to design, produce and operate artefacts, goals that can be taken to define engineering. (. . . ) Scientists, by contrast, use knowledge primarily to generate more knowledge” [ibid., p. 226]. 3 Engineering science is also a risky starting point, because of all technological practices, it is probably closest to science, and therefore least likely to be autonomous — or even in need of autonomy.

314

Wybo Houkes

engineering scientists are supposed to be content with theories and models that are practically useful, but known to be incorrect. This way of studying technological knowledge is not only evidently connected to the TU-intuition. It also stands a good chance of non-trivially developing it, because it generates some specific research questions — which test both the TU-intuition and the ideal of strong emancipation. I give only two examples of such questions here, in the form of specific hypotheses. Firstly, if practical usefulness is the central value of technological practice, one would expect this to affect the validation of theories and models in engineering science, given their obvious relation to technological practice. To put it roughly, one would expect theories and models in engineering science to be valued if — although perhaps not only if — they are usefulness-tracking, unlike theories and models in natural science. If this abstract difference, based on the TU-intuition, is not manifested in concrete evaluations of theories and models, it makes little sense to call technology epistemically autonomous, at least in this important respect. Secondly, engineers frequently employ theories from the natural sciences. If we suppose that these theories were previously evaluated for their verisimilitude and that engineering scientists value them for their usefulness, one would expect that changes (if any) to these scientific theories and models reflect this shift of values — and that not all such changes are valuable within the natural sciences. If no such changes are made, or if every change by engineering scientists is hailed as simultaneously advancing scientific research, the distinction between scientific and technological knowledge has no normative bite. Neither these nor other, equally specific hypotheses regarding technological knowledge have been investigated. Most authors who address epistemic differences between science and technology are content to state the TU-intuition, giving some illustrations — usually of artefacts that were developed on the basis of false theories or in the absence of theories. The validation of theories and models used in the engineering sciences is seldom studied. Still, some papers identify or even illuminate the issues raised above. I discuss four in some detail. The first three papers address the first issue, that of the validation of theories and models developed within the engineering sciences. Both Ronald Laymon [1989] and Vincent Hendricks, Arne Jakobsen and Stig Andur Pedersen [2000] relate the development of specific models to the central value of practical usefulness. Their main aim is to show that this value is at work and that engineering is therefore different from science, but they also offer material for the more valuable analysis of how the value affects the evaluation of models. Ronald Laymon examines the role of as-if theories, or fictitious models in engineering science. More specifically, he studies the history of models of a swinging pendulum, as they might be used in instrument building. Such models have to account for buoyancy effects: the textbook harmonic-oscillator idealization is of little use for practical purposes. One way to provide such an account is to correct for the mass of displaced air, and then to correct this by means of an experimentally determined correction factor — which accounts for all non-hydrostatic effects of

The Nature of Technological Knowledge

315

the presence of air. Such empirical data raise a projectibility issue: will they apply to slightly different pendulums in slightly different circumstances? Laymon discusses various responses to this question that may be typical for engineering. One is that, in technological practice, the projectibility issue may be largely avoided by rebuilding successful devices and/or (artificially) recreating the circumstances of successful performance. This does not add significantly to the body of technological knowledge, making this response uninteresting for our current purpose. Another response is more interesting. To understand the behaviour of pendulums in new circumstances, the experimental correction factors may be analysed for continuities and correlated changes, and one may seek explanations of such correlations. The engineering scientist appears to have considerable freedom in seeking explanations: because of the ultimate goal to produce practically useful artefacts, clearly fictitious or as-if theories are just as welcome as realistic ones. Laymon mentions Airy’s theory of “adhesive air” as an example: the presence of air may be accounted for by supposing that a quantity of air adheres to the pendulum while moving — adding to its buoyancy without changing its weight. This theory can be taken seriously within engineering science if it has instrumental value. Moreover, it leaves open a more realistic explanation in terms of the viscosity of air, which can again be corrected for its “viscosity bias”. This cycle of idealized model, correction, explanation and refinement of the model is probably familiar from other experimental sciences. Yet the role that as-if theories play in the cycle presented by Laymon may be characteristic for the engineering sciences.4 And, what is more important to the topic of this paper, by means of his concrete example, Laymon gives considerably more content to the claim that engineers do not seek “true” theories, and are primarily interested in “usefulness”. Another methodological feature of engineering science, the existence of “lumpedparameter models”, is examined in some detail in [Hendricks et al., 2000]. In these models, the behaviour of a system is described by analyzing it as a complex of subsystems, for which idealized models are available. These models may not be realistic. They may even be transferred from a different domain altogether. A mechanical system may, for instance, be modelled in such a way that an isomorphism with the model of an electrical system is revealed. The point of this method of decomposition-cum-isomorphism cannot be veracity. Thus, lumping is another example of the way in which the central value of practical usefulness affects the evaluation of models: because engineering science aims at usefulness rather than truth, (more) lumped-parameter models may be acceptable. This reveals an evaluative difference between the natural and engineering sciences, albeit one that calls for more detailed analysis: as Hendricks, Jakobsen and Pedersen notice, lumping4 Laymon’s expression of this difference relies on the TU-intuition: “The problem created by the use of idealizations for science . . . is to determine whether failures to achieve experimental fit to within experimental error are due to the falsity of the theory or of idealization. (. . . ) For the engineer the problem seems altogether different. If [the closeness of predictive fit achieved by theory and idealization] is good enough for some practical purpose then the engineer’s job is done . . . ” [Laymon, 1989, p. 354].

316

Wybo Houkes

parameter models are also found in some parts of physics. They mention models in solid-state physics; the liquid-drop model of nuclear physics may provide another example. A closer comparative study of the roles these models play, and the conditions for accepting or rejecting them may lead to more insight in the relation between usefulness and the engineering sciences. The liquid-drop model of nuclear physics is, for example, not regarded as merely a predictive instrument: it is commonly supposed that nuclei are structurally similar to drops of liquid, and that this explains some aspects of their behaviour. Presumably, engineers do not take the successes of an “electrical” model of a mechanical system to show such a structural similarity. Whether this means that lumped-parameter models in engineering sciences are less tightly constrained, because there are no requirements of truth-likeness, remains to be seen.5 Peter Kroes [1992] takes another perspective on the development of theories in the engineering sciences. Rather than stressing the role of idealized or fictional models, he studies what he calls “engineering theories”, i.e., formally or mathematically structured, experimentally validated systems of knowledge that explain the technological function of a particular class of technical artefacts or technical-artefact-related materials in terms of their design or construction.6 Using Pambour’s theory of the steam engine as an illustration, Kroes argues that design considerations confer a “distinctly technological flavour” [1992, p. 70] on engineering theories. This flavour shows in three features. Firstly, as the characterization already makes clear, the domain of application of an engineering theory is a designable technical artefact or artefact-related material, not a physical phenomenon: Pambour’s theory is about piston-operated steam engines, not about all heat engines. Secondly, engineering theories may contain basic principles related to the design or construction of technical artefacts. These principles, such as Pambour’s principle of the conservation of steam, may be reformulated in terms of physical boundary conditions, but they involve more than an application of physical principles: design considerations, not physical considerations, explain why these conditions are relevant. Thirdly and finally, engineering theories employ technical concepts as well as physical ones. Technical concepts are again related to design characteristics. To confuse matters, some theoretical concepts may be homonyms, referring to either technical or physical characteristics. Examples of concepts with such “dual significance” [Kroes, 1992, p. 91] are “resistance” and “pressure”. 5 Hendricks and his co-authors regard engineering science as combining the values of truthfulness and usefulness: “(...) the objective for engineering science is an optimal degree of theoretical correctness (typically limited by time and resources) combined with pragmatic considerations of practical usability.” [Hendricks et al., 1999, p. 302]. This combination view seems to minimize the difference between natural and engineering science, since the former also seems to combine truthfulness and usefulness. See also section 3. 6 Kroes [1992, p. 69] grafts this characterization on Staudenmaier’s [1985, p. 107] definition of an engineering theory. He modifies it to focus on technical artefacts, and technological functions rather than behavioural characteristics; both modifications are indeed called for, since: (a) many theories in the experimental sciences describe artefacts, viz. artifically induced phenomena; (b) the behaviour of artefacts can be described in physical or chemical terms.

The Nature of Technological Knowledge

317

Of the three papers discussed, Kroes’s is the most specific. It focuses on a clearly circumscribed subset of the total body of engineering knowledge, and identifies several distinguishing features. Moreover, it relates these features to some of the most basic concepts used to describe technology, “artefact”, “(technical) function” and “design”. As I shall argue in the next section, this gives the approach an analytic edge over that in the other two papers, which discuss more general features of engineering models, and appeal merely to the TU-intuition to distinguish the models from scientific ones. Yet Kroes’s approach also has drawbacks, partly because of its specificity. For one thing, Kroes’s approach might only emancipate a very small part of technological knowledge. Some knowledge may have a “distinctly technological flavour” without being related to a specific type of technical artefact; Vincenti’s control-volume analysis, discussed immediately below, comes to mind as an example. Secondly, the basic concepts invoked by Kroes, such as “design” and “function”, are in need of further analysis. If, for instance, “design” may refer to the selection of physical objects for practical purposes, and function to physical behaviour, the distinction between engineering theories and scientific theories may evaporate. Thirdly, Kroes’s focus on distinctions between concepts is innovative and initially plausible, but at further inspection problematic. If, for instance, “pressure” indeed has a dual significance, should Pambour’s theory be disambiguated so that it only contains design parameters? Doing so seems necessary to argue that engineering theories differ from scientific theories “in substance”, as Kroes suggests [1992, p. 93]. However, once disambiguated in this way, it is not clear how engineering theories “exploit scientific theories in solving technical problems” [Kroes, 1992, p. 92], since their content is, strictly speaking, different from that of scientific theories. The second issue, the adoption and adaptation of scientific theories within engineering science, is even more rarely addressed. It is, however, the topic of one of Vincenti’s case studies [1990, Ch.4; the original paper is from 1982]. Vincenti examines the development of control-volume analysis, a technique for solving problems regarding fluid flow by selecting a hypothetical surface and calculating the values of physical quantities on its boundaries. This technique is compatible with thermodynamics and does not add irreducible concepts to it, and it is a standard part of many engineering curricula. It is not, however, found in thermodynamics textbooks for physicists — Vincenti mentions a textbook that presented controlvolume analysis in an edition for physicists and engineering students, but omitted it in a later edition for physicists alone. The reason is that the technique is global. Control-volume analysis only yields overall results regarding the behaviour of a system; the inside of the hypothetical control volume may be regarded as a physical black box.7 Within the confines of this black-boxing, control-volume analysis is a powerful technique, which can be used to describe the behaviour of all kinds of devices that involve fluid flow — including rocket motors and pipes in installa7 A physicist might want to use such a global calculation, if she is interested in predicting fluid flow. It would, however, be remarkable if physicists would develop a systematic technique for such calculations.

318

Wybo Houkes

tions. For such systems, the control volume and the relevant quantities are easily determined by the context of use: “what goes in” and “what comes out” are far more relevant to the performance of artefacts than “what goes on inside”. Thus, control-volume analysis shows how engineering scientists adopt a physical theory and make it suitable for their, presumably, different purpose. From this brief review, it can be concluded that the evaluation of theories and models in engineering science points out several interesting, possibly distinctive features of technological knowledge — but that the literature does little more than point out these features, and that emancipatory arguments often appeal to the TU-intuition without explicating it. Furthermore, all efforts to examine evaluative differences between natural and engineering science have been isolated: the papers reviewed have not given rise to sustained discussion or further refinement; they do not even build upon each other.

3

THE INSTRUMENTALIST OBJECTION

The discussion above has shown that the TU-intuition is a recurrent theme in the literature on technological knowledge. As stated above, the TU-intuition understands the difference between natural science and technology (or, more narrowly, the engineering sciences) in terms of a difference in goals: the former aims at finding out true theories, whereas the latter aims at practical usefulness. In this section, I point out that merely appealing to this intuition is not sufficient to emancipate technological knowledge. For the difference in goals appears to presuppose a realist conception of science, on which scientific theories ought to be interpreted as descriptions of (the structure of) reality, and science as a continuing enterprise to construct more accurate theories. There are, of course, many ways of developing this realist view of science and scientific theories,8 and a one-line description may not be representative for all of them. Still, the broad spectrum of realist conceptions can be contrasted with another view of science: instrumentalism.9 Instrumentalists seek to decouple scientific inquiry from truth, and instead emphasize its connection to usefulness. There are several ways to achieve this. Some instrumentalists argue for a re-appraisal of the notion of truth that is relevant to scientific inquiry: instead of the traditional correspondence theory, they propose a “pragmatic” theory of truth. Other instrumentalists prefer an epistemic route to the semantic one. They accept the realist idea that scientific theories are candidates for being true in a correspondence sense, but they deny that scientists may justifiably accept or reject a theory because of its truth-likeness. Instead, they say that theory choice 8 See, for instance, Ladyman’s [2007] review of traditional and contemporary varieties of realism and instrumentalism. 9 The discussion of instrumentalism as an alternative to scientific realism does not reflect an opinion that instrumentalism is the only viable anti-realist conception of science. Rather, instrumentalism is the anti-realist conception that most directly undermines the TU-intuition.

The Nature of Technological Knowledge

319

ought to be dictated by the usefulness of theories for solving the empirical and theoretical problems of science.10 That the instrumentalist conception of science conflicts with the TU-intuition is clear enough. If, like technology, science is concerned with usefulness instead of truth (in the correspondence sense), forging an epistemic distinction between the two activities in terms of their goals seems a questionable enterprise. More specifically, the primary epistemic virtues of science and technology would be the same, making it impossible to emancipate technology from science through distinguishing their primary epistemic virtues. One might try to overcome this obstacle by arguing directly against the instrumentalist conception of science, or at least to decrease its plausibility by attacking the arguments supporting it. Such a maneuver would lead us into the territory of the general philosophy of science, so I do not consider it here. However, its effectiveness seems doubtful. Instrumentalism is a minority position in the philosophy of science, but the arguments used to sustain it, such as the pessimistic induction and underdetermination thesis, are plausible and remain defensible despite numerous attempts to invalidate them. It would be interesting to see whether technology offers a fresh perspective on the entrenched realism debate, but it is hard to feel optimistic about the possibilities of a major breakthrough.11 Another response might be to accept the main thrust of the argument, but to remove its sting by arguing that technological knowledge is appraised, not in terms of usefulness in general, but in terms of practical usefulness. Technology concerns deliberate changes that serve more or less immediate practical purposes, like transportation and hygiene. To these purposes, engineers primarily produce (designs of) technical artefacts, including systems and processes, and they are aided in this by theories. Scientific theories may be understood as instruments, just like technical artefacts, and the construction of theories may be an instrumental activity, just like design. Yet these instruments serve “theoretical” purposes such as predicting or capturing data, rather than the “practical” purposes that shape technology. This response might go some way towards dispelling the instrumentalist objection. Yet it appears that, by accepting the gist of the objection, the goal of epistemic emancipation becomes unattainable. If science and technology are subordinate to the same primary epistemic virtue — namely usefulness — establishing strong emancipation by focussing on more specific goals seems difficult. Theories in particle physics and microbiology serve different specific purposes, e.g., to predict the behaviour of mesons and of enzymes, but since the primary epistemic virtue is the same for both types of theories, we might not want to say that they answer to their own sets of rules; instead, physical and microbiological knowledge 10 [Stanford, 2005] is a recent overview of historical, current and possible instrumentalist conceptions of science. 11 One may build upon Hacking’s [1983] suggestion that scientists treat those objects as real which they can manipulate, and to examine the role of technology in shaping this manipulability, and of engineering science in describing it.

320

Wybo Houkes

are typically regarded as species of one epistemic kind, namely scientific knowledge. This worry increases once an additional feature of the response is noticed. No one would want to deny that technical artefacts, such as cars and hand soap, serve immediate practical purposes. But artefacts do not constitute technological knowledge, although their design and production might be based on it. It seems that, to keep the strong-emancipation ideal alive, the TU-intuition must be explicated by focussing on the epistemic products of technology, such as theories and models in the engineering sciences. At this level, making a principled distinction in terms of specific goals is less plausible. The liquid-drop model is known to be unrealistic, but still used to predict the behaviour of nuclei. Airy’s adhesive-air theory is known to be false, but used to predict the behaviour of pendulums. If there are any epistemic differences, they remain to be discovered, below the surface. This renewed objection suggests a third response, which is to bite the bullet. For the moment, instrumentalism regarding scientific theories seems a viable position, which reduces the epistemic contrast between science and technology to the vanishing point. Therefore, philosophers of technology who seek epistemic emancipation cannot rely on a realist image of science — despite multiple attempts, this image has not been shown to be sufficiently reliable, and the naive version that appears to be presupposed in the TU-intuition certainly needs significant refinement. Still, detailed studies into the acceptance of theories and models by engineers — such as those discussed in Section 2 — may lead to additional arguments for a realist image of science, or to an instrumentalist image that retains some contrasts with technology. If it could be shown, for instance, that the contexts in which engineers accept unrealistic models are qualitatively different from the contexts in which scientists would be prepared to do so; or that engineers accept more blatantly false theories than any scientists would be prepared to do, then the apparently contrast-reducing statement that “Both scientists and engineers use theories as instruments” could be explicated into different statements about science and technology. Such a sophisticated response has, to the best of my knowledge, never been given. As indicated above, Laymon considers the need for such a response, offers material that may be helpful, but ultimately relies on the TU-intuition and a realist image of science himself. Alternatively, one could follow Kroes’s example and try to specify the instrumental role of engineering theories and models by more closely circumscribing the practical purpose, e.g., in terms of the design and construction of technical artefacts. This strategy seems promising, in the sense that it might explicate the TU-intuition in terms of several concepts that are fundamental to our descriptions of technology. However, these concepts, such as “design” and “technical artefact” are in need of further analysis. Furthermore, narrowing down the practical context of technology runs the risk of narrowing the scope of one’s analysis of technological knowledge — as pointed out in Section 2, Kroes’s analysis of engineering theories might address only a small portion of what might be called technological knowledge.

The Nature of Technological Knowledge

321

Still, by developing arguments and analyses along the lines suggested here, philosophers of technology could examine the role of theories and models in the engineering sciences and simultaneously contribute to the philosophy of science, instead of (perhaps unconsciously) applying insufficiently sophisticated ideas from the philosophy of science. 4 TAXONOMIES OF TECHNOLOGICAL KNOWLEDGE Inventorizing the contents of technological knowledge would improve our understanding of it. This need not involve an explicit contrast with another type of knowledge, just like an inventory of bears need not involve contrasting them with wolves. Thus, the taxonomical way of analyzing technological knowledge is at least prima facie different from the contrastive analysis outlined in the previous two sections. Yet, like this analysis, making an inventory can serve the purpose of epistemic emancipation:12 if the items on this inventory are sufficiently different from those on an inventory of scientific knowledge, one may take this as evidence that they embody different types of knowledge. The traditional distinction between fields within the engineering sciences is an obvious starting-point for a classification of technological knowledge. In engineering schools and elsewhere, e.g., in library cataloguing systems, we find taxa such as mechanical engineering, chemical engineering, and bio-medical engineering. These disciplines and bodies of knowledge appear to be named after the kind of scientific knowledge that they are thought to apply. Moreover, we find taxa such as software engineering and maritime engineering, which appear to be based on the kinds of artefacts produced within the fields. Neither way of classification seems epistemically informative, and the former might even strike those interested in epistemic emancipation as misleading. It is therefore hardly surprising that attempts at classification seldom start from existing distinctions between engineering fields and sciences. They are even seldom presented as attempts at reconstructing or revising these distinctions. Rather, most classifications present categories that cut across the boundaries between fields and disciplines. Several authors have proposed taxonomies of technological knowledge. I shall give an overview of four efforts: those made by Vincenti [1990], Ropohl [1997], Faulkner [1994], and de Vries [2003].13 Not all of these authors explicitly state the purpose of epistemic emancipation.14 Nevertheless, given the context of this 12 The taxonomies may serve other purposes, for instance aiding engineers in classifying and storing their knowledge. Broens and De Vries [2003] note that engineers find Vincenti’s taxonomy most useful for this purpose — which is compatible with any conclusion regarding the usefulness of this taxonomy for emancipatory purposes. 13 My presentation in the remainder of this section has profited from Broens and De Vries [2003], but differs from it in some details and criticisms. 14 The doubts I raise (especially the general doubts presented in Section 5) might strike some as unfair criticisms of proposed taxonomies. One might reasonably doubt whether a taxonomy could even in principle be used for emancipatory purposes, i.e., to determine the (autonomous) nature of the knowledge that is classified. Still, existing work on technological knowledge often

322

Wybo Houkes

paper, I shall review all four in this light. Moreover, I shall assess the taxonomies with regard to their formal merits: as taxonomies, they ought to present categories that are mutually exclusive and jointly complete; every item in the domain should be classified in one and only one category. In the next section, I go on to consider the viability of the taxonomical way of emancipating technology from science. The Table (pp. 324–325) provides an overview of the categories of knowledge introduced by the four authors, along with subcategories, a one-phrase clarification, and/or some examples. Not all labels may be self-explanatory; indeed, key notions in all four taxonomies are in need of further analysis. For the moment, I postpone further clarification and comments. In the remainder of this section, I focus on three aspects of the taxonomies: their formal characteristics; their mutual differences; and the way(s) in which they bring to light the relation between technological and scientific knowledge. Let us start with the formal characteristics, i.e., exclusiveness and completeness. Here, Vincenti’s classification performs badly — as he admits before starting his presentation [1990, p. 208]. To give just one example, his scheme is partly guided by the distinction between codifiable theoretical tools and quantitative data on the one hand, and uncodified practical considerations on the other.15 However, practical considerations may be codified [1990, p. 219], without thereby turning into either tools or data. A similar observation may be made regarding Faulkner’s taxonomy, since she incorporates Vincenti’s distinction, rephrasing it as one between “practical experience” and “engineering theory”. Furthermore, she grounds her distinctions in the possible subjects of technological knowledge, whereas one element of knowledge may have multiple subjects (e.g., performance data about and specifications of material properties). The taxonomies of Ropohl and De Vries seem to fare better in this respect. Neither includes a distinction between knowledge and skills, or between variously codifiable elements of knowledge. Instead, both authors refer, in different ways, to the distinction between structure and function. At first glance, this seems sufficiently principled to support mutual exclusiveness of categories. Yet problems ensue as soon as one looks for a more detailed understanding. For one thing, the notion of artefact function is far from uncontested, as Preston’s contribution to this handbook makes clear; on some views, such as Robert Cummins’ [1975], the function of an artefact may not be distinct from structural features, such as dispositions and other physical behaviour. These views may be contested qua theories of artefact functions, but this holds the two taxonomies hostage to an unresolved philosophical debate. A second set of remarks concerns the manifest differences between the taxonomies, which roughly divide into two pairs. The systems of Vincenti and takes the form of constructing a taxonomy, and is frequently motivated by the quest for epistemic emancipation. It therefore makes sense to evaluate the taxonomical work in the light of this quest. 15 Vincenti distinguishes these practical considerations from both tools and data because they “frequently do not lend themselves to theorizing, tabulation, or programming into a computer” and “they are hard to find written down” [1990, p. 217].

The Nature of Technological Knowledge

323

Faulkner, and those of Ropohl and De Vries, seem similar, but also show some notable differences. For the Vincenti-Faulkner pair, the similarities are easily explained, because Faulkner used Vincenti’s (earlier) work as an explicit guiding line for her own investigation into innovation. Still, she added categories (e.g., knowledge about knowledge) and subcategories (e.g., new product ideas), removed others (e.g., quantitative data), and reshuffled still others (e.g., by combining in one subcategory both operational principles and normal configurations). Matching the taxonomies of Ropohl and De Vries is harder, given their terminological differences. Ropohl’s functional rules, for instance, appear to match De Vries’ process knowledge rather than his functional-nature knowledge. Still, that both authors distinguish a ‘functional’ category makes their taxonomies more alike to each other than to any of the other two. These differences partly reflect differences in guiding principles. As is routinely noted in textbooks that deal with classification and categorizations, items can be grouped together in arbitrarily many ways. Cars, for instance, can be classified in terms of ownership (privately owned, rental, leased, etc.), fuel (gasoline, electrical, hybrid, etc.), engine type, colour, ownership history (first-hand, second-hand, third-hand, etc.) number of dents, etc. To curtail this arbitrariness, some guiding principle should be invoked. For many scientific classifications, it is required that its classes “function in, or facilitate the formation of, scientific laws”.16 This requirement is pointless in the present context, not only because the four taxonomies are reviewed for their emancipatory success, but also because some of them feature a (sub-)category of scientific theories and laws. Ropohl and De Vries instead use perspectives from the philosophy of technology: their taxonomies are guided by systems philosophy and the dual-nature thesis respectively. The other two taxonomies have no clear guiding principle: Vincenti’s taxonomy seems largely the result of personal reflection on a large number of case studies in one, designoriented discipline, namely aeronautical engineering, whereas Faulkner’s additions and adaptations to Vincenti’s system mainly stem from her studies into technological innovation. I will return to this difference in guiding principles, or lack thereof, in the next section. For the moment, I note that this underlying difference means that one resolution of the manifest differences is unavailable. If two biologists agree on the criteria for speciation, but one distinguishes five species of dog, and the other distinguishes six species, a straightforward solution is that the former has overlooked one species. This resolution is probably not available for taxonomies of technological knowledge: the four systems cannot be merged into one super-taxonomy by distinguishing every category that is listed by at least one taxonomy. For one thing, this super-taxonomy would share the formal flaws of any original taxonomy; for another, it would require some possibly arbitrary decisions. De Vries, for instance, does not distinguish competences and know-how from theoretical or propositional knowledge. Given the other systems, he might have done so in two different ways: he might have followed Ropohl’s example in listing know-how as a 16 David

Hull, “Taxonomy”, in the Routledge Encyclopaedia of Philosophy.

324

Wybo Houkes

Vincenti [1990]

Ropohl [1997]

Fundamental design concepts

Structural rules

• operational principles • normal configurations

on the assembly and interplay of the components of a technical system

Criteria and specifications

Technological laws

• general, qualitative goalsa • specific, quantitative goals • goal-to-specification translationsb

Theoretical tools

transformation of natural laws with regard to technical processes

Functional rules

• models and theories • intellectual concepts (e.g., ‘boundary layer’)

Quantitative data • descriptive (e.g., operational conditions, human behaviour) • prescriptive (e.g., safety factors)

Practical considerations

what to do if a certain result is to be attained under given circumstances

Technical know-how (implicit knowledge and skills)

Socio-technical understanding

• experience from production, operation, accidents • design rules of thumb

systematic knowledge about the relation between artefacts, natural environment and social practice

Design instrumentalities • structured design procedures • ways of thinking (e.g., controlvolume thinking) • judgemental skills a This b See

subcategory and the next are only implicitly distinguished by Vincenti. Marc de Vries’ contribution to this Handbook for a closer analysis of this subcategory.

Table 1.a

The Nature of Technological Knowledge

Faulkner [1994]

De Vries [2003]

Related to natural world

Physical-nature knowledge

325

• scientific and engineering theory • material properties

Related to design practice • • • • •

Functional-nature knowledge

criteria and specifications instrumentalities fundamental design concpets competence practical experience

Related to experimental R & D • • • •

experimental and test procedures research instrumentalities research competence experimental and test data

Related to final product

Knowlege of physics-function relations

Process knowledge

• new product ideas • operating performance • production competence

Related to knowledge • location of knowledge • availability of equipment, materials, facilities or services

Table 1.b

326

Wybo Houkes

separate category, or he might have included the distinction in the form of subcategories. Furthermore, only Ropohl explicitly lists “socio-technical understanding” as a category; this kind of knowledge is either missing from the other taxonomies, or very covertly included. The third and final set of remarks concerns the way in which the taxonomies incorporate possible differences between scientific and technological knowledge. All taxonomies list categories or subcategories that largely, or even exclusively, appear to consist of run-of-the-mill scientific knowledge. Therefore, if the taxonomies serve the purpose of epistemic emancipation at all, they do so by incorporating scientific knowledge, rather than by contrasting an elaborately classified system of technological knowledge with a system of scientific knowledge. So, Vincenti includes models and techniques from mathematics and physics among his examples of theoretical tools; Ropohl’s category of structural rules might, and De Vries’ structural-nature knowledge definitely does, include many statements about physical or geometrical relations between artefact components; and Faulkner explicitly distinguishes scientific and engineering theory as a subcategory. This incorporative strategy seems reasonable, if “technological knowledge” is taken to be the body of knowledge used in engineering science, design and/or practice; after all, engineers routinely use scientific theories and models. Still, the strategy creates at least two problems: one of a formal nature, and the other with respect to the goal of epistemic emancipation. Formally, as soon as one of the four taxonomies (or the super-taxonomy that results from combining them) is combined with a taxonomy of scientific knowledge, a taxonomy results that does not satisfy the demand of mutual exclusivity: some (sub-)categories will feature both in the technological and in the scientific part of the encompassing taxonomy. With regard to emancipation, it makes little sense to include categories of knowledge that answer to scientific standards — the resulting body of technological knowledge will certainly not be (completely) autonomous if these standards apply to part of it. One may think to solve both problems at once by excluding from one’s taxonomy of technological knowledge all categories that feature in a taxonomy of scientific knowledge. In this way, double entries are avoided, and one may still claim that the resulting science-less body of technological knowledge answers only to its own rules. The resulting taxonomy would remain silent on the nature of those rules — making the autonomy claim uninformative. What is worse, it would make the autonomy claim trivially true, by constructing technological knowledge as an epistemic system that is different from science. Thus, this solution might offer only formal consolation, without furthering emancipatory ends.17

17 This can be avoided if the identification of “genuinely technological” elements of technological knowledge is followed by an analysis of their epistemic character. Even then, however, one might do no more than make explicit one’s intuititions regarding the epistemic differences between science and technology, since these intuitions might be presupposed in the identification of the “genuinely technological” elements.

The Nature of Technological Knowledge

327

Another, more roundabout solution would be to argue that the inclusion of knowledge from scientific disciplines such as mathematics and physics is only apparent. One might maintain that this knowledge is either selected from that available within the discipline by criteria that are unique to engineering science and design — meaning that the distinctive nature of technological knowledge features in the criteria of selection; or that scientific knowledge is adapted to engineering purposes — meaning that the distinctive nature of technological knowledge features in the content of every (sub-)category. Both of these options are familiar from Section 2: they are two ways in which the contrastive strategy for epistemic emancipation may be developed. This does not mean that this roundabout solution must come to naught. Yet it does mean that, as soon as “scientific knowledge” is included among the (sub-)categories of a taxonomy of technological knowledge — as in the four reviewed taxonomies — the taxonomical strategy for emancipation reduces to the contrastive strategy. 5 THE DOUBLE-DEMARCATION PROBLEM Apart from the problems with individual taxonomies discussed in the previous section, there is a more general problem that needs addressing. This problem affects both the contrastive and the taxonomical strategy for epistemic emancipation. To appreciate this general problem, it is worthwhile to consider the other end of the epistemic-emancipation problem: scientific knowledge. Suppose someone is interested in making a list of types of scientific knowledge, for instance to distinguish possible contributions to an encyclopaedia of science. There are various ways of organizing this classification, requiring some kind of principled decision, as discussed in the previous section. Suppose this decision is guided by the results of science, such as the lawlike regularities that form the backbone of scientific theories, or the theories themselves. Thus, one obtains entries about Newton’s laws of motion or classical mechanics, about chemical bonds, or the regularities guiding supply and demand. In addition, a second decision is required, one that concerns the boundaries of scientific knowledge. One needs to decide why (not) to include controversial regularities, such as homeopathy’s laws of similars and infinitesimals or the correlation between fossil-fuel consumption and climate change. And one needs to decide whether to include models and phenomenological laws, which merely describe and do not explain by referring to some underlying causal mechanism. Without these decisions, one about the guiding classificatory principle and two about the boundaries of knowledge, a list of scientific knowledge would be arbitrary. Yet at least one of these decisions is notoriously hard to make in a principled way: the decision to exclude, for instance, the central tenets of homeopathy amounts, of course, to the familiar problem of demarcating science from pseudoscience, or unscientific knowledge. The failure of various purported demarcation criteria forms the backbone of many introductions into the philosophy of science. There may be characteristics that many sciences have in common, and some that

328

Wybo Houkes

most do not have in common with pseudo-science, nonsense or common sense; but if there is an acceptable, clear-cut criterion to be had, no-one has been able to formulate it. Fallibility, confirmation, prediction and explanation seem central elements, but philosophers cannot even agree on these elements, let alone on a slogan that captures them all. This problem affects taxonomies of technological knowledge, if they are used for epistemic emancipation. Firstly, the negative experiences with demarcation in science provide inductive support for pessimism about similar inquires into technological knowledge. Thus, compiling an encyclopaedia of technological knowledge seems at least as arbitrary as the encyclopaedia-of-science project. A complicating factor is that ‘technological knowledge’ is, to some extent, a technical term. Whereas ‘scientific knowledge’ sees a considerable amount of everyday usage, one seldom comes across descriptions of a model or research result as ‘technological knowledge’. This means that determining the boundaries of this type of knowledge may be, in a sense, easier than determining those of scientific knowledge. There may be entries that are beyond controversy, and some of these have been used as examples of technological knowledge in the literature: Vincenti’s [1990] control-volume analysis, Constant’s [1999] material-balance analysis, finiteelement analysis, and Smith’s [1960] metallurgy all come to mind. Beyond the domain of those examples there is, undoubtedly, a grey area, but if ‘technological knowledge’ is indeed a technical term, this part of its extension may be determined by stipulation. Although stipulations are, in this case, a legitimate way of solving boundary problems, they make fully explicit the arbitrariness of this constitutive rule for compiling a list of technological knowledge. To give two examples: all the paradigmatic entries mentioned above concern knowledge that is produced and employed by engineers, but technological knowledge might also conceivably include the instrumental knowledge that users possess about their cars and computers; and all entries mentioned above concern knowledge that can be expressed verbally, whereas much of our knowledge about technology appears to consist of know-how and competences. To be sure, one might resolve the latter issue by distinguishing ‘technological’ knowledge from ‘technical’ knowledge, where the latter consists of non-codified or non-codifiable techniques for achieving practical purposes. This is not only a stipulation, but also a distinction that does not seem to guide any of the taxonomies of technological knowledge currently on offer — all the examples reviewed in the previous section either explicitly include know-how and competences or, in De Vries’ case, do not exclude them. Thus, a taxonomy based on this distinction would be idiosyncratic, even if there is no rich tradition with which it would break. To make things worse, the demise of demarcation as a philosophical research project affects the taxonomical strategy in another way. If constructing a taxonomy of technological knowledge is to serve the purpose of emancipating it from scientific knowledge, it should at least be clear in what principled way the taxonomy distinguishes both types of knowledge. There should, in other words, be a

The Nature of Technological Knowledge

329

reason why some item is included in one list rather than another. It is possible to construct a list without first explicitly stating some criterion for including items: the list may be constructed extensionally, by including knowledge that is produced in an engineering context, or knowledge that concerns the use and design of technical artefacts. Still, if such a taxonomy of technological knowledge is to serve emancipatory purposes at some point, it encounters a double demarcation problem. It should put clear boundaries to the term ‘technological knowledge’ and, simultaneously, distinguish it from the equally vague term ‘scientific knowledge’. Even philosophers without great sceptical inclinations might feel cagey about such an enterprise. There may be clear and uncontroversial examples of scientific and technological knowledge, and these may serve as prototypes for distinguishing the two epistemic categories. However, the mere existence of paradigmatic examples does not solve a demarcation problem. Hardly any philosopher of science would deny that Newtonian mechanics and the knowledge compiled in your local phonebook may serve as paradigms for scientific and non-scientific knowledge. Still, a criterion is needed for evaluating borderline or otherwise disputed cases. Paradigmatic examples may be used to check candidate criteria, they do not supply them. If one seeks to establish that technological knowledge is autonomous, and if a taxonomy is to be useful for that purpose, one needs to determine what should be classified as technological knowledge, and what should not.18 As an illustration of the double-demarcation problem, consider the Carnot engine. This hypothetical artefact was first introduced by Carnot in his R´eflexions sur la Puissance Motrice du Feu (1824). Our present-day description of it is largely based on Clausius’ work in the 1860s. The engine has played a pivotal role in the development of, in particular, the concept of entropy, and it is a standard element of introductory textbooks on thermodynamics. Like any heat engine, the Carnot engine involves the conversion of heat transfer into mechanical work, in a completely reversible cycle (the Carnot cycle). Since, in reality, heat-engine cycles always create entropy and thus cannot be completely reversible, the Carnot engine is an idealization. It is, however, a useful idealization: it increases our fundamental understanding of heat-transfer processes, and it can be used to determine the maximal efficiency of thermodynamic engines. If we were to construct a taxonomy of human knowledge, both scientific and technological, it is not clear how to classify the Carnot engine. That it should be classified is beyond reasonable doubt, since Carnot’s work is generally regarded as a major intellectual achievement. Still, the engine is an idealization, putting Carnot’s work squarely in the gray area of thought experiments. Moreover, it is 18 An alternative would be to examine whether the paradigmatic examples of technological knowledge, say Pambour’s theory of the steam engine, is autonomous from scientific knowledge. This may be a viable and much-needed epistemological project, but it is much less ambitious than examining the autonomy of the entire category of technological knowledge. At most, studies into specific types of technological knowledge yield hypotheses about what might be epistemically distinctive about all technological knowledge. But to check this hypothesis, a complete inventory of technological knowledge would be needed, leading back to the (double) demarcation problem.

330

Wybo Houkes

both the cornerstone of an important scientific theory, thermodynamics, and a (fictional) artefact that provides guidelines or limitations for the design of technical artefacts. As such, it seems to have earned its place in both the body of scientific knowledge, and that of technological knowledge. Nevertheless, Kroes [1992] classifies Carnot’s theory about heat engines as scientific, contrasting it with Pambour’s “engineering” theory on the basis of his definition. One may have a principled reason to classify all our knowledge about Carnot engines as either scientific or technological, or some as scientific and some as technological — but the list itself does not make this reason explicit: one needs something like Kroes’s definition. In this sense, a taxonomy requires a solution to the double-demarcation problem rather than providing it. Moreover, it seems that a principled reason should be, or can only be, derived from an in-depth study into the use and structure of idealizations in science and technology, or natural and engineering science. If the Carnot engine would be presented in exactly the same way in textbooks for physicists and engineers, and if statements regarding the engine would have the same epistemic value in both domains, classifying this knowledge as either scientific or technological would be an arbitrary decision: nothing would be at stake. In this sense, the taxonomical strategy for emancipating technological knowledge depends on the contrastive strategy — which was earlier shown to be underdeveloped. 6 TECHNOLOGICAL KNOWLEDGE AS TACIT Using ideas and notions developed by Michael Polanyi [1958; 1966], some authors have emphasized the importance of tacit knowledge in engineering and technology.19 They have argued, or stated, that part of the knowledge produced by technological practice is hard or even impossible to make fully explicit in declarative statements, but can only be acquired through personal experience. Some make tacitness part of their characterization of technological knowledge, e.g.: “(. . . ) the knowledge of techniques, methods and designs that work in certain ways and with certain consequences, even when one cannot explain exactly why.” [Rosenberg, 1982, p. 143] Others use technological practice to characterize tacit knowledge, e.g.: “(. . . ) the implicit, wordless, pictureless knowledge essential to engineering judgement and workers’ skills.” [Vincenti, 1990, p. 198] This emphasis on tacit knowledge is not exclusive to the philosophy of technology. In fact, most work on tacit knowledge and technology is done outside of philosophy. One field where this relation is especially prominent is that of knowledge management, where the communication and sharing of knowledge is a central concern (e.g., [Nonaka and Takeuchi, 1995; Choo, 1998; Baumard, 1999; Firestone and McElroy, 2003]). Other fields where tacit knowledge is an important point of concern are the design of expert systems (e.g., [Berry, 1987]) and studies of 19 Nightingale’s contribution contains more details of the literature on tacit knowledge, and focuses on its possible importance for understanding engineering design, rather than for understanding the nature of technological knowledge.

The Nature of Technological Knowledge

331

technological innovations and technology transfer (e.g., [Nooteboom et al., 1992; Senker, 1993; Howells, 1996; Leonard & Sensiper, 1998; Nightingale, 1998; Wong and Radcliffe, 2000; Salter & Gann, 2003]). Here, the appeal to tacit knowledge is typically used to indicate the firm-specificity and person-dependence of knowledge. It is stressed, for instance, that successful implementation of new technologies requires detailed and specific knowledge about a particular situation, which is — at least at first — only available through personal experience and rules of thumb (e.g., [Arora, 1996]). Other authors point out that, although codified and explicit knowledge is available for more established technologies, effective use still relies on skilled operators and maintenance personnel, arguably showing that there is an irreducibly tacit component in technological knowledge (e.g., [Noble, 1978]). That many contributions to the literature on technological knowledge appeal to tacitness is beyond question; moreover, most do so by pointing out that it has been generally overlooked, because of an exclusive focus on codified knowledge, and that it is essential to a full account of knowledge. Thus, the existing literature seems based on the idea that there is something distinctive about tacit knowledge, and perhaps also something distinctively tacit about technological knowledge. This makes the appeal to tacit knowledge potentially interesting for the epistemic-emancipation project. Yet to see how and to what extent the appeal to tacit knowledge could improve our understanding of the nature of technological knowledge, and emancipate it from scientific knowledge, two questions need to be answered. Firstly: what is the relation between tacit and technological knowledge? Secondly: is the tacitness of technological knowledge more prominent or encompassing than that of other types of knowledge? Insofar as these questions have (implicitly) been answered in the literature, the answer to the crucial second question appears to be negative. Let us tackle them each in turn. Virtually everyone who writes on tacit knowledge, even those who do not ultimately use the term, agrees about one conceptual issue — that there is something about human knowledge that standard, justified-true-belief or propositional, accounts do not capture. Beyond this stage, however, there is considerable disunity about the appropriate concepts, concerning both the phenomenon of “tacitness” and the standard view(s) with which it supposedly contrasts. At least three distinctions are at issue in the literature. These distinctions are related, but different, and they are seldom distinguished as carefully as they should be. Firstly, there is, what might be called, the psychological distinction between implicit and explicit knowledge (e.g., [Dienes and Perner, 1999]; see also [Reber, 1993]).20 One way to phrase this distinction is as follows: when we know a fact, we have an accurate representation of it. On the basis of its functioning and its accuracy, this representation may be identified as “knowledge” (rather than a desire). If we are 20 By calling this distinction ‘psychological’, I do not mean that it is a unanimously accepted part of contemporary cognitive or developmental psychology. This distinction is, however, mainly discussed by cognitive psychologists, and concerns the functioning of representations rather, e.g., than the justification of statements.

332

Wybo Houkes

not aware of this representation, and its accuracy, the knowledge is implicit. By contrast, our knowledge concerning this representation is fully explicit. Phrased in this way, the psychological distinction is gradual, and all distinguished states involve representations. Secondly, there is a grammatical or linguistic distinction between two types of statements involving ‘knowing’: knowing that something is the case, and knowing how something can be done. This distinction is, of course, primarily associated with work by Gilbert Ryle [1949], and it is clearly language-relative. In some languages, like German and Dutch, this distinction is expressed by means of similar-sounding words (‘kennen’, as in ‘Ich kenne Der Zauberberg nicht’; and ‘k¨ onnen’, as in ‘Ich kann Schlittschuhlaufen’, respectively) rather than one word; other languages may even use completely different words. Thirdly, types of knowledge may be distinguished social-epistemically, with regard to their communicability: knowledge that can be transferred exclusively through verbal communication, and knowledge that is not or cannot be so transferred — for instance, because it can only be acquired through personal experience. I will call the first type ‘verbal’ knowledge, and the second ‘non-verbal’. When introduced, the term “tacit knowledge” is typically used in the latter sense. However, the interconnections with Ryle’s primarily grammatical distinction are particularly strong in the philosophical literature, so that the actual use of “tacit knowledge” is at least ambiguous in this respect. Even authors who do not explicitly refer to Ryle often use terminology reminiscent of his, and refer to the same stock examples, e.g., of riding a bicycle. Furthermore, connections are forged with the (folk-) psychological distinction between knowledge and skills. This is frequently equated with Ryle’s distinction, and “tacit knowledge” is taken to refer to skills and “know-how”. Much more may be said about this, but I will cut some corners in calling this distinction a red herring. The reason is that, as soon as the unicity and autonomy of technological knowledge is sought by assimilating it to skills, the epistemic-emancipation project becomes open to the objection that it is based on a category mistake. After all, if skills are contrasted with knowledge, and the difference between technology and science is based on this contrast, the sought (and perhaps found) difference cannot be epistemological: it is not a distinction between types of knowledge, but between knowledge and something else, e.g., action. Thus, the frequent appeals to tacitness, and discussions of this phenomenon with regard to technology, suffer from multiple ambiguities in the very notion of “tacit knowledge”, which affect its usefulness for the epistemic-emancipation project. One may think that, while these ambiguities are being sorted out, a preliminary epistemological distinction may be made between fully explicit, propositional, verbal knowledge on the one hand, and the overlooked “tacitness” phenomenon on the other hand. This would, however, be naive, since a major epistemological distinction is concealed beneath the conceptual distinctions.21 21 This epistemological distinction is only occasionally made in the literature; see, e.g., [Gorman, 2002].

The Nature of Technological Knowledge

333

One possible understanding of tacit knowledge is as a supplement of the traditionally analyzed body of propositional knowledge. The associations with Ryle’s sharp grammatical distinction, and with the knowledge-skill distinction facilitate this understanding. From a psychological perspective, this understanding is highly problematic: one may think up all kinds of cognitive processes that share characteristics of both types of knowledge, undermining the idea of supplementary, but autonomous bodies of knowledge. The internalization of calculation rules is, for instance, a process that shows how explicit, and highly verbalized procedures can turn into implicit routines through frequent exercise. This does not mean that the distinction is conceptually indefensible, but it does not have the immediate plausibility of Polanyi’s original appeal to tacitness. Moreover, this understanding has the disadvantage of inviting the category-mistake objection mentioned above: if tacit knowledge is this skill-based supplement to propositional knowledge, why call it “knowledge” at all? Therefore, some psychologists — and researchers in other disciplines who take psychological studies into account — prefer another understanding of the appeal to tacitness (e.g., [Wagner, 1987]). On it, our body of knowledge contains a tacit element, in all senses distinguished above: explicit knowledge must be based on implicit knowledge, which is at least conceptually prior; knowledge-that always involves knowing-how, since it involves, among other things, competence in reasoning; and verbal knowledge presupposes non-verbal knowledge, if only in the trivial sense that we cannot make fully explicit our speech patterns, including rules for appropriate utterances and other pragmatic aspects of language. Some passages in the writings of both Ryle and Polanyi suggest this understanding of tacitness — as a general aspect, component or ‘dimension’ of knowledge. And although this view requires substantial elaboration, it does not have the abovementioned disadvantages of the first understanding. Gradualism can be captured by analyzing knowledge as having a more or less prominent tacit component; and since tacitness is an integral part of all knowledge, it is an appropriate subject for epistemology. This understanding of tacit knowledge answers the two questions posed earlier. Firstly, technological knowledge may be said to have a strong relation to tacitness. Both the knowledge possessed by designers and that possessed by users, and even the more theoretical models of engineering sciences involve a tacit component. Indeed, some examples in the general literature on tacit knowledge, such as Ryle’s bicycle riding, are derived from the technological domain broadly conceived (albeit mainly from the user’s perspective); and both design and use are clearly competence-based activities, easily described in terms of knowing-how. That they also involve knowing-that, and can in various degrees be verbalized does not run counter to the appeal to tacitness in this sense.22

22 On a gradualist understanding of tacit knowledge it is problematic to make in one’s taxonomy a sharp distinction between competences and know-how on the one hand and “theoretical” knowledge on the other hand.

334

Wybo Houkes

However, on this understanding of tacitness, there is almost by definition no special relation between it and technological knowledge. If the arguments of Ryle and Polanyi are sound, they would show that all knowledge contains a tacit component. And their general arguments have been supported by a host of more specific studies in cognitive psychology (e.g., [Reber, 1993]) and the epistemology of science. Like engineers, scientists are said to rely on rules of thumb in designing experiments and interpreting data, and to require personal experience in addition to theoretical education (e.g., [Collins, 1973, 1982; Senker, 1993; Sternberg and Horvath, 1999]). The works of Donald Sch¨ on [1983; 1988] also illustrate this generality. Some of Sch¨ on’s examples are drawn from domains that might be called technological; and he often phrases his general claims by referring to “technical” on’s claims regarding the importance problems or “design” contexts.23 Yet Sch¨ of personal experience and improvisation concern professional practice in general, not engineering design in particular. In sum, the literature on tacitness in technological knowledge shows a lack of conceptual clarity. Furthermore, insofar as clarity can be obtained, appealing to tacitness does not further the end of epistemic emancipation. Instead, it increases the burden of proof resting on those who want to establish emancipation through pointing out the role of tacit knowledge in technology: rather than showing that such knowledge plays a role, they should show that it plays a distinctive role. 7 SOCIAL SCEPTICISM In this section, I will follow up my observations about ambiguities and lack of emancipatory arguments in the current literature with a general argument. This argument concerns the social-epistemic understanding of tacit knowledge, i.e., as knowledge that is not communicable by verbal means. This incommunicability may, in itself, not be a distinctive characteristic of technological knowledge; if all knowledge contains a tacit component, it is all impossible to make fully explicit by verbal means. What is more, there seems to be hardly any knowledge that cannot be made partially explicit. Even in the standard example of cycling, it is possible to state some rules concerning the use of a bicycle (e.g., “Sit on the saddle, and put your feet on the pedals”). Thus, there appears to be a spectrum, ranging from knowledge that can be almost fully expressed verbally to knowledge that is virtually inexpressible by verbal means. All knowledge claims, scientific, technological and other, are somewhere on this spectrum. One might argue that technological knowledge is, on the average, more toward the inexpressible end of this spectrum than scientific knowledge; or that it occupies an interval more to the inexpressible end. Then, tacitness would be more characteristic for technological knowledge than for scientific knowledge, even though it 23 E.g., “It is not by technical problem solving that we convert problematic situations to wellformed problems; rather, it is through naming and framing that technical problem solving becomes possible.” [Sch¨ on, 1988, p. 5; emphasis added]. Here, the context makes clear that Sch¨ on refers to problem solving in domains such as medicine and law as well as engineering.

The Nature of Technological Knowledge

335

is not a distinctive mark. To establish this, one might point to empirical research regarding technological innovations or technology transfer. In the literature, examples of which were referred to above, it is often stressed that tacitness creates transfer problems — problems that are key issues for knowledge management and the development of expert systems. If similar empirical results have been found for scientific knowledge, they have neither gained the same prominence nor set a similar agenda for new subdisciplines. Thus, one might conclude that these empirical studies show at least a gradual distinction between the two types of knowledge. This empirical conjecture might very well be falsified. But let’s accept it for the sake of the argument. For even if technological knowledge were relatively illexpressed and scientific knowledge relatively well-expressed, this does not establish epistemic emancipation. The reason is that this empirical difference might not be the result of the nature of technological and scientific knowledge, but of the social organization of science and technology. If every bit of knowledge is verbally expressible to some extent, verbally expressing it becomes not just a matter of degree, but also of practical interest. Take, again, the example of riding a bicycle. If someone is the only bicycle rider in the world, expressing one’s knowledge of how to ride a bicycle would be of no, or at most of personal, interest. However, as soon as someone wants to learn another person how to ride a bicycle, verbal expression becomes relevant. Yet how relevant it is depends on a number of factors, including the capacity of the educated person to respond to verbal instructions, the difficulty of acquiring the competence without any verbal instructions (if anyone could ride a bicycle on first trial, verbal instructions for it become as useful as breathing instructions), and the educator’s willingness to teach the competence without trying any verbal “shortcuts”. The extent to which cycling know-how is expressible enters the equation somewhere, but it is hard to say where exactly. Assume that someone lives in a society where there is a high demand for cycling instruction manuals. In these circumstances, verbal expression of cycling competence is a socially, perhaps even financially, rewarding enterprise. It would be reasonable for cyclists to invest considerable time and effort into moving their knowledge of cycling further towards the fully-expressed end of the knowledge scale; if someone would succeed in making her implicit knowledge slightly more explicit, she might acquire an edge over competitors on the market for cycling manuals. This example is fictional and rather trivial.24 It does show, however, the close connection between epistemic, social and even economic aspects of the tacit component of knowledge. This connection was to be expected, since “tacit knowledge” can be defined as a social-epistemic concept. For this concept, verbal expressibility of knowledge, its actual degree of expression, the social need for expressing it 24 To take another example: “being a successful manager” is a difficult skill (if it is even one skill) to express verbally. Yet there is a substantial market for even the most partial verbal expressions, in the form of lectures and books about management. Thus, the amount of “expression attempts” may say little about the expressibility of skills and knowledge, and much about economic viability and social need of attempts.

336

Wybo Houkes

(given a division of labour or of expertise), and economic interests in expressing it can and should be kept apart analytically. Yet it is difficult to decide which of the four factors must be invoked to decide where a knowledge claim is to be placed on the tacitness scale. This difficulty arises in full force if one wants to explain the different places that scientific and technological knowledge take on this scale (remember that above, we assumed that they take different places, or occupy different intervals). Technological knowledge is acquired in a certain social context, in which it is more or less profitable to express this knowledge verbally. To take an extreme example: if engineering were an exclusively one-person enterprise, and if practically useful items were a highly valued commodity, verbally expressing one’s knowledge of how to design these items would not be worthwhile and might even be disadvantageous. Suppose that, by contrast, there were no scarcity: all material needs were fulfilled by means of imperishable or very easily replaced artefacts, and human beings were virtually immortal. Then, the design of new artefacts could be an activity for artificers who merely want to satisfy their curiosity. These artificers might verbally describe to each other their design knowledge in excruciating detail — supposing they had nothing better to do. To put it very roughly: the current social circumstances of engineering involve considerable scarcity, a marked division of labour between professional designers and end-users, heavy commercialization, an increasing amount of teamwork in design, a decreasing loyalty of employees to companies, and heavy competition between companies that design new artefacts. On the one hand, in these circumstances, verbally expressing design know-how (an important element of technological knowledge) is advantageous to companies to some extent, since it facilitates teamwork, and improves the continuity of design work despite job-hopping employees. Hence, knowledge management is an economically interesting enterprise. On the other hand, there is a point at which further verbal expression of design knowledge becomes economically uninteresting — the costs of further expression outweigh its benefits — or even potentially disadvantageous, because another company could conceivably steal the entire body of design knowledge. Thus, the actual degree of expression (or codification) of technological knowledge may be largely due to socio-economic circumstances, not to the nature of the knowledge involved. The same argument may be given for scientific knowledge. Science shares many of the features of technology indicated above: there is scarcity of (epistemic) resources, a division of labour between researchers and laypeople, at least some commercialization, an increasing amount of teamwork in most disciplines, transfer of researchers between institutions, and competition between researchers and institutions. Yet there may also be differences. Following Merton’s (1973) identification of instutional norms in science, one could maintain that scientists should communicate their results and the way in which they achieved them. Furthermore, the market for scientific research results probably has a different structure from the market for technical artefacts, especially if (again following Merton) one thinks that scientists cannot claim ownership of knowledge. As a result, the competition

The Nature of Technological Knowledge

337

involved in science would be different from that in engineering. Consequently, the cost-benefit analysis for the verbal expression of scientific knowledge may also be different: the professional obligation to share results, combined with a possibly milder form of competition, may suffice to pull scientific knowledge towards the well-expressed end of the tacitness scale. Again, this place would then be explained by appealing to social circumstances rather than to any epistemically distinctive features of science. A similar argument has been expressed by economists who are interested in tacit knowledge (e.g., [Cowan et al., 2000; Balconi, 2002]),25 and criticized by others [Johnson et al., 2002]. In this section, I have stated this in a more general form, as a counterargument to epistemically emancipating technology from science by appealing to tacitness. This “social-skepticism” argument is vulnerable to several objections, including charges of misrepresenting and oversimplifying the sociology of both science and technology. Although it is probably guilty of those charges, the argument does not require empirical adequacy: it only purports to show that, even if science and technology might be on different ends of the sliding “tacitness” scale, this difference might be a result of the social organization of science and technology. Some sociological storytelling suffices to show this. As a consequence, this alleged difference in tacitness does not entail that science and technology are epistemically different. To counter this argument, one needs to show that the alleged prominence of tacitness in technology is not only real, but also a matter of epistemic necessity rather than a social contingency. Given the state of confusion concerning tacit knowledge and the unwillingness of many sociologists of science and technology to make a clear distinction between social and epistemic matters, such a counterargument may be a long time in coming.

8

PRESCRIPTIVE KNOWLEDGE

One existing strategy for epistemically emancipating technology from science remains to be discussed. Consider the following quote: The engineer, and more generally the designer, is concerned with how things ought to be — how they ought to be in order to attain goals, and to function. [Simon, 1981, p. 7] Science is allegedly descriptive because it is aimed at truth or empirical adequacy; by contrast, engineering is supposed to be at least partly prescriptive because it is aimed at changing reality: “(. . . ) The modal mood of a pure scientist is largely 25 The former paper includes the following telling quote: “Any individual or group makes decisions about what kind of knowledge activity to pursue and how it will be carried on. Should the output be codified or remain uncodified? Are the inputs to be made manifest or latent in the production process? For an economist, there is a simple, one-line answer: the choices will depend on the perceived costs and benefits” [Cowan et al., 2000, p. 214].

338

Wybo Houkes

descriptive, while the mood of engineering is generally prescriptive” [Hendricks et al., 2000, p. 278]. Some specifications of this difference reveal puzzlement and confusion rather than characteristics of technological knowledge. To give an example, Vladimir Hubka and W. Ernst Eder [1990] present a variety of types and forms of what they call “design knowledge” — an epistemic category that seems to overlap significantly with technological knowledge. Besides presenting a list of types and a diagram depicting connections between design knowledge and other areas, Hubka and Eder classify the types by means of two distinctions: that of product versus process, and that of descriptive versus prescriptive statements. These two distinctions are both useful and relevant.26 However, Hubka and Eder undermine the quality of their analysis by next presenting “maps” of statements and knowledge [1990, Figs. 4 and 6] in which the distinctions are represented by orthogonal continuous lines, and individual contributions to design knowledge by areas within the graph. Representing the distinction between descriptive and prescriptive statements as a sharp dichotomy might be an oversimplification, but representing it as a continuous scale without any argumentation or even examples of intermediate cases “resolves” some thorny philosophical issues with, literally, a single stroke.27 Furthermore, some authors presume that pointing out the presence of prescriptive statements in technological knowledge suffices to differentiate it from scientific knowledge. Taken literally, this is incorrect. Paradigmatic examples of scientific knowledge, such as physics, comprise prescriptive as well as descriptive statements. A widely used textbook on electrodynamics,28 for instance, contains prescriptive statements such as: “It is useful to keep track explicitly of the total fields propagating in the two directions”, “Because of the generality of the contribution from the shadow region, it is desirable to consider it separately” [Jackson, 1975, p. 372, p. 448; emphasis added]. One might object that engineering texts contain a greater proportion of prescriptive statements, or more prominent ones. Indeed, the statements above were collected from a substantial sample of a large textbook. Yet making this supposed feature of technological knowledge into a topic of textual statistics is not exactly clarifying the issue at hand. Alternatively, one might attempt to convert the prescriptive statements in the physics textbook into descriptive ones, such as: “An accurate description of the propagation depends on A(ω) as a function of complex ω” or “A general model is obtained once one considers separately the contribution from the shadow re26 This essay does not consider the product-process distinction. Yet an analysis of technological knowledge is bound to include it, given the product-process ambiguity of the central notion of “design” and of “technology” itself. 27 To do Hubka and Eder justice, it should be remarked that their [1990] paper is a brief summary of a significant body of work on design knowledge. Yet the continuous-line diagrams also appear in other work, e.g., [Hubka and Eder, 1988], without lengthier arguments for choosing this particular representation. 28 A handbook on a highly theoretical part of physics was chosen to prevent the objection that all sample prescriptive statements are engineering intrusions in science, related to the design of experiments or the interpretation of their results.

The Nature of Technological Knowledge

339

gion”. This conversion does not show that these statements given above were pseudo-prescriptive, i.e., that they can be reduced to descriptive statements. For to capture their meaning fully, explicitly prescriptive statements concerning accuracy and generality should be added: knowledge of A(ω) as a function of complex ω is only required if accuracy is a guiding value, and separate consideration of the shadow region is only desirable if generality is a desideratum. Hence, closer analysis of the prescriptive statements from physics shows that they presuppose scientific values such as empirical adequacy and generality. Therefore, analysis of prescriptive statements from physics and engineering may well return us to familiar grounds, namely the TU-intuition that science is directed towards truth and technology towards usefulness (see Section 2). Because this intuition is primarily one of values, it is only to be expected that handbooks from both physics and engineering contain prescriptive statements, but that these are related to the different central values of the disciplines. To go beyond restating the TU-intuition, one should do more than note prescriptive statements in technology, or their relation to the goal of usefulness or changing reality. One way to do this is to analyze the fact that technological knowledge is not about just any change in reality, including the diffusion of gases or the construction of theories, but about deliberate changes that serve practical purposes. This analysis starts from the seemingly trivial observation that technology is related to human, intentional actions. Most technical artefacts and processes do not occur naturally, but need to be designed and manufactured. Few artefacts realise their functions automatically, but require active manipulation by a user. And even artefacts that function more or less automatically, such as fire alarms or assembly-line robots, require monitoring and maintenance. Because technology is intimately action-related, it makes sense to assume that technological knowledge is related to designing, using and other actions as well. Moreover, since the goal of technology is to make useful changes in reality, these actions cannot just be described, but they must also be prescribed. To employ a car or an assembly-line robot, a user has to know not only for which purposes the artefact may reliably be used, but also which actions he or she should take, might profitably take when certain situations arise, or how to recognize undesirable behaviour of the artefact. In short: the practical aim of technology implies that technological knowledge prescribes and recommends intentional actions. It does not just describe what is the case, or what is desirable, but also what human beings should do to bring this desirable state of affairs about. This forges an intuitive distinction between technological knowledge and knowledge gained in the behavioural and social sciences, which seem primarily descriptive. Given this starting point, a closer analysis of technological knowledge may employ (and require) action-theoretical resources rather than notions and perspectives borrowed from traditional epistemology or philosophy of science. Hence, one may look for teleological notions such as “goal” and “function”;29 one may 29 See

Preston’s contribution to this Handbook.

340

Wybo Houkes

study the role of practical reasoning;30 and one may investigate the status and justification of rules and recommendations in technological knowledge. This shift of perspective is non-trivial, especially given the goal of epistemic emancipation. Studying the distinction between scientific and technological knowledge by using notions and perspectives developed for understanding the former leads to questions like those posed at the end of section 2. As said there, few enough attempts have been made to answer these questions. Moreover, emancipation may require a different perspective altogether: action-theoretical terminology might be more appropriate for an understanding of technological knowledge than it is for an understanding of scientific knowledge. One attempt to develop such an understanding is made by Mario Bunge. According to Bunge, one characteristic product of the engineering sciences is a technological rule, “an instruction to perform a finite number of acts in a given order and with a given aim” [1967, p. 132]. An example would be “If you are interested in comfortable private transportation, drive a car”, where driving a car is a specific series of actions: getting in the driver’s seat, starting the car, etc. Similar rules may be specified for other goals and action types, including design and maintenance. As it stands, this way of characterising the prescriptive content of technological knowledge is rather broad and non-specific. The description given in the quote above applies to all practical rules, including: “When you are caught in a thunderstorm, avoid trees and large bodies of water and roll up in a ball”. Taking technological knowledge as a part or a continuation of such common-sense practical knowledge may be correct as a first approximation, like taking science to be the continuation of common sense, but much work remains to be done to go beyond this first approximation. indent Bunge does that by explaining how technological rules are, in the engineering sciences, grounded in scientific knowledge and elaborately tested, leading to a tremendous growth in reliable and productive rules after the Industrial Revolution. In this way, technological knowledge may indeed be distinguished from run-of-themill practical rules,31 but Bunge’s choice has a high price: the “grounding” claim regarding technological rules has made Bunge’s work a standard target in the applied-science debate. Moreover, it seems to have made people so suspicious of the notion of technological rule that critical analyses have been screened-off by criticisms of Bunge’s supposed applied-science thesis. Yet it may be possible to employ the notion of technological rule without accepting Bunge’s claims regarding grounding. One possibility is to consider the role of artefacts in such rules. Many practical rules, like the one concerning thunderstorms, involve only our own body; others, like “Do not drink salt water, even if 30 See

Hughes’ contribution to this Handbook. if being grounded in scientific knowledge is to serve as a distinguishing characteristic, it should subsequently be clarified how instructions for driving a car are so grounded, whereas instructions for avoiding death by lightning are not. It is not clear whether even a gradual distinction may be gained in this way. 31 Still,

The Nature of Technological Knowledge

341

you are very thirsty”, involve our bodies and natural objects. Whether these rules are grounded in scientific knowledge or not, they seem to involve techniques, rather than technology. By introducing artefacts, Bunge’s characterization of technological rules may be amended into “instructions to perform a finite set of actions, including manipulations of one or more artefacts, in a given order and with a given aim”. This idea of technological rules has been developed, in different terms from Bunge’s, by Pieter E. Vermaas and myself. The central notion in this line of research is that of use plans, “goal-directed series of considered actions, including manipulations of one or more artefacts” [Houkes and Vermaas, 2004]. Both use — the execution of use plans — and design — the construction and communication of use plans — can be analysed in terms of this notion [Houkes et al., 2002]. The resulting perspective on technology and technological knowledge emphasizes goal-directed, intentional actions and the standards of (instrumental) rationality for these actions rather than the objects employed in such actions. It considers descriptive knowledge only insofar as it plays a role in intentional actions. Consequently, it provides action-theoretical resources for analysing the prescriptive content of technological knowledge. The use-plan account provides a picture of prescriptive technological knowledge that is richer than the notion of technological rules alone. Knowledge regarding use plans need not consist only of instructions: they might carry both stronger and weaker normativity. Artefacts may be used in many different ways, not all of which may or can have been envisaged by their designers. The use-plan account incorporates this by a liberal notion of design. Everyone, engineers and consumers alike, can design in the sense of constructing and communicating use plans. One need not have a degree in engineering to use an empty milk bottle as a vase — use that is as effective and efficient as it is widespread. Knowledge regarding this use may be regarded as technological, in the minimal sense that it concerns use of an artefact for a practical purpose. The corresponding knowledge, that milk bottles can be used for holding flowers, is normative [Houkes, 2006], but involves a recommendation in some circumstances rather than an instruction.32 Other knowledge regarding artefact functionalities involves requirements, which are considerably stronger than instructions. To give an example: some use is regarded as (im)proper, meaning that it is privileged over other ways of using an artefact. Such privileges may be analysed by referring to the fact that, although many agents are capable of designing, only some of them are professionally engaged in it. Their use plans are standardized and often even embedded in legal systems: many warranties, for instance, are declared void in cases of improper use. Thus, the use of artefacts is embedded in a (largely un-analyzed) system of rules, recommendations and requirements that is far richer than mere sets of instructions for attaining a goal. 32 See Franssen’s contribution to this Handbook on artefacts and normative judgements for a more detailed analysis.

342

Wybo Houkes

Furthermore, the use-plan account may be employed to study the relation between prescriptive and descriptive statements regarding artefacts. That there is such a relation seems beyond doubt: prescriptive statements that are not, somehow, related to accurate, propositional knowledge are at best very risky recommendations. Moreover, professional designers, engineers in particular, often possess knowledge about the physicochemical composition of artefacts, and design artefacts on the basis of this knowledge. On the use-plan analysis, one way in which prescriptive and descriptive statements regarding artefacts are related is by means of a specific type of explanation for the function of an artefact [Houkes, 2006; De Ridder, 2006]. In such ‘technological’ explanations, descriptions of the structure of an artefact are related to descriptions of the actions included in the use plan of the artefact, to show that these actions can be rationally expected to lead to realization of the goal state. That there are these explanations does not mean that prescriptive statements should be grounded in scientific knowledge, let alone that they are little more than “applications” of this knowledge. Some communicated use plans are, for instance, based on successful tests in a variety of circumstances, on trial-and-error, or simply handed down through generations of users [Vermaas and Houkes, 2006]. To conclude, the distinction between descriptive and prescriptive statements is in itself insufficiently specific for epistemic emancipation of technology. However, a closer analysis of some prescriptive statements made within a technological context — technological rules or recommendations and requirements regarding artefact use — might reveal a connection to intentional actions and practical (instrumental) rationality specific to technology. This analysis of prescriptive statements is still rudimentary, and it warrants further attention, even independently from the quest for epistemic emancipation. 9 OUTDATED EMANCIPATIONISM The review of the existing literature in the previous sections shows that there are several ways in which authors have tried to establish epistemic emancipation. Few ways are developed beyond the embryonic stage, none have given rise to elaborate discussions and refinement of points of view and arguments. What is perhaps most important, all have so far failed to establish strong emancipation. For some attempts, general arguments can be offered that appear to show that they are bound to fail; for others, analysis shows that specific issues need to be addressed — more specific issues than those covered by existing efforts. The results of the review are, in short, not encouraging. Establishing epistemic emancipation appears to require a concentrated, collective effort, aimed in part at overcoming some general counterarguments. It might, therefore, be understandable that historians and philosophers of technology have shifted their attention towards other topics: substantial effort would be needed to get the topic of the nature of technological knowledge off the ground, and the benefits might be so small that research time is more efficiently spent otherwise.

The Nature of Technological Knowledge

343

Additional discouragement is given by an increasingly powerful movement in the literature on the history, philosophy and sociology of science and technology. The 1970s and 1980s not only saw a decline in the interest for technological knowledge, and the conceptual and epistemic distinctions between science and technology. During these decades, an alternative perspective and research agenda was promoted in the newly developed field of Science and Technology Studies (STS). This chapter is not about the history, central characteristics and many divergent results and approaches within this field. Yet it is beyond doubt that traditional epistemic issues are not high on the agenda of research in STS,33 and that many conceptual distinctions are typically regarded as outdated, or topics for deconstruction. Of particular interest here is that the distinction between science and technology has been subjected to criticism and revision, on the basis of both empirical, sociological research and more conceptual and methodological concerns. Many authors, including Don Ihde [1979; 1991], Bruno Latour [1987; 1993] and Andrew Pickering [1995], have pointed out or argued that scientific knowledge is not just historically and socially situated, but that it is acquired, distributed and defended in an increasingly intricate technological context. Scientists use technology to perform experiments, to manipulate and store data, to write research papers, and to communicate with other scientists. Many of these technological aspects of science are not merely contingent characteristics, but appear to be essential for science as it is conducted nowadays.34 Since the 17th and 18th century, science has been experimental and mathematical — but since the 1950s experimentation and mathematization increasingly depend on technologies such as lasers, computers, and satellites. For the authors mentioned above, and many other STS researchers, the role of technologies in scientific research is so prominent and inalienable that they prefer to speak of “technoscience” rather than “science”.35 Suppose that the main idea behind this neologism is correct, and that scientific knowledge can indeed not be studied in isolation from its technological context, because it is necessarily embedded in it. Then it may still be possible to emancipate technological knowledge with respect to scientific knowledge. After all, technology is not equated with science. There may be reason, also from a sociological point of view, to assume that technological knowledge is acquired, distributed and defended independently from scientific research.36 There may be institutional 33 Here, “traditional epistemic issues” means the issues regarding (among other things) truth, justification and epistemic virtues that characterize epistemology as studied in the AngloAmerican analytical tradition, and as reviewed in introductory books such as Audi [2002] and the essays in [Greco and Sosa, 1998]. Parts of the STS research agenda are and can be labeled as “epistemology” as well; take, e.g., many of the papers published in a journal such as Social Epistemology. 34 One clear expression of this sentiment is: “Modern Science, in contrast to its ancient and more contemplative origins, [is] essentially and necessary embodied in technologies, instruments.” [Ihde, 1993, p. 74; original emphasis). 35 Here, “technoscience” indicates specifically a system in which scientific research cannot be studied in isolation from its technological context. The notion is used in a broader variety of senses in the literature. 36 Many technoscience scholars claim that there is a reciprocal dependence relation between

344

Wybo Houkes

cross-connections, but if anything, these show that technology deserves closer epistemological attention, because it is so important for scientific knowledge. Yet epistemic emancipation of technology from science no longer makes much sense: the model of science as epistemically autonomous from technology needs to go, along with any epistemology based on this model. Therefore, there is no standard epistemology left with which to contrast an analysis of technological knowledge. To put it crassly: why argue that technological knowledge is autonomous from science, when scientific knowledge is thoroughly technological? The same point may be made by looking at the thesis that technology is applied science, which shaped so much of the literature on technological knowledge. If research on technoscience is correct, the thesis is at least unilluminating. The thesis conceals that (techno-)scientific research consists of the application of technologies, and may be shaped to a large extent by promises and expectations of future technological rewards. Technology may, in turn, be based in part of applying scientific theories, but this feature cannot be used as its most basic characteristic. The typical argument offered against the applied-science thesis is that some technologies have been developed without the aid of scientific theories. From the technoscience perspective, the argument is correct, but it ignores the deeper insight that scientific research is thoroughly technological — and it might in this way reinforce the mistaken epistemology that regards science as autonomous. These observations offer plenty of reasons to revise our understanding of the relation between science and technology, but no reasons to ignore the study of technological knowledge. On the contrary, they make analyzing the acquisition, distribution and defence of this kind of knowledge far more important than it would be on either the applied-science image or the autonomy image. After all, understanding the epistemology of contemporary technology, together with that of science, would be crucial to understanding technoscience. However, the quest for emancipation, which shapes virtually all work on technological knowledge, should be abandoned: it involves a false assumption about present-day scientific and technological research, and is therefore outdated. This line of thought offers a sociological or “empirical” counterpart to the more analytical counterarguments and problems presented in earlier sections. Together, I think they give ample reasons to abandon the quest for epistemic emancipation: whatever technological knowledge is, and from whatever perspective one wants to study it, one should not try to understand it as an epistemic category that is different from that of scientific knowledge. Before I tentatively suggest an alternative research agenda in the concluding section, let me address a question that might be raised by the previous reflections on technoscience. The question is: why did technoscience scholars not start to study technological knowledge afresh, given its importance for understanding the very phenomenon that they describe? They may have reasons to abandon contemporary science and technology, so that modern technology cannot be studied in isolation from scientific research. This claim may be true, but it is stronger than the earlier claim about science alone.

The Nature of Technological Knowledge

345

the applied-science thesis, the ideal of epistemic emancipation, and perhaps traditional epistemology. Yet they have not replaced these views with an empirically informed, up-to-date epistemology of technology. Indeed, the very notion of “technological knowledge” is sorely lacking in the technoscience literature. Given the previous reflections, this is an oversight. One might speculate about the causes of this oversight. Perhaps the analysis of technological knowledge was so firmly associated with the misguided applied-science debate and the isolationist model of science that, in promoting a different perspective, STS researchers unwittingly threw away the baby with the bathwater. Surely, the abandonment of other traditional philosophical views, such as the fact-value distinction, makes it difficult or outright impossible to develop some of the more promising routes considered in this chapter. Understanding technological knowledge as prescriptive would, for instance, become as misguided as the applied-science thesis. Whatever the causes may be, I do not think that they offer sufficient reasons: technoscience scholars ought to analyze both technological and scientific knowledge, and their mutual dependence, just like philosophers of science and philosophers of technology should. It is time to give some indication how this analysis is still possible and useful in spite of the criticisms levelled at previous attempts. 10

AFFIRMATIVE ACTION AND FUTURE PROSPECTS

To conclude, I review the possibilities for making a fresh start in studying the nature of technological knowledge. I firmly believe that these possibilities exist. Besides criticisms, the preceding sections already contained some suggestions for future research. This section lists them again, by way of recapitulation, and adds several more general topics of research as well. Work on one or more of these topics might achieve weak epistemic emancipation of technology. The problems raised in the previous sections may not amount to fatal counterarguments and, in any case, mainly raise obstacles for strong emancipation. This leaves room for establishing a weaker claim. Even if one feels justified in abandoning the emancipation project altogether, there is still sufficient reason to develop the topics below. Not putting emancipation as first — or even only — item on the research agenda, but showing that epistemically interesting results may be gained by studying technology would constitute affirmative action. It would show philosophers that technology has been unwisely ignored, not because it is fundamentally different from science, but because good philosophical work can be done on it. Topics for further research proposed in previous sections include: • The role of practical usefulness (rather than truthlikeness) in validating theories and models developed in the engineering sciences (Section 2). • The role of practical usefulness in explaining the way in which theories and models from the natural sciences are adapted for use in the engineering sciences (Section 2).

346

Wybo Houkes

• Possible distinctions between scientific concepts and technical concepts in engineering theories (Section 2) • The different role of idealizations and hypothetical objects such as the Carnot engine in the natural and engineering sciences (Section 2; Section 5). • The relation between features of technological knowledge (e.g., tacitness) and the social organization of engineering (Section 7). • The distinction between technological rules and everyday techniques (Section 8). • The inherently prescriptive or, more broadly, normative content of technological knowledge (Section 8). In addition, the following two topics may be explored: Technology and the nature of knowledge The epistemology of technology has mainly been studied by considering technology as knowledge. Yet this does not exhaust the possible relations between technology and knowledge. As the sketchy review of research in science and technology studies in section 9 shows, technology is also related to knowledge, scientific and other, by providing much of the context in which knowledge is acquired, distributed and defended. These roles may be regarded as belonging to the context of discovery, and therefore rejected as a proper subject of epistemological studies. Yet they also require evaluation. Some of this evaluative work is done in what has been called the “philosophy of scientific experimentation” (e.g., [Radder, 2003]), in which the epistemic role of experiments and the technological devices used therein are studied. An even less explored topic is the extent to which new technologies allow researchers to acquire and support knowledge in hitherto unprecedented ways. The sciences nowadays do not only rely on technologically ever more complicated experiments. Scientific observation is not just theory-laden, but has become, perhaps irreversibly, technology-laden as well. Software is used to gather, manipulate and graphically represent data, and both natural and social scientists are trained in using computers to solve mathematical problems. Some of these problems might have been solved, with considerable effort, by some unaided human brains; in other cases, computers apply approximation techniques on a scale that is at least practically impossible to achieve for human beings; and in a growing number of cases, computers solve problems that have proved humanly intractable. This epistemic use of technology resembles its use for, e.g., transportation: in some cases, it is merely convenient, like driving to the supermarket instead of walking; in others, it is clearly more effective, like crossing the Channel — which some gifted individuals can do swimming, but most of us cannot; in still others, like flying to the moon, it is indispensable.

The Nature of Technological Knowledge

347

The use of technology in scientific research might lead to particular methodological questions, e.g., concerning automated proof systems. It also leads to epistemological issues. One such issue concerns possible new, technologically generated sources of justification. A prominent example is the use of simulations to support scientific hypotheses, e.g., about climate change. Intuitively, computeraided simulations do not provide observational data, like radio telescopes; nor do they involve approximation techniques that could, in principle or to a limited extent, be applied by human researchers. Instead, simulations are partly based on the theories and models used to construct them; but they also offer new insight into, and possibly evidence about, complex phenomena like turbulence [Winsberg, 2001; 2003]. Yet since simulation techniques occupy some middle ground between theory and observation, whatever evidence they offer is of an ill-understood type. More attention to the mathematics and technology of simulations is needed to clarify their epistemological status. The present neglect of simulation techniques in the philosophy of science jars with their increasing importance in all kinds of sciences. The grounding of technological knowledge In the philosophy of technology, studies of the relation between science and technology have been dominated by the applied-science debate. One unfortunate consequence of this domination, noted above, is that the notion of “technological rule” has become firmly associated with the thesis that engineers merely apply scientific knowledge. This has screened off this notion from further development. It has also precluded the development of alternative models of the relation between science and technology — models that might incorporate the fact that engineers frequently do apply scientific knowledge, or are at least trained in understanding and applying theories like thermodynamics and classical mechanics. Outside of philosophy, however, interest in such models continues. To give one example, in The Gifts of Athena [2002], Joel Mokyr seeks to explain the sustained economic growth since the Industrial Revolution — a project that is squarely outside philosophy. However, the basis of his explanation is that science and engineering have since the early 19th century undergone an unprecedented period of mutually re-inforcing progress. To develop this explanation, Mokyr uses both evolutionary terminology, which need not concern us here, and an epistemological model. In this model, he distinguishes two types of knowledge, in a way that is reminiscent of both Ryle and Bunge: knowledge can be propositional or prescriptive, where the former involves any proposition, and the latter both rules and skills. Moreover, prescriptive knowledge can be grounded in propositional knowledge, either minimally (we support adding some old leavened dough to fresh dough because we know that this procedure has successfully produced leavened bread in the past) or more elaborately (we know that there are starter cultures of yeast in the old dough, which cause fermentation). Mokyr’s hypothesis is that the Industrial Revolution came about when more prescriptive knowledge was grounded in

348

Wybo Houkes

an elaborate way, leading to new techniques (e.g., the isolation of pure cultures of yeast), which in turn led to new scientific developments, which allowed additional grounding of techniques, etc. Let us assume that this model fulfills a real need in its field. Then, a closer philosophical analysis of technological knowledge contributes directly to non-philosophical aims. For Mokyr’s model does not analyze the grounding relation in detail. Moreover, it suffers from overly tolerant definitions of both propositional and prescriptive knowledge, making the problem of their relation almost into an artefact of the classification.37 Because of these unclarities and idiosyncracies, the model is vulnerable to the objection that it merely revives the applied-science thesis, and that it is based on epistemologically problematic distinctions, such as Ryle’s. This should not lead epistemologists and philosophers of science and technology to ignore models such as Mokyr’s. Instead, I think these models show that there is a need to develop a realistic, epistemologically refined analysis of the grounding relation — and of the ways in which technological knowledge and rules may not be grounded in scientific knowledge. This analysis should not be grafted on either the applied-science thesis or the epistemic-emancipation ideal. The previous section may have shown that these influences are hard to avoid. Yet this does not make the analysis any less needed: it only makes it more of a philosophical challenge. ACKNOWLEDGMENTS Research by Wybo Houkes was supported by the Netherlands Organization for Scientific Research (NWO). BIBLIOGRAPHY [Arora, 1996] A. Arora. Contracting for tacit knowledge. Journal of Development Economics, 50, 233—256, 1996. [Audi, 2002] R. Audi. Epistemology: A Contemporary Introduction. 2nd ed., Routledge, 2002. [Balconi, 2002] M. Balconi. Tacitness, codification of technological knowledge and the organization of industry. Research Policy, 31, 357—379, 2002. [Baumard, 1999] P. Baumard. Tacit Knowledge in Organizations. Sage, 1999. [Berry, 1987] D. C. Berry. The problem of implicit knowledge. Expert Systems, 4, 144—151, 1987. [Broens and de Vries, 2003] R. C. J. A. M. Broens and M. J. de Vries. Classifying technological knowledge for presentation to mechanical engineering designers. Design Studies, 24: 457—471, 2003. [Bunge, 1966] M. Bunge. Technology as applied science. Technology and Culture, 7, 329—347, 1966. [Bunge, 1967] M. Bunge. Scientific Research II: The Search for Truth. Springer, 1967. [Choo, 1998] C. W. Choo. The Knowing Organization. Oxford University Press, 1998. [Collins, 1973] H. M. Collins. The TEA set: Tacit knowledge and scientific networks. Science Studies 4: 165—186, 1973. [Collins, 1982] H. M. Collins. Tacit knowledge and scientific networks. In: Science in Context: Readings in the Sociology of Science, B. Barnes and D. Edge, eds., MIT, 1982. 37 Mokyr admits that his model might not stand up to critical analysis, which only makes the need for such an analysis more apparent.

The Nature of Technological Knowledge

349

[Constant, 1999] E. W. Constant. Reliable knowledge and unreliable stuff. Technology and Culture, 40, 324–357, 1999. [Cowan et al., 2000] R. Cowan, P. A. David, and D. Foray. The explicit economics of knowledge codification and tacitness. Industrial and Corporate Change, 9, 211—253, 2000. [Cummins, 1975] R. R. Cummins. Functional analysis. Journal of Philosophy, 72, 741—765, 1975. [Diens and Perner, 1999] Z. Dienes and J. Perner. A theory of implicit and explicit knowledge. Behavioral and Brain Sciences, 22, 735—755, 1999. [Faulkner, 1994] W. Faulkner. Conceptualizing knowledge used in innovation. Science, Technology and Human Values, 19, 425—458, 1994. [Firestone and McElroy, 2003] J. M. Firestone and M. W. McElroy. Key Issues in the New Knowledge Management. Butterworth-Heinemann, 2003. [Florman, 1992] S. Florman. Review of: Walter Vincenti, What Engineers Know and How They Know It. Technology & Culture, 33, 140—142, 1992. [Gorman, 2002] M. E. Gorman. Types of knowledge and their roles in technology transfer. Journal of Technology Transfer, 27, 219—231, 2002. [Greco and Sosa, 1998] J. Greco and E. Sosa, eds. The Blackwell Guide to Epistemology. Blackwell, 1998. [Hacking, 1983] I. Hacking. Representing and Intervening. Cambridge University Press, 1983. [Hendricks et al., 2000] V. F. Hendricks, A. Jakobsen, and S. A. Pedersen. Identification of matrices in science and engineering. Journal for General Philosophy of Science, 31, 277— 305, 2000. [Hindle, 1966] B. Hindle. Technology in Early America. University of North Carolina Press, 1966. [Houkes, 2006] W. Houkes. Knowledge of artefact functions. Studies in the History and Philosophy of Science, 39, 102—113, 2006. [Houkes and Vermaas, 2004] W. Houkes and P. E. Vermaas. Actions versus functions. The Monist, 87, 52—71, 2004. [Houkes et al., 2002] W. Houkes, P. E. Vermaas, K. Dorst, and M. J. de Vries. Design and use as plans. Design Studies, 23, 303—320, 2002. [Howells, 1996] J. Howells. Tacit knowledge, innovation and technology transfer. Technology Analysis & Strategic Management, 8, 91—106, 1996. [Hubka and Eder, 1988] V. Hubka and W. E. Eder. Theory of Technical Systems. Springer, 1988. [Hubka and Eder, 1990] V. Hubka and W. E. Eder. Design knowledge. Journal of Engineering Design, 1, 97—108, 1990. [Ihde, 1979] D. Ihde. Technics and Praxis. Reidel, 1979. [Ihde, 1991] D. Ihde. Instrumental Realism. Indiana University Press, 1991. [Ihde, 1993] D. Ihde. Philosophy of Technology. Paragon, 1993. [Jackson, 1975] J. D. Jackson. Classical Electrodynamics. 2nd ed., John Wiley, 1975. [Jarvie, 1972] I. C. Jarvie. Technology and the Structure of Knowledge. In: Philosophy and Technology. Readings in the Philosophical Problems of Technology, C. Mitcham and R.C. Mackey, eds., pp. 54–61, The Free Press, 1972. [Johnson et al., 2002] B. Johnson, E. Lorenz, and B. A. Lundvall. Why all this fuss about codified and tacit knowledge? Industrial and Corporate Change, 11, 245—262, 2002. [Kroes, 1992] P. A. Kroes. On the role of design in engineering theories. In: Technological Development and Science in the Industrial Age, P. A. Kroes and M. Bakker, eds., pp. 69–98, Kluwer, 1992. [Ladyman, 2007] J. Ladyman. Ontological, epistemological, and methodological positions. In: General Philosophy of Science: Focal Issues. Handbook of the Philosophy of Science, vol.1, T.A.F. Kuipers, ed., pp. 303–376. North Holland, 2007. [Latour, 1987] B. Latour. Science in Action. Princeton University Press, 1987. [Latour, 1993] B. Latour. We Have Never Been Modern. Harvard University Press, 1993. [Laymon, 1989] R. Laymon. Applying idealized scientific theories to engineering. Synthese, 82, 353—371, 1989. [Layton, 1974] E. Layton. Technology as knowledge. Technology and Culture, 15, 31—41, 1974. [Leonard and Sensiper, 1998] D. Leonard and S. Sensiper. The role of tacit knowledge in group innovation. California Management Review, 40, 112—132, 1998.

350

Wybo Houkes

[Merton, 1973] R. Merton. The normative structure of science. In: The Sociology of Science, pp. 267—278. University of Chicago Press, 1973. [Mitcham, 1978] C. Mitcham. Types of technology. Research in Philosophy and Technology, 1, 229—294, 1978. [Mokyr, 2002] J. Mokyr. The Gifts of Athena. Princeton University Press, 2002. [Nightingale, 1998] P. Nightingale. A Cognitive model of innovation. Research Policy, 27, 689— 709, 1998. [Noble, 1978] D. F. Noble. Social choice in machine design. Policy & Society 8, 313—347, 1978. [Monaka and Takeuchi, 1995] I. Nonaka and H. Takeuchi. The Knowledge-Creating Company. Oxford University Press, 1995. [Nooteboom et al., 1992] B. Nooteboom, C. Coehoorn, and A. van der Zwaan. The purpose and effectiveness of technology transfer to small business by government-sponsored innovation centres. Technology Analysis & Strategic Management, 4, 149—166, 1992. [Pickering, 1995] A. Pickering. The Mangle of Practice. University of Chicago Press, 1995. [Polanyi, 1958] M. Polanyi. Personal Knowledge. University of Chicago Press, 1958. [Polanyi, 1966] M. Polanyi. The Tacit Dimension. Routledge & Kegan Paul, 1966. [Radder, 2003] H. Radder, ed. The Philosophy of Scientific Experimentation. Pittsburgh University Press, 2003. [Reber, 1993] A. S. Reber. Implicit Learning and Tacit Knowledge. Oxford University Press, 1993. [de Ridder, 2006] J. de Ridder. Mechanistic artefact explanation. Studies in the History and Philosophy of Science, 37, 81—96, 2006. [Ropohl, 1997] G. Ropohl. Knowledge types in technology. International Journal of Technology and Design Education, 7, 65—72, 1997. [Rosenberg, 1982] N. Rosenberg. How Exogenous is Science? In: Inside the Black Box: Technology and Economics. Cambridge University Press, 1982. [Ryle, 1949] G. Ryle. The Concept of Mind. Hutchinson, 1949. [Salter, 2003] A. Salter and D. Gann. Sources of ideas for innovation in engineering design. Research Policy, 32, 1309—1324, 2003. [Sch¨ on, 1983] D. A. Sch¨ on. The Reflective Practitioner. Basic Books, 1983. [Sch¨ on, 1988] D. A. Sch¨ on. Educating the Reflective Practitioner. Jossey-Bass, 1988. [Senker, 1993] J. Senker. The contribution of tacit knowledge to innovation. AI & Society, 7, 208—224, 1993. [Simon, 1981] H. A. Simon. The Sciences of the Artificial, 2nd ed. MIT, 1981. [Solimowski, 1972] H. Skolimowski. The Structure of Thinking in Technology. In: Philosophy and Technology: Readings in the Philosophical Problems of Technology, C. Mitcham and R.C. Mackey, eds., pp. 42–49. The Free Press, 1972 [Smith, 1960] C. S. Smith. A History of Metallurgy. University of Chicago Press, 1960. [Stanford, 2005] P. K. Stanford. Instrumentalism. In: The Philosophy of Science, S. Sarkar and J. Pfeifer, eds. Routledge, 2005. [Staudenmaier, 1985] J. M. Staudenmaier. Technology’s Storytellers. MIT, 1985. [Sternberg and Horvath, 1999] R. J. Sternberg and J. A. Horvath, eds. Tacit Knowledge in Professional Practice. Lawrence Erlbaum, 1999. [Vermaas and Houkes, 2006] P. E. Vermaas and W. Houkes. Technical Functions. Studies in the History and Philosophy of Science, 37, 5–18, 2006. [Vincenti, 1990] W. G. Vincenti. What Engineers Know and How They Know It. Johns Hopkins University Press, 1990. [de Vries, 2003] M. J. de Vries. The nature of technological knowledge. Techn´ e, 6, 1–21, 2003. [Wagner, 1987] R. K. Wagner. Tacit knowledge in everyday intelligent behavior. Journal of Personality and Social Psychology, 52, 1236–1247, 1987. [Winsberg, 2001] E. Winsberg. Simulations, models, and theories. Philosophy of Science 68, S442–S454, 2001. [Winsberg, 2003] E. Winsberg. Simulated Experiments. Philosophy of Science, 70, 105–125, 2003. [Wise, 1985] G. Wise. Science and technology. Osiris 2nd series, 1, 229—246, 1985. [Wong and Radcliffe, 2000] W. L. P. Wong and D. F. Radcliffe. The tacit nature of design knowledge. Technology Analysis & Strategic Management, 12, 493—512, 2000.

TACIT KNOWLEDGE AND ENGINEERING DESIGN Paul Nightingale

1

INTRODUCTION

In 1752 a French official lamented the slow diffusion of technology from England during the first industrial revolution by noting that “the arts never pass by writing from one country to another, eye and practice alone can train men in these activities” [Harris, 1988, p. 42]. In doing so, he was emphasising the importance of un-codified, person-embodied tacit knowledge to the engineering arts. This short chapter briefly reviews the role of this tacit knowledge in engineering design, highlighting a series of issues of importance to the philosophy of technology. The chapter aims to show how tacit knowledge as a concept is used: firstly, as an empirical description of knowledge that is impossible or difficult to articulate and codify; secondly, to explain phenomena not accounted for in other ways of thinking about engineering design; and, lastly, as a way of thinking about engineering design that is linked to broader and potentially more interesting concepts within the philosophy of technology. Understanding what tacit knowledge is, and particularly how the concept is used, is important for philosophers of technology because it is now a central concept in policy discussions related to engineering. It is used to explain why knowledge production is localised, cumulative and path-dependent, and therefore why designers, design teams, firms and regions differ in their technological performance. Given the impact of public policy related to the ‘knowledge economy’ there is a legitimate role for philosophers of technology to investigate the foundations of these ideas in more detail. This is particularly important because the terminology of tacit knowledge is applied very widely, but is rarely explicitly explained [Tsoukas, 2003]. Just what tacit knowledge is, and how it is valuable during the development of technology, is often itself a ‘tacit’ concept. This is unfortunate, because, as this chapter will argue, while tacit knowledge is a useful empirical descriptor, it is probably too broadly defined to carry the theoretical weight thrust upon it. All the same, the concept usefully points to interesting problems with the dominant conception of technology within modern culture. The remaining part of this introduction defines engineering, while section two explores what tacit knowledge is and how it is used to explain technological change in the social sciences. Section three proposes an alternative way of thinking about Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

352

Paul Nightingale

tacit knowledge that is argued to be more in tune with its philosophical origins. This is used to explore the process of design. Finally the conclusion points to some of the strengths and weaknesses of the concept of tacit knowledge. Within this chapter engineering is defined as the art of organising and negotiating the design, production, operation and decommissioning of artefacts, devices, systems and processes that fulfil useful functions by transforming the world to solve recognised problems. This hopefully highlights the practical, creative nature of engineering, with a clear connection to judgements and choices about solutions that achieve a balance between potentially conflicting outcomes in terms of their aesthetic, economic, environmental, technical and other criteria [Tang and Leifer 1988; Sch¨ on, 1982; Bucciarelli, 1994]. For an elaborate account of how to define technology and the engineering sciences, see Mitcham and Schatzberg’s chapter in Part I of this Volume. The emphasis on organisation differentiates engineering from other tasks in the production of artefacts [Vincenti, 1990, p.7]. During this production process designing is only one among many roles played by engineers. While design may be one of the most glamorous of engineers’ roles, and an emphasis on creativity helps legitimise engineers as professionals, their other tasks remain important even if they are not addressed in this chapter. This definition is similar to, but slightly more comprehensive than, Dym’s in which “Engineering is a systematic, intelligent process in which designers generate, evaluate, and specify concepts for devices, systems or processes whose form and function achieve clients’ objectives or users’ needs while satisfying a specified set of constraints” [1993, p.17]. It is also similar to G. F. C. Rogers’ definition of engineering as “the practice of organising the design and construction of any artefact which transforms the physical world around it to meet some recognised need” (quoted in [Vincenti, 1990, p.5]). Within all three definitions is a shared focus on a temporal process of creating solutions to problems, assessing and selecting them and bringing them to fruition in order that they might effect some change. As such, these definitions reflect academic interests, and may differ from practitioners’ perceptions or the reality of engineers’ day to day activities identified in ethnographic studies [Jagodizinski et al., 2000]. The specific concern in this chapter is design — widely seen as a central core of engineering practice — which refers to both the content of a set of plans and the process that produced those plans [Vincenti, 1990, p.7]. For Herbert Simon, design simply involves “changing existing situations into preferred ones” [1969, p.111] which blurs the distinction between designing and building a technology. However, the concern in this chapter is specifically with engineering design which Ferguson [1977; 1978; 1993] highlights is differentiated from artisan design by its use of drawings that now mediate the previously direct link between the artisan’s mind and the materials they are working with. This introduction of visual diagrams has had profound implications for engineering design and has led to new kinds of visual thinking, new tools, new forms of communication, and a greater division of labour between the people who design

Tacit Knowledge and Engineering Design

353

and the people who build technology [Ferguson, 1993; Arnheim, 2004]. Once diagrams had opened up a space between designers’ minds and their artefacts, many more processes of design became possible. In particular, technology itself could be more easily applied to design to help formulate, analyse, communicate and test designs. As Constant [1980] argues, what distinguishes modern engineering from the 15th century engineering of Filippo Brunelleschi is the development of regimes of testing that further intermediate between a designer’s mind and the final product. In craft production, improvements in technology occur slowly and in a haphazard fashion, while with modern engineering the specialisation and professionalisation of testing allows a faster, more accurate and much more public comparison of alternatives [Constant, 1980, p.23]. Two changes were vital here: first, the emergence of specialised academic engineering science, such as chemical and electronic engineering, in the early 20th century. These new academic disciplines engaged in research that generated new theories, frameworks, data, tools and particularly a new generation of professionally trained engineers who were able to use new testing technologies [Rosenberg, 1998]. The second important change was the development and widespread use of testing-technology that was often provided as a service by the new engineering consultancies that emerged at the turn of the 20th century. While this might appear at first as a simple Weberian shift from local, tacit knowledge to more global, scientific, visual and articulated technology, the rest of the chapter will argue that such changes have not been so simple. 2

TACIT KNOWLEDGE: FROM THE MARGINS TO THE MAINSTREAM

The notion that tacit knowledge was something more important than just unarticulated elements of conversations appears in the work of the Hungarian doctor and chemist Michael Polanyi (1891–1976). Polanyi moved into the philosophy of science in response to the dominance of positivism, and in particular the potentially totalitarian dangers that he saw in its legitimisation of the centralised control of science. In doing so, he drew on his experience of hands-on experimentation in physical chemistry to argue against conceptions of knowledge that saw it as abstract, mechanical, deterministic and therefore possible to centrally plan. Instead Polanyi stressed how all knowledge is centred on an agent and her body that is constantly interacting with the world [Polanyi, 1969, p.147]. This interaction, including the use of words and symbols, requires creativity, skill, imagination and personal knowledge. These are essential to our ability to learn through unconscious trial and error when we “feel our way to success” [Polanyi, 1958, p.62]. More importantly, he suggests that our conscious actions are dependent on creative, preconscious processes of integration that produce new emergent cognitive phenomena that were not previously present in its components. Consequently, our knowledge is more than the sum of its parts and, while it can be described by rules, it cannot be reduced to rules, with the implication that “we know more than we can say” [Polanyi, 1969; Nisbett and Wilson, 1977].

354

Paul Nightingale

To explain these ideas Polanyi used an example of the new stereo-image formed by looking at two stereoscopic photographs with different eyes. He argues we become focally aware of this stereo-image by being subsidiarily aware of the two separate pictures. Subsidiary awareness functions by bearing on the focus of our attention and making us conscious of merged meanings. Tacit knowing therefore involves a process of integration rather than a [reversible] inference or deduction: it is “knowing a focal object by attending subsidiarily to the clues that bear on it” and this knowledge is lost by focusing on clues in isolation [Polanyi, 1965, p.799]. All aspects of knowing, for Polanyi, share this anti-reductionist character and are based on bodily interactions and creativity. For Polanyi, this applies as much to tools as it does to ideas and concepts [Polanyi, 1969, p.148; 1968]. While tool-users initially have to focus their attention on their tools, after a period of practice they develop the subsidiary awareness that allows them to use the tools with skill. Focusing on particular features of our experience, such as turning when cycling or on the hammer when hammering, brings them out of subsidiary awareness into focal awareness. This isolates them from our wider tacit understanding and destroys the coherence and meaning of our actions [Polanyi, 1966a, p.10]. This is why focusing on words when speaking, or finger movements when playing the piano, disrupts the flow of these actions [Polanyi, 1969, p.144]. As a consequence, description of comprehensive entities based only on their parts, or on the laws of nature which apply to their parts, can never reveal the operation of the higher principles that define what they are [Polanyi, 1965; 1968]. Polanyi [1965, p.799] argues that: to go back to the premises of a tacit inference brings about its reversal. It is not to retrace our steps, but to efface them. Suppose we take out the stereo-pictures from the viewer and look at them with both eyes. All the effects of the integration are cancelled; the two pictures no longer function as clues, their joint meaning has vanished. Because such tacit knowledge is holistic and non-reducible it cannot be simply built up from components or learnt by following rules [Polanyi, 1966a; 1968]. Polanyi [1966b; 1967; 1969] therefore places great emphasis on what he calls ‘indwelling’ for comprehension and learning. When we learn, we have to dwell within the concepts we are using for a period of time until they move into subsidiary awareness. This enables us to creatively see the broader coherence of what we are studying and appreciate that body of knowledge as a whole. This can be seen in apprenticeships where students must initially take everything on trust and follow examples until they build up the knowledge needed to understand the activity as a coherent whole. As Polanyi put it: An art that cannot be specified in detail cannot be transmitted by prescription, since no prescription for it exists. It can be passed on only by example from master to apprentice. This restricts the range of diffusion to that of personal contacts, and we find accordingly that

Tacit Knowledge and Engineering Design

355

craftsmanship tends to survive in closely circumscribed local settings. [1958, p. 52] Using more modern terminology, we might say that rules and descriptions of how to perform actions are imposed from the outside, rather than being intrinsic to the actions. As a consequence, they can never fully transmit knowledge without the mediation of a background of cognitive dispositions [Searle, 1995]. As a result, “all knowledge is either tacit or rooted in tacit knowledge. A wholly explicit knowledge is unthinkable” [Polanyi, 1969, p.144].

2.1

Cognitive and social scientists on tacit knowledge

While empirical observations about the difficulty of transmitting some kinds of knowledge may seem trivial, Polanyi argues that they show the implausibility of ‘objective’ knowledge that is detached from human action and of various theories built on such ideas [Polanyi, 1962; 1969]. Given that explaining how such knowledge is possible has been a central focus of the philosophy of science, Polanyi has had an important, if not always positive, influence on a number of philosophers of science such as Feyerabend, Lakatos, and Agassi. More recently, Searle [1995] has argued that a range of implicit cognitive dispositions, much like tacit knowledge, that he terms the Background provides structure to our thoughts and actions and prevents them from being reducible to rules. Similar ideas have been important in critical attacks on the largely over-inflated claims of proponents of Artificial Intelligence [Dreyfus and Dreyfus, 1986; Collins, 1974; 1990; 2001]. Given how positivist ideas about knowledge have been foundational to many social sciences such as psychology and economics, the concept of tacit knowledge would seem to have the potential to be widely applied [Gill, 2000; Lakoff, 1987]. However, within psychology it is not widely used [Reber, 1989; 1993; Marcel, 1983] and is often considered to be too broad to be analytically useful. It does, however, help explain implicit learning, for example, how experimental subjects learn to anticipate electric shocks without being able to articulate what triggered them and types of knowledge that can only be recalled by doing [Lazarus and McCleary, 1949; Reber, 1989; Underwood, 1996; Lewicki and Czyzewska, 1992; Schacter, 1992]. A considerable amount of empirical work supports Polanyi’s view that much of our learning and problem-solving ability is tacit [Sternberg, 1986; Lihlstrom, 1987; Reber, 1989; Dixon, 1971; Merikle, 1992; Berry, 1994; 1997; and Buckner, 1995] as well as his assertions about the roles of cognitive gestalts in structuring perception [Pylyshyn, 1981]. These allow parts of an image to be seen as a whole (as when we recognise a face) even though our eyes only focus on one bit at a time [Reber, 1989]. Recent advances in genetics and neurology seem to support Reber’s [1989] conjecture that tacit knowledge is an older, more primitive form of ‘knowledge’ that supports later evolutionary developments like consciousness and language [Damasio, 1994; 2000]. Much of our cognition is tacit in the sense of not being accessible

356

Paul Nightingale

by the mind, and conscious thought is dependent on neural systems that either cannot be, or are not, part of consciousness. These neural systems generate images of you changing in response to an object, allowing you to feel changes produced by external objects as a subjective, inner, qualitative state [Damasio, 2000]. This seems to fit Polanyi’s account — “I shall say that we observe external objects by being subsidiarily aware of the impact they make on our body and of the responses our body makes to them” — very well [1965, p. 805]. Such neural systems allow images to be brought from subsidiary awareness to focal awareness [Posner, 1994] enabling concentrated attention that can be linked to memory and categorisation to allow learning from errors (see Tononi and Edelman, [1999] for a mechanism). Brain imaging technology has shown that as we learn neural images are gradually moved to areas of the brain that cannot be accessed by consciousness. This functional isolation produces a “gain in speed and precision, but a loss in context-sensitivity, accessibility, and flexibility” [Tononi and Edelman, 1998, p. 1847] and makes expert knowledge generated by repeated practice difficult to articulate. Beyond the cognitive sciences there is also a substantial literature on tacit knowledge that begins to address technology. Again it plays a supporting role for heterodox approaches that contest more positivistic paradigms. For example, tacit knowledge has been a central idea for many years within the heterodox economics literature that places emphasis on technological learning [Nelson and Winter, 1982; Freeman, 1982]. Nelson and Winter [1982, p. 77-79] for example, in a very influential work, highlight the importance of procedural tacit skills in the design and development of technology, and the consequent difficulties involved in creating, diffusing and using technology. In doing so, they build on a body of work by writers on engineering, such as [Constant, 1980, p. 22-27; 2000; Court et al., 1997; Donovan, 1986; Ferguson, 1977; Gille, 1986, p. 1156-61; Stapleton et al., 2005] and [Rogers, 1983] who have reflected on empirical examples of the tacit nature of engineering knowledge, with [Vincenti, 1990] as the seminal work on engineering knowledge. Because engineering knowledge is partly tacit, it tends to be private [Dosi, 1988, p. 242] and mainly transmitted through face-to-face interaction [Leonard-Barton, 1995; Leonard and Sensiper, 1998]. Its specificity to particular technologies and environments enables firms to develop capabilities that differentiate them from their peers [Pavitt, 1986; 1996; Freeman, 1982; Nelson and Winter, 1982; Pavitt, 1984, p. 343; Dosi et al., 1989; Nelson, 1991; Dosi, 1988, p. 224]. Since these capabilities are associated with improved performance, tacit knowledge is a central focus of the organisational learning literature [Argyris and Schon, 1974; Tsoukas, 1996; Spender, 1995; 1998; Lam, 2000]. Professional organisations, such as engineering design offices, are particularly dependent on the accumulation of tacit knowledge [Becher, 1999; Howells, 1996; Benner, 1984; Eraut, 1999; Megginson, 1996; Veshosky, 1998]. Sch¨on [1982] has highlighted that professional learning involves building up tacit knowledge through critical reflection on actions. This, he argues, makes the practice of design in-

Tacit Knowledge and Engineering Design

357

herently interactive. These ideas have been influential within the management literature, which has sought to understand how tacit knowledge can be built up and used for economic advantage [Teece, 2000; Dougherty, 1992; Leonard-Barton, 1995; Leonard and Sensiper, 1998; Tsoukas, 2003; Brown and Duguid, 2000; Kogut and Zander, 1992]. These ideas have also been applied at the regional and national level within the economic geography literature where the difficulties of transmitting tacit knowledge, and its importance to technological development, are used to explain regional diversity and the geographic clustering of industries [Pavitt, 1996; Asheim and Gertler, 2005; Gertler, 2003; Howells, 2002; Lawson and Lorenz, 1999; Audretsch and Feldman, 1996; Maskell and Malmberg, 1999]. As tacit knowledge became an important concept within economics, management and geography, more critical voices began to emerge that questioned its empirical and theoretical value. Tsoukas [2003] is supportive of the analytical value of tacit knowledge, but suggests that the notion (prominent in the knowledge management literature) that tacit knowledge can be codified misunderstands what tacit knowledge is. As Tsoukas [2003; 416] noted, “tacit and explicit knowledge are not the two ends of a continuum but the two sides of the same coin: even the most explicit kind of knowledge is underlain by tacit knowledge.” Breschi and Lissoni [2001] similarly argued that just because tacit knowledge can explain regional agglomeration it does not follow that it is in fact the correct explanation. Cowan et al. [2000] expressed extreme scepticism that tacit knowledge was a strong enough concept to explain every deviation from the predictions of neo-classical theory in economics, while Nightingale [2003] likened tacit knowledge to physicists’ “dark matter” that explains away the empirical failures of existing theory, but is rarely critically explored. These criticisms suggest that tacit knowledge has been useful for highlighting the empirical failures of social sciences that build on objectivist conceptions of knowledge, such as neo-classical theory in economics, but the idea itself covers a range of distinct features of cognition that are probably better kept distinct. Even within the literature just reviewed, tacit knowledge covers the embodied nature of knowledge; unconscious knowledge; implicit learning; subsidiary (and focal) knowledge; knowledge that is simply unsaid; knowledge that can never be articulated; and gestalts that structure cognition. Similarly, neurologists distinguish between neural mechanisms; neural mechanisms that produce neural images; neural images that can be potentially brought to conscious attention, i.e. preconscious or potentially conscious mental images; and mental images that are currently being consciously attended to. Being such a broad concept, tacit knowledge has tended to be used as the name for empirical counter-examples to theories of learning or technical change that reduce knowledge to easily transmittable information. This, however, does not exhaust Polanyi’s ideas and potentially, as the next section will argue, overlooks a more insightful side of Polanyi’s thought.

358

Paul Nightingale

3 AN ALTERNATIVE VIEW OF TACIT KNOWLEDGE From the perspective in this chapter on engineering design, tacit knowledge is interesting because Polanyi suggests it is a component of technology, rather than just a kind of knowledge needed to create technology. In much of the social science literature just reviewed, technology and tacit knowledge are distinct and one (tacit knowledge) plays a role in the development of the other. However, for Polanyi tacit knowledge is part of technology in the sense that the function of a technology (which is what a technology is) is realised through a process of tacit inference and, like a stereoscopic image, ceases to be what it is in the absence of tacit knowledge. Being a technology is an imposed rather than intrinsic property [Searle, 1995]. Polanyi writes [1958, p. 52] in a quote picked up by Nelson and Winter [1982, p. 119] that “even in modern industries the indefinable knowledge is still an essential part of technology. I have myself watched in Hungary a new, imported machine for blowing electric lamp bulbs, the exact counterpart of which was operating successfully in Germany, failing for a whole year to produce a single flawless bulb.” In this quote Polanyi says that indefinable knowledge is “an essential part of technology” rather than “is needed to get technology to work”. While we must be cautious of taking phrases out of context, seeing tacit knowledge as part of technology, in the strong ontological sense that tacit knowledge makes technologies what they are, fits with Polanyi’s non-reductionist view of the world and his emphasis on creativity. This is more than the weak epistemological sense in which tacit instrumental knowledge is just needed to get technologies to function. As a chemist Polanyi understood the inherent implausibility of reductionism [1965; 1968], more recently, see [Dupr´e, 1993]. For chemists, reductionism is misleading because many qualities exist within chemistry that cannot be reduced to, let alone explained by, the behaviour of their component parts. This is why you cannot explain why Gold (the metal) is gold (the colour) or why mercury is a liquid using only quantum mechanics [Scerri and McIntyre, 1997; Dupr´e, 1993]. Such emergent phenomena do not contradict the laws of nature [Barrow, 1988], but exist within Polanyi’s [1965] “boundary conditions” of potential behaviour that is consistent with those laws. Tacit knowledge adds something to artefacts in the ontological sense because in some instances these boundary conditions can be governed from above: the possibilities opened up by the rules of chess, for example, can be controlled by the strategies of the players. Similarly, the laws of mechanics may be controlled by the operational principles of a machine which are imposed by designers and are not reducible to the machine’s components. These higher principles make technologies what they are, and are distinct from the lower principles which remain in operation even if the machine is smashed up. This again highlights Polanyi’s point that comprehensive entities, in this case technologies, are more than the sum of their parts. For Polanyi, the property of being a technology, like the property of being a beautiful painting, is not purely intrinsic. It reflects, in part, a coherence the

Tacit Knowledge and Engineering Design

359

viewer imposes on an object. Just as paintings are more than blobs of paint, so technologies are more than their components. For Polanyi these additional features are created through a process of tacit inference generated by indwelling. The same ideas are used by Polanyi to explain why science is inherently creative: because complex entities cannot be reduced to their parts, scientists have to dwell within their subjects to build up understanding and creatively come up with theories that explain them. With technology, however, designers don’t just understand features of the world that cannot be reduced to lower order principles — they actively create those features. They create new solutions to problems through a process of tacit inference and then change the world to impose those solutions on technological artefacts to create new behaviour that is not reducible to its components. The idea that higher operational principles, imposed by designers, define what technologies are, is similar to ideas presented more recently by Searle [1995], Kroes and Meijers [2006] and Vermaas and Houkes [2003; 2006]. For Searle [1995, p. 19] technologies have an intrinsic physics — that appropriates Polanyi’s boundary conditions — and an imposed function that determines how the technology should behave (i.e. drugs should cure diseases and umbrellas should keep you dry) — that approximates Polanyi’s operational principle. This imposed function is ontologically prior to the intrinsic physics and determines what a technology is [Searle, 1995, p. 19]. This is why a safety valve is still a safety valve with the function of stopping explosions, even if it malfunctions and fails to do so [ibid ]. Because technical functions are not intrinsic, technologies can have multiple functions — which is why a computer disc can both store data and stop a coffee cup marking the table. However, the range of possible functions of a given technology is constrained, as a technology’s physics has to be able to match its imposed function. If the epistemic idea that tacit knowledge is needed to get technology to work is the first step away from just seeing technology as artefacts and the Searlean idea that imposed functions are ontologically prior to technologies’ intrinsic properties is the second step, then Polanyi makes a further much more controversial step. Polanyi suggests that technologies’ intrinsic properties come to embody imposed higher order principles that are generated by tacit inference. Presumably for Searle engineers would understand a function and impose it on the world by changing the world until the technology’s intrinsic properties matched the desired function. As a simple theory, this has much to recommend it, but from Polanyi’s perspective it doesn’t address his concerns about reductionism and would work in a world where reductionism was true. For example, in a world where technological artefacts could be reduced to their component parts, knowledge of those components and their interactions would be sufficient to generate a desired function. Polanyi’s position is more contentious and suggests that because reductionism doesn’t hold, the function of the artefact isn’t implicit in the functions of its components. Instead, higher order boundary conditions define the function and have to be creatively developed through a process of tacit inference. Once the world is changed to match this function, the tacitly created boundary conditions become embodied in the technology. In more Searlean language, the

360

Paul Nightingale

intrinsic properties of the technology are modified to match an imposed function that is not implicit in the intrinsic properties of its components. The technology therefore comes to embody tacitly created boundary conditions. While dollar bills function as money because society accepts the institutional fact of their value [Searle, 1995], technologies embody tacitly created functions in their physical make up. The resulting behaviour, unlike being money, can continue even if society stops believing in it. For example, currencies become worthless when societies stop trusting them, but an unmanned space probe sent out from earth continues to behave in ways that match an imposed function even when it is out of sight. If in millions of years time, long after the earth has been engulfed by the sun, the probe was found by an alien anthropologist, they might decipher something about our culture from its behaviour because part of our culture is embodied in what the thing actually is.

3.1 The difference between science and technology Polanyi is particularly interesting to those concerned about engineering design because he extends his ideas about tacit knowledge, the imposed nature of functions, and the irreducibility of comprehensive entities to draw out the differences between science and technology. He writes [1958, p. 177]: [T]he beauty of an invention differs . . . from the beauty of a scientific discovery. Originality is appreciated in both, but in science originality lies in the power of seeing more deeply than others into the nature of things [i.e. the non-reducible emergent order in chemistry that cannot be reduced to physics, yet is not incompatible with it], while in technology it consists in the ingenuity of the artificer in turning known facts to a surprising advantage. The . . . technician . . . follows the intimations, not of a natural order, but of a possibility for making things work in a new way for an acceptable purpose, and cheaply enough to show a profit. In feeling his way towards new problems, in collecting clues and pondering perspectives, the technologist must keep in mind a whole panorama of advantages and disadvantages which the scientist ignores. He must be keenly susceptible to people’s wants and able to assess the price at which they would be prepared to satisfy them. A passionate interest in such momentary constellations is foreign to the scientist, whose eye is fixed on the inner law of nature. As this passage shows, when it comes to science Polanyi is a realist and for him scientific theories and explanations are meant to be true. However, when it comes to technology, to use anachronistic terminology, Polanyi is much more of a constructivist [Polanyi, 1967; 1969]. This is because technologies are meant to be useful and usefulness reflects inherently subjective, time-dependent assessments of value. As a consequence, the particular trade-offs made during design are entirely alien to his (very purist) view of science. Moreover, they give design a particular

Tacit Knowledge and Engineering Design

361

cognitive element not found in science that helps distinguish the philosophy of technology from the philosophy of science. Today such a clear cut separation between science and technology overlooks the role of design in experimental sciences, in both the design of experiments and the design of experimental apparatus, and also the increasing role played by scientific knowledge in design processes. The idea that science and technology are distinct but closely interacting finds support in the work of scholars of technology such as Pavitt [1998] and Layton [1974; 1976] who distinguish technology from science because technical behaviour has to be (1) reliably created, (2) for users, and (3) in the complexity of the outside world, rather than in the atypical purified conditions of the laboratory as a one-off, largely private, and not necessarily reliable phenomenon. This means that engineers (defined as professionals who are held legally responsible for producing products that are ‘fit for use’) have to understand the environment in which products are used [Parnas, 1999, p. 3]. This differentiates them from scientists and is why engineers focus on what works reliably rather than on new knowledge, require a broad understanding of how their products will be used, and normally rely on a legal process of accreditation, based on an established and formalised body of knowledge, to ensure the quality of their work, unlike scientists who need to be up-to-date with the latest findings in their field, can be narrow in their specialisation and can let external referees determine the quality of their work [Layton, 1979, p. 77–78; 1976; Parnas, 1999]. As Pavitt notes [1998, p. 795] this creates important differences between the purposes of science and technology and the nature of the knowledge they generate: One of the main purposes of academic research is to produce codified theories and models that explain and predict natural reality. To achieve analytical tractability, this requires simplification and reduction of the number of variables (e.g., ‘Under laboratory conditions . . .’, ‘Other things being equal . . .’). On the other hand, the main purpose of business R&D is to design and develop producible and useful artefacts. These are often complex, involving numerous components, materials, performance constraints and interactions, and are therefore analytically intractable (i.e. theory and formal models are an insufficient guide to, and predictor of, practice). Knowledge is therefore accumulated through trial and error. These differences, in turn, relate to the nature and location of the knowledge production processes: Academic research is mainly basic research; business research is mainly the development and testing of prototypes and pilot plants. Academic institutions dominate in the publication of scientific papers, and business firms in the granting of patents. And despite examples of spectacularly close links between basic research and technology (i.e. biotechnology), basic research builds mainly on basic research (scientific papers cite other scientific papers much more frequently than patents)

362

Paul Nightingale

and technology builds mainly on technology (e.g., patents cite other patents much more frequently than scientific papers). [Pavitt, 1998, p. 795] Polanyi’s conceptual framework of a non-reductionist view of nature, an emphasis on tacit inference and creativity in generating both new scientific theories and new technologies, and the corresponding emphasis on in-dwelling, have important implications for understanding engineering design that are much deeper than a simple empirical observation that some knowledge used in engineering cannot be easily articulated. Dividing design processes into understanding problems, formulating solutions and testing, provides a way to explore how some of the existing history and philosophy of technology relates to Polanyi’s ideas.

3.2 The process of design: understanding problems and negotiating solutions Focusing first on framing problems, most design — even for simple technologies — involves very complex and often conflicting demands that have to be negotiated and clarified: a process that has been nicely illuminated within the history and sociology of technology literatures [Nye, 2006; Hughes, 2004]. These multiple and potentially conflicting demands form part of designers’ subsidiary awareness and are often unstated. For example, if I was asked to design a hammer, and produced one made from the horn of the last black rhino calf, there is a very real sense in which I did not understand what was intended, even though at an explicit level my response perfectly matches the requirements. More importantly, the unstated background assumptions are not fixed and change as engineers creatively merge conflicting and often open-ended requirements. This often involves understanding the wider impacts of their proposed solutions. Gardiner and Rothwell [1990; Rothwell and Gardiner, 1988] for example, highlight the importance of considering manufacturability in early design, and how sharing components within a family of designs can simplify production and generate economies of scope. Rothwell [1992] found that the ability to consider these factors, while also paying attention to consumers’ needs (which may not be the same as what they think they need) is a vital part of successful design-led innovation. Formulating design problems is therefore open-ended and cannot be reduced to simple rule following [Dym, 2000, p. 17]. It requires the integration of knowledge, as judgements have to be made about which problems to address and what relative weights to give to conflicting demands [Hacker, 1997]. Many of these multiple criteria will typically have to be considered, merged and explored during the design process. The difficulties of sharing tacit understanding of problems and the uncertainties associated with their exploration make design a negotiated process rather than a simple creative event [Burcarelli, 1994]. Designers will have subsidiary awareness of many of these issues and bring them in and out of focal awareness as they explore different design options and make explicit their concerns to other members of the design team [Henderson, 1999]. This makes design more

Tacit Knowledge and Engineering Design

363

complex than simply recognising a problem, matching a solution to that problem, and creating that solution. There is ample empirical evidence that this process involves knowledge that is difficult to articulate [Vincenti, 1990]. However, Polanyi’s framework suggests tacit knowledge plays a role in structuring the design process. Because “being a problem” is an imposed rather than intrinsic property, it is understood contextually, which for Polanyi involves a process of tacit inference. For Polanyi, technologies’ coherence is understood through “indwelling”, as when Vincenti’s [1990] aircraft designers had to get into aeroplanes and sit on pilots’ laps because they had been unable to understand pilots’ experience of stability without experiencing it as a coherent whole. This knowledge was something that could not be reduced to information. This may help to explain Cross’ [2004, p. 432] findings in his review of studies of design choices which show that experienced designers often approach design tasks through ‘solution conjectures, rather than through problem analysis’. Rather than working through the problem in great detail to generate a solution, they use their experience to conjecture design solutions that might work and then try them, using the results of their experiments to better understand the problem they are faced with and how potential solutions might address it. In doing so they select particular features of the problem to attend to and identify potential solutions that they wish to explore. This “imposes on the situation a coherence that guides subsequent moves” [Cross, 2004, p. 423]. Because design choices are open-ended, designers have a degree of choice in how problems are framed, and expert designers have been observed to deliberately define problems in difficult and challenging ways [Cross and Clayburn, 1998; Ho, 2001]. Given the inherent uncertainty of design implicit in Polanyi’s non-reductionist ontology, and his emphasis on indwelling, it does not seem surprising that expert designers might proceed in this solutionled trial and error way. Or rather, it would be surprising if they only approached design through the analysis of problems, as by breaking comprehensive entities into parts, analysis loses the imposed coherence that designers are trying to impose.

3.3

Generating solutions

During their training, engineers pick up an understanding of various design options and a contextual understanding of when and where tried and tested solutions can be applied [Nightingale, 1998]. Despite substantial investments, these choices have not been reduced to technical rule-following. Partly, this is because engineers rely on what Vincenti [1990] calls Fundamental Design Concepts that sit in the back of designers’ minds and are implicit in their design choices. The first of these are operational principles that show how the components of a design will “fulfil their special function in combining to an overall operation which achieves the purpose of the device” [Vincenti, 1990, p. 208; Polanyi, 1958, p. 328]. A classic example of such an operational principle would be Sir George Cayley’s definition of the operational principle of an aeroplane involving making “a surface support a given

364

Paul Nightingale

weight by the application of power to the resistance of air” [Vincenti, 1990, p. 9]. Once designers have this idea in the back of their mind they no longer have to consider creating aircraft that flap their wings. Polanyi argued that these operational principles define what technologies are and exist outside scientific knowledge. As a result, “the complete [scientific] knowledge of a machine as an object tells us nothing about it as a machine” [Polanyi, 1958, p. 330]. This is supported by Vincenti who notes that operational principles “originate outside the body of scientific knowledge and come into being to serve innately technological purposes. The laws of physics may be used to analyze such things as airfoils, propellers and rivets once their operational principles have been devised, and they may even help in devising it; they in no way, however, contain or by themselves imply the principle” [1990, p. 209]. Scientific knowledge can explain why a particular solution produces the result it does, but, because imposed functions are linked to the intentional plans of technologies’ designers, scientific knowledge that is divorced or unconnected to these plans will not provide those solutions [Nightingale, 1998]. Vermaas and Houkes [2006, p. 16] make a similar point when they highlight how “technological functions ... create a conceptual bridge between the intentional and structural natures of artefacts; function ascriptions connect the intentional description of the use plan [what the technology will do] with a physical description of the artefacts themselves via the physical capacities of the artefacts that explain why this plan is effective”. When scientific theories are used to understand technology they can help explain why a particular design produces the effects it is intended to. However, they cannot explain why those particular effects were intended in the first place. Vincenti’s [1990] second fundamental design concept is the normal configuration of a device which refers to the general arrangement of components that allows artefacts to generate their operational principle [1990, p. 209, 102–110]. Car designers, for example, will be able to draw on a paradigm case of a car with four wheels, a front-mounted, water-cooled, petrol-driven engine, and four doors [Vincenti, 1990, p. 209]. Again, such concepts are implicit and rarely articulated during design. These fundamental design concepts define the structure and direction of the problem-solving process by addressing certain key problems, while leaving a penumbra of flexibility to address the wide variety of other design issues that arise. In doing so, their application re-defines the design problem and makes it more specific, setting up the conditions for the next round of design. The iterative application of operational principles can therefore generate a hierarchy of structurally related, increasingly specific sub-problems that form the basis for the design process [Nightingale, 1998; 2000]. Vincenti [1990, p. 9] nicely highlights this process in which design moves from very general problem definition that translates ill-defined problems into more concrete technical problems, after which the process shifts to overall design which provides an overarching layout of the system, then moves to the design of major components, which is then followed by further subdivision of the project (see also [Bucciarelli, 1994]).

Tacit Knowledge and Engineering Design

3.4

365

Testing and modification

Analysis and testing are important in modern engineering design because operational principles only provide rough guidance, and mark the first step in a long trial and error journey to construct a predictable final product. As Dupr´e [2001, p. 171] notes of internal combustion engines, a first approximation of the operational principle is that: a mixture of air and petrol is exploded in a cylinder, pushing a piston down the cylinder; the cylinder is connected to a shaft which is rotated by the moving piston. A number of similar cylinders are connected to this shaft, and a sequence of explosions keeps the shaft rotating continuously . . . But if, on the basis of this explanation, someone lined up some coffee cans partially filled with petrol on the kitchen floor, stuck toilet plungers in the cans, tied the ends of the plungers to a broomstick, and then posted lighted matches through the holes in the sides of the coffee cans, they would certainly not have built an internal combustion engine. Initial designs are therefore only potential solutions and as Constant [1980; 2000] has argued the mediation of regimes of testing, based around widely-used testing technologies, has transformed engineering and the ability of designers to produce complex technology. A considerable amount of modern engineering design involves working out criteria and specifications that help define how a technological system will achieve its desired function in more detail. The production of specifications involves translating very “general, qualitative goals for the device into specific, quantitative goals couched in concrete technical terms” [Vincenti, 1990, p. 211]. Typically this is a complex process involving the production of diagrams, models, mock-ups and back of the envelope calculations. These artefacts allow knowledge to be shared between the various actors involved in design, and the negotiation (or not) of conflicts within the inherent trade-offs between different design choices. As such, the model or mock-up acts as a ‘boundary object’ [Henderson, 1998; 1999] to allow shared understanding of the design and design process. This helps mediate between different groups’ understanding of the design, and the validity of the ‘facts’ that make it up. Such models also play a key role in facilitating learning during design. The complexity of many engineered artefacts, together with their interactions with a changing environment, make working out the effects of many design changes either analytically intractable or analytically very difficult [Pavitt, 1984; Nightingale, 2004]. It is therefore misleading to see design as a simple linear process, particularly with multi-component systems where the appropriate design of one component is sensitive to the design of others. These interdependencies mitigate against trying to change many components at once [Nelson, 1982, p. 463]. Consequently, the design, development and production of complex artefacts involves learning, experimentation, testing, and numerous modification and feed-back loops.

366

Paul Nightingale

Henderson [1995; 1998; 1999] has shown in a series of wonderful case studies how sketches and models are used interactively at both the individual and group levels “to work out and negotiate various perspectives and to draw in, literally and figuratively, a wealth of tacit knowledge” [1998, p. 141]. Component designers, for example, can show production engineers their designs, who can in turn then articulate their ‘gut feelings’ why particular parts might be hard to machine, and what design changes might improve them, without having to articulate exactly why. D’Adderio [2001] similarly reflects on the very visual nature of the knowledge used in these negotiations, and the way graphical tools are used by designers to communicate with one another. While Henderson’s sociological approach focuses on social groups, her Actor Network Theory approach is consistent with seeing these models as part of a negotiation with nature, in which nature refuses to negotiate on designs that do not work. As a consequence, a lot of engineering design work involves finding out what behaviour nature finds acceptable. While it is possible to rely on purely empirical methods and unguided changes to produce improvements to designs, such approaches tend to be costly and timeconsuming. Instead, design is guided by tacit understanding and rules of thumb that are specific to local situations and technological configurations [Vincenti, 1990]. Given the complexity of most designs, the experimental processes involved in engineering design typically involve creating simplified (i.e. artificially predictable) conditions where the assumptions underpinning these local explanations are true [Nightingale, 2004]. This allows explanations that are too simple to work in the real world to be used to guide the design process. As knowledge is accumulated, the simplifying conditions can be relaxed and the design process can proceed from ‘laboratory conditions’ to models, prototypes, field tests and eventually real-world applications. This guidance (hopefully) reduces the number of experimental dead-ends and improves final designs. As this process proceeds, designers take practical considerations, such as the clearance needed for maintenance, or the idiosyncrasies of the staff that will eventually operate the technology, into account. Much of this practical knowledge is unarticulated, context-dependent and defies codification, making testing prototypes an essential part of design [Vincenti, 1990]. In carrying out this testing and modification, designers rely on shared, but unarticulated, ways of thinking and implicit models and analogies. These analogies and models — for example, thinking about the stability of an aircraft about its vertical axis as a ‘weathercock’ — are again not always easily expressible in words. They often involve a very visual form of thinking, and need to be articulated on diagrams and drawings to be worked on and transmitted [Vincenti, 1990; Henderson, 2000]. Such models are analysed to produce descriptive information about how the design will behave as well as prescriptive data about what is needed for the design to achieve its desired function. Academic and industrial engineering research has developed a series of theories, theoretical tools, mathematical methods and intellectual concepts for analysing designs. Like Polanyi’s operational principles, some of these intellectual tools are specific to engineering, for example, concepts

Tacit Knowledge and Engineering Design

367

like propulsive efficiency and feedback enable quantitative analysis, but are not scientific terms [Mayr, 1976, p. 882; Vincenti, 1990, p. 216; Ferguson, 1978, p. 450]. Such tools allow engineers to investigate how well designs and design options match, or mismatch, design criteria and specifications. During the process of testing artefacts and components, engineers switch between seeing technologies in functional terms as part of a wider system of use, and seeing them in terms of their intrinsic physics which can be subject to empirical analysis. In each instance, the alternative is left implicit and the new knowledge generated through testing integrated back into the process of design. Designers therefore have to reflect on their designs and the results of tests, negotiate changes to inter-dependent components, and work out prescriptive performance criteria, often using models and diagrams as tools for what Hutchins [1995] has called “external cognition”, that are modified in an attempt to capture implicit, background understanding and tacit knowledge [Henderson, 1995]. The role played by tacit knowledge in Polanyi’s thought contrasts with a strong tradition of understanding engineering design in terms of a means-ends practical reason. Simon [1969], for example, is an influential exponent of the view that design is a “science of the artificial” in which decomposable problems are analysed and fitted back together. For Polanyi design can’t be about taking problems apart and fitting them back together again because coherent entities cannot be reduced to the sum of their parts. Instead, as Sch¨ on [1982] has shown, it is inherently creative and involves interactions, practice and reflection on actions. For Polanyi, design can’t only be about adapting means to well-defined ends because those ends and means are not always at hand. They will often have to be created, and this creative process involves tacit inference. Seeing design as a clean “science of the artificial” often misses the inherently creative, messy and open-ended processes of developing and adjudicating between conflicting demands and benefits. The tacit nature of the knowledge involved in creating the novel boundary conditions that make technologies behave in particular ways cannot be reduced to a simple calculation. Reducing design to a science of design leaves un-explored the complex, creative processes used by designers, and the role of diagrams, models and visual thinking in exploring design options. 4

CONCLUSION AND SYNTHESIS

This chapter has hopefully shown that tacit knowledge is a useful, but probably over-encompassing, concept that nevertheless helps illuminate important features of engineering design. While most of the literature that uses the concept of tacit knowledge does little more than report that there are features of engineers’ knowledge that are difficult, if not impossible, to articulate, this chapter has highlighted that Polanyi’s original ideas are substantially more interesting. Polanyi begins with an ontology that rejects reductionism and asserts that many entities are more than the sum of their parts. This, he implies, has implications for how we understand the world, as coherent entities cannot be understood by understanding

368

Paul Nightingale

their components; hence Polanyi’s insistence that we know more than we can tell. Instead, Polanyi stresses the importance of tacit inference for perceiving coherence. Learning, therefore, often requires a process of indwelling that builds up knowledge and gradually moves it into subsidiary awareness to enable tacit inference to take place. Polanyi extends these ideas beyond science to the design of technology to highlight how scientific knowledge does not encompass the entirety of what can be known. Because designed technology has a coherence beyond its component parts, the design and production of technologies involves knowledge that is distinct from scientific understanding of those components. Scientific understanding, for Polanyi, focuses on truth, but technological knowledge instead focuses on usefulness. As such, it reflects inherently social concerns about practical applications and judgements about the inherent trade-offs that must be made during design. Operational principles, that imply how a technology will achieve its function, are inherently technological. Their selection and application closes down the number of possible alternative design routes and focuses the design process in a particular direction. In doing so, their selection structures the design process by making the design problem more specific. For Polanyi, tacit knowledge is therefore an essential feature of design and is what allows designers to creatively generate new solutions. It helps explain the creative nature of design, the limited success of attempts to automate design (and weaknesses with the outputs of AI more generally), the importance of diagrams and visual knowledge, and why good design practice is so hard to learn, articulate and teach. In applying Polanyi’s ideas to design, it is difficult to avoid the feeling that tacit knowledge is too broad a concept for the theoretical burdens that have been imposed on it. The cognitive sciences have broken tacit knowledge into a series of distinct, but interacting, phenomena. Similarly, Vincenti and other historians of technology have tended to use more precise and more applicable concepts like operational principles, engineering research, implicit knowledge, etc. While these concepts often draw heavily on Polanyi’s original ideas, they allow a deeper exploration of design. For example, they help us understand how flexible Polanyi’s operational principles are, and how much additional testing has to be undertaken to move from ideas in designer’s minds to final, working artefacts [Vincenti, 1990]. Much of this more recent work, particularly by authors such as Bucciarelli and Henderson, also adopts a much more social understanding of design than the often very individualistic approach taken by Polanyi. While Polanyi’s philosophy of science often presents a historically misleading picture of the heroic lone scientist, his philosophy of technology similarly too often presents design as something that occurs within one person’s head. The ability of tacit knowledge to explain a host of very diverse phenomena, which on closer inspection actually turn out to involve something else, suggests a substantial weakness in how the concept is used. For example, tacit knowledge might explain the localised nature of design capabilities, or localisation might be the result of specialised designers simply having to interact

Tacit Knowledge and Engineering Design

369

more often with one another. Too often the fact that tacit knowledge might explain a phenomenon is used to draw the incorrect inference that it provides the correct explanation. There is no doubt that the concept of tacit knowledge provides useful ways of thinking about design, particularly empirically important aspects of design that are often overlooked elsewhere (visual knowledge, for example). All the same, more work needs to be done towards clarifying what tacit knowledge is and how or if it plays a role in design. Such a conclusion is open to two substantial criticisms. On the one hand, it allows the concept of tacit knowledge to get away with too much. To anyone trained within the Anglo-Saxon analytical tradition, Polanyi’s ideas can be difficult to follow as he jumps between different meanings of the term tacit knowledge. Too often, one gets the feeling that difficult problems are being explained away, rather than explained. Concepts like tacit knowledge, tacit inference and indwelling are rarely clearly defined and it is often difficult to see what they do, and, more importantly, what they do not, encompass. Polanyi might respond that he is correct and many of the problems that seem to exist are simply metaphysical hangovers from assuming that reductionism is true and knowledge is a ‘mirror of nature’, to use Rorty’s phrase. Admittedly, if one thought that all entities in the universe were reducible to the sum of their parts then Polanyi’s ideas may seem magical or mysterious, but he knows as a scientist that the universe isn’t like that. He is therefore simply explaining empirical events. Such a response would seem to be provided with substantial empirical support by historians of technology: much engineering knowledge is difficult to articulate, codified information is rarely sufficient to generate technology, many design concepts are implicit and much of the knowledge used in design involves interaction with material objects, such as drawings, and reflection upon their changed meaning. On the other hand, an alternative critique might be that this chapter has not gone far enough. In trying to explain tacit knowledge and engineering design the chapter has dissolved, and therefore lost, the inherent interconnections between the two. The two have to be understood together, through a process of tacit inference, in order to be understood at all. Like the stereoscopic images, by bringing each into focal awareness the coherence that links them has been lost. Such a criticism should not be dismissed too easily, as intellectual figures as diverse as Raymond Aron and Charles Taylor have found Polanyi’s ideas extremely profound. In response, hopefully this chapter at least hints at this possibility; however, a full integration is beyond the capabilities of the author and the length constraints of an introductory chapter. To reach a conclusion that would placate both sides does not seem easy. There do seem to be good grounds for scepticism about the value of the concept of tacit knowledge. Where it is used, it tends to be used to explain empirical phenomena that are not explained within existing frameworks in the social sciences. However, the explanations often don’t seem particularly robust. Rather than providing a way to change or radically reformulate existing ways of thinking about technical

370

Paul Nightingale

change, economics or design, tacit knowledge seems to plug the gaps in existing theories and allow them to proceed onwards unchallenged. Rarely, if ever, are questions raised about the deeper compatibility between the resulting theoretical chimeras [Tsoukas, 2003]. All the same, tacit knowledge as a concept does hint at something more substantial. It was put forward by Polanyi as part of a very radical attempt to challenge the foundations of 20th century social thought. Hopefully, this chapter has shown that, rather than reinforcing existing ways of thinking, Polanyi’s ideas can help understand their very real limitations. By highlighting the emergent nature of phenomena it stresses the unpredictable nature of the world we inhabit, and the failures of reductionism and strong determinism. Polanyi’s ideas can be used to attack the legitimisation of ‘scientism’ without being anti-science [Gill, 2000]. By showing that technological phenomena cannot be reduced to scientific phenomena, even if they can be explained by science, tacit knowledge as a concept can help to highlight the distinct nature of the philosophy of technology [Vincenti, 1990]. Given the ever-increasing importance of technology to society, this suggests a continuing and growing importance for the philosophy of technology in helping society understand what technology is, how it generates unintended consequences, and how it can be directed along more fruitful paths. Polanyi’s ideas may raise more questions than they answer for the philosophy of technology, but those questions are important enough to deserve more time than they have received so far.

ACKNOWLEDGEMENTS The research for this chapter was supported by the ESRC, EPSRC and NESTA.

BIBLIOGRAPHY [Argyris and Sch¨ on, 1974] C. Argyris and D. A. Sch¨ on. Organizational Learning: A Theory in Action Perspective. Addison-Wesley, 1974. [Arnheim, 2004] R. Arnheim. Visual Thinking, University of California Press, 2004. [Asheim and Gertler, 2005] B. T. Asheim and M. S. Gertler. Regional Innovation Systems and the Geographical Foundations of Innovation. In The Oxford Handbook of Innovation, Fagerberg J, Mowery D., and Nelson R., eds., pp. 291-317. Oxford University Press, 2005. [Audretsch and Feldman, 1996] D. B. Audretsch, M. P. R. and Feldman. R&D Spillovers and the Geography of Innovation and Production. The American Economic Review, 86, 3, 630640, 1996. [Barrow, 1988] J. D. Barrow. The World Within the World. Oxford University Press, 1988. [Becher, 1999] T. Becher. Professional Practices: Commitment and Capability in a Changing Environment. Transaction Publishers, 1999. [Benner, 1984] P. Benner. From Novice to Expert: Excellence and Power in Clinical Nursing Practice. Addison-Wesley, 1984. [Berry, 1994] D. C. Berry. Implicit Learning: Twenty-Five Years on: A Tutorial. In Attention and Performance 15: Conscious and Non-conscious Information Processing, Attention and Performance Series, M. M. Carlo Umilta, ed., pp. 755-82. MIT Press, 1994. [Berry, 1997] D. C. Berry. How Implicit is Implicit Learning? Oxford University Press, 1997.

Tacit Knowledge and Engineering Design

371

[Breschi and Lissoni, 2001] S. Breschi and F. Lissoni. Localised knowledge spillovers vs innovative milieux: knowledge ‘tacitness’ reconsidered. Papers in Regional Science, 80, 255-73, 2001. [Brown and Duguid, 2000] J.-S. Brown and P. Duguid. The Social Life of Information. Harvard Business School Press, 2000. [Bucciarelli, 1994] L. L. Bucciarelli. Designing Engineers, MIT Press, 1994. [Buckner, 1995] R. L. Buckner. Functional Anatomical Studies of Explicit and Implicit Memory. Journal of Neuroscience, 15, 12-29, 1995. [Collins, 1974] H. M. Collins. The TEA Set: Tacit Knowledge and Scientific Networks. Science Studies, 4, 2, 165-185, 1974. [Collins, 1990] H. M. Collins. Artificial Experts: Social Knowledge and Intelligent Machines. MIT Press, 1990. [Collins, 2001] H. M. Collins. What is tacit knowledge? In The Practice Turn in Contemporary Theory Schatzki T. R., Knorr Cetina, K. and von Savigny E. eds., pp. 107-119. Routledge, 2001. [Constant, 1980] E. Constant. The Origins of the Turbojet Revolution. Johns Hopkins University Press, 1980. [Constant, 2000] E. Constant. Recursive practice and the evolution of technological knowledge, in Technological Innovation as an Evolutionary Process, J. Ziman ed., pp. 219-233. Cambridge University Press, 2000. [Court et al., 1997] A. Court, S. Culley, and C. McMahon. The Influence of IT in New Product Development: Observations of an Empirical Study of the Access of Engineering Design Information. International Journal of Information Management, 17, 359-375, 1997. [Cowan et al., 2000] R. Cowan, D. Foray, and P. A. David. The Explicit Economics of Codification and the Diffusion: The Economics of Codification and the Diffusion of Knowledge. Industrial and Corporate Change, 9, 211-53, 2000. [Cross, 2004] N. Cross. Expertise in Design; an overview. Design Studies, 25, 427-441, 2004. [Cross and Clayburn, 1998] N. Cross and A. Clayburn. Expertise in Engineering Design. Research in Engineering Design, 10, 141-149, 1998. [D’Adderio, 2001] L. D’Adderio. Crafting virtual prototype: how firms integrate knowledge and capabilities across organisational boundaries. Research Policy, 30, 9, 1409-1424, 2001. [Damasio, 1994] A. Damasio. Descartes’ Error: Emotion, Reason, and the Human Brain, Putnam Books, 1994. [Damasio, 2000] A. Damasio. The Feeling of What Happens: Body and Emotion in the Making of Consciousness, Harvest, 2000. [Dixon, 1971] N. F. Dixon. Subliminal Perception: The Nature of a Controversy, McGraw-Hill, 1971. [Donovan, 1986] A. Donovan. Thinking about Engineering. Technology and Culture, Special Issue: Engineering in the Twentieth Century, 27, 4, 674-679, 1986. [Dosi, 1988] G. Dosi. The nature of the innovation process. In Technical Change and Economic Theory, G. Dosi, ed., Pinter, pp. 221-238, 1988. [Dosi et al., 1989] G. Dosi, D. J. Teece, and S. G. Winter. ‘Towards a theory of corporate coherence: Preliminary Remarks’ unpublished paper, Centre for Research in Management, University of California at Berkeley, 1989. [Dougherty, 1992] D. Dougherty. A Practice-Centred Model of Organizational Renewal Through Product Innovation, Strategic Management Journal, 13, 77-92, 1992. [Dreyfus and Dreyfus, 1986] H. L. Dreyfus and S. E. Dreyfus. Mind over Machine: The Power of Human Intuition and Expertise in the Era of the Computer. Basil Blackwell, 1986. [Dupr´ e, 1993] J. Dupr´ e. The Disorder of Things. Harvard University Press, 1993. [Dupr´ e, 2001] J. Dupr´ e. Human Nature and the Limits of Science, Oxford University Press, 2001. [Dym, 1993] C. L. Dym. Engineering Design A Synthesis of Views, Cambridge University Press, 1993. [Eraut, 1999] M. Eraut. Professional knowledge, practice and modes of cognition. In Proceedings of International Conference on Self-Directed Learning, R.A. Foucher. ed. Montreal University Press, 1999. [Ferguson, 1978] E. S. Ferguson. Elegant Inventions: The Artistic Component of Technology. Technology and Culture, 19, 450-460, 1978. [Ferguson, 1993] E. S. Ferguson. Engineering and the Mind’s Eye. MIT Press,1993.

372

Paul Nightingale

[Ferguson, 1977] E. S. Fergusson. The Mind’s Eye: Non-Verbal Thought in Technology. Science, 197, 827-836, 1977. [Freeman, 1982] C. Freeman. The Economics of Industrial Innovation. 2nd ed. Pinter, 1982. [Gardiner and Rothwell, 1990] P. Gardiner and R. Rothwell. Design management strategies. In Technology Strategy and the Firm: Management and Public Policy, Dodgson, M. eds., pp. 45-65. Longman, 1990. [Gertler, 2003] M. S. Gertler. Tacit knowledge and the economic geography of context, or The undefinable tacitness of being (there). Journal of Economic Geography, 3, 75-99, 2003. [Gill, 2000] J. Gill. The Tacit Mode: Michael Polanyi’s Postmodern Philosophy. SUNY University Press, 2000. [Gille, 1986] B. Gille. Essay in Technological Knowledge. In The History of Techniques, B. Gille, ed., pp. 3-30. The Montreux Press, 1986. [Hacker, 1997] W. Hacker. Improving engineering design — contributions of cognitive ergonomics. Ergonomics, 40, 1088-1096, 1997. [Harris, 1988] J. R. Harris. The Diffusion of English Metallurgical Methods to Eighteenth Century France. French History, 2, 22-44, 1988. [Henderson, 1995] K. Henderson. The Political Career of a Prototype: Visual Representation in Design Engineering. Social Problems, 42, 274-299, 1995. [Henderson, 1998] K. Henderson. The Role of Material Objects in the Design Process: A Comparison of Two Design Cultures and How They Contend with Automation. Science, Technology and Human Values, 23, 139-174, 1998. [Henderson, 1997] K. Henderson. On Line and On Paper: Visual Representations, Visual Culture and Computer Graphics in Design Engineering. MIT Press, 1997. [Ho, 2001] C.-H. Ho. Some phenomena of problem decomposition strategy for design thinking: differences between experts and novices. Design Studies, 22, 27-45, 2001. [Howells, 1996] J. Howells. Tacit Knowledge. Technology Analysis & Strategic Management, 8, 91-106, 1996. [Howells, 2002] J. Howells. Tacit Knowledge, Innovation and Economic Geography. Urban Studies, 39, 5-6, 871-84, 2002. [Hughes, 2004] T. P. Hughes. Human Built World. University of Chicago Press, 2004. [Hutchins, 1995] E. Hutchins. Cognition in the Wild. MIT Press, 1995. [Jagodzinski et al., 2000] R. Jagodzinski, F. Reid, and P. Culverhouse. Enthnographic Approaches to the Study of Design. Design Studies, 21, 4, 333-355, 2000. [Kogut and Zander, 1992] B. Kogut and U. Zander. Knowledge of the Firm, Combinative Capabilities, and the Replication of Technology. Organization Science, 3, 383-397, 1992. [Kroes and Meijers, 2006] P. Kroes and A. Meijers. The Dual Nature of Technological Artefacts. Studies in the History and Philosophy of Science, 37, 1–4, 2006. [Lakoff, 1987] G. Lakoff. Women, Fire and Dangerous Things. University of Chicago Press, 1987. [Lam, 2000] A. Lam. Tacit Knowledge, Organizational Learning and Societal Institutions: An Integrated Framework. Organization Studies, 21, 487-513, 2000. [Lawson and Lorenz, 1999] C. Lawson and E. Lorenz. Collective Learning, Tacit Knowledge and Regional Innovative Capacity. Regional Studies, 33, 305-317, 1999. [Layton, 1974] E. T. Layton. Technology as Knowledge. Technology & Culture, 15, 1, 31-41¸1974. [Layton, 1976] E. T. Layton. American Ideologies of Science and Engineering. Technology and Culture, 17, 688-701, 1976. [Layton, 1979] E. T. Layton. Scientific Technology, 1845-1900: The hydraulic turbine and the origins of American industrial research. Technology and Culture, 20, 64-89, 1979. [Lazarus and McCleary, 1949] R. S. Lazarus and R. A. McCleary. Autonomic discrimination without awareness: An interim report. Journal of Personality, 18, 171-179, 1949. [Leonard and Sensiper, 1998] D. Leonard and S. Sensiper. The Role Of Tacit Knowledge In Group Innovation. California Management Review, 40, 3, 112-132, 1998. [Leonard-Barton, 1995] D. Leonard-Barton. Wellsprings of Knowledge: Building and Sustaining the Sources of Innovation. Harvard Business School Press, 1995. [Lewicki et al., 1992] P. Lewicki, T. Hill, and M. Czyzewska. Non-conscious Acquisition of Information. American Psychologist, 47, 796-801, 1992. [Lihlstrom, 1987] A. Lihlstrom. The Cognitive Unconscious. Science, 237, 1445, 1987.

Tacit Knowledge and Engineering Design

373

[Marcel, 1983] A. J. Marcel. Conscious and Unconscious Perception: An Approach to the Relations between Phenomenal and Perceptual Processes. Cognitive Psychology, 15, 238-300, 1983. [Maskell and Malmberg, 1999] P. Maskell and A. Malmberg. Localised learning and industrial competitiveness. Cambridge Journal of Economics, 23, 167-185, 1999. [Mayr, 1976] O. Mayr. The Science-Technology Relationship as a Historiographic Problem. Technology and Culture, 17, 4, 663-673, 1976. [Megginson, 1996] D. Megginson. Planned and emergent learning: Consequences for development. Management Learning, 27, 411–428, 1996. [Merikle, 1992] P. M. Merikle. Perception without Awareness: Critical Issues. American Psychologist, 47, 792-95, 1992. [Nelson, 1982] R. R. Nelson. The Role of Knowledge in R&D Efficiency. Quarterly Journal of Economics, 97, 453-470, 1982. [Nelson, 1991] R. R. Nelson. Why do firms differ, and how does it matter? Strategic Management Journal, 12, 61–74, 1991. [Nelson and Winter, 1982] R. R. Nelson and S. G. Winter. An Evolutionary Theory of Economic Change. Harvard Univ. Press, 1982. [Nightingale, 1998] P. Nightingale. A Cognitive Theory of Innovation. Research Policy, 27, 689709, 1998. [Nightingale, 2000] P. Nightingale. The product–process–organisation relationship in complex development projects. Research Policy, 29, 913–930, 2000. [Nightingale, 2003] P. Nightingale. If Nelson and Winter are half right about tacit knowledge, which half? Industrial and Corporate Change, 12, 149-183, 2003. [Nightingale, 2004] P. Nightingale. Technological capabilities, invisible infrastructure and the un-social construction of predictability: the overlooked fixed costs of useful research. Research Policy, 33, 1259–1284, 2004. [Nisbett and Wilson, 1977] R. E. Nisbett and T. D. Wilson. Telling more than we can know: Verbal reports on mental processes. Psychology Review, 84, 231–259, 1977. [Nye, 2006] D. L. Nye. Technology Matters: Questions to Live With. MIT Press, 2006. [Parnas, 1999] D. L. Parnas. Software engineering programmes are not computer science programmes. IEEE Software, 16, 19-30, 1999. [Pavitt, 1984] K. Pavitt. Sectoral patterns of technological change: towards a taxonomy and a theory. Research Policy, 13, 343–374, 1984. [Pavitt, 1986] K. Pavitt. International Patterns of Technological Accumulation. In Strategies in Global Competition, N. Hood ed., Wiley, 126-157, 1986. [Pavitt, 1996] K. Pavitt. National policies for technical change: where are the increasing returns to economic research? Proceedings of the National Academy of Science USA, 93, 12693–12700, 1996. [Pavitt, 1998] K. Pavitt. The social shaping of the national science base. Research Policy, 27, 793–805, 1998. [Polanyi, 1958] M. Polanyi. Personal Knowledge: Towards a Post-Critical Philosophy, Harper Torchbooks, 1958. [Polanyi, 1962] M. Polanyi. Tacit Knowing and Its Bearing on Some Problems of Philosophy. Reviews of Modern Physics, 34, 601-16, 1962. [Polanyi, 1965] M. Polanyi. The Structure of Consciousness. Brain, 68, 799-810, 1965. [Polanyi, 1966a] M. Polanyi. The Logic of Tacit Inference. Philosophy, 41, 1-18, 1966. [Polanyi, 1966b] M. Polanyi. The Creative Imagination. Chemical and Engineering News, 44, 85-93, 1966. [Polanyi, 1967] M. Polanyi. The Tacit Dimension. Doubleday, 1967. [Polanyi, 1968] M. Polanyi. Life’s Irreducible Structure. Science, 160, 1308-12, 1968. [Polanyi, 1969] M. Polanyi. Knowing and Being. Routledge, 1969. [Posner, 1994] M. I. Posner. Attention: The Mechanism of Consciousness. Proceedings of the National Academy of Sciences of the United States of America, 91, 7398-403, 1994. [Pylyshyn, 1981] Z. W. Pylyshyn. The imagery debate: Analogue media versus tacit knowledge. Psychological Review, 88, 16-45, 1981. [Reber, 1993] A. S. Reber. Implicit Learning and Tacit Knowledge: An Essay on the Cognitive Unconscious. Oxford University Press, 1993. [Reber, 1989] A. S. Reber. Implicit learning and tacit knowledge. Journal of Experimental Psychology: General, 118, 219–235, 1989.

374

Paul Nightingale

[Rogers, 1983] G. F. C. Rogers. The Nature of Engineering: A Philosophy of Technology. Palgrave Macmillan, 1983. [Rothwell and Gardiner, 1988] R. Rothwell and P. Gardiner. Re-Innovation and Robust Designs: Producer and User Benefits. Journal of Marketing Management, 3, 372-389, 1988. [Rothwell, 1992] R. Rothwell. Successful Industrial Innovation: Critical Success Factors for the 1990 s. R&D Management, 22, 221-239, 1992. [Rosenberg, 1998] N. Rosenberg. Technological change in chemicals: the role of universityindustry relations. In Chemicals and Long-term Economic Growth, Arora, A., Landau, R., and Rosenberg, N. eds., pp. 193-230. Wiley, 1998. [Scerri and McIntyre, 1997] E. R. Scerri and L. McIntyre. The case for the philosophy of chemistry. Synthese 111, 213-232, 1997. [Schacter, 1992] D. L. Schacter. Implicit Knowledge: New Perspectives on Unconscious Processes. Proceedings of the National Academy of Sciences of the United States of America, 89, 11113-17, 1992. [Sch¨ on, 1982] D. A. Sch¨ on. The Reflective Practitioner, Basic Books,1982. [Searle, 1995] J. R. Searle. The Construction of Social Reality. Allen Lane, 1995. [Searle, 2001] J. R. Searle. Rationality in Action. MIT Press, 2001. [Simon, 1969] H. A. Simon. The Sciences of the Artificial. MIT Press, 1969. [Spender, 1995] J.-C. Spender. Organizational knowledge, learning and memory: Three concepts in search of a theory. Journal of Organizational Change Management, 9, 63-78, 1995. [Spender, 1998] J.-C. Spender. The dynamics of individual and organizational knowledge. In Managerial and Organizational Cognition. C. Eden, and J.-C. Spender eds., pp. 13–29. Sage, 1998. [Stapleton et al., 2005] L. Stapleton, D. Smith, and F. Murphy. Systems engineering methodologies, tacit knowledge and communities of practice. AI & Society, 19, 159-179, 2005. [Sternberg, 1986] R. J. Sternberg. Intelligence Applied, Harcourt, 1986. [Tang and Leifer, 1988] J. C. Tang and L. J. Leifer. A framework for understanding the workspace activity of design teams. Proceedings of the 1988 ACM conference on Computersupported cooperative work CSCW, 1988. [Teece, 2000] D. J. Teece. Strategies for Managing Knowledge Assets: the Role of Firm Structure and Industrial Context. Long Range Planning, 33, 35-54, 2000. [Tononi and Edelman, 1998] G. Tononi and G. Edelman. Consciousness and Complexity. Science, 282, 1846-51, 1998. [Tsoukas, 2003] H. Tsoukas. Do we really understand tacit knowledge? In The Blackwell Handbook of Organizational Learning and Knowledge Management. Easterby-Smith and Lyles eds., pp. 411-427. Blackwell Publishing, 2003. [Tsoukas, 1996] H. Tsoukas. The Firm as a Distributed Knowledge System: A Constructionist Approach. Strategic Management Journal, 17, 11-25, 1996. [Underwood, 1996] G. Underwood. Implicit cognition. Oxford University Press, 1996. [Vermaas and Houkes, 2003] P. E. Vermaas and W. Houkes. Ascribing functions to technical artefacts: A challenge to etiological accounts of function. British Journal of the Philosophy of Science, 54, 261–289, 2003. [Vermaas and Houkes, 2006] P. E. Vermaas and W. Houkes. Technical functions: a drawbridge between the intentional and structural natures of technical artefacts. Studies in the History and Philosophy of Science, 37, 37-41, 2006. [Veshosky, 1998] D. Veshosky. Managing innovation information in engineering and construction firms. Journal of Management in Engineering 14, 58-66, 1998. [Vincenti, 1990] W. Vincenti. What Engineers Know and How They Know It, Johns Hopkins Press, 1990.

PRACTICAL REASONING AND ENGINEERING Jesse Hughes

1

INTRODUCTION

Engineering is to science as practical reasoning is to theoretical reasoning. This is a crude analogy, but it has the kernel of truth to it. Roughly, science aims to learn facts about the world around us, while engineering aims to change that world in order to further our aims and goals. Science is successful if it generates true beliefs — in other words if it is successful as a method of theoretical reasoning — and engineering is successful to the extent that it provides means to satisfy our needs and desires, i.e. if it is successful as a method of practical reasoning. Philosophy of science has accordingly adapted epistemology to its analysis of science. If we want to analyze scientific practice, we must understand it in terms of its ultimate goal: the discovery of facts about the world around us. The scientific method is a prescription for reliable judgments about scientific hypotheses and so it is natural to analyze and evaluate this method, both in its general form and in particular applications, in epistemological terms. Similarly, engineering is a method for creating practical devices and processes. Indeed, a recent text [Dym and Little, 2004] offers the following definition: “Engineering design is the systematic, intelligent generation and evaluation of specifications for artifacts whose form and function achieve stated objectives and satisfy specified constraints.” The engineering method — that is, the design process1 — has an explicit practical end. And it makes sense to understand the process, both in its general terms and in specific applications, in terms of practical philosophy. To put it crudely, the design process is analogous to the scientific method and an engineering need — the problem to be solved — is analogous to a scientific hypothesis. The scientific method works to the extent that it generates correct judgments on individual (scientific) hypotheses and the design process works to the extent that it solves the problem at hand.2 1 It

is a bit naive to refer to the design process, since each author offers his own version of it. Similar concerns could be raised about the scientific method, but let us pass over these issues. Whatever design process we have in mind, the aim is the same: to produce useful artifacts or processes. 2 The design process also works if it correctly judges that the problem at hand is unfeasible, either in principle or with current technology. Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

376

Jesse Hughes

In particular, instrumentalist terminology fits remarkably well with engineering, on at least three levels. First, the design process is the method engineers use in order to create artifacts (or processes, methods, etc.). This process is divided into steps, each of which aims at a particular goal through the use of various means. This is quite explicit, for example, in [Dym and Little, 2004]: “Each phase requires an input, has design tasks that must be performed, and produces an output or product.” As they describe each phase in detail, they list the “means and methods” used to produce the desired output. Thus, the design process itself is presented in steps, each of which is described in terminology familiar to an instrumentalist. Second, the tools engineers use in design process (that is, the methods in Dym and Little’s terminology) are often instrumental in nature. To take one example, function-means trees are used to generate early proposals for solutions. These trees identify tasks which must be accomplished and alternative ways to achieve these tasks, i.e. it is a matter of identifying ends and sufficient means to those ends. Thus, engineers analyze their problems in instrumental terms. Finally, the final end of the design process is the creation of an artifact (typically) and artifacts are evaluated primarily in terms of their instrumental value (see [Franssen, 2006]). The artifact itself is a means to ends its user may have. If we know how to use the artifact, then we know how it may help us realize certain ends we have now or may adopt later and in this way, artifact functions induce associated means-end claims [Hughes, 2008]. Knowing that fire extinguishers are for putting out fires provides one with a practical consequence. If you happen to have a fire you’d like extinguished, you can use a fire extinguisher to do so.3 This is what artifact functions are about: they are about how and when to use the artifact and for what purpose. Artifacts are typically categorized according to function. As Karen Neander [1995] writes, “a brake is a brake in virtue of what it is supposed to do — was intended or designed to do — not in virtue of having some specific structure or disposition.” Roughly, then, artifacts are categorized according to the ends for which they are intended as means. Furthermore, artifacts — both types and tokens — are evaluated in terms of their effectiveness and reliability, that is, in terms of their efficacy as means. An artifact token that cannot do what it should do is malfunctioning. A type that does not perform as well as it should is badly designed. In both cases, our judgments regarding artifacts are in practical terms.4 In sum, engineering is a process for creating certain kinds of means. We will focus our attention on the design process rather than its end products, since engineering design is so very clearly related to practical reasoning. This is because both the design process and its tools are naturally understood in the familiar language of instrumentalism. It is also true, of course, that the end results 3 Provided, of course, that the fire is not too large or the wrong kind of fire for the extinguisher at hand. Knowing the artifact’s function also suggests when it should not be used. 4 Franssen [2006] states this point most explicitly when he argues that a token is malfunctioning just in case one has a (practical) reason not to use it.

Practical Reasoning and Engineering

377

(the artifacts created by engineers) are related to practical reasoning for the reasons noted above, but this relationship has been explored elsewhere [Franssen, 2006; Hughes, 2008]. Philosophy of engineering thus restricted is interested in what it is that engineers do (when designing artifacts), why they do it and whether we philosophers can tell them how to do it better. Practical philosophy is interested in more or less the same issues, but with particular emphasis on personal rather than professional action. That is, instrumentalist accounts typically focus on desires and their motivational efficacy, but it is not at all clear that desires are the right sort of pro-attitude for a philosophy of engineering.5 Engineering design is about accomplishing certain stable, clearly specified ends, not satisfying personal desires. So, we will have to do a bit of work projecting current practical philosophy onto philosophy of engineering, but in the end, we will see that the adjustments are not drastic and the benefits are clear: our application of practical philosophy will be natural and insightful. Practical philosophy serves two distinct purposes. In its explanatory role, the theory gives reasons why agents behave as they do, while in its normative aspect, it offers an evaluation of an agent’s behavior in terms of practical rationality. Philosophy of engineering similarly has two roles. In its so-called descriptive form, philosophy of engineering provides both descriptions and explanations for engineering practice, while in its normative role, we aim to evaluate engineering practice according to some standard of practical rationality. We will focus primarily on the explanatory role in what follows, since explanation seems a more modest and feasible goal than evaluation. In Section 2, we will give a fairly lengthy introduction to instrumentalism, the dominant theory of practical reasoning. We argue that there is a natural fit between means-end reasoning and engineering and we will provide some rudimentary development in this direction. We will discuss coherence and rationality as they apply to practical goals in Section 3. Sections 4 and 5 discuss promising alternatives to instrumentalism as they apply to engineering. Specificationism is the view that all the real work in deliberation occurs in specifying one’s vague goals more precisely, in particular in the context of other, conflicting goals. We discuss Bratman’s theory of planning in Section 5, indicating similarities it shares with models of the design process. 2

2.1

INSTRUMENTALISM

A brief primer on instrumentalism

The de facto theory of practical reasoning today is instrumentalism.6 We will also argue that instrumentalism is eminently applicable as a theory of practical 5 Doing philosophy of engineering in terms of desire would be something like doing philosophy of science with an epistemology based on hearsay and prejudice. 6 Sometimes called Humeanism, but see [Millgram, 1995] for a discussion of the aptness of this term.

378

Jesse Hughes

reasoning in engineering design. We begin with a summary of the characteristic features of instrumentalism and a discussion of their application to engineering. Following this, we will distinguish beliefs and desires in terms of direction of fit and discuss their role in the design process. In Section 2.2, we distinguish the different kinds of ends by means of a simple example and in Section 2.3, we use practical syllogisms to introduce a typology of means. The fundamental thesis of instrumentalism is that practical reasoning consists of nothing but means-end reasoning. That is, practical reasoning is fundamentally about using one’s knowledge of causal relations, constitutive relations, etc., in order to achieve one’s goals and desires — and there is nothing more to practical reasoning than this. As [Audi, 2006] describes Hume’s theory, “Roughly, desire, guided by belief is what produces action.” An agent aims to realize his desires — or at least as many as he can — and applies reason in order to do so. Put differently, practical reasoning is about selecting a course of action given certain beliefs about causal relations and certain prior desires. As Aristotle famously says [Ross, 1908], “For the end cannot be a subject of deliberation, but only the means.”7 To be sure, as a result of such deliberation, we may select new ends, namely those instrumental for (or constitutive of) our prior ends: if I want to earn a comfortable living, I may choose to do so by pursuing a business career. In order to get a good job in business, I may choose to enter a university with a well-regarded business school, and so on. Thus, my original end (to have a comfortable living) has led to the adoption of new ends (to work in business, to enter a good university, etc.). But some ends must be given at the start of this process. Thus, another essential claim of instrumentalism is that there are certain ends (the so-called final ends) which are not adopted for instrumental value but rather motivate the agent prior to deliberation. As Hume [1777/1975] (§244) writes: Ask a man why he uses exercise; he will answer, because he desires to keep his health. If you then enquire, why he desires health, he will rapidly reply, because sickness is painful. If you push your enquiries farther, and desire a reason why he hates pain, it is impossible he can ever give any. This is an ultimate end, and is never referred to any other object. . . . It is impossible there can be a progress [of reasons] in infinitum; and that thing can always be a reason why another is desired. Something must be desirable on its own account, and because of its immediate accord or agreement with human sentiment. For Aristotle [1908] (Book I, Ch. 7), there was ultimately one final end, namely happiness. This is the only end that is chosen “always for self and never for the sake of something else.” Other ends are chosen for their ability to lead to happiness. 7 Although [Millgram, 1995] disputes the claim that Hume was an instrumentalist and [Wiggins, 2001] argues that Aristotle is a specificationist, we will nonetheless take Hume and Aristotle as standard figures in instrumentalism.

Practical Reasoning and Engineering

379

Even though ends such as honor and reason are chosen for themselves, they are also chosen “also for the sake of happiness, judging that by means of them we shall be happy.” Peter Railton [1997] calls this approach, in which all action aims for some single, final end — the good — the High Brow account. Hume, by contrast, is the exemplary proponent of the Low Brow account: Some things other than the good are desired for their own worth. Satisfying such “intrinsic” desires (as [Audi, 2006] calls them) provides pleasure, but nonetheless these things are pursued for their own sake and not for the consequent pleasure. These include such desires as to punish one’s enemies and provide happiness for one’s friends [Hume, 1739]. In summary, then, the two characteristic claims of instrumentalism are: (1) all practical reasoning is means-end reasoning and (2) some ends admit no further justification. As one might expect, these claims are also the source of much controversy. In particular, many authors ([Millgram, 1997; Kolnai, 1978; Schmidtz, 2001], among others) dispute the claim that chains of ends necessarily come to a rationally unjustified “final” end. Since the argument for such final ends follows from the instrumentalist claim that all practical reasoning is ultimately means-end reasoning, such criticisms aim at the heart of instrumentalism.8 Instrumentalism and engineering By and large, however, philosophers of engineering may sidestep the controversies regarding instrumentalism. For the “pure” practical philosopher, whether practical justifications necessarily come to unjustified final ends is an essential question, but it’s not so relevant in understanding engineering practice. If we want to understand how a mechanical engineer chooses the materials for her project, say, we would certainly stop asking about her motivation before she answers in terms of personal pleasure or pain. For our purposes, a final-in-context end will suffice. We take for granted that our engineer has chosen to design, e.g., a springloaded bar mousetrap (the common mousetrap found in any hardware store). We investigate her later decisions given the project at hand.9 As Aristotle says of the professional classes, “They assume the end and consider how and by what means it is to be attained” [Ross, 1908](Book III, Ch. 3). Similarly, we may avoid the difficult question of whether all practical reasoning is means-end reasoning. We are not interested in all practical reasoning, but rather practical reasoning of a narrow sort: the kind that engineers use in designing 8 Schmidtz is an exception here, since he works in an instrumental setting while denying the necessity of unjustified final ends. 9 Of course, one may be interested in how the project at hand was selected as well. For instance, many issues in engineering ethics are concerned with the morality of designing certain kinds of artifacts. Regardless, the point is that our enquiry into motive will be limited and so we will not fall into Hume’s infinite descent.

380

Jesse Hughes

their products. And it seems undeniable that at least some practical reasoning is means-end reasoning. In fact, [Millgram, 2001a] calls means-end reasoning the “least controversial form” of practical reasoning: at least sometimes, we reason about how to achieve some goal we have. So let us skip broad questions regarding whether this least controversial form is the only form and instead see whether an analysis of engineering design in terms of means-end reasoning bears fruit.10 It is tempting to suggest that means-end reasoning plays the same role in understanding engineering as it does for understanding any occupation. According to Aristotle, every profession assumes an end and deliberates about the best means to that end. Nonetheless, we are not arguing that instrumentalism is the best way to understand what financial consultants, poets or economists do. But engineering practice is inextricably bound with the practical methods of engineering, including the design process. Indeed, the point of studying engineering design is to acquire the skills to reliably deliberate about what should be done and to select the design that will best satisfy our needs. To be sure, an engineer’s education includes other topics. A bulk of the curriculum is devoted to learning engineering science, but this is understandable in practical terms as well: reliable deliberation requires true beliefs as well as a clearly stated end to be accomplished, and engineering science is aimed at the former. In summary, the various design methods are methods of practical deliberation and engineering is an inherently practical discipline. And since it is natural to interpret both the process and end products of engineering in terms of means-end relations, instrumentalism seems an appropriate first consideration for a theory of practical reasoning in engineering. Beliefs and desires in instrumentalism Instrumentalism is appealing for its emphasis on basic folk psychology that most of us accept in our daily reasoning. In seeking explanations for actions, it is usual to refer to an agent’s beliefs and desires at the time. She chooses to fetch a key, say, because she wants to unlock the door and she believes that the key provides a means to do so. These two kinds of mental states play different roles in our explanation: she chooses her action in order to realize her end, but her choice is informed by her beliefs. The most popular current analysis of the difference between beliefs and desires is in terms of direction of fit, due to [Anscombe, 1989]. A belief aims at being true, and if one’s beliefs are not true, then they should be changed to “fit” the world. False beliefs are discarded and beliefs which more closely fit the world are adopted. A desire, on the other hand, is not discarded just because it does not fit the world. Rather, if one desires ϕ, one wants to change the world to make ϕ true. As [Platts, 1979] says, “beliefs should be changed to fit with the world, not vice versa. Desires aim at realization and their realization is the world fitting with them.” 10 One may nonetheless ask a narrower form of the instrumentalist question: Is means-end reasoning the only kind of practical reasoning that engineers use?

Practical Reasoning and Engineering

381

This intriguing metaphor is considered in some detail by [Smith, 1987], [Humberstone, 1992] and elsewhere. Humberstone gives a useful clarification of direction of fit in terms of conditional intentions. It is constitutive of beliefs, he says, that one intends not to believe ϕ given that ϕ is false. Symbolically, Intend(¬Bϕ | ¬ϕ),

(1)

but note that the conditional intention operator cannot be contraposed on this account. That is, (1) is not equivalent to Intend(ϕ | Bϕ). The latter expresses that I intend that ϕ is true, given that I believe it, but this is not a feature of belief. Rather, it is a feature of desires: I intend ϕ given that I want that ϕ. Consequently, Humberstone says that the following condition is constitutive of desire: Intend(ϕ | W ϕ). (2) Because science is a theoretical endeavor and engineering a practical one, there is a loose analogy of directions of fit for the science/engineering distinction. Put crudely, science aims to learn about the world and hence takes the world as input to produce a system of beliefs. Engineering aims to manipulate the world, and hence takes ends as input to make changes in our circumstances. As [Simon, 1969] says, “The natural sciences are concerned with how things are. . . Design, on the other hand, is concerned with how things ought to be, with devising artifacts to achieve goals.” The dictionary definition of engineering also reflects this picture: engineering: . . . the art and science by which the properties of matter are made useful to man, whether in structures, machines, chemical substances, or living organisms. [Webster, 1913] This rough adaptation of direction of fit is intended as metaphor, of course, and we have used a fairly naive view of both science and engineering. It is nonetheless suggestive of the crux of the theoretical/practical divide. How, then, does the belief/desire terminology fit into an analysis of engineering? The role of beliefs in engineering is reasonably clear: the beliefs that an engineer brings to his deliberations include knowledge of engineering science (including material science, physics and other causal laws and particular results deduced from relevant scientific theories) as well as beliefs regarding the desires of his client or potential consumers. Desires, on the other hand, are a personal attribute, a feature of one’s psychological state, and that is an awkward level on which to analyze professional performance. We say that the engineer engaged in building a better mousetrap is directed towards an end or goal, rather than that she is driven by her desires for a better mousetrap. Some authors (e.g. [Audi, 2006; Millgram, 1997]) use the term “desire” broadly enough to include the pursuit of ends, while others (e.g. [Richardson, 1994]) take desires to involve weaker commitments, so that “a desire is a particular psychological state with motivational efficacy” while an end “is something for the sake of which an action is to be done.” To put it differently, “although one simply has

382

Jesse Hughes

desires, one adopts or pursues ends.”11 Thus, while it may be rational to desire something while being unwilling to do what is necessary to attain it, one cannot have an end without doing what is necessary to realize it. This distinction is plausible and useful in the traditional domain of practical philosophy, but Richardson’s notion of desire will play little role in our understanding of engineering. Thus, we will follow Robert Audi’s lead and use the term desire to indicate any sort of want — including the pursuit of an end. When we use the term in an engineering context, we will generally interpret “desire” as the pursuit of a relevant engineering goal.

2.2

A crash course in applied instrumentalism

In this section, we will introduce a basic typology of ends by offering an account of mousetrap design. Our engineer has decided to design a springloaded bar mousetrap. This is, broadly, the end of her current deliberation, but there are many related (and conflicting) goals. She would like the mousetrap to be durable, reliable, inexpensive and safe. She may not yet have a clear idea how these objectives conflict with one another or how she ought to strike a balance, but she seeks to realize her end in a way that satisfies each of these qualities as far as possible.12 She begins the process now by identifying certain steps which advance her final end. For instance, to design a standard springloaded bar mousetrap, among other goals, one must choose the materials to be used and design the spring mechanism. In order to design the spring mechanism, our engineer must decide how hard the spring should strike the mouse. It should strike hard enough to kill the mouse (i.e., it should be reliable) but not so hard that it poses a danger to the person setting the trap (i.e., it should be safe). It should also at least give the appearance of humanely killing the mouse. Our engineer must determine, then, what the optimal force is for her needs. Assuming that she does not know this at the start, this will require investigation.13 Thus, she acquires a new goal: investigate to find the proper force for the springloaded bar. To do this, she may decide that a literature search is appropriate, and so she adopts another end: consult the literature. She continues in this manner until she reaches, at last, a task that she can perform at will — one which needs no further preparation. Thus, from her initial end of designing the mousetrap, our engineer has found means that advance her end and come to acquire at least some of these means as new ends to pursue. The situation is represented in Figure 1. An arrow from one task to another indicates that the former is done in order to advance the latter, 11 The

emphasis in this and all other quotes appear in the originals. specificationist terms, her end is currently only vaguely identified and will become more sharply defined as she proceeds. See Section 4 for an overview of this alternative theory of practical reasoning. 13 Let us hope that research into published studies and maybe some calculations are sufficient, so that she does not have to experiment with live rodents and vulnerable human fingers! 12 In

Practical Reasoning and Engineering

383

Design a mousetrap 7 dII II ppp p II p p II c c ppp II p p II p pp Choose materials

Design spring mechanism O i

Investigate proper force

Figure 1. Designing a mousetrap generates new ends. i.e. is a means to the latter. The topmost task (“Design a mousetrap”) has no outgoing arrows and is final in this context. The bottommost task (“Investigate proper force”) is instrumental to the task of designing the spring mechanism. It is neither necessary nor sufficient, but investigation will promote the design of the mechanism: our engineer will achieve better results by performing this task than otherwise. In contrast, neither of the tasks on the second level are instrumental. One does not choose the materials in order to facilitate the end of designing the mousetrap. On the contrary, these tasks are part of that goal. They are constitutive means,14 as in Ackrill’s putting example [1980]: “One does not putt in order to play golf as one buys a club in order to play golf. . . It will be ‘because’ you wanted to play golf that you are putting.” A game without putting is not a golf game. A means is constitutive just in case it is a constituent of or an ingredient in its goal (ibid.). Just as an instrumental means can be necessary or sufficient (or neither), so too with constitutive means. One cannot design a springloaded-bar mousetrap without designing the spring mechanism, and hence it is necessary to our task. Clearly, it is not sufficient, since one must also (among other things) choose materials. Going for a jog15 is a sufficient constitutive means for exercising. By jogging, one has exercised; there is nothing more to do. Castling while playing chess is an example of a constitutive end which is neither necessary nor sufficient. We will discuss sufficiency and necessity in more detail in the next section.

14 See 15 An

also [Schmidtz, 2001] for a discussion of the constitutive/instrumental distinction. example by [Schmidtz, 2001].

384

Jesse Hughes

Adapting the notation from [Schmidtz, 2001],16 we have labeled the arrows in Figure 1 with “i” if the source is instrumental for the target and “c” if it is constitutive. These are the two primary kinds of means generated from more final ends. Each means serves a dual role as an end for further means, until we reach an action which can be performed straightaway (say, reaching for a book in order to pursue our investigation). When an instrumental (constitutive, resp.) means is adopted as an end, then we also refer to it as an instrumental (constitutive, resp.) end. Thus, we have the following three kinds of end: • Final (in context): The ultimate end of the project underway (e.g., designing a mousetrap). • Instrumental: An end which may be pursued because it advances another (more final) end (e.g. investigating the proper force needed). • Constitutive: An end which may be pursued because it is a constituent of or ingredient in another (more final) end (e.g. designing the spring mechanism). This will be our crude model of an engineer at work, then. But means-end reasoning is relevant to engineering in another way: for an artifact to work properly, its components must serve as means to the functional end. For instance, a television remote must deliver signals to the television set. There are various ways this could be done, including infrared or radio transmissions or through a wire. An engineer designs an artifact by selecting appropriate means to perform these various functions. See the discussion of function-means trees in [Dym and Little, 2004] for the importance of this kind of analysis. Maieutic ends in engineering There are various complications to this basic model, but perhaps the most interesting for our purposes is found in [Schmidtz, 2001]. Schmidtz accepts the basic instrumentalist assumption — practical reasoning is about means to ends — while arguing that some final ends are nonetheless rationally chosen. There is, says Schmidtz, a fourth kind of end, namely a maieutic end — a goal of coming to acquire new final ends. To take Schmidtz’s example, as Kate nears young adulthood, she may feel a need to settle on a career path. Once Kate settles on a career in medicine, her maieutic end has been realized while her career goal is now pursued for its own sake. It is not instrumental to the maieutic end, since that goal was to choose a career and pursuing a medical career is not a means to choosing a career. Neither is it constitutive, for the same reason: pursuing and choosing are different activities. Notice that a maieutic end, in turn, could be either final, instrumental or constitutive. It may be that Kate feels pressure to choose a career path just to satisfy 16 This article is a shorter adaptation of [Schmidtz, 1994], omitting many examples of reasoning with maieutic ends, but the later version includes the diagrams that motivate our discussion here.

Practical Reasoning and Engineering

385

her parents and school counselors. In that case, she has adopted a maieutic end for its instrumental value. Schmidtz’s interest here is to show that certain ends — those mandated by maieutic ends — are both final and rational. If adopting an end satisfied a maieutic end had at the time, then it is rational. This is true even if the end was not consciously chosen for this purpose. Perhaps Kate simply found herself drawn to a medical career independently of any felt need to choose a career at all. Nonetheless, the adoption of this career goal is rational, since she had a need to choose a career. Indeed, Schmidtz argues that one may reach a state in which every final end one possesses is rational in this sense: chosen to satisfy a maieutic goal she had at the time. Do maieutic ends occur naturally in engineering design? Arguably, they occur regularly, at the start of the design process. Each year, automobile manufacturers find ways to change their offerings so that the 2008 GMW Tranquestor attracts attention that is waning for the 2007 model. Similar “progress” is evident in software. New versions come out with new features (and occasional bug fixes) in order to entice new purchases. Each of these projects begins with a basic question: How shall we update our product to encourage sales? Thus, software engineers are faced with a maieutic goal: finding something to do to make Turbo Word++ 2008 better than Turbo Word++ 3.0. And once the changes are selected, they get to work on the new product.

GF

@A

76 /01 Existing knowledge 54 23

GF

?> /89

State of the art

=< :;

?> 89

 Scientific curiosity

=< :;

 /. ()Identification of need-, *+

76 01

 Hypothesis

54 23

 76 Conceptualization 23 54 01

?> 89

 Logical analysis

=< :;

 76 Feasability analysis 23 54 01

/. ()

 Proof BC

-, *+

@A

/. ()

 Production BC

-, *+

Figure 2. Hill’s comparison of the scientific and engineering methods.

386

Jesse Hughes

Indeed, the process of engineering, especially in its entrepreneurial forms, begins with broad maieutic goals: find a need and satisfy it. This is how software engineering projects, both proprietary and free, often start: with a search for a problem to be solved. For commercial efforts, these maieutic goals are motivated at least in part by the desire for monetary reward, but some commercial (and perhaps most free) software developers also desire projects for the intellectual challenge and the joy of creation. The Linux operating system famously began as a hobby for then graduate student Linus Torvalds [Torvalds, Linux, 1991]. Philosophers of engineering, however, start their investigation after a need has been found. They follow the lead of engineering educators in this respect, since design models also assume at the start that a need has been identified, at least in rough form. For instance, [Dym and Little, 2004] discuss four descriptive and one prescriptive model of the design process, but each of these take for granted that a need has been identified. Similarly, Dieter’s [1983] simpler model of the design process begins with the recognition of a need. An exception is Hill’s [1970] comparison of the scientific and engineering methods (Figure 2, adapted from [Dieter, 1983]). This cyclical model includes an “identification of needs” step. At least for Hill, identification is something that happens during engineering rather than prior to it. One wonders whether there is a “recipe for rational choice” (as in [Schmidtz, 2001]) with an engineering focus to apply to the mostly overlooked task of identifying a need.17

2.3 Syllogisms and means In the previous section, we distinguished different kinds of ends according to how they contribute to more final ends. In this section, we introduce a classification of means in terms of their motivational efficacy. We will do so by examining the roles of means in practical syllogisms. Means-end reasoning seems particularly well-suited to a syllogistic presentation,18 since it is fairly uniform in its structure. An instance of means-end reasoning typically includes the following features: 1. a statement that a particular condition is desired or aimed for (the major premise), 2. a proposition or belief, either about an instrumental or constitutive means to the desired position (the minor premise), 3. an intention or pro-attitude towards an action or an action itself (the conclusion). This is the way in which means-end reasoning combines beliefs and pro-attitudes in order to generate practical consequences. 17 Such inquiries have analogues in philosophy of science, namely in the logic of scientific discovery (i.e. hypothesis generation). 18 This point is not without controversy. See, e.g., [Richardson, 1994].

Practical Reasoning and Engineering

387

Roughly, then, a practical syllogism has the form: A wants to realize ϕ. α is a means to ϕ. Therefore A has reason to do α.

(3)

Variations on this form are due to the exact relation of the means to its end. Of course, a means may be either instrumental or constitutive, but this is not the distinction we have in mind here. Rather, we are primarily interested in whether the means is necessary, sufficient, optimal or none of the above.19 A means α to an end ϕ is necessary if ϕ cannot be realized without doing α. In this case, the conclusion of (3) should be in the strongest normative terms: in this situation, A has overwhelming reason to do or must do α. See, for instance, von Wright’s [1963] example of a valid syllogism: A wants to make the hut habitable. Unless A heats the hut, it will not become habitable. Therefore A must heat the hut.

(4)

Von Wright expresses the normative force of the conclusion in very strong terms: “if action does not follow, we should have to describe the subject’s case by saying either that he did not in fact want his professed object of desire or did not, after all, think it necessary to do the act in order to get the wanted thing.” A cannot recognize the truth of the premises without thereafter heating the hut. We must take care with this conclusion on two points, however.20 First, it is always possible that I did want ϕ until I realized that I must do α. Once I realized that α is necessary, I may change my mind about wanting ϕ after all — at least assuming that my desire for ϕ is the sort that I can change my mind about. This is analogous to the situation with theoretical syllogisms. When I realize that my beliefs in A and A → B require that I accept B (which I find plainly false), I may adjust my beliefs about A or A → B. The same adjustment may occur in any practical syllogism which gives strong reason to perform an action I am unwilling to perform. It would be irrational, of course, to decide that, because I don’t want to do α, I will refuse to believe that it is necessary, but I have the freedom to revise (at least some of) my desires and goals. Audi [2006] raises the second issue regarding the strong consequences for this kind of syllogism, namely that the conclusion is not valid without some further assumptions on the means α. The fact that ϕ won’t be realized unless we do α is not enough to motivate me to do α. I should also be convinced that, having done α, there is a real chance to bring about ϕ. I am unable to realize my goal of 19 These

distinctions are suggested in [Audi, 2006]. [1994] criticizes the syllogisms on other grounds. The term “unless” must be interpreted in the logical sense or else the syllogism is not valid, but it is extremely rare that one has sufficient evidence that a means is necessary in this strong sense. Thus, the syllogism is valid only if it is almost never applicable. 20 Richardson

388

Jesse Hughes

drinking the entire Atlantic Ocean through a straw unless I first acquire a straw, but that should not give me an overriding reason to acquire a straw.21 My goal ought to be attainable and doing α should put me in a reasonable position for achieving ϕ before I accept von Wright’s syllogism.22 But necessary means are the easy case. If one wants to accomplish his goal, he has little choice but to do what is necessary. In other cases, the conclusion is rather more vague. A typical example regarding sufficient means is found in [Railton, 1997]23 :

ϕ is an end of mine. Means α would secure ϕ.

(5)

There is that much to be said deliberatively in favour of my doing α, or against my having ϕ. In this case, we assume that doing α will bring about ϕ, i.e. that α is a sufficient means to ϕ. This is a fairly good reason to do α: it would realize the goal under consideration. Of course, I may have many other options which would bring about ϕ, so this reason is not overwhelming. In fact, I might choose a means which is not sufficient, but which is nonetheless preferable for other reasons. I want to avoid illnesses found in contaminated meats. I can be sure to do so by eliminating meat from my diet. Any other alternative is less reliable, but nonetheless I choose a less onerous alternative, such as exercising care in the selection and preparation of my meats and choosing reputable, clean restaurants. These other means (which are not guaranteed to secure my aim) are preferable to the sufficient means of avoiding meat. This brings us to optimal means. I have decided against the sufficient means (stop eating meat) because it is not as good in my estimation as an alternative. For various reasons, I prefer to keep eating meat but to change my habits to decrease risk. Let us suppose that I have some criteria in mind so that I can compare two actions and evaluate whether one is better or preferable to another. This criteria may include effectiveness, reliability, cost in some broad sense and coherence with other ends I may hold. In case a means α is superior to all others, then it seems plausible that I should do α. A syllogism to this effect is found in [Churchland, 1970]:24

21 Perhaps, of course, I should give up my desire to quaff an ocean, but not for the reasons discussed above. Getting a straw is not onerous to me and presumably neither is the act of drinking the ocean. The problem is that my end is irrational, but an instrumentalist cannot complain about irrationality unless I have chosen to drink the ocean as a means to some other end. For final ends, we want what we want. 22 See also [Hughes et al., 2007] for a semantics of necessary means that includes attainability. 23 Variables changed to match mine. 24 Adapted here from [Audi, 2006], substituting my variable names.

Practical Reasoning and Engineering

I want ϕ. Doing α is a way for me to bring about ϕ under these circumstances. There is no other way to bring about ϕ now which is preferable to me as, or more preferable to me than, doing α.

389

(6)

There is no sufficient reason for me not to bring about ϕ under these circumstances. Therefore, let me do α. The third premise, missing from our previous examples, is an important consideration. As Audi [2006] points out, one should not choose an optimal means to an end ϕ if there is an overriding end ψ in conflict with ϕ. Such an overriding end would provide sufficient reason not to do α and hence Churchland’s syllogism would not apply. We often reason about means which are rather weaker than any of these examples. We may be faced with a situation in which no means is clearly optimal, necessary or sufficient. In these cases, we generate weaker reasons to act: each means which advances our end produces a motivational (and perhaps normative) reason to act accordingly.25 It is difficult to spell out clearly how one measures the strength of each of these reasons and, indeed, whether practical reasoning is really captured by an internal calculation based on comparisons of reasons’ strength. Let us pass by these more complicated arguments at present and examine practical syllogisms in engineering. Certainly, engineers deal with each of the different kinds of means discussed above. The function-means trees found in [Dym and Little, 2004] are a tool for enumerating alternative sufficient means to the project goal, for instance. Physical restrictions create certain necessary means: any internal combustion engine must include a device that mixes oxygen and fuel. And engineering design books are chock full of different measurements of optimality as well as methods for calculating the optimal outcome. As [Dieter, 1983] puts it, “in engineering design we have a situation in which there is a search for the best answer. In other words, optimization is inherent in the design process.” If something like the above syllogisms are correct models of human reasoning, it follows that engineers implicitly use syllogistic reasoning in their deliberations. But these syllogisms apply only when the real intellectual work has been done. As Richardson [1994] says, “Once it has endorsed an action as both the best means to an end and not productive of too much countervailing bad, then deliberation has largely done.” Indeed, [Velleman, 1996] suggests that such syllogisms cannot 25 In such situations, we may rely on Simon’s “satisficing”: a solution which is good enough is good enough — especially in cases where the costs of searching for a better solution are significant and the likelihood of failure non-negligible. See Chapter Five on The Science of Design in [Simon, 1969].

390

Jesse Hughes

Design objectives

Weight (%)

Polythelene bottle with twist cap

Mylar sack with straw

Eval.

Score

Eval.

Score

Environmentally benign

33

0.9

29.7%

0.1

3.3%

Easy to distribute

09

0.5

4.5%

0.6

5.4%

Preserves taste

22

0.9

19.8%

1.0

22%

Appeals to parents

18

0.8

14.4%

0.5

9.0%

Permits marketing flexibility

04

0.5

2.0%

0.5

2.0%

Generates brand identity

13

0.2

2.6%

1.0

13%

Totals

99

73.0%

54.7%

Table 1. A numerical evaluation matrix adapted from Dym and Little. count as practical reasoning at all, since they explain reflexive behaviors as well as “full-blooded actions”. Nonetheless, these syllogisms provide a good account of the motivational content of the different kinds of means encountered in engineering and other practical deliberations. Furthermore, we can explain the role of various evaluative tools used by engineers. Recall the difficulty in interpreting the consequence of the sufficient means syllogism (5). If α is sufficient, then we have some reason to do α, but this conclusion gives no hint whether it is a strong or weak reason or how one should choose from among the many different sufficient means available. What one wants is a way to evaluate these different options, so that one may be chosen as superior to the others, that is, one wants to be able to apply reasoning like that of the optimal means syllogism (6). And this is precisely what engineers are taught to find. When the design process suggests several different options, designers find a standard of evaluation by which to compare them so that they can, as [Dym and Little, 2004] say, “pick a winner”.26 The authors suggest three methods for selecting the best design. A typical example of such methods, the numerical evaluation matrices method (Table 1) rates the alternatives as follows: 1. Assign weights wo to each of the objectives o ∈ O (low cost, environmentally benign, appealing, etc.) identified earlier in the design process. The weights indicate importance.27 26 Interestingly, the standard of optimality is chosen after the set of alternative means has been identified! 27 Dym and Little give some advice [2004] on how these weights are chosen. These methods for minimizing subjectivity deserve some examination themselves, but we will not do so here.

Practical Reasoning and Engineering

391

2. Evaluate each alternative design d with respect to each objective, assigning a value vod indicating the degree to which the design meets the objective.  3. For each alternative d, calculate o∈O vod · wo . This is the value v d of the design. The best alternative is that design d with the highest value v d . These methods aim at removing subjective judgments from the selection process, at least as far as possible. Engineers must decide the importance of each of the previously identified objectives and the degree to which each design meets these objectives. Note that these tasks are not deliberation, in the sense of practical reasoning. The engineers are not choosing what to do (nor what ends to pursue) when they rank the objectives or evaluate the designs. In the first case, they are ascertaining the degree to which the objectives are relevant for the problem being solved and in the second, they are judging (hypothetical) performance with respect to well-defined criteria. Both tasks are strictly in the realm of theoretical reasoning. The evaluation techniques in standard design theory, then, give one tools to transform a decision involving several competing sufficient means to a situation in which one means is optimal. This practice is motivated by basic instrumental reasoning. A means-end analysis thus serves to explain this method. Practical philosophy thus provides an insight into this piece of what engineers do and why they do it. 3 COHERENCE We turn now from a purely instrumentalist account to consider three alternative views of practical coherence and rationality. The first is a familiar and welldeveloped topic, decision theory. The second, due to [Edvardsson and Hansson, 2005], proposes four criteria a goal must meet in order to count as rational. The final account is offered by [Millgram and Thagard, 1996], and evaluates the coherence between means and end in terms of facilitation. We will examine each of these theories with attention to its relevance to engineering. The numerical evaluation matrices described at the end of the last section resemble decision theory, an influential theory of practical reasoning. Decision theory begins by assuming that an agent has a utility function u, mapping circumstances to real numbers. The function encodes the degree to which the agent desires to be in that circumstance. I prefer a hot bath to being roasted alive, so my utility function assigns a rather higher value to the hot bath than to the painful ordeal at the stake. In order to determine which action is the best course to take, I determine the probabilities of the outcomes of the various actions and take a weighted average of the utilities of these outcomes.28 This calculates the expected utility of each action and I should perform that action with the highest expected utility. 28 Since decision theory is not our primary interest here, I will describe only the simplest version. I will ignore the much more usual case, in which we do not know these probabilities with any specificity.

392

Jesse Hughes

One of the charms of decision theory is that it provides a clear account of rational preferences. There is a small set of axioms that one’s preference relation must satisfy so that it generates a sensible utility function.29 This set of axioms gives a well-motivated notion of coherence for preferences, by ensuring, for example, that if one prefers ϕ to ψ and ψ to ϑ, then one prefers ϕ to ϑ. One would like to have similar criteria on the goals engineers aim for. One option, then, is to consider whether decision theory is a reasonable tool for analyzing the engineering process. Unfortunately, because the theory gets its start after the preferences have been assigned, it seems to miss an important part of the process.30 The hard parts of engineering are discovering what options are available and deciding among these options. In the latter part, engineers often use a tool (the numerical evaluation matrices) that mimicks decision theory, but again, the real decision-making occurs when one chooses the objectives, “sets the weights” and evaluates each design with respect to the objectives. The remainder is mere calculation. To adapt a metaphor from [Wiggins, 2001], decision theory is well-suited for describing a game of snooker, rather than a process in which we decide how to solve a practical problem by constructing a new or improved artifact. Instead of demanding coherence for our set of preferences, let us retain focus on our instrumental analysis and ask whether there is a notion of rationality for a set of ends. To be sure, if the hardline instrumentalist is correct, there may be little sense in aiming for a coherent set of ends. After all, when it comes down to it, we want what we want. Our final ends are simply given at the start and if they are incoherent in some sense, so be it.31 But this negative response needs not deter our interest, since we are considering the instrumental ends that arise in a particular context, namely engineering. Engineers deliberate on which objectives are desirable for their constructions and there is a clear selection process involved. We may ask whether there is a set of rational constraints on the outcome of this process. A minimal constraint is that the ends selected be consistent. One should not aim to pursue two mutually inconsistent goals. Our mousetrap engineer would be irrational to insist that her product should not harm or restrict the freedom of the mouse in any way. One cannot solve a mouse problem without at least displacing some mice.32 But even this minimal requirement needs some clarification. Engineering objectives, such as low cost and high performance, are often in opposition. In general, success at lowering cost will have an adverse effect on performance and vice versa. The consistency criterion is best reserved for those situations in which 29 See,

e.g., [Luce and Raiffa, 1957]. [1969] makes a similar, if more modest, point when he writes, “When we come to the design of systems as complex as cities, or buildings, or economies, we must give up the aim of creating systems that will optimize some hypothesized utility function.” For simpler projects, on the other hand, he evidently supports decision theory as the proper model of design theory. 31 See also Section 4 for an alternative reply (specificationism) to this response. 32 Let us suppose that customers are not willing to move to a new location so that the mice continue to enjoy their current living arrangements. 30 Simon

Practical Reasoning and Engineering

393

the goals are stated precisely (the cost should be no more than $x and the vehicle should accelerate at a rate of no less than y). Edvardsson and Hansson [2005] have given an analysis of goal rationality that is well applicable to engineering practice. They identify four criteria — precision, evaluability, approachability and motivity — that a goal must meet in order to be rational, i.e. achievement-inducing. Let us take these in increasing order of importance for philosophy of engineering. The motivity of a goal is the degree to which it motivates those involved to work towards it. It is easier to meet a goal if everyone involved is committed to it. This feature is not much discussed in design literature, since it is more naturally an issue for management33 than engineering and we will not dwell on it here. A goal is approachable if it is feasible to either accomplish the goal or come close enough to do some good. Engineers pay explicit attention to this feature via feasibility studies, a standard tool for ensuring approachability. Precision is a measure of the detail in the goal’s specification. This feature is prominent in engineering, especially in the problem definition phase. Models of design process begin with the recognition of a need, whether requested by a client or independently discovered. This serves as the broad goal of the engineering project, but before any work can be done (even on the conceptual level), this goal must be clarified. The problem definition stage is an essential part of design processes presented in both [Dym and Little, 2004] (see Figure 3) and also [Dieter, 1983]. As Dieter says [1983], “The true problem is not always what it seems to be at first glance.” The engineering goal must be refined to include such information as design objectives, user requirements, constraints and functions, per Dym and Little’s presentation. The authors also list the particular sources of information (including literature on the state of the art, regulations, etc.) and methods (including function-means trees, requirement matrices and so on) that are particular to this important stage of development. A goal is evaluable if the agents can determine the degree to which it has been met. It is related to precision, inasmuch as a goal cannot be evaluable if it is not precise. The problem definition and conceptual design stages produce evaluable goals for the subsequent stages. Engineers verify that the design-in-progress meets these goals and that the goals are reasonable, thus forming a feedback loop (shown in Figure 3) that allows further refinement in the design. So, like precision, evaluability is a valued feature in engineering. Millgram and Thagard [1996], (see also [Thagard, 2001]) give a rather different account of coherence. They propose a model that clarifies the “instrumental relation of facilitation”. Coherence between an action A and a goal G measures the degree to which A facilitates G. Watching a rerun of “World’s Wildest Police Chases” facilitates my goal of enjoying mindless entertainment, but it impedes my goal of finishing this article. Thus, it coheres well with the former but not with the latter. Because the theory does not distinguish means from ends, the 33 Edvardsson and Hansson have come to their subject through management theory rather than engineering.

Jesse Hughes

Client statement on ml hi (Need) jk

onFinal designmlo hi jk (specs and docs)

?> / 89

Problem definition

?> 89

 Conceptual design

?> 89

 Preliminary design

?> 89

 Detailed design

=
Verification

394

=< :;

 =< ?> Design 89 communication :;

Figure 3. Dym and Little’s five-stage descriptive model of the design process. relation is symmetric: my goal of mindless entertainment also coheres will with my watching the rerun. One decides what to do based on how well the various options cohere with my settled goals. Since writing the article coheres well with my aim of earning a living, which coheres well with a whole slew of my desires, I will miss the episode of World’s Wildest.34 What, then, about coherence applied to engineering? Many objectives found in a typical engineering project (low cost, high performance, long lifespan, etc.) evidently cohere very poorly. Success at keeping costs low, for instance, makes it unlikely that high performance will be achievable. Indeed, if coherence is an overriding concern in engineering projects, then we would have a reason to emphasize one of these two objectives to the neglect of the other. Certainly this is sometimes the path taken, but generally, engineers try to accomplish both objectives as far as possible. In fact, this is an integral part of the design process: seeing how objectives conflict with one another, determining the relative importance of each objective (e.g., assigning weights for the numerical evaluation matrices) and achieving the evidently incoherent objectives to the extent one can (i.e., optimization). Put bluntly, engineers try to be as incoherent as they can get away with. 34 Or

at least I would miss it, except for the very facilitating PVR that records it.

Practical Reasoning and Engineering

395

To be fair, a broader analysis may show the real coherence in engineering design. The objectives are incoherent only when we restrict our attention to the development of the project at hand, that is, when we ignore the point of the design. Our engineer designs mousetraps so that consumers will purchase them. She will sell more mousetraps if she designs them so that they are reliable and cheap, so trying to maximize these two objectives coheres well with her larger end: selling many units of her design. The evident incoherence disappears when we consider design objectives in this larger context. Indeed, both contexts are useful for understanding engineering: design tends to be difficult because engineers aim for apparently incoherent objectives, but they adopt these objectives because by optimization, their designs will compete better in the marketplace.35 Thus, by considering the rationality of engineering goals, we can come to appreciate certain features of the design process. The theory of goal rationality proposed by [Edvardsson and Hansson, 2005] is reflected in typical models of the design method, suggesting that such methods implicitly recognize at least some of Edvardsson and Hansson’s criteria. Millgram and Thagard’s theory of coherence, on the other hand, gives a theory to judge the conflict between design objectives as well as a motivational account for why such apparently incoherent objectives are selected: they cohere with the engineer’s goals in the larger context of marketplace competition.

4 SPECIFICATIONISM We turn now to an alternative to instrumentalism, but one which still embraces the emphasis on means-end reasoning. As we have said, a purely instrumentalist account of engineering design begins with some ends given (i.e. final-in-context). The engineer then researches the problem thoroughly, so that he can identify a set of means that are relevant to its solution. Once these means are found, they become ends sought for the sake of our original end. Again, research and existing knowledge are brought to bear to find appropriate means to realize these ends and so it goes until, at last, a design has emerged. This account obscures an essential part of the design process, however, namely the problem definition phase (also discussed in the previous section). It is imperative that engineers refine the original statement of need so that they can begin work on a clearly defined problem. A considerable effort is made to replace the original engineering goal (the “client statement” in Figure 3, page 394) with a more precise goal. This important step is non-trivial: Dym and Little devote an entire chapter to outlining how one refines a client statement into a clearly specified need. According to specificationism, refinement of one’s ends and their place in the hierarchy of ends is an essential part of practical deliberation, Aristotle’s famous 35 Similar considerations apply to, say, civil engineering projects in which different firms compete in the bidding process.

396

Jesse Hughes

dictum that deliberation is of means,36 not ends, notwithstanding. Even in personal practical reasoning, our ends are largely ambiguous and roughly stated and we often cannot reason about particular means to these ends without first refining them, adding details and clarification until we may see how to approach our goal. As [Wiggins, 2001] writes, “Deliberation is still zetesis, a search, but it is not primarily a search for means. It is a search for the best specification. Till the specification is available there is no room for means. When this specification is reached, means-end deliberation can start.” This observation is even more essential to an account of professional practical reasoning. The best developed account of specificationism is Richardson’s [1994], but it is not so well-suited for our purposes. Richardson’s focus is on end-norms, broad statements regarding how an agent should behave “in general” or “in most situations”, but these kinds of ends are rather different than the engineering ends we have in mind here. An engineering problem such as, “Build a better pogo stick,” is narrow in focus and duration, unlike, say, “Do what is necessary to achieve success.” The latter is a guiding principle rather than a task to be accomplished. Consequently, we will focus our attention on [Kolnai, 1978], whose presentation more closely matches our interest. Kolnai adapts Aristotle’s physician to illustrate specificationism. We will compare his account with Audi’s [2006] recent summary of the same example. A physician’s end is to heal. Consequently (per Audi’s presentation of Aristotle), he deliberates about how to accomplish this goal, first deciding on applying medicine and then deciding that penicillin is the best medicine for his patients needs and so on, until at last the chain of means to ends reaches an action (reach for the medicine) that can be performed at will. Kolnai counters that the real deliberation comes from clarifying the vague end “to heal” in a situation in which many other ends influence the physician’s options. For instance, it may be that the most effective treatment now also has the unfortunate side effect of shortening the patient’s life. The current, vaguely stated goal of healing the patient does not determine whether this treatment should be chosen or avoided. This end must be specified more precisely and the physician’s other ends inform this specification (so that it is properly a process of practical reasoning and not arbitrariness). It is precisely this kind of deliberation that occurs in the problem definition phase. One must reflect on the need to be fulfilled in context of many various other ends, desires and requirements in order to get a clearer statement of the problem. The conflict between various desired features is indeed common to most engineering projects: how to balance, say, cost and safety? Settling such issues require not only investigation, but deliberation. Yotaro Hatamura [2006]37 reports on a project to develop a telescopic arm clamshell digger — a piece of heavy 36 This is W.D. Ross’s influential translation. Wiggins [2001] suggests that the appropriate translation is not “means”, but “what is toward the end” and, moreover, argues that Aristotle is a specificationist and not an instrumentalist. 37 This book is invaluable to philosophers of engineering. It contains dozens of reports from engineers detailing their decision making on various projects. This includes false starts and revised goals and so gives a unique insight in how engineers do what they do.

Practical Reasoning and Engineering

397

construction equipment. The engineers’ first step was a user survey to discover problems with their competitors’ products so that they could avoid the same. They also reflected on the consequences of entering the market late and thus identified further constraints and objectives due to this situation. Clearly, the engineers were motivated to please their potential customers and also turn a profit. These goals put constraints on their engineering goal, but these constraints could be known only through research and investigation. Now, as Kolnai (echoing Aristotle) say, research is not deliberation, but rather input into the deliberative process. One requires factual claims as well as a multiplicity of background ends in order to deliver a rational specification. We may also reevaluate our account of numerical evaluation matrices from Section 2.3. We claimed that the assigning of weights to project objectives was not deliberation per se, since it was not about discovering what one should do. Instead, it was a matter of finding appropriate numbers for the weights which reflect the importance of each objective, and this is more akin to investigation than deliberation. But the situation is surely more complicated than merely finding the right values. Rather, one must compare the various objectives involved (and their motivating ends) and come to a decision about appropriate ranking. And once that decision is made and the designs evaluated, their rankings are fixed. As Kolnai [1978] says, “Deliberation is in fact an exercise of freedom in some sense aiming at the restriction of freedom.” This is an apt description of the design evaluation strategies in [Dym and Little, 2004]: engineers aim for as much objectivity as possible in their evaluations, so that they avoid arbitrary preferences. But the deliberation cannot be in the final calculations nor in applying the optimal means syllogism. Instead, it occurs in the weighting of the objectives. Specificationism is also compatible with a broadening, rather than a tightening, of one’s ends. As [Richardson, 1994] says, “Sometimes we move to a less specific formulation of an end-norm.” While the goal of an engineering project is not an end-norm in Richardson’s sense, this observation applies equally well to engineering practice. Another example from [Hatamura, 2006] illustrates this kind of revision. Section 5.1.4 describes a project for automating the junction hood connecting two cars of a passenger train, so that coupling and decoupling cars would be easier and faster. The engineers’ first proposal modified the existing junction hood by adding an automated mechanism, but this solution was deemed too complex and unreliable. Their clients suggested that the alteration did not have to be compatible with the existing hardware; it was acceptable to replace the hardware with all-new hoods. Thus, the new project goal was considerably more liberal than the previous goal and the engineers devised a radically different and simpler solution. The new information allowed a restatement of the problem definition and hence an improved final design. This example is instructive in another manner. The new specification was arrived at only after considerable work had been done on the project. This is a typical feature of engineering: as flaws or imperfections are found in proposed solutions, one revisits the engineering requirements. Perhaps, as in this case, the

398

Jesse Hughes

client is willing to alter the requirements so that a better solution can be found. The shifting of attention from solution to problem specification and back again is common to engineering. Indeed, in many cases, the problem specification is fully developed only by working on a partial or preliminary specification: to discover what the problem is, one has to start working on the solution! In closing, specificationism gives a fruitful model of some of the most important aspects of engineering. Engineers do more than select appropriate means for the goal at hand. They deliberate on what the goal should be and they refine the goal as the process continues. These features are missing from a purely instrumentalist account of engineering.

5 PLANNING Michael Bratman [2001] (see also [1987]) provides another alternative to instrumentalism of particular interest to philosophers of engineering. Instrumentalism takes the basic units of deliberation as the means to some considered end, but this is, according to Bratman, a shortsighted variety of reasoning. Intentions can be reconsidered and altered without warning. Humans are interested in longer-term strategies: they plan. And no one spends more time planning than engineers. In Bratman’s theory, plans differ from other pro-attitudes, such as desires and ends, in three primary ways. First, they are stable. A plan is a settled course of action. An agent will not reconsider his plan unless difficulties crop up. Indeed, cooperation requires such stability.38 When we cooperate, we rely on others to do as agreed. On Monday, I agree to meet with Krist on Tuesday. This agreement makes sense only if I am reasonably sure that, whatever happens to Krist between today and tomorrow, he will still be at the meeting we both are planning.39 One cannot cooperate effectively with an agent in the habit of changing his plans willynilly (though I’m sure we’ve all tried). The second important feature of plans is that they are incomplete. Bratman mentions three different ways in which plans are incomplete, but the third holds special interest for us. Plans are typically formulated in vague terms to begin with. I plan to go to the university tomorrow, but I have given no thought about how I will get there. That I will think about tomorrow, when I need a more clearly specified plan. Just as Kolnai, et al., consider further specification of ends an essential part of practical reasoning, so too with Bratman’s plans: they come with details to be filled in later. This feature is also apparent in the Dym and Little’s design process in Figure 3. In Bratman’s terms, the design process is a process for selecting a plan. An 38 Cooperation also requires another feature of plans, namely that they are, like intentions, conduct-controlling rather than conduct-influencing, like mere desires. 39 Such plans usually include implicit agreements to contact the other in case of unforeseen difficulties. But if Krist is fickle in following his plans to begin with, I cannot expect such a call either.

Practical Reasoning and Engineering

399

engineering design is a plan for solving the engineering need at hand. Thus, Figure 3 (on page 394) is a description of how engineers generate plans. The middle three steps — the design steps — consist of specifying the results of the previous step in greater detail and testing the result. It goes from conceptual design (an “outline solution”) to detailed design, in which the specification includes the individual part types and dimensions. In the end, one specifies a plan to the level of detail appropriate for one’s own needs. I have a seventeen item checklist that I must follow in order to get my son ready for school. My wife’s plan seems to consist of the imperative: Get the boy ready and out the door. From that one sentence, everything seems to flow for her. An engineer’s plans are clearly on the more, rather than less, detailed end of the scale, but for different reasons than mine. Engineers develop a plan that will most likely be implemented by others. In order to avoid error, it is essential that the plan is stated as explicitly as possible. The third and final feature of planning theory relevant to engineering regards the relation between plans and subplans. An individual plan has a fairly obvious linear structure — one step is done before another — but plans are also related to one another in a hierarchical fashion. I plan to spend this weekend with my family. This plan includes subplans, such as fishing and doing yard work. Each of these subplans are full-fledged plans themselves, but they are also part of the grander scheme. In the same way, engineering solutions typically include breaking a problem down into subproblems, in a top-down approach (see, e.g., [Dym and Little, 2004]). This can only work if there is a guiding plan, that is, a higher-level (or background ) plan including the lower level plans as steps, so that the solutions to the subproblems fit together to form a solution to the encompassing problem. If the one of the subplans should fail to be realized, the engineers must return to deliberate on the guiding plan. They must investigate how to replace the failed step and still accomplish their goal. As Bratman writes, “the background plan constrains what options are relevant options,” so that deliberation is reasonably focused rather than open-ended. If the engineer’s spring mechanism will not work as it should, she deliberates on how to alter her plans for the springloaded-bar mousetrap. She would not return to the question of whether the springloaded-bar is the right kind of mousetrap — unless the failure is so damaging that she abandons that background plan as well and goes “one level up”, to the plan that she should build some kind of mousetrap. In the face of failure, we do not give up all our plans and start the deliberations afresh. Instead, we consider new plans within the context of the narrowest background plan that still “works”. The revision of plans in the context of an engineering project is indeed very common. An engineering project rarely (never?) proceeds from the initial stages through greater clarity and detail to the final project without some revision in the plans. Ideas that appear promising often turn out unworkable as the conceptual comes closer to realization. As well, one may discover that the initial design goals were less compatible than thought, and hence return to the drawing board.

400

Jesse Hughes

Such back-and-forth between subplan and background plan seems a characteristic feature of engineering design. Finally, it is worth noting that the design process itself is a plan in Bratman’s sense. Namely, it is a plan for generating plans. As one might expect, it is stated in quite general terms in the various engineering texts, since this plan is meant to be adapted to a wide variety of engineering settings. Dym and Little, for instance, suggest tools and methods to be used in implementing the design process, but the working engineer must decide how best to implement the design process, i.e., how to fill in details omitted from Dym and Little’s specification. These tools and methods serve as subplans to be “plugged into” the background plan of the design process, as appropriate. In closing, Bratman’s theory provides the following features for a practical analysis of engineering: (1) plans are stabler than arbitrary ends and in this respect they resemble engineering goals, (2) the initial incompleteness of plans is similar to the development of an engineering proposal, which becomes more detailed as the project proceeds, and (3) a hierarchical relation between plans and subplans which serves to model the top-down approach common to engineering. On the other hand, an instrumentalist account of engineering does not require that ends are generally as stable as Bratman’s plans; it suffices that some ends, namely engineering ends, exhibit stability. We can also model feature (2) without losing the simplicity and clarity of traditional means-end reasoning, by moving from a purely instrumentalist account to specificationism. Neither instrumentalism proper nor specificationism include an explicit hierarchical arrangement “out of the box”, so to speak, and so feature (3) does seem most naturally expressed in terms of Bratman’s plans. Whether this is a sufficient reason to choose planning theory over more familiar and easily motivated means-end reasoning is a matter that should be decided only when we see more clearly the fruits of each approach in analyzing engineering design and methods. 6 CONCLUDING REMARKS This has been a very brief survey of some current (and not so current) topics in practical philosophy, with an indication of how they apply to engineering. The aim has been to take an established branch of philosophy and see what fruit it can bear when adapted to a newer study. The results are promising, even if the fit is not perfect. Practical philosophy in general and instrumentalism in particular is geared towards personal, rather than professional, action. Motivations are given in terms of desires rather than clearly stated goals. Nonetheless, the taxonomy of means and ends provides a framework for an analysis of engineering. As well, alternative theories such as specificationism and planning theory can provide disparate frameworks for modeling engineering practice and thereby focus on different features of engineering design. But this examination has been cursory. If it has been minimally successful, then philosophers of engineering will have a new conceptual toolkit from which to

Practical Reasoning and Engineering

401

draw occasional terminology and schemata. If, as I believe, practical philosophy has more to offer than mere organizational principles, then we may find specificationist, say, analyses of engineering design. In the best outcome, there will be real debate about which theory of practical reasoning best reflects the engineering process. Engineering is, at its heart, about deliberation and action. Thus, we have good reason to expect that the most natural existing framework for philosophy of engineering is an adaptation of practical philosophy in some form. BIBLIOGRAPHY [Ackrill, 1980] J. L. Ackrill. Aristotle on Eudaimonia. In Am´ elie Oksenberg Rorty, editor, Essays on Aristotle’s Ethics, pages 15–35. University of California Press, 1980. [Anscombe, 1989] G.E.M. Anscombe. Von Wright on practical inference. In P.E. Schlipp and L.E. Hahn, editors, The Philosophy of Georg Henry von Wright, pages 377–404. Open Court, 1989. [Audi, 2006] Robert Audi. Practical Reasoning and Ethical Decision. Routledge, 2006. [Bratman, 1987] Michael E. Bratman. Intentions, Plans and Practical Reason. Harvard University Press, 1987. [Bratman, 2001] Michael E. Bratman. Taking plans seriously. In Millgram [2001b], pages 203– 221. [Churchland, 1970] Paul M. Churchland. The logical character of action-explanations. Philosophical Review, 79, 1970. [Dieter, 1983] George E. Dieter. Engineering Design: A Materials and Processing Approach. McGraw-Hill, Inc., 1983. [Dym and Little, 2004] Clive L. Dym and Patrick Little. Engineering Design: A Project-Based Introduction. John Wiley and Sons, Inc., 2004. [Edvardsson and Hansson, 2005] Karin Edvardsson and Sven Ove Hansson. When is a goal rational? Social Choice and Welfare, 24(2): 343–361, 2005. [Franssen, 2006] Maarten Franssen. The normativity of artefacts. Studies in History and Philosophy of Science, 37, 2006. [Hatamura, 2006] Yotaro Hatamura, editor. Decision-Making in Engineering Design: Theory and Practice. Springer-Verlag London Limited, 2006. Translated from the Japanese by Kenji Iino. [Hill, 1970] Percy H. Hill. The Science of Engineering Design. Holt, Rinehart and Winston, 1970. [Hughes et al., 2007] Jesse Hughes, Peter Kroes, and Sjoerd Zwart. A semantics for means-end relations. Synthese, 158(2), 2007. [Hughes, 2008] Jesse Hughes. An artifact is to use: an introduction to instrumental functions. Synthese, 2008. Currently available online, print publication forthcoming. [Humberstone, 1992] I.L. Humberstone. Direction of fit. Mind, 101(401): 59–83, Jan. 1992. [Hume, 1739] David Hume. A Treatise of Human Nature. McMaster University, 1739. Available at http://socserv2.socsci.mcmaster.ca/ econ/ugcm/3ll3. [Hume, 1777/1975] David Hume. Enquiry Concerning the Principles of Morals. Oxford University Press, 1777/1975. [Kolnai, 1978] A. Kolnai. Deliberation is of ends. In Ethics, Value and Reality. Hackett, 1978. [Luce and Raiffa, 1957] R. Duncan Luce and Howard Raiffa. Games and Decisions: Introduction and Critical Survey. John Wiley & Sons, Inc., 1957. [Millgram and Thagard, 1996] Elijah Millgram and Paul Thagard. Deliberative coherence. Synthese, 108: 63–88, 1996. [Millgram, 1995] Elijah Millgram. Was Hume a Humean? Hume Studies, 21(1): 75–93, 1995. [Millgram, 1997] Elijah Millgram. Practical Induction. Harvard University Press, 1997. [Millgram, 2001a] Elijah Millgram. Practical reasoning: The current state of play. In Varieties of Practical Reasoning [2001b], pages 1–25. [Millgram, 2001b] Elijah Millgram, editor. Varieties of Practical Reasoning. MIT Press, 2001.

402

Jesse Hughes

[Neander, 1995] Karen Neander. Misrepresenting & malfunctioning. Philosophical Studies, 79(2): 109–141, August 1995. [Platts, 1979] Mark de Bretton Platts. Ways of meaning : an introduction to a philosophy of language. London, Routledge and Kegan, 1979. [Railton, 1997] Peter Railton. On the hypothetical and non-hypothetical in reasoning about belief and action. In G. Cullity and B. Gaut, editors, Ethics and Practical Reason, pages 53–79. Oxford, 1997. [Richardson, 1994] Henry Richardson. Practical Reasoning About Final Ends. Cambridge University Press, 1994. [Ross, 1908] W.D. Ross, editor. Aristotle’s Nicomachean Ethics. Clarendon Press, 1908. Available at http://classics.mit.edu/Aristotle/nicomachaen.html. [Schmidtz, 1994] David Schmidtz. Choosing ends. Ethics, 104(2): 226–251, 1994. [Schmidtz, 2001] David Schmidtz. Choosing ends. In [Millgram, 2001b], pp. 237–257. [Simon, 1969] Herbert A. Simon. The Sciences of the Artificial. The MIT Press, 1969. [Smith, 1987] Michael Smith. The Humean theory of motivation. Mind, 96(381): 36–61, Jan. 1987. [Thagard, 2001] Paul Thagard. How to make decisions: Coherence, emotion and practical inference. In Millgram [2001b], pp. 355–372. [Torvalds, Linux, 1991] What would you like to see most in minix? Usenet post 25 Aug 1991, 1991. Available at http://www.linux.org/people/linus post.html. [Velleman, 1996] J. David Velleman. The possibility of practical reason. Ethics, 106(4): 694–726, Jul. 1996. [von Wright, 1963] Georg Henrik von Wright. Practical inference. The Philosophical Review, 72(2): 159–179, Apr. 1963. [Webster, 1913] Webster’s revised unabridged dictionary. http://machaut.uchicago.edu/websters, 1913. [Wiggins, 2001] David Wiggins. Deliberation and practical reason. In Millgram [2001b], pp. 279–299.

Part III

Philosophy of Engineering Design

This page intentionally left blank

INTRODUCTION TO PART III Peter Kroes, associate editor The designing and making of material artifacts or the designing and providing of services that involve the manipulation of technical objects is one of the core professional activities of engineering. In order to remain competitive design engineers will have to ensure that they are knowledgeable in the engineering sciences. They will have to be able to conduct research in a setting that is practice or mission oriented, which means that they will have to be able to analyze the properties of the processes and things they are dealing with. That is why in engineering curricula a great deal of time is devoted to the teaching of engineering sciences. In order to become a good design engineer, however, it is not sufficient to just possess analytical research skills. Apart from and in addition to having these skills, engineers need to have synthetic design skills: when designing new technical artifacts, they must be able to combine elements (components or processes) in inventive, creative ways so that they can satisfy practical means-end or functional requirements. The designing of technical artifacts is considered to be primarily a synthetic rather than an analytic activity. Apart from teaching research competencies, engineering curricula are therefore geared to teaching the competencies that are required to become a good design engineer. This prima facie appealing view of designing as a synthetic activity, as opposed to research as an analytic activity, requires further elucidation if it is to be of any use in understanding the nature of engineering design. After all, researchers also have to be skilful designers, not of technical artifacts or services but of theories, experiments and the equipment needed to perform such experiments. For these purposes they also need to have synthetic skills; theories, experiments as well as experimental equipment are composed of different elements (like, for instance, laws, actions and physical components) and they result from researchers putting these elements together in specific ways to satisfy requirements, cognitive and otherwise. Moreover, analytic and synthetic methods have always been part and parcel of research, ever since the early days of scientific research. The distinction between analytical forms of reasoning, in which one proceeds from effects to causes and synthetic forms, in which one proceeds in the opposite way from causes to effects, may be traced back to Greek antiquity. Causal explanations, for instance, typically involve the composition (synthesis) of chains of cause-effect relations leading up to the phenomenon to be explained. Within mathematics, the distinction between the analytic and synthetic methods, already defined by Euclid, received its canonical form in the distinction between analytic and synthetic geometry; the analytical approach assumes what solution is to be given and Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

406

Peter Kroes, associate editor

reasons from it to arrive at propositions that may be taken to be true, whereas the synthetic approach reasons from true propositions to the sought solution. So the characterization of engineering design as a synthetic activity does not, as such, help very much when clarifying what kind of activity it is. Many problemsolving activities, including engineering design and scientific research, may be considered to be of an analytic and synthetic nature, since they employ analytic as well as synthetic methods. Does this mean that the analytic-synthetic distinction is irrelevant to an understanding of the nature of engineering design, especially in comparison to the nature of scientific research? That is a matter that remains to be seen. Even though at a very general level of abstraction design and research may be characterized as forms of problem solving that involve analytic and synthetic methods, there may be differences between the kinds of problems involved in the two fields leading to different roles of the analytic and synthetic methods used to solve these different kinds of problems. Consider a typical engineering research problem such as the matter of determining whether a given construction will be able to bear a certain load with the help of mechanics and compare that to a typical engineering design problem such as that of finding a construction that satisfies a list of requirements, one of which being that it must be able to bear a given load. It is not obvious at all that these kinds of problems require the same kinds of analytical and synthetic skills. Similarly, it is not clear whether what Hacking [1983] has called the creation of phenomena in scientific research is comparable, as a synthetic activity, to the creation of technical artifacts. It is not difficult to point to differences between these two kinds of creative activities which might make them different as synthetic activities. The creation of phenomena in science may be analyzed as the synthesis of cause-effect chains with initial and boundary conditions which are such that the synthesized whole exhibits a certain phenomenon. The creation of technical artifacts, however, involves the synthesis of functional components that together realize the overall function of a technical artifact. Moreover, in the case of the creation of technical artifacts the object created or synthesized may malfunction; the notion of malfunction, does not, however, make sense for a phenomenon created in scientific research. Both kinds of synthetic activities result in the creation of different kinds of entities. So, the designing of technical artifacts appears to be a synthetic activity with distinctive features of its own. From a philosophical point of view little is known about what these distinctive synthetic features of engineering design are. That is because engineering design or design in general has not received much attention within the field of philosophy. Whenever it is discussed it is usually within the context of the argument from design, a context that is only obliquely related to engineering practice. A systematic philosophical analysis of engineering design is lacking, which is not surprising given that the philosophy of making and the philosophy of technical artifacts, to which that is closely related, are marginal fields within mainstream philosophy. The notion of design is in urgent need of further philosophical explication. The brief remarks made above indicate that such a philosophical explication will tie

Introduction to Part III

407

this notion to traditional problems in philosophy. Though the intuitive characterization of designing as a synthetic activity, as opposed to research as an analytical activity, may be very attractive this specific conception of the synthetic-analytic distinction is in need of further clarification. It may well be that just as the distinction between analytic and synthetic statements has come under attack [Quine, 1951], so the distinction between analytic and synthetic methods will have to be reconsidered (for a discussion on this distinction see, for instance, [Beaney, 2007]). Another traditional philosophical topic to which the notion of design is related is the distinction between artificial and natural objects, since artificial objects are often characterized as the material realization, the ‘embodiment’ of human designs (or more generally, human ideas). Here we encounter the notion of design as understood in noun form, as something that is the result of design in a verbal sense, a notion which, although familiar to any engineer, also requires further philosophical clarification. In view of this state of affairs, this part of the handbook does not present a neat, systematized and exhaustive overview of the most important topics that fall under the philosophy of design and how those topics have so far been dealt with by philosophers. It is much more an attempt to explore the uncharted domain of engineering design philosophy by discussing various topics in this field. It starts with a historical overview of the development of design thinking by Richard Buchanan. This overview presents a general background to the discussion of a series of more specific topics. The first one of these, analysed by Kees Dorst and Kees van Overveld, concerns the problem of how to classify various design practices into different types. This is followed by a discussion by Marc de Vries on a problem encountered in any kind of engineering design practice, namely how to go from customer requirements to technical specifications. Next comes an analysis by Peter Kroes of a number of conceptual problems concerning the notion of design viewed as a verb and as a noun. Also included is a contribution by William Wood on a problem that is of growing importance to engineering design practice, namely how to represent functions (i.e. the functional properties) of technical artifacts in formal or computational settings. It is followed by an analysis of the role and limits of rationality in engineering design by Peter Kroes, Maarten Franssen and Louis Bucciarelli. A discussion of the design of socio-technical systems by Johannes Bauer and Pauline Herder closes this part of the handbook. Most of the contributions are explorative and the topics discussed pose interesting challenges for engineers as well as for philosophers. That is the reason why we have approached both engineers and philosophers to contribute to this part. It is our strong conviction that the philosophy of engineering design is a field that must be developed through close collaboration between engineers and philosophers. Finally there are some topics that are conspicuously absent, despite our efforts to include them. One of the most important ones is the notion of trade-offs. It plays a central role in engineering design thinking and raises fundamental methodological questions (some aspects of trade-offs are briefly discussed in the chapter by van de Poel in part V of this Volume). Another omission is an overview of

408

Peter Kroes, associate editor

the most important results of design methodology. Design engineers and methodologists have developed a wealth of methods and techniques that are intended to support the solution of design problems. A systematic survey of the philosophy of design certainly ought to include these topics. Other topics of interest for a philosophy of design, such as the method of functional decomposition or means-end reasoning are not addressed in this part but appear elsewhere in the handbook (see Pieter Vermaas and Pawel Garbacz, this Volume part II, and Jesse Hughes, this Volume, part II). BIBLIOGRAPHY [Beaney, 2007] M. Beaney. Analysis. The Stanford Encyclopedia of Philosophy (Fall 2007 Edition). E. N. Zalta. (editor), http://plato.stanford.edu/archives/fall2007/entries/analysis/, 2007. [Hacking, 1983] I. Hacking. Representing and Intervening: Introductory Topics in the Philosophy of Natural Science. Cambridge, Cambridge University Press, 1983. [Quine, 1951] W. V. Quine. Two dogmas of empiricism. The Philosophical Review 60(1): 20-43, 1951.

THINKING ABOUT DESIGN: AN HISTORICAL PERSPECTIVE Richard Buchanan The accomplishments of design methodology in the twentieth century reflect a coherent practical inquiry into the nature of the human-made world. This inquiry is based on a philosophical pluralism that relates the subject matters, methods of thought and action, and guiding principles of design. The inquiry follows three major strategies of investigation: Dialectic, Rhetorical Inquiry and Productive Science, and Design Science. The origin of these strategies may be traced to the ancient world, but the development of these strategies from the Renaissance to the nineteenth century are a conditioning influence on twentieth century discussions. In the twentieth century, the traditional design methods of craft and drawing were elevated through practice, education, and philosophical reflection to address many of the complex problems of production that emerged with new technologies and new social circumstances. The problems of engineering design parallel the problems in areas such as industrial design, urban planning, and architecture. Methodological investigations focused on the sequence or phases of design, such as analysis, synthesis, and evaluation, as well as on the principles and goals of design. The result was not a single theory or system of design but a pluralism of approaches that represent the ecology of design culture. 1

INTRODUCTION

Every art and every science of theoretical, practical, or productive inquiry may be distinguished by its subject matter, characteristic methods of thought and action, and goals or guiding principles. The common subject matter of design is variously described as the artificial or the human-made or products that support human beings in all of their individual and collective activities. From this perspective, the history of design in the twentieth century represents a coherent practical inquiry into the subject matter of the human-made world, unfolding in a sequence that is shaped by fundamental problems of thinking and making whose ongoing exploration continues to affect virtually every aspect of daily life. This is reflected in the expanding concept of “product.” At the beginning of the twentieth century, “product” typically meant a physical artifact — the result of industrial design, engineering, or architecture. By the end of the century, the concept of “product” had expanded to mean any result of the creative work of designers, and the concept of “designer” had correspondingly expanded to include any individual whose work Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

410

Richard Buchanan

involved forethought in the conception and planning of any aspect of the humanmade world. Products were generally divided into four broad classes, sometimes known as the Four Orders of Design, reflecting the fundamental problems that designers were called upon to address in their work: communication through signs and symbols; construction and fabrication of artifacts of any scale; deliberation in planning actions, activities, services, and processes; and integration or systematization in encompassing wholes such as social organizations, physical, human, and symbolic environments, and cultures. Exploration of these problems led to the further particular or proper subject matters of design, manifested in a proliferating array of design professions, each guided by its own characteristic methods of practice. However, there is a difficulty embedded in such an account that complicates our understanding of design. Subject matters do not come ready made, even though the problems on which they are based may suggest that this is the case. Subject matters are made by methods of inquiry, whether the inquiry is formal and theoretical or pragmatic and practical. From this perspective, the history of design represents not only a pluralism of subject matters and products but also a pluralism of arts, methods, and techniques employed by designers and those who investigate design. The relationship of subject matter and method makes any effort to assess the accomplishments of design methodology a philosophical problem as well as a practical or historical problem, because “method” is open to sharply different interpretations based on philosophic assumptions, whether those assumptions are openly acknowledged or tacitly held. Thus, instead of studying the subject matters of design and how they led to the establishment of methods of practice, it is also reasonable to study the methods of thought and action in design, because it is the function of method to discover problems for inquiry, identify potential subject matters for exploration, investigate those subject matters in thought and action, and establish the principles that explain different subject matters, giving them form and purpose in the human community. The concept of “method” in Western culture has its origins in Greek philosophy [McKeon, 1998, p. 166] (see also [Mitcham, 1994]). Plato was the first philosopher to use the term methodos or “way after.” Along with a variety of related terms such as “way,” “reason,” “mode,” “treatment,” and “art” used by earlier philosophers, he employed methodos in the context of dialectic, the art by which he explored opinions and ideas about all subjects, including human-made products. In turn, Aristotle also used the variety of earlier terms, but he employed methodos as a technical term in his philosophy, using it to treat both the intellectual arts (rhetoric, dialectic, and logic) and the sciences (theoretical, practical, and productive). According to Aristotle, there was a third method or way in Greek philosophy, attributed to Democritus as the “physical” way, where ideas are derived from sensations and where atomic simples are combined in larger compositions and syntheses in a materialist reality. Plato is generally silent on this method, but Aristotle remarks that his own method is midway between the dialectical way of Plato and the physical way of Democritus.

Thinking about Design: An Historical Perspective

411

Aside from philosophical discussion of human-made products in the ancient world, the surviving technical treatises on architecture, building, engineering, and other human-made products were typically guided by a combination of rhetoric and productive science rather than by dialectic or by the physical method of the atomists.1 Rhetoric focused on the ethical and inventive character of the artist or designer; productive science focused on the discipline and methods of making and the properties of the made-thing. A prime example is Vitruvius’ Ten Books on Architecture, which is considered the oldest surviving treatise on engineering as well as architecture in western culture. This treatise is grounded in Cicero’s conception of rhetoric and, in turn, focuses on architecture as a discipline for making buildings, instruments for measuring time, and weapons of war. In China, however, technical treatises tended to combine dialectic and rhetoric, placing artifacts and other human-made products in a context of practical life organized in social hierarchy and guided by duties and obligations to higher authority. All of the terms and distinctions employed in the discussion of design and design methodology in the twentieth century were already invented and the subject of debate in the ancient world with regard to a wide variety of applications, including human-made products. The ongoing exploration of these terms throughout the history of Western culture serves as a background and conditioning influence for understanding their application to design. These terms include analysis and synthesis, invention and judgment, division and composition, form and matter, element and structure, common and proper, probability and a priori certainty, theoretical and practical, demonstration and proof, induction and deduction, technology, the logical modals (necessity, contingency, possibility, and impossibility), and visualization through schemata. The exploration of design methods in the twentieth century places these terms and distinctions in the context of three major strategies of inquiry that echo the three methods or “ways” of the ancient world: Dialectic, Inquiry, and Design Science. For all of the varieties of dialectic that have occurred throughout history, ranging from idealist to materialist to skeptical and pragmatic dialectics, the common strategy of Dialectic is to overcome conflicts, contradictions, and oppositions encountered in everyday life by bringing them within a system or ordered whole, regarded variously as Being, material and historical determinism, phenomenological consciousness, or experience. As Jaspers says, “The truth begins with two” [Jaspers, 1951, p. 124]. The dialectical strategy in design is to identify and overcome conflicting opinions and oppositions on issues such as form and matter, the values held by producers and the values held by consumers, or the various conflicting requirements that affect products created by designers. In contrast, the strategy of Inquiry is to seek the resolution of theoretical, practical or productive problems and move toward the advancement of knowledge in the various branches of human learning and activity. It may take the form of Rhetorical Inquiry into the inventive and creative power of the designer and his or her ability to effect 1 A notable exception is St. Augustine’s treatise on music, De Musica, which employs a dialectical strategy of inquiry.

412

Richard Buchanan

social change through argument and communication. Or, it may take the form of a Poetics or Productive Science of the human-made world, focusing on analysis of the essential elements of products (e.g. form, function, materials, and manner of production) and the creative synthesis of these elements in the various branches of design, with due regard for how products are designed, produced and distributed as well as how products evolve in use to support the human community. Finally, in contrast to the strategies of Dialectic and Inquiry, the strategy of Design Science is to seek basic elements that underlie the complexities of the material world and the workings of the mind and, then, to investigate the processes and mechanisms by which those elements are combined to yield the world of experience and the cognitive processes of designing.

Figure 1. The interplay of these strategies helps to explain the otherwise bewildering diversity of design practice and design research in the twentieth century as well as the common tendency to merge the problems of design method with the problems of technology. It provides a framework for assessing the accomplishments of design methods and methodology among the many allied branches of design. However, while these broad strategies provide the underpinnings of design methods and methodology in the twentieth century, it is important to note that their pursuit in most cases was not informed with any significant awareness of the history, philosophy, and intellectual resources that otherwise may have contributed to the development of design thinking and design practice. Hence, we witness both original explorations as well as the gradual rediscovery of intellectual traditions in establishing the foundations of design. Contained within the relationship of subject matter and method in design is the further problem of the diverse traditions within which design has been practiced in the past and within which design and its products have received attention. These traditions seriously affected the methods employed by designers, and they also point toward two of the philosophical problems that are significant issues in design practice: the problem of parts and wholes and the problem of means and ends (also referred to as the problem of the useful (utilia) and the good (honesta) or the problem of use and enjoyment).

Thinking about Design: An Historical Perspective

413

There is a long tradition of associating design with the fine arts, since the creation of practical products is similar to the creation of works of art such as paintings or sculptures, which are experienced, understood, and appreciated in their wholeness. For this reason, design was often regarded as a form of applied or commercial art. This association continues in all cultures, affecting the understanding of design method. However, there is also a long tradition of associating design with technology, understood in a broad, ambiguous sense as the “art of science” or the “science of art,” derived from the ancient roots of the word (“techne” and “logos”). In this tradition, design and engineering are also associated with the natural sciences and mathematics, since the creation of artifacts, whether simple or complex, requires some form of calculation and forethought about materials, energy, and how the parts of a product may be fabricated, combined, and assembled to bear loads, distribute stress and heat, and so forth. For this reason, design and engineering were often regarded as a form of “practical mechanics” or applied science, with an emphasis on the analysis of parts prior to their composition or synthesis into wholes. Both of these traditions are evident in design at the beginning of the twentieth century, representing an approach from wholes to parts or from parts to wholes. But two other traditions, less often remarked in technical discussions of design, are also evident. They represent alternative explorations of the problem of means and ends, which is sometimes recast as the problem of the useful and the good. One is the tradition of thinking about design and the products of design in the context of political science, economics, and social behavior. The wealth of nations, the division of labor, the social implications of work, and the just distribution of goods: these and many related themes provided a background and context for considering the implications of industrialization in the modern world. They led to political discussions about design and technology in the twentieth century, and they also led, eventually, to explorations of design and technology in the context of the social and behavioral sciences, management, and organization theory. The other tradition was perhaps even less remarked at the beginning of the twentieth century, but by the end of the century it had risen at least to equal influence in the development of design. This is the tradition of philosophy, humanism, and spiritual reflection on the place of design in human culture and spiritual life. By the middle of the twentieth century, this tradition was already exerting considerable influence on the development of design methods and methodology. Though the early influence of this tradition was sometimes subtle and easily overlooked, one may argue that it eventually exerted a decisive influence in shaping the investigation and development of design methods. The interplay of these traditions provides a theoretical and practical matrix within which the development of design methods and methodology may be better understood without reducing that development to unfolding material conditions, the creative power of individuals, or broad cultural ideas and values. The matrix points toward an ecology of design culture that provides a rich context for understanding how the rise of new technologies influenced design methods; how

414

Richard Buchanan

education and the schools helped to advance the formal study of design methods; how the deeply ambiguous concept of “science” became enmeshed in our efforts to understand design thinking; and how pragmatic design practice was increasingly influenced by design research. Furthermore, the matrix helps to explain the complex relationship between design studies and the practical work of designers who explore design thinking through action and production. Design studies — the history, criticism, and theory of design, as well as design research through empirical investigation and philosophic speculation — developed gradually throughout the twentieth century, exerting progressively greater influence on the study of design methods and methodology. The accomplishments of design and the accomplishments of design methodology in the twentieth century are intimately related. Among those accomplishments, several stand out with potential for lasting significance in the future. One is, indeed, a deeper awareness of the nature of the artificial in our lives and the scope of what should be considered artificial or human-made, with a concomitant sense of human responsibility in its creation. Another is a legacy of serious reflection on the arts, methods, and techniques of designing, manifest in a pluralism of approaches to the problems and processes of thought and action in design. Closely related to this is the effort to reunite the arts of words with the mathematical arts of things: to remove the separation between theory and practice by bringing philosophic discussion about how we should lead our lives back into close relationship with the practical, concrete actions we may take in creating the human-made world. Finally, one of the most important accomplishments is a growing recognition of the dignity of design in human culture. What once was considered a servile activity without intellectual or philosophical grounding — a minor practice in society — has been transformed into a subject that is worthy of inquiry both in thought and in action, recognized for its service to human beings and the advancement of culture. 2

TRADITIONAL DESIGN METHODS AND THE RISE OF DESIGN EDUCATION

Although the philosophical roots of design may be traced to the Renaissance and the work of individuals such as Pico della Mirandola, Francis Bacon, and Galileo Galilei, design emerged in the twentieth century with two methods of practice. The first was the craft method, based on traditional practices of trial and error in the making of artifacts and the gradual evolution of product forms adapted to particular circumstances. This was an experimental method, first embedded in craftwork as a whole and then gradually separated from machine manufacture because of industrialization. Crafting prototypes, with anticipation of their eventual manufacture by machine methods, provided a way of exploring the forms and materials as well as the parts and wholes of products. The second method of design focused on drawing and draftsmanship. In this method, the designer sketches possible product forms that satisfy the needs of manufacturers and the market-

Thinking about Design: An Historical Perspective

415

place and then develops detailed scale drawings that can be used as instructions or specifications to guide manufacture and construction. Visualization through drawing has a long history in design, playing an important role in architecture, engineering, industrial design, and other branches of design. The strengths and weaknesses of this method are well described by John Christopher Jones in one of the most important works on design methods in the twentieth century [Jones, 1992]. In essence, drawing elevates the task of design thinking out of the hands of the craftsman, who is limited to the physical working of materials in a single case, and places the trial-and-error of craftwork in a medium that allows rapid experimentation and change. The advantages are several. By specifying parts and wholes, the drawing allows production to be divided among a variety of steps or processes. It also allows the designer to conceive of products that are too large and complex to be made in crafted prototypes. Finally, as Jones further points out, the drawing allows simultaneous production of different parts prior to assembly, thus speeding up the rate of production. One could add that drawing also allows careful consideration of the various engineering factors that bear on production, including physical laws, material properties, and, eventually, issues of power and electronics. The method of drawing supported standardization, upon which twentieth century mass production and manufacturing depended for its efficiency and economic advantage. Drawing was also essential — along with other forms and processes of the visual arts, including technological advances in areas such as photography — for the development of typography, printing and, eventually, mass communication as it came to be understood in the twentieth century [Meggs, 1983]. The two traditional methods of design remained important throughout the twentieth century in many branches of design, but they were gradually shaped, transformed and developed by different strategies of inquiry. One example emerged even in the late nineteenth century, and proved to be a harbinger of the social critique of design and technology in the twentieth century. Reacting against inferior design and the loss of quality in industrial products — and, by extension, the loss of quality in the practice of the industry trained designer — William Morris criticized the entire capitalist commercial system that put greed above quality and social responsibility. Poor quality, for Morris, was not a technical or artistic problem. Rather, it was a consequence of “the great intangible machine of commercial tyranny, which oppresses all of us” [Morris, 1914, p. 352]. This view led to the Arts and Crafts movement, which urged a return to the craft tradition of the past, based on intuition and individual artistic creation of whole products. It was not a pure return to the pre-industrial past, because industry was clearly part of the new circumstances of society and culture. Instead, craftsmanship became an opposing voice to industrialization. In this sense, it was an expression of a dialectical strategy of inquiry initially drawn from Marx and then applied through socialist ideas in a skeptical dialectical critique of the effects of industrialization and mass production on the human spirit in general and British culture in particular. The critique of the commercial system initiated by John Ruskin and William Morris and sustained by the Arts and Crafts movement led to new ways of think-

416

Richard Buchanan

ing about design and design methods in the context of industrial culture. It led to craftsmanship as an opposing voice to industrialization, but it also led to new efforts to incorporate art within industry, both in design practice and in educational programs that gave artists practical training for work in industry. This was the response of designers such as Henry Van de Velde and Peter Behrens, and it was also the response of educational efforts such as the Deutscher Werkbund that were directed toward the cultivation of applied art, with the various idiosyncratic methods and techniques that artists may employ in creation. This manner of treatment persisted throughout the twentieth century in many art schools and art academies and eventually in the ‘art and design’ departments of many universities. For others, however, the industrial and commercial system was not, in itself, the cause of poor quality in design and manufacture, and the answer to the problems of industrialization was not a return to craftsmanship or revived artistic design. Instead, the answer lay in improved design education, informed with the humanism of ethical and social purpose. And the key to improved education lay precisely in method. Walter Gropius, founding director of the Bauhaus, points toward a “synthetic method” that seeks to combine technical knowledge with artistic vision: The best kind of practical teaching is the old system of free apprenticeship to a master-craftsman, which was devoid of any scholastic taint. Those old master-craftsmen possessed practical and formal skill in equal measure. But as they no longer exist it is impossible to revive voluntary apprenticeship. All we can substitute for it is a synthetic method of bringing practical and formal influences to bear on the pupil simultaneously by combining the teaching of first-rate technicians with that of artists of outstanding merit. [Gropius, 1965, pp. 72–75] The method proposed by Gropius, and implemented at the Bauhaus, is clearly intended to be strategic in nature — an “approach” — as opposed to method as skill or technique. The teaching of a method of approach is more important than the teaching of skills. It should be a continuing process which must grow concentrically like the annual rings of a tree. In all its stages the scope should be all-embracing instead of sectional, increasing slowly in intensity and detail in all fields of discipline simultaneously. The integration of the whole range of knowledge and experience is of the greatest importance right from the start; only then will the totality of aspect make sense in the student’s mind. He will easily absorb all further details and place them where they belong if he progresses from the whole to the details, and not vice versa. [Gropius, 1970, p. 49–50] Furthermore, the method is an “architectonic art” for the modern world, where the art is “architectonic” in the ancient Greek sense of the word as it is used by Aristotle: providing organizing principles for all aspects of human activity, whether in thought or action.

Thinking about Design: An Historical Perspective

417

Thus the Bauhaus was inaugurated in 1919 with the specific object of realizing a modern architectonic art, which like human nature was meant to be all-embracing in its scope. It deliberately concentrated primarily on what has now become a work of imperative urgency — averting mankind’s enslavement by the machine by saving the massproduct and the home from mechanical anarchy and by restoring them to purpose, sense and life. This means evolving goods and buildings specifically designed for industrial production. Our object was to eliminate the drawbacks of the machine without sacrificing any one of its real advantages. . . Experiment once more became the center of architecture, and that demands a broad, co-ordinating mind, not the narrow specialist . . . Our guiding principle was that design is neither an intellectual nor a material affair, but simply an integral part of the stuff of life, necessary for everyone in a civilized society. [Ibid, pp. 19-20] The essential features of Gropius’s method are easy to discern in the curriculum of the Bauhaus. The famous Preliminary Course focused on experiments in practical issues of materials and manner or technique, taught by “technicians” and formally synthesized by “artists of outstanding merit.” As practical mastery developed, the student moved toward creating his or her own designs in the full context of a commercial environment, combining the elements of materials, manner, form and function in a synthetic or creative work suitable for production. In short, the intellectual and the material aspects of design were brought together in a pragmatic strategy of design as Inquiry, directed toward a practical outcome in a specific product. To support this strategy, however, Gropius also notes a further feature that he regarded as one of the most important contributions of the new Bauhaus method and its research. This was a grammar of design, based on the phenomenon of human sight and psychological experiences with form, space, and color. Establishing such a grammar was, for Gropius, a prerequisite for design, serving as “the controlling agent within the creative act” [Ibid., p. 43]. The grammar of design focused on the visual language employed by all designers, and it provided a common ground and understanding of the elements that are synthesized in form and turned to communicative and practical purpose by the designer. The formal expression of this method came in various speeches by Gropius throughout his career, in explanations of the new Bauhaus curriculum, and, later, in two books published in 1955 and 1956, after he had moved to the United States and joined the faculty of Harvard University. It was also expressed in Laszlo Moholy-Nagy’s Vision in Motion (1947). However, one of the most succinct expressions of the method is an article by Moholy-Nagy published in 1944 as “Design Potentialities” and later included in Vision in Motion [Moholy-Nagy, 1947].2 This short article is an important document for understanding the development of design methods in the twentieth century because it presents a clear illustration of the strategy of Inquiry based on the idea of design as poetics or productive science. 2 For

an abbreviated version, see [Kostelanetz, 1970].

418

Richard Buchanan

The article begins with the statement of a problem: the principle of “form follows function” and its degradation into a cheap slogan. The hypothesis is that advances in the natural sciences have provided new materials and methods of production (technology) that have confused and blurred the meaning of functionalism, requiring a new inquiry into its meaning in new circumstances. The remainder of the article is divided into two parts: analysis of the elements of design in the context of new knowledge, focusing on changing materials, manner of production, form, and function; and the problem of synthesis, based on the artist’s use of logic to understand the separate elements and intuition to grasp their relationships in the whole product as it is situated in society. The article concludes with a review of the tasks of the designer in the new circumstances of society and culture. The later Bauhaus, reconstituted amid the departure of Gropius and MoholyNagy and the turbulent political climate in Nazi Germany in the late 1920s, turned for a short time under the direction of Hans Meyer toward a Marxist dialectical approach that began to emphasize sociology, political theory, and the service of design to mass-production while eschewing the role of architectonic art in the sense developed by Gropius and Moholy-Nagy. The idea of design as poetics or productive science was put aside in the late Bauhaus in favor of design as part of a dialectical materialist science. Many of the ideas about design method developed by Gropius and others in the early Bauhaus found fertile ground in the United States, but there were few new formal developments in the theory of design methods. One reason for this was the tendency of designers and design educators in American schools to avoid theory in favor of highly practical, operational, and pragmatic approaches to practice. Ideas of design as ‘applied art’ or design as ‘applied science’ commonly provided a weak theoretical framework for design education and little opportunity for development. Indeed, the growing influence of the idea of applied science, Herbert Simon argued, almost drove design out of the curriculum of engineering schools [Simon, 1976, pp. 354–355]. Nonetheless, important design schools were established in the United States in the 1930s and 1940s, including programs at the Carnegie Institute of Technology and the New Bauhaus. In addition, two new philosophical and theoretical inquiries were underway that would have significant impact on the investigation of design methods. 3 THE STRATEGIES OF INQUIRY AND DESIGN SCIENCE Foremost among the new philosophical inquiries that would affect design was the work of John Dewey, regarded as one of the leading figures in the American philosophical movement known as Pragmatism. He published Art as Experience in 1934 and Logic: The Theory of Inquiry in 1938. The former book soon found use in art and design schools, including Laszlo Moholy-Nagy’s New Bauhaus venture in Chicago. The chapter “Having an Experience” later became one of the foundational frameworks for the development of the new practice of “interaction design” when design thinking was applied to human-computer interface and interaction

Thinking about Design: An Historical Perspective

419

at the Xerox Palo Alto Research Center (PARC), established in 1970 to create “the architecture of information.” Art as Experience employed the strategy of Inquiry to investigate the human-made world, yielding a new form of poetics or productive science. Instead of analyzing the factors involved in works of art and other products, as such, Dewey analyzed the factors involved in the experience of such works. Again, the factors were form, matter, function, and manner, but applied to experience. Manner, for Dewey, is both the creative thinking of the artist as well as the “reconstructive doing” of those who experience products in use and enjoyment. Form is the pattern of inception, development, and fulfillment in any experience; and the materials of form are signs and symbols, overt doings, and emotional qualities. Finally, emotion provides the unifying quality that distinguishes an aesthetic experience from a practical or intellectual experience, although all experiences involve intellectual signs and symbols and overt doings as well as emotional qualities [Dewey, 1958, pp. 54–55]. In Logic: The Theory of Inquiry, Dewey investigated logical method in general. However, the work provided insight into the deeper structure of method behind Art as Experience and laid a philosophic groundwork for design thinking and design as inquiry in the latter decades of the twentieth century, evident in the work of authors such as Donald A. Sch¨ on. In the Logic he distinguished between common sense inquiries and formal or scientific inquiries. I shall designate the environment in which human beings are directly involved the common sense environment or “world,” and inquiries that take place in making the required adjustments in behavior common sense inquiries . . . . the problems that arise in such situations of interaction may be reduced to problems of use and enjoyment of the objects, activities and products, material and ideological . . . of the world in which individuals live. Such inquiries are, accordingly, different from those which have knowledge as their goal. The attainment of knowledge of some things is necessarily involved in common sense inquiries, but it occurs for the sake of settlement of some issue of use and enjoyment, and not, as in scientific inquiry, for its own sake. [Dewey, 1964, pp. 60-61] The difference between common sense inquiries and formal scientific inquiries, Dewey argued, “resides in their respective subject-matters, not in their basic logical forms and relations; that the difference in subject-matters is due to the difference in the problems respectively involved; and, finally, that this difference sets up a difference in the ends or objective consequences they are concerned to achieve” [Ibid., pp. 114–115]. However, both kinds of inquiry follow a common pattern and are characterized by a single definition: “Inquiry is the controlled or directed transformation of an indeterminate situation into one that is so determinate in its constituent distinctions and relations as to convert the elements of the original situation into a unified whole” [Ibid., pp. 104–115]. Inquiry proceeds by identifying an inde-

420

Richard Buchanan

terminate or problematic situation, determining a problem and possible solution, developing a hypothesis through reasoning and experimentation, and reaching the objective or object of inquiry, which is to transform the indeterminate situation and its elements and relations into a unified whole [Ibid., pp. 105–119]. Dewey’s characterization of a problem is significant for later discussions of design method, because it points toward the importance of problem finding as well as problem solving. A problem is not a task to be performed which a person puts upon himself or that is placed upon him by others — like a so-called arithmetical “problem” in school work. A problem represents the partial transformation by inquiry of a problematic situation into a determinate situation. It is a familiar and significant saying that a problem well put is half-solved. To find out what the problem and problems are which a problematic situation presents to be inquired into, is to be well along in inquiry. To mistake the problem involved is to cause subsequent inquiry to be irrelevant or to go astray. Without a problem, there is blind groping in the dark. The way in which a problem is conceived decides what specific suggestions are entertained and which are dismissed; what data are selected and which rejected; it is the criterion for relevancy and irrelevancy of hypotheses and conceptual structures. On the other hand, to set up a problem that does not grow out of an actual situation is to start on a course of dead work, nonetheless dead because the work is “busy work.” Problems that are self-set are mere excuses for seeming to do something intellectual, something that has the semblance but not the substance of scientific activity [Ibid., p. 108]. For Dewey, common sense and scientific inquiries both employ symbols, but the symbols of common sense inquiries are part of a practical system that is constituted by “the traditions, occupations, techniques, interests, and established institutions of the group” [Ibid., p. 115]. This may be applied, for example, to the various branches of design practice, each with its own symbol system for design visualization, crafting, and so forth. However, scientific inquiries arose only when the meanings of symbols were freed from the concerns of a limited group and when problems of common sense required disinterested inquiry. Then, intellectual interest led to the investigation of meanings and their relations in systems of a general or abstract character, common to many or all communities of practice and use, and independent of particular times and places — though, of course, often with great potential for reference to and use in actual existential situations. In contrast to Dewey’s strategy of Inquiry, Herbert A. Simon pursued a positivist strategy that became one of the clearest expressions of the strategy of Design Science in the twentieth century. Instead of Dewey’s distinction between the problems of common sense inquiries and the problems of formal scientific inquiries — a distinction, for example, between the inquiries of design practice and the formal

Thinking about Design: An Historical Perspective

421

inquiry of poetics and productive science or any of the other natural, social, or behavioral sciences — Simon argued that methods are organized in a hierarchy that descends from treatment of the most complex phenomena to the simplest. At the highest level are the common practices of daily life, shaped around desires, values, preferences, and intentions. This is the area of complex human performance, involving problem solving, concept attainment, and a variety of other intelligent behaviors, including and often identical with the practices of design. While these behaviors may be intelligent, they are typically guided more by “rules-of-thumb” and “methods of the cookbook” than by logical analysis and proven methods of reasoning. Below the methods of complex daily behavior are the successive levels of social, behavioral, and natural sciences, with the symbol systems of mathematics and logic as the foundation of thinking. Simon’s distinctive contribution to design methodology lies in the identification of methods that are midway between the high level practices of daily life and the lower level methods that constitute the natural sciences. He called the middle area the domain of the “sciences of the artificial,” concerned with the theory of design, the logic of design thinking, and the variety of methods or tools that designers employ or may employ in their daily work. In essence, the sciences of the artificial seek to explain the complex behaviors and reasoning of designers, without entirely reducing them to the laws of natural science. The difference between the sciences of the artificial and the natural sciences involves the modality of their reasoning. The natural sciences are based on the modality of necessity, but the sciences of the artificial are based on contingency and possibility as well as necessity. Contingency comes precisely from the motivations of human beings that must be considered in the transformation of existing situations into preferred situations. The natural sciences do not and cannot adequately explain the domain of the artificial, which is a combination of natural laws and human preferences and motivations. Simon’s interest in design did not come directly from an interest in engineering, although it did eventually reveal new methods that could be applied particularly well to engineering design. Instead, it came from an interest in the social sciences, mathematics, and organizational behavior, beginning with his studies at the University of Chicago, where Dewey’s influence remained strong in many departments. While at Chicago his mentor was Henry Schultz, the econometrician and mathematical economist, but he also studied logic with the philosopher Rudolf Carnap, a leading proponent of logical positivism. During his undergraduate studies Simon formed an interest in decision-making in organizations, and this matured into his doctoral dissertation, later published as Administrative Behavior: A Study of Decision-Making Processes in Administrative Organization (1947). This work is significant in the development of design and design methods for two reasons. First, it focused explicit attention on organizations as products of design thinking, offering an alternative to the classics of organization theory and the methods of organization design developed in the first part of the twentieth century.

422

Richard Buchanan

Organization design is not unlike architectural design. It involves creating large, complex systems having multiple goals. It is illusory to suppose that good designs can be created by using the so-called “principles” of classical organization theory . . . To design an organization requires balancing competing and conflicting considerations and desiderata. No single design principle can be applied without taking account of the consequences of the numerous alternative principles that are also applicable. The examples provided in this book show how attention to the decision-making and communication processes provides a viable alternative approach to organization design, which dispenses with the classical “principles.” [Simon, 1945, p. xxii] In essence, a new problem area and a new subject matter for design thinking emerged from the book, opening the way for a variety of new design methods that would, throughout the remainder of the twentieth century, have significant influence on management practices — although the design methods were not always those advanced by Simon and his followers. Second, the book provided the genesis of Simon’s approach to design and the sciences of the artificial. As he remarks, the original problem posed in the book led him in steps to deeper and more basic questions, reminiscent of Dewey’s account of the rise of scientific inquiries out of the problems of common sense inquiries. When I tried, beginning forty years ago, to find answers to some questions of municipal organization . . . I discovered that no existing theory could provide the answers, and I was forced into an analysis of the ways in which organization affects human choice. Finding no better answers to this new stratum of questions, I began to reexamine the theory of rational decision-making. The latter task required me, in turn, to settle in my own mind some basic problems of logic. [Ibid., p. xv] From this work, Simon turned toward human cognition, decision-making, problem solving, artificial intelligence, and computer simulation, gradually leading to a new approach to understanding design and the sciences of the artificial. Simon’s most important work on design is The Sciences of the Artificial (1969). In a sense, this book could be regarded as a modern variation of Aristotle’s Poetics. Indeed, “the sciences of the artificial” is an unusual but not mistaken translation of “poetics” in Aristotle’s meaning of productive science or science of made-things. Like the Poetics, Simon’s book begins with the identification of the domain of the artificial, whose boundaries are set by four indicia, echoing the four causes that are employed by Aristotle in the Poetics to analyze made-things [Simon, 2001, pp. 3–5]. • Artificial things are synthesized by human beings (the efficient cause or genesis of made-things)

Thinking about Design: An Historical Perspective

423

• Artificial things imitate the appearance of natural things through physical symbol systems while lacking the reality of natural things (the material cause or means of made-things) • Artificial things involve synthesis and adaptation (the formal cause or object of imitation in made-things); • Artificial things involve imperatives (the final cause or function and purpose of made-things). These indicia and their interrelationships provide the organization or “method” of the book. The first chapter addresses the genesis of the artificial in the efforts of human beings to adapt, survive, and achieve in a complex outer environment that is governed by natural laws. The next three chapters address the means: the use of symbol systems in economics and psychology, resting “squarely on the hypothesis that intelligence is the work of symbol systems.” The next chapter addresses the science of design and man’s relation to his complex outer environment, focused on the problem of form and synthesis. Simon explains that this chapter could also be construed as an extension of the discussion of psychology, but in its detail it is a survey of modern design tools employed for synthesis and adaptation through the creation of functional forms in the artificial world. The final chapters address the extension of the design tools to further problems of complexity: social planning and the design of evolving systems, the general problem of complexity, and the architecture of complex hierarchical systems. If the broad framework of The Sciences of the Artificial echoes Aristotle’s approach to made-things in the Poetics, the strategy of investigation and the particulars depart radically from Aristotle — and from Dewey — and turn toward the positivist idea of the sciences and the reductionist method that Simon characterizes as science.3 For Simon, the science of design is a complement to the natural sciences, but its investigation is properly an investigation of psychology and how human beings relate and adapt to the complex outer environment of the natural world in which they seek to survive and achieve [Ibid., pp. 135-136]. The major function of invention and design is to describe an artifice in terms of its organization and functioning, and the best description of the functional synthesis created by the designer comes from “imitation.” Imitation is the central principle of Aristotle’s Poetics, because it provides the basis for methodical scientific inquiry into the nature of made-things. However, while Simon regards imitation as the basis of understanding of the artificial, he notes that imitation is better called simulation 3 Ibid., p. 172. “. . . we can adhere (and I will adhere) to reductionism in principle even though it is not easy . . . to infer rigorously the properties of the whole from knowledge of the properties of the parts. In this pragmatic way, we can build nearly independent theories for each successive level of complexity, but at the same time, build bridging theories that show how each higher level can be accounted for in terms of the elements and relations of the next level below. This is, of course, the usual conception of the sciences as building upward from elementary particles, through atoms and molecules to cells, organs and organisms.”

424

Richard Buchanan

in the modern world. Again, the principle and method of investigation depart radically from Aristotle, because it turns from a functional analysis of the elements of made-things toward a reductionist analysis of the act of simulation in physical symbol systems and psychological processes and mechanisms. If the interface between the inner and outer environment of a complex entity can be simulated, then one can claim that the interface is understood, at least in its actional behavior, independent of the materials in which the simulation is conducted. In this sense, the designer’s crafting of prototypes and efforts at visualization are all simulations that test the designer’s understanding of an artifact and the artifact’s ability to adapt the inner natural environment to the outer natural environment. However, Simon argued that computers and computer programs offer a new and fundamentally important tool for simulation because of their ability to rapidly manipulate symbols and trace the dynamic consequences of different artifact systems. Indeed, for Simon, the development of computers and artificial intelligence is a central feature of design theory and the basis of his hope for a science of design that is analytic, partly formalizable, partly empirical, and offers teachable doctrine about the design process [Ibid., p. 113]. In substantial part, design theory is aimed at broadening the capabilities of computers to aid design, drawing upon the tools of artificial intelligence and operations research . . . The need to make design theory explicit and precise in order to introduce computers into the process has been the key to establishing its academic acceptability — its appropriateness for a university. [Ibid., p. 114] Computers allow not only the simulation of artificial systems but also the simulation of the processes of thought that are involved in creating artificial systems. The connection between design method and the cognitive behavior of human beings lies in the use and manipulation of symbols and symbol systems in information processing. The computer is a member of an important family of artifacts called symbol systems, or more explicitly, physical symbol systems. Another important member of the family . . . is the human mind and brain. It is with this family of artifacts, and particularly the human version of it, that we will be primarily concerned in this book. Symbol systems are almost the quintessential artifacts, for adaptivity to an environment is their whole raison d’etre. They are goal seeking, informationprocessing systems, usually enlisted in the service of the larger systems in which they are incorporated. [Ibid., p. 22] This leads to analysis of cognitive behavior and the information processing system that lies at the core of intelligent behavior. Simon decomposed information processing into simpler components and mechanisms that could be modeled in computer programs, and then he sought to show how the components could be gradually unified to give coherence to Thinking Man and his work of designing.

Thinking about Design: An Historical Perspective

425

In contrast to the analytic techniques of the scientist, who explains phenomena by dissecting them and pulling them apart into simpler elements, the fundamental problem of design method is synthesis. “The techniques of the practitioner are usually called ‘synthetic’. He designs by organizing known principles and devices into larger systems” [Simon, 1945, p. 353]. Therefore, to transform design from an art into a science, the task is to develop “an explicit, abstract, intellectual theory of the processes of synthesis and design, a theory that can be analyzed and taught in the same way that the laws of chemistry, physiology, and economics can be analyzed and taught” [Ibid., p. 354]. Along this path, Simon identifies seven initial topics for a design curriculum and a theory of design, applicable in the first instance to engineering design but ultimately, he believed, applicable to all areas of design and all design problems. The topics are methods or tools of analysis and synthesis. They consist of heuristics, algorithms, and procedures of designing that decompose the overall design process into simpler steps or mechanisms, enabling the designer to make decisions and arrive at a coherent synthetic solution to a design problem. The topics include methods for evaluating designs (optimal or satisfactory solutions — “satisficing”), searching for alternative solutions, structuring artifacts and the design process, itself, into hierarchies of systems and sub-systems, and representing design problems. Taken together, the topics provide a formal theory of design or, where further research is needed to develop formal theory, they simply provide for treatment that is more pragmatic and more empirical than the vague methods typically employed by designers. For Simon, these methods replace the common understanding of design synthesis as intuitive, judgmental, and not fully explicit. But nowhere do we need to return or retreat to the methods of the cookbook that originally put design into disrepute and drove it from the engineering curriculum. For there exists today a considerable number of examples of actual design processes, of many different kinds, that have been defined fully and cast in metal, so to speak, in the form of running computer programs: optimizing algorithms, search procedures, and special-purpose programs for designing motors, balancing assembly lines, selecting investment portfolios, locating warehouses, designing highways, diagnosing and treating diseases, and so forth. Because these computer programs describe complex design processes in complete, painstaking detail, they are open to full inspection and analysis, or to trial by simulation . . . The programs are the tangible record of the variety of schemes that man has devised to explore his complex outer environment and to discover in that environment the paths to his goals. [Simon, 2001, p. 135] The ambiguity of “science” is evident in the relationship between the work of Dewey and Simon. For Dewey, science is a formal inquiry into symbols, meanings, and their relations, yielding a unified understanding of the phenomena that are the subject of inquiry. Analysis provides the functional elements of a problematic

426

Richard Buchanan

situation and synthesis provides the essential relations and connections that lead to unified understanding. For Simon, too, science involves formalization through analysis and synthesis, but these terms have distinctly different meaning from those of Dewey. Analysis of the elements for Simon leads to the simplest parts and mechanisms of phenomena, not the functional elements of Dewey. Synthesis, then, is the aggregation of the parts derived from analysis as they are gradually combined and unified in larger and larger systems, giving coherence to the understanding of phenomena. For Dewey, science resolves a problem of inquiry, contributing to the advancement of theoretical understanding and practical action. For Simon, science gradually builds explanations from the simplest phenomena to the more complex, making the wonderful and complex comprehensible to the mind and showing “how complexity was woven out of simplicity” [Ibid., p. 2]. The implications for design method are distinctly different. Dewey provides a philosophical foundation for productive science as well as the “common sense inquiries” of individuals such as Gropius and Moholy-Nagy and many others who pursue a strategy of Inquiry in exploring design in theory and practice. Science — e.g. “productive science” or any of the other natural and social-behavioral sciences — is relevant to practice, but the problems of science and the problems of design practice are different; the practicing designer uses the results of the various scientific inquiries to understand factors in the problematic situation, preliminary to the use of design method in the creation of solutions to practical problems. In contrast, Simon provides a theoretical foundation for those who pursue the strategy of Design Science and the application of cognitive psychology and artificial intelligence to the formation of a design theory that may reduce the complexities of design practice and put design thinking on an intellectual, rational pathway toward managing the non-rational, contingent motivations and intentions of human beings. Both strategies would soon be manifest in the Design Methods Movement in the United States and Europe, but also in later developments of design methodology. 4

DESIGN METHODOLOGY AND CRITICAL CONSCIOUSNESS

The Bauhaus idea of a modern architectonic art served a deeper purpose than providing a practical response to the problems of industrialization at the beginning of the twentieth century. It served a moral and ethical purpose that took shape in consciousness of the contradictions and horror of the First World War. Gropius writes, [T]he full consciousness of my responsibility as an architect, based on my own reflections, came to me as a result of the First World War, during which my theoretical premises first took shape . . . After that violent eruption, every thinking man felt the necessity for an intellectual change of front. Each in his own particular sphere of activity aspired to help in bridging the disastrous gulf between reality and ide-

Thinking about Design: An Historical Perspective

427

alism. It was then that the immensity of the mission of the architect of my own generation first dawned on me. [Gropius, 1970, p. 19] A similar consciousness emerged among many designers after the Second World War, during the reconstruction of Europe and the reintegration of industry into a peacetime economy in the United States and other countries around the world. The new advances in science, technology, and engineering that had been used in war were now available to peacetime industry, but the question was how they could be turned toward new social and cultural purposes in the design and construction of the human environment. Nowhere was this question more focused among designers than at the Hochschule f¨ ur Gestaltung Ulm, formally founded in 1953 but in planning since 1947 amid a scene of “utter devastation” yet with “unlimited curiosity” about all of the new disciplines of science and new ideas in the philosophy of science and mathematics [Maldonado, 1991, p. 222]. As at the Bauhaus, the central issue was method. Tom` as Maldonado, one of the leaders of the new school, writes: “The mainspring of all of our curiosity, our reading, and our theoretical work was our determination to find a solid methodological basis for the work of design” [Ibid., p. 222]. However, the strategic approach that shaped HfG Ulm reflected neither the strategy of Inquiry that informed the Bauhaus idea of an “architectonic art” and Dewey’s idea of common sense and formal scientific inquiries, nor the strategy of Design Science advanced by Simon in his idea of the “sciences of the artificial.” Instead, it was a strategy of Dialectic, in the form of a pragmatic, skeptical dialectic that was distinctly different from either idealist or materialist dialectics. Elements of both of the latter forms of dialectic were present from time to time in the discussions and controversies that are one of the hallmarks of the school, but the skeptical dialectic of HfG Ulm was practiced without prior commitment to any particular ideology. Nonetheless, it was grounded in one shared concept that served as the unifying principle for the otherwise exceptionally diverse community of the school. It was the idea “which we all shared in spite of disagreeing on absolutely everything else: the idea that industry is culture, and that there exists the possibility (and also the necessity) of an industrial culture” [Ibid., p. 223]. At HfG Ulm, industry was the active principle of culture, and the strategic challenge was to turn the active power of that principle in a new direction. The goal of the Ulm dialectic was not to merge theory and practice in a single, all encompassing system, as one finds in idealist and materialist dialectics. Rather, the goal was to overcome the dichotomy between theory and practice that became evident after the Second World War and was the source of many contradictions and divisions in social and cultural life before the war. The designers and design theorists at HfG Ulm wanted to bring theory to bear on practice in order to influence the actions of others, reducing theory from ideological rigidity and mistaken syntheses to new forms of practice so that scientific knowledge and critical consciousness could inform the work of design in addressing the problems of action in industrial culture. In short, the goal was to turn ideas toward practical action.

428

Richard Buchanan

The mere enunciation of new ideas — or supposed new ideas — is not enough. When we speak of defining our task in the age of struggle against food and housing destitution, it should be clear that the important fact is not to find an abstract (or literary) definition but an operational one: namely, a definition which is adequate to the demands of reality and which may help to guide our action successfully. Notwithstanding, an operational definition must not only clearly determine our objectives; it must also point out the methods for reaching them. And this is yet not all. Both objectives and methods, even when defined with the utmost empiric rigour, become transempiric, if from the start we do not question our technical, scientific or merely professional capacity to perceive those objectives and apply those methods. [Maldonado, 1989, p. 43] Such a skeptical, questioning dialectic required a new approach in design education. It would synthesize science and design in a new scientific humanism that recognized the pluralism of methods and methodological perspectives needed by the designer in addressing the new problems of industrial culture.4 In practice, the skeptical dialectic of HfG Ulm operated on two levels. On a theoretical level, discussion attempted to use the balance of alternative doctrines about design and design method to overcome conventional beliefs about design, industry, society, and culture that had become obstacles to innovation and action. This was the work of “critical consciousness” in raising the designer’s awareness of social and cultural responsibility, values, and new objectives. An array of theoretical positions mingled in the turbulent atmosphere of the school, but no unifying theoretical vision or system of design methodology emerged. What did emerge, however, was the beginnings of an operational view of design science — what Maldonado sometimes called a “scientific operationalism” — drawing on particular methods from many disciplines, gathered in an eclectic array. The final test of a particular method was not its philosophical coherence or the strength of its theoretical premises in basic science but its effectiveness in offering innovative ways to analyze design problems, leading in turn to innovative syntheses in products. This testing was the task of the practical level of the Ulm dialectic, where curricular discussion and studio experimentation attempted to introduce new methods and techniques in the studio and classroom. This was the work of operational, action-oriented design, informed with scientific knowledge and methods that could be applied in different ways to the solution of design problems in the human environment. Indeed, the skeptical dialectic gave way naturally to a strategy of 4 Ibid., p. 55: “This implies a challenge for the design educators. In the near future we will have to revise our position, not only our position, however, but our methods too: that is, we must develop our specific working methods, adjusting them to the specific types of problems that we shall have to solve. Thus we can successfully face the task entrusted to us by society: the reconstruction of human environment in the new era of scientific humanism. This will mean overcoming the contradictions between theory and practice, between knowledge and action, between consciousness and reality, between freedom and necessity.”

Thinking about Design: An Historical Perspective

429

rhetorical inquiry, based on the practical ability of designers to create a new human environment and influence the attitudes and behavior of human beings. The problems of design were repositioned as problems of industrial design, offering one of the broadest definitions of industrial design in the twentieth century. In dialectical terms, industrial design comprised two categories at HfG Ulm: objects (the means of mass-production, evident in the departments of Product Design and Building) and visual and linguistic signs (the means of mass-communication, evident in the departments of Visual Communication and Information). However, also in dialectical terms, both means were means of communication, broadly understood. The designer who commanded these means and the methods or techniques of their formation would have the power to influence behavior, attitudes, and opinions in industrial culture — provided he understood the conditions of power and management in industrial organizations [Gugelot, 1989, p. 16]. The HfG we are building in Ulm intends to redefine the terms of the new culture. Unlike Moholy-Nagy in Chicago, it does not merely want to form men who would be able to create and express themselves. The school at Ulm . . . wants to indicate what the social goal of this creativity should be; in other words, which forms deserve to be created and which do not. That is, generic modernity and creativity hold no place in its program . . . It is a widespread belief, at least, in certain circles, that the industrial designer, the planner who works for mass production, has only one function: that of catering to the sales program of large scale industry, while stimulating the mechanism of commercial competition. In contrast to this view, the HfG proposes that the designer, even while working for industry, must continue to absolve himself of his responsibilities with regard to society. (Maldonado, quoted in [Frampton, 1973, p. 35]) Methods from the natural and behavioral sciences, engineering, and mathematics were combined with concepts and methods drawn from rhetoric and semiotics, and all of these were further supplemented by ideas from political science and philosophy — particularly positivism but also various forms of social philosophy, including ideas from the Frankfurt School of critical theory and Marxist thought. In a sense, then, HfG Ulm sought to become a “university of methods,” in Maldonado’s phrase, borrowed from the American philosopher of pragmatism, C. S. Pierce [Maldonado, 1989, p. 55]. In fact, however, the broadening of the design curriculum at HfG Ulm followed the earlier broadening of the theoretical and scientific aspect of the design curriculum attempted by Moholy-Nagy at the New Bauhaus and supported by distinguished scientists and philosophers from the University of Chicago — an attempt cut short by financial difficulties and, subsequently, the untimely death of Moholy-Nagy [Buchanan, 1995, p. 39] (see also [Findeli, 1991]). The legacy of HfG Ulm for the development of design methods has several features. The school provided a meeting ground for individuals with widely different theoretical and practical perspectives, where new ideas could be discussed, tested,

430

Richard Buchanan

and developed in the classroom or in research. Individuals such as Bruce Archer and Horst Rittel, both central figures in the emerging Design Methods Movement, participated in the school, as did many others — students, teachers, design theorists, and others — who would carry the spirit of the inquiry into other environments. In addition, the school brought into a practical design context many concepts and methods developed in other disciplines, stimulating designers to experiment and consider their relevance to new design practices. For the most part, these were methods of analysis, but they provided the elements for new operational syntheses in studio projects. The new disciplines introduced at HfG Ulm included cybernetics, information theory, systems theory, communication theory, semiotics, and ergonomics. Also important were methods of planning in general and urban planning in particular, mathematical modeling, and operations research, with growing recognition of the potential of the computer to inform planning and decision-making. With the introduction of all of these ideas and methods, the school also gave prominence to the idea of design research as part of the culture of design thinking. Unlike Simon’s basic research in cognitive behavior, the research at HfG Ulm tended to be highly practical and related to industrial funding. Nonetheless, the school gave a rationale for a particular kind of research in design that grew elsewhere in the later decades of the twentieth century. Perhaps most important, the school raised an alternative possibility for a science of design. The idea of such a science at HfG Ulm was not based on Simon’s positivist sciences of the artificial, although positivist ideas did emerge in various ways. Similarly, it had no direct connection to Dewey’s understanding of the relationship between common sense inquiries and formal scientific inquiry, although it did have an echo of the American philosophy of pragmatism, mostly in reference to C. S. Pierce’s treatment of methodology and semiotic. Instead, it pointed toward scientific operationalism and a pragmatic strategy of rhetorical inquiry, informed with skeptical or critical consciousness of social and cultural purpose. However, the “scientific operationalism” of HfG Ulm was not unique to that institution. A pragmatic operational method — the testing of ideas by what they could accomplish in practice — was evident in engineering design from its earliest establishment and continued in the United States and other parts of the world throughout the twentieth century. Despite Herbert Simon’s argument that scientific theory and the natural sciences had nearly driven design out of the engineering curriculum in the twentieth century, one could still argue that engineering design was typically practiced through the operational application of scientific knowledge and mathematics. Indeed, an operational method was common throughout the design community and its many branches in the twentieth century, partly explaining the disdain for theory and the emphasis on style that many designers expressed. However, the skeptical attitude at HfG Ulm allowed a questioning of design and its theories and methods in a way that invigorated the field; it encouraged innovation and inquiry at a time when design faced significant challenges in each of its major problem areas.

Thinking about Design: An Historical Perspective

5

431

THE DESIGN METHODS MOVEMENT

The design school at Ulm provided an institutional setting for ideas that were, in fact, emerging in many places and among many individuals in Europe and the United States in the postwar period. In this sense, HfG Ulm was only part of a wider movement to understand design methodology, motivated by new problems and circumstances in the practice of design. The movement continued throughout the remaining decades of the twentieth century, gradually expanding from the narrow investigation of design methods and systematic design methodology into the broader enterprise of design studies and design research. However, if HfG Ulm provided an institutional setting for these ideas, a more important setting was provided by a series of conferences that began in 1962, marking the beginning of what came to be known as the Design Methods Movement. Conferences represent a different method of collective inquiry than institutional programs. Institutional programs are typically guided by a sustaining vision such as that of the Bauhaus, the New Bauhaus, or HfG Ulm. In contrast, conferences typically take shape around a theme or issue that provides a common ground for meeting and discussion. The pattern of the conference then follows the pattern of an idealized rhetorical debate, where individuals present their ideas and then discussion of the different perspectives represented among the participants contributes toward the advance of common understanding — although seldom to a single shared vision or system. Institutional discussions may be pluralistic to some degree, but conferences are pluralistic in essence because they include different perspectives with no obligation to shape a single dialectical or other monistic system. Conferences may be a step toward a system, but usually they are not. Rather, they attempt to stimulate innovative thinking that advances community inquiry. The conferences on design methods that began in 1962 were of this type.5 The common ground of these conferences was general agreement among the many participants that the practical work of designers should be made explicit in order to deal with the new problems and circumstances of design, particularly in the areas of industrial design, architecture, and urban planning. It was no longer adequate that design should be practiced in isolation by intuition and imagination, unsupported by the resources of new technology and scientific knowledge. Toward this end, they also thought it was important to find a new “scientific” basis for designing that would lead to systematic methods and, hence, to a methodology of design. Finally, they all expressed, in different ways, a concern for the context of design thinking, whether that context was management and organizational practices, new developments in technology and the sciences, or the individual and social life of the people to be served by design. However, the common ground of the conferences concealed differences in strategy and principle that would later lead to a reconsideration of the movement and its value. These differences were not entirely evident in the beginning, but they emerged gradually, reflecting the 5 For a collection of some of the key papers in the Design Methods Movement, see [Cross, 1984].

432

Richard Buchanan

pluralism of the field of design and design studies. This is well illustrated in the work of four individuals who were central figures in what became known as the Design Methods Movement: John Christopher Jones, Bruce Archer, Horst W. J. Rittel, and Christopher Alexander. While they worked from the common ground of the movement and often shared similar terminology, they also offered distinctly different views of design method and methodology. In “A Method of Systematic Design” (1963), the British designer John Christopher Jones sought a unified system of design that would integrate the two approaches that he observed to be in conflict in design in the postwar period [Jones, 1984]. One was the traditional method of intuition and experience, based in craft, drawing, and imagination. The other was the trend that emerged in the 1950s toward rigorous logical and mathematical methods. His goal was to develop a method that would reduce errors and delay in design and lead to more imaginative and advanced designs, particularly under special conditions where there were large quantities of information, the design team was free of routine design work and could focus on development, and significant departures from existing designs were required. To this end, he focused on the conflicting operations of the mind represented by creative, imaginative thought and logical analysis. Imagination, he believed, ranged over the whole of the design problem, while logic proceeded in a step-bystep method. The new method, he argued, would keep these two operations apart by external means, allowing each to work in its own natural way toward a design solution. The method had two elements. The first element involved leaving the mind alone to do its creative work. The second element involved a “system of notation” for recording all relevant information outside of memory, in essence separating design solutions from design requirements. Potentially supported at least in part by the memory capability of the new computer technology, recording followed three stages in the development of a design solution: Analysis, Synthesis, and Evaluation. Each stage was further divided into sub-stages, and the remainder of his essay presented the details for each stage, organized by a variety of topics, methods, and techniques that may be useful at each stage. This essay — as well as his later important book on the subject, Design Methods (1970) — was a virtual copia of methods and techniques that could aid the practicing designer in discovering and recording ideas that could support the imagination in solving a design problem. Like a Renaissance rhetorical copia, the topics or methods provided the means for amplification of the thinking of the designer; and like a rhetorical copia it was widely read by those seeking new practical ways of exploring design problems. It provided useful tools and distinctions that could be employed in many contexts. However, it is important to remember that for Jones, this was a discovery and recording aspect of design methodology intended to support creative imagination. Again like a rhetorical copia, it served as a background for the discovery of design arguments and solutions by the operation of the imagination. In contrast to Jones’s strategy of rhetorical inquiry, the British industrial designer Bruce Archer pursued a strategy of productive science that was strongly

Thinking about Design: An Historical Perspective

433

influenced by his emphasis on creative actional principles. It led to a variation of productive science that could easily be interpreted as the kind of operational science that Maldonado and others at the HfG Ulm sought to explore. Indeed, Archer taught at HfG Ulm from 1960 to 1962, developing and articulating the practical design methodology that was commonly associated with the school until it closed in 1968. The approach was published in “Systematic Method for Designers” (1963 and 1964) and became one of the cornerstones of the Design Methods Movement. Instead of focusing on the operations of the mind, as Jones did, Archer focused on the act of designing and the possibility of a method that would systematically organize the act in the face of growing complications found in the diversity of new materials and processes of production. In this sense, the problem with which Archer begins is strikingly similar to the problem with which Moholy-Nagy began his article, “Design Potentialities,” decades earlier — the growing diversity of new materials and new processes of production. However, the argument is significantly updated in the new circumstances. In the past, Archer argues, the art of design was close to the art of sculpture, based on a limited selection of materials and the shaping of those materials to meet the requirements of aesthetics and function as well as limited methods of production. In the postwar period, he argues, design has shifted its focus from the sculptural to the technological. Furthermore, in addition to new materials and more flexible methods of production, the new circumstances of design involve knowledge gained from new disciplines such as ergonomics, cybernetics, marketing, and management science — disciplines that had evolved after the war. Finally, he observes a trend toward an approach that is, he believes, characteristic of most technology: a “systems approach” rather than an “artifact approach.” The method of inquiry that Archer then employs is structured around the four functional elements of productive science — namely, the elements whose creative synthesis is the goal of productive science. It is instructive to compare Archer’s four elements with the four elements identified in Moholy-Nagy’s “Design Potentialities” and Simon’s four indicia of artificial things — and thence, of course, to the four causes of Aristotle. Each set of four is a variation of the elements of productive science. However, it is important to observe how the treatment of these elements in Archer is shifted toward the act of designing and the actions of the designer: • Formulating a prescription or model suitable for eventual embodiment — this is the genesis of a design idea. • Intention to embody the model in the means or materials called for in the situation. • Synthesis and the form of the solution, which may result from a “creative leap” or merely from a “non-creative” process of calculation.

434

Richard Buchanan

• Definition of the specific need or problem, based on the nature of the constraints of the situation — this, of course, is the goal or purpose of the design. Combined together, these elements yield Archer’s formal definition of design. So we find that design involves a prescription or model, the intention of embodiment as hardware, and the presence of a creative step. This definition embraces the central activities of architecture, most forms of engineering (including some systems engineering), certain sciences, all industrial design, and most applied art and craft. It implies a purposeful seeking after solutions rather than idle exploration. [Archer, 1984a, p. 59] This formal definition provides the basis, in turn, for a formal analysis of the method of systematic design. The method is divided into three phases of action: Analytic, Creative, and Executive. For Archer, the analytic phase is inductive, primarily involving data gathering. It begins with an “art of reconciliation” that seeks to overcome conflicts among the many factors that arise from the three main aspects of industrial design: function, marketing, and manufacture. Reconciliation also includes solution of all of the sub-problems into which a design problem may be decomposed, sometimes aided by the computer — as Archer observes, “Computer science is a very good means for setting out a problem in a systematic and logical way” [Ibid., p. 63]. Following the analytic phase, Archer then turns to the operational and procedural phases of the act of designing: getting the brief, examining the evidence, the creative leap, and the “donkey work.” The “creative phase” itself involves actions of analysis, synthesis, and development, described in a pattern that is strikingly similar to Dewey’s description of the response of a living creature to problems encountered through interaction with the environment [Ibid., pp. 65-6]. A key feature of Archer’s creative phase is the use of “heuristic” methods for discovery. Heuristic, as Archer points out, “is concerned with plausible rather than exact reasoning.” Referring to the source of his idea of heuristics, Archer notes: “Professor Polya emphasizes that, although plausible reasoning can and does produce solutions to problems, it cannot be represented as proof. If proof is needed, it must be worked out retrospectively” [Ibid., p. 66]. With the analysis made possible through observation, heuristic methods of questioning, evaluation, and judgment, it is then possible to begin a synthesis. Synthesis, for Archer, involves forming a judgment that may be subjective or based on deductive reasoning, and it yields a decision which is then subject to development through various types of modeling: drawing, crafting of prototypes, and other more abstract kinds of model making by computer or other tools. The final phase of the method is Executive, based on execution of the design and the need to communicate the design solution to management, production specialists, and the marketing function of an organization. Many readers of Archer’s essay focus their attention on the procedures and phases of the act of design, treating Archer’s lists as a kind of rhetorical copia

Thinking about Design: An Historical Perspective

435

or handbook similar to J. Christopher Jones’s work. However, Archer himself believed that the significance of his work was not in the identification of procedures or an array of methods and techniques for design. Rather, he believed that the significance was in how the discussion of methodology demonstrated and ensured that qualitative issues, ethics, beauty, and human values in general were a defensible part of the act of designing [Ibid., p. 75]. In short, the exploration of design methodology for Archer was the development of a principle of creative action, standing against mechanistic and unchanging principles of causality. Indeed, he argued against the idea of fixed values and unchanging laws of nature, preferring instead to celebrate the transience of design and the variety of values held by designers in all of their uses of methodology. Today, we can suspect that all the laws of nature are arbitrary, or even that there are no laws at all. . . . With scientists taking this sort of view of science, designers should be unembarrassed at accepting the transience of design. Human values, fashion, and public taste may well be describable in the same terms, if not on the same time scale, as probabilistic phenomena in physics. So may the mechanism of the creative leap. [Ibid., p. 76] Like Archer, Horst W. J. Rittel was a professor at the HfG Ulm during a key period of transformation of the school, between 1958 and 1963. With a background in mathematics, physics, and sociology, he introduced students and colleagues to a somewhat different perspective through teaching in methodology, theory of science, and operations research. There is no evidence of any direct conflict between Archer and Rittel, but there was a significant difference in their strategies of inquiry and design methodology. Indeed, Rittel explains that the genesis of his own approach to design methodology came when he tried, around 1960, to apply the new methods of design in the area of planning and found that they did not yield satisfactory results. This led him to reconsider the foundations of design methodology. Rittel’s work emphasized the early phase of planning and preparation for design work, based on operations research, systems thinking, and the use of information to support decision-making when dealing with complex problems. However, this also involved significant emphasis on human participation in the social process of decision- making within organizations. It was this latter aspect of design, the human perspective and social process, which gradually emerged to distinguish his approach to design and design methodology from that of Archer and others. To clarify his position, Rittel introduced a distinction between first- and secondgeneration design methods in systems analysis and design in general [Rittel, 1972]. Many in the design methods movement came to embrace this distinction for their own reasons, but for Rittel, the distinction allowed him to shift the ground of design theory and methodology, changing the central issues for discussion. It allowed him, first, to question the assumptions and beginning points of designers, and, second, to include a wider range of participants in the design process. As we have seen, others did not ignore these issues in the design methods movement. Jones and

436

Richard Buchanan

Archer, each in their own way, recognized and addressed the new complexity of design problems and the importance of including diverse knowledge and opinion in design methodology. However, from Rittel’s perspective, the other approaches did not regarded these issues deeply enough in the social context of design. Initially at least, Rittel’s view of design methodology seemed to stress the theoretical and philosophical dimensions of design thinking over the practical matter of addressing concrete problems. The main purpose of design methodology seems to be to clarify the nature of the design activity and of the structure of its problems. This role of design methodology seems to me to be much more important than its practical use in dealing with concrete problems. [Rittel, 1984, p. 317] This was in accord with the skeptical dialectic of HfG Ulm, where the goal was to bring theory to bear on practice. Yet, Rittel’s approach became more operational and action-oriented. He believed that discussion of the various assumptions and perspectives of those involved in the design process was, in itself, highly practical. Submitting an idea to organized criticism through debate and discussion, he argued, tends to insure that essential considerations are not forgotten. Furthermore, it tends to insure that when an idea is agreed upon, those who will be affected by the idea will also support it. With the goal of improving both the understanding and the practice of design, Rittel pursued what, in retrospect, was clearly a strategy of rhetorical inquiry. The strategy had some similarities with the rhetorical strategy of Jones but it also departed significantly from his work. Jones focused on the operations of the mind and the conflict between imagination and logic. Rittel, too, focused on the operations of the mind, but he did not recognize an inherent conflict between imagination and logic. He had a rhetorical understanding of logic rather than the commonplace idea of induction and deduction held by most of his colleagues. For Rittel, logic “is the rules of asking questions, generating information, and arriving at judgments” [Ibid., p. 323]. He turned toward design as argumentation, and he wanted to understand how designers actually reason and form judgments. In this context, imagination plays a central role: imagination is the place of designing, the place where arguments are invented and take shape before action.6 6 “Instead of immediately and directly manipulating their surroundings by trial and error until these assume the desired shape, designers want to think up a course of action thoroughly before they commit themselves to its execution. Designing is plan-making. Planners, engineers, architects, corporate managers, legislators, educators are (sometimes) designers. They are guided by the ambition to imagine a desirable state of the world, playing through alternative ways in which it might be accomplished, carefully tracing the consequences of contemplated actions. Design takes place in the imagination, where one invents and manipulates ideas and concepts instead of the real thing — in order to prepare the real intervention. They work with models as means of vicarious perception and manipulation. Sketches, cardboard models, diagrams and mathematical models, and the most flexible of them all, speech, serve as media to support the imagination” [Rittel, 1988, p. 1].

Thinking about Design: An Historical Perspective

437

For design methodology to move forward, Rittel believed that two areas required further work. One was the study of the logic of design reasoning and the development of an argumentative model of the design process. The other was the development of an instrumental version of the model, suited to implement a practical version that designers could use. Most of his work centered on the first area, developing a model of design argumentation, but he also had many suggestions about the instruments needed for practical design work and he developed a specific method that is employed in planning and some areas of design. Instead of identifying steps or phases of design methodology in the manner of Jones or Archer, Rittel identified four broad issues around which an argumentative model should take shape. The first issue was the determination of the problem to be addressed. Earlier methodology, he argued, made assumptions about the nature of design problems instead of investigating them more carefully. In his own work, he distinguished between “tame” or benign problems and “wicked” or malignant problems [Rittel and Webber, 1969, p. 160]. Others, such as Herbert Simon, had already made a similar distinction between “well-structured” and “ill-structured” problems, but Rittel’s colorful naming of “wicked” problems attracted attention and helped to elevate the issue within the planning community and then the broader design community. More important, his discussion of wicked problems pointed toward a radical indeterminacy in this type of problem that was not evident in the concept of “ill-structured” problems treated by Simon or others. Wicked problems are not merely underdetermined; they are essentially indeterminate, never yielding to the determinacy that belongs to “tame” or “well-structured” problems. This is because wicked problems are located in the conflicting perspectives, values, and interests of human beings. In contrast to tame problems, which yield to single solutions that may be true or false, wicked problems have no definitive formulation and yield to solutions that may only be judged good or bad. Rittel and Webber identified ten characteristics of wicked problems and discussed them in detail [Ibid., p. 161–7]. The second issue in the argumentative model of design followed from the distinction between tame and wicked problems. Tame problems, he says, are characteristic of many of the problems in the natural sciences, mathematics, and engineering, and they yield to solutions through classic experimentation, logic, and calculation — and to design methods of the first generation. However, tame problems, to the degree that they depend on social and human perspective, lie within deeper wicked problems of human interest, intent, purpose, and philosophic world-view. That is, all tame problems, when pushed far enough, reveal deeper wicked problems. Wicked problems, though often ignored or unrecognized, are common in design, and they require a different strategy of reasoning and a different understanding of logic. If there is a logic of design, it is not a formal logic. Rather, it is “a certain way of reasoning, a ‘philosophy’ guiding a mode conduct” [Rittel, 1988, p. 2]. Reasoning pertains to all those mental operations we are aware of, can even communicate to others. It consists of more or less orderly trains of thought, which include deliberating, pondering, arguing, occasional

438

Richard Buchanan

logical inferences. Imagine a designer thinking aloud, arguing and negotiating with himself (or with others), trying to explain or justify what he is proposing, speculating about the future consequences of his plan, deciding the appropriate course of action. [Ibid., p. 2] In this reasoning, he says, there is no separation of phases such as problem definition, synthesis, and evaluation. They occur together all of the time, and determining what the problem is is the problem. From the beginning, the designer has an idea of the ‘whole’ resolution of his problem which changes with increasing understanding of the problem, and the image of its resolution develops from blurry to sharp and back again, frequently being revised, altered, detailed and modified. [Ibid., p. 2] The focus of the second issue of the argumentative model, then, is an investigation of the common recurring patterns of design reasoning, the typical recurring issues and positions held by participants, and the typical “meta-issues” that lurk in the background of design reasoning [Ibid., p. 3]. Along with this comes study of the “fine-structure” of reasoning on individual issues, with particular emphasis on the invention of ideas and arguments. Finally, there is an understanding of the “epistemic freedom” of designers: there are no limits to the conceivable, and there is seldom a sufficient reason to dictate a particular course of action. The third issue of the argumentative model is an investigation of the varieties of reasoning that lie behind the great diversity of styles of designing and products in the world. Design, he argues, depends on the world-view of the designer, and the world-view is shaped by many factors, all of which should be subject to further investigation. The factors include beliefs and values, degree of power, what is accepted as a constraint, what is regarded as a reliable source of knowledge, professional training, personality, and cognitive style. And behind all of these factors lies the social context of designing and all of the stakeholders in the design process, each of whom are influenced by the same factors as the individual designer. The fourth issue of the argumentative model is the possibility of a science of design and the tasks of such a science. For Rittel, a science of design is possible, and it rests on developing new theories of design and deeper investigation of the reasoning of designers. This should include, he argues, empirical inquiries into how plans arise and how the effects of plans may be compared with their intended effects. Finally, the science of design should seek methods, tools, and instruments that bring the ideas of argumentation into practical action, amplifying human ability and protecting the design process from falling victim to covert individual values that may be smuggled into design projects without examination through debate and discussion. Rittel identified a variety of principles that characterized his “second-generation” design methodology [Rittel, 1984, pp. 324–326; 1972]. They are all principles in the sense of beginning points and guides to the conduct of argumentation and collaboration, similar to the guidelines that one finds in books on rhetorical method

Thinking about Design: An Historical Perspective

439

and process. They are, for example, distributed expertise, levels of issues, transparency of arguments, objectification, control of delegated judgment, the “conspiracy” model of planning, and systematic doubt. He also developed a particular instrumentality, called “Issue-based Information Systems” (IBIS), to provide method in addressing wicked problems in support of planning and political decision processes [Kunz and Rittel, 1970]. The method begins with an unstructured problem area, identified by a “topic” and “subtopics.” Out of the topic a discourse emerges, and in the course of discussion issues are identified, arguments are constructed for or against positions on the issues, and questions are raised, providing focus for further issues. In general, Kunz and Rittel distinguish four types of issues, expressed as questions. There are factual issues (“Is X the case?”), deontic issues (“Shall Y become the case?”), explanatory issues (“Is this the reason for that?”), and instrumental issues (“What are the appropriate means to accomplish an objective in a particular situation?”). Argument proceeds until opponents are convinced of a solution or until a formal decision procedure is activated. The IBIS methodology was a more robust version of Jones’ “system of notation.” It provided a way of recording or documenting the preliminary phase of design, when crucial decisions were reached. However, more than Jones, Rittel saw the methodology of discussion and debate as an intimate part of design process, where issues and arguments were not merely recorded but also invented in the contrast of positions and reasons offered to explain a proposed decision. “Issues” were the elements of the method, functioning as the “atoms” of design process for Rittel, in contrast to the functional elements of Bruce Archer and those who pursued a strategy of Productive Science. Indeed, Rittel may well have argued that functional elements could be decomposed into the smaller elements of issues — something that was consistent and already understood by Archer and others. The IBIS approach was applied to planning and political process, and it has had some use in engineering, architecture, and software development, particularly for large projects with conflicting interests among stakeholders, where it was often desirable to review the reasons for earlier decisions rather than reinvent the debate late in the design and development process. It subsequently influenced further work in the area of “design rationale,” the documentation of reasons for design decisions. However, IBIS was a labor-intensive method that even Rittel recognized did not have to be employed directly in many kinds of design projects.7 Nonetheless, it represented a step toward the form of a science of design that would have new instruments and encourage empirical research into the relationship between design intention and design realization. The work of Jones, Archer and Rittel represent variations in the strategy of Inquiry that played a significant role in the Design Methods Movement. Jones pursued a rhetorical strategy that emphasized memory and invention; Archer pursued a strategy of productive science that emphasized the act of designing; and Rittel pursued a rhetorical strategy that emphasized design as argumentation. 7 Developing suitable software was part of Rittel’s work on IBIS, and the effort has continued with various investigators and private enterprise.

440

Richard Buchanan

Christopher Alexander, however, pursued an entirely different strategy, the strategy of Dialectic. This was a challenging strategy, partly because many of the terms that Alexander employed were common in the design community, but used in the context of other strategies and methods. Because of this, his use of the terms was easily misinterpreted and misapplied. Alexander himself soon reacted against the mistaken translation of his ideas and approach into other ways of thinking about design. He repudiated the reduction of the central idea of his work — a rational, symbolic method for understanding the relationship between form and context, expressed in pattern diagrams — into mere “design methods” for creating those patterns. However, the strategy was also challenging because few individuals were familiar with the use of dialectic as a method of investigation and design. Indeed, little ingenuity was employed in the early period of the Design Methods Movement to understand the difference between a strategy of inquiry — an intellectual art that provides strategic guidance in thought and action — and the particular tactical methods and techniques employed along the way of an inquiry. Without such an understanding, it is easy to overlook the coherent development of Alexander’s work from the early period of his doctoral research and dissertation, Notes on the Synthesis of Form, to his later work, such as A Timeless Way of Building and A Pattern Language. Alexander begins not with the operations of the mind or the act of designing but with conflict and contradiction. The fundamental conflict is between form and context in the practice of design. Designer’s focus on form as the ultimate object of their work, but their understanding of the program or intended function of objects is often inadequate because it is merely intuitive or even arbitrary. Without adequate grounding in the true context of use, designers create products that are not fit for use either in the short term or in the long term. To overcome the conflict, Alexander begins with the general origins of form. There would be no form, he argues, if the world were regular and homogeneous; but since it is not, form emerges as a compensation for irregularity. “[A]n irregular world tries to compensate for its own irregularities by fitting itself to them, and thereby takes on form” [Alexander, 1964, p. 15]. This leads to the characterization of design problems and the approach of the inquiry in a search for patterns. The following argument is based on the assumption that physical clarity cannot be achieved in a form until there is first some programmatic clarity in the designer’s mind and actions; and that for this to be possible, in turn, the designer must first trace his design problem to its earliest functional origins and be able to find some sort of pattern in them. It is based on the idea that every design problem begins with an effort to achieve fitness between two entities: the form in question and its context. The form is the solution to the problem; the context defines the problem. In other words, when we speak of design, the real object

Thinking about Design: An Historical Perspective

441

of discussion is not the form alone, but the ensemble comprising the form and its context. Good fit is a desired property of this ensemble which relates to some particular division of the ensemble into form and context. [Ibid., p. 15–6] When the problems are simple, designers rely on intuition to grasp the fit between form and context. However, when the problems are complex, involving many conflicting needs, activities, and requirements, the problem cannot be grasped intuitively. A new method of analysis and synthesis is needed. The method that Alexander proposed had strong similarities with Plato’s dialectical strategy for approaching problems. This is evident even in the epigraph that Alexander selected for Notes on the Synthesis of Form: a quotation from the Phaedrus that describes the two procedures of division and collection that characterize Plato’s dialectical method. For Alexander, as well as for Plato, analysis and synthesis have an identity; they are two faces of the same method, one ascending from parts to the whole and the other descending from the whole to the parts. What does make design a problem in real world cases is that we are trying to make a diagram for forces whose field we do not understand. Understanding the field of the context and inventing a form to fit it are really two aspects of the same process. [Ibid., p. 22] From one perspective, then, design method is divided into three stages, following a common structure of dialectical inquiry in ascending from parts to the whole. The first stage is the deliberate and inventive search for conflicts in an environment. The second stage is the definition of individual geometrical relations or patterns that prevent those conflicts. The third stage is the combination of those relations to form a cohesive whole. In this approach, the context is a field of infinite constraints, demands, needs, and requirements. To narrow the field and make it manageable for design, Alexander used mathematical and logical structures to analyze and represent design problems, because “we cannot go on accepting the intuitive method innocently” [Ibid., p. 8]. However, he provided an interpretation of both mathematics and logic that is distinctly different from popular understanding. For example, he pointed out that modern mathematics “deals at least as much with questions of order and relation as with questions of magnitude” [Ibid., p. 7]. As such, it is a powerful tool for exploring conceptual order and pattern. In turn, he pointed out that logic is concerned with more than deduction and inference, which are little help to designers when they attempt to determine physical form. But, in speaking of logic, we do not need to be concerned with processes of inference at all. While it is true that a great deal of what is generally understood to be logic is concerned with deduction, logic, in the widest sense, refers to something far more general. It is concerned with the form of abstract structures, and is involved the moment we make pictures of reality and then seek to manipulate these pictures

442

Richard Buchanan

so that we may look further into the reality itself. It is the business of logic to invent purely artificial structures of elements and relations. Sometimes one of these structures is close enough to a real situation to be allowed to represent it. And then, because the logic is so tightly drawn, we gain insight into the reality which was previously withheld from us. [Ibid., p. 8] For Alexander, the nature of an object is determined by its components, not by an abstract or transcendent idea of form. The components that Alexander seeks, however, are not the functional elements that one finds in the work of Bruce Archer or in the strategy of Productive Science. He argues, for example, that the four functional elements of a city identified by the Congr`es International d’Architecture Moderne (CIAM) — work, dwelling, recreation and transportation — are not fundamental components at all. Rather, they are “macro-components” composed of still smaller elements. Design that remains at the level of functional elements, he says, merely leads to the arrangement and rearrangement of those elements, without creating new structures. To find the true elements — the “atoms” of environmental structure — it is necessary to decompose a system into the needs, active forces, and tendencies of the people for whom the system is designed. For Alexander, “need” is a vague and passive concept. He preferred to interpret need as an active force or tendency, an operational version of the concept of need; the idea of a need is better replaced with the idea of “what people are trying to do.” Identifying such tendencies is a painstaking business, because “no tendency can be stated in any absolute or final form.” Consistent with dialectical thinking, the statement of a need is always a hypothesis, subject to refinement over a long period of time and through the work of many independent observers. However, once the active tendencies are discovered in a situation or environment, the next task is to identify potential conflicts among them and then find geometrical patterns that characterize a physical “relation” that will prevent conflict from occurring. To this end, Alexander employs nonnumerical mathematics, graphs, and topology, yielding the “pattern language” and diagrams for which his work in architecture and urban planning is known. In practice, however, the two tasks of design — the identification of tendencies and conflicts and the discovery of patterns or geometrical relations that prevent such conflicts — do not follow in a strict linear sequence. In dialectical fashion, they develop together, ascending from conflicting parts or human tendencies, through patterns and combinations of patterns, to greater and greater wholes, ultimately leading to a dynamic whole that is rational, constructive, and evolutionary. The concept of wholeness is central to dialectic, and it is no less central for Alexander. Even in the early work associated with the design methods movement, he was at pains to explain the unifying idea that lay behind the decomposition of systems and the search for the atoms of environmental structure. We believe that all values can be replaced by one basic value: everything desirable in life can be described in terms of the freedom of peo-

Thinking about Design: An Historical Perspective

443

ple’s underlying tendencies. Anything undesirable in life — whether social, economic, or psychological — can always be described as an unresolved conflict between underlying tendencies. Life can fulfill itself only when people’s tendencies are running free. The environment should give free rein to all tendencies; conflicts between people’s tendencies must be eliminated. [Alexander and Poyner, 1984, p. 133] The work of decomposition and the gradual integration of individual patterns into larger structures is the physical expression of the ethical and scientific idea that lay at the heart of Alexander’s work. The implications of this idea become even clearer in Alexander’s later work, when the integration and interdependence of the atomic patterns discovered or invented by means of the decomposition and ascending method of his dialectic is placed in the context of design practice. In short, no pattern is an isolated entity. Each pattern can exist in the world, only to the extent that is supported by other patterns: the larger patterns in which it is embedded, the patterns of the same size that surround it, and the smaller patterns which are embedded in it. This is a fundamental view of the world. It says that when you build a thing you cannot merely build that thing in isolation, but must also repair the world around it, and within it, so that the larger world at that one place becomes more coherent, and more whole; and the thing which you make takes its place in the web of nature, as you make it. [Alexander et al., 1977, xiii] Notwithstanding Alexander’s explanation that patterns are hypotheses for discovering ways of solving physical problems in architecture and urban planning, it was easy to misinterpret Alexander’s pattern language as a rigid system of fixed categories for construction. This is part of what he objected to in his repudiation of the design methods movement, when he complained that so many of his readers focused on the method that leads to the patterns or diagrams rather than on the diagrams themselves. What Alexander discovered were not logical or grammatical categories, in the literal sense that readers may have assumed. Rather, what he discovered was a set of dialectical problems and dialectical places for dealing with those problems. In the intellectual tradition of dialectical reasoning, one would say that he discovered dialectical topoi or “topics.” The pattern diagrams are essentially places of discovery and invention for creative design practice, open to unlimited variation in practical situations. The elements of this language are entities called patterns. Each pattern describes a problem which occurs over and over again in our environment, and then describes the core of the solution to that problem, in such a way that you can use this solution a million times over, without ever doing it the same way twice. [Ibid., p. x] Alexander’s dialectical pattern diagrams — the “topics” and “methods” of his dialectic — are clearly different from the rhetorical copia of topics and design

444

Richard Buchanan

methods developed by John Christopher Jones and the issues of Horst Rittel. Yet, they are all intended to provide the designer with new tools for invention and discovery in practice. Their differences represent, in pr´ecis, the traditional difference — and sometimes conflict — between dialectic and rhetoric in Western culture. For Jones, the topics of his copia offer innovative perspectives on design problems, and their use strengthens the imagination of the individual designer and enhances his or her expressive voice. For Rittel, issues offer the focus of design argumentation. For Alexander, the patterns or topics of his approach offer hypotheses on problems that are embedded in the context to which the designer’s form is a response. Their use and exploration elevates the designer from personal and subjective opinion to an objective point of view that harmonizes with “the web of nature,” becoming “scientific.” The study and use of topics, therefore, leads in different directions. For Jones, it leads to a diverse design community, with each member pursuing different imaginative lines of thought and work. For Alexander, it leads to collective participation by those who are to benefit from design as well as the body of designers and design researchers who work toward the improvement of design. The body of known relations must therefore grow and improve. Design, if understood as the invention and development of relations, is no longer merely a collection of isolated and disconnected efforts. It becomes a cumulative scientific effort. [Alexander, 1984, p. 133] 6 DESIGN METHODS, METHODOLOGY, PRINCIPLES, AND CAUSALITY Despite the enthusiasm that animated the search for design methods and systematic methodology in the 1960s, by the 1970s each of the central figures of the Design Methods Movement had found reasons to be dissatisfied with the effort. Some cited the lack of impact on design practice; others cited remoteness from the actual practices of designers. Many were distressed by the emphasis that seemed to be placed on procedures and step-by-step, mechanical instructions — as if design could be practiced with a recipe. Others expressed concern that the methods identified paid too little attention to human values and creativity. All of these complaints had some legitimacy, but they were signs of a deeper consideration. There is a paradox in the various criticisms of the design methods movement, particularly by those who participated in the early stage of the movement. On the one hand, the writers appear to repudiate the entire enterprise of the movement, sometimes in scathing terms. On the other hand, the criticisms are precisely a continuation of the movement and perhaps even an elevation of the movement toward its ultimate goal. The reason for this apparent contradiction lies in the nature of method and methodology. Many of the participants and those who followed developments in the movement were motivated by a narrow pragmatic or operational interest in what could be done within the practice of design rather than motivated by an

Thinking about Design: An Historical Perspective

445

interest in the principles and goals of design. In short, they were more interested in finding pragmatic methods and techniques than in understanding the deeper issues of methodology and principle that guided design in thought and practice. Methods and techniques are about process and procedure, and methods possess an interpretive framework or direction of purpose that provides guidance in the use of techniques. Methodology, however, is the study of sets and systems of methods and techniques. And more than this, methodology is the study and analysis of the principles that organize a field and give purpose and value to its exploration. While the early papers of the movement appeared to be focused on methods and techniques, it was issues of principle, purpose, and value that guided the later criticisms of the movement. Indeed, it is the discovery and articulation of principles that is one of the important accomplishments of design methodology in the design methods movement. From this perspective, it is important to review the principles sought and established by the central figures of the movement. The widest difference of principles was between Alexander and Archer, who held strikingly different positions on the question of whether principles are unchanging or changing. Alexander, for example, used the strategy of dialectic and the method of decomposition to find the needs or active forces and tendencies that may come into conflict, requiring the designer’s intervention with pattern language to resolve conflicts, leading to the sane, constructive, and evolutionary attitude that is required for supporting human beings. The active forces and tendencies are the principles in his vision of design, since their unfolding play constitutes human freedom. He argues that all values can be replaced by one basic value: everything desirable in life can be described in terms of the freedom of people’s underlying tendencies — life can fulfill itself only when people’s tendencies are running free [Ibid., p. 133]. For Alexander, the underlying tendencies are unchanging over time — hence, the possibility of a “timeless way of building.” The unchanging nature of Alexander’s principles contrasts sharply with the principles that Bruce Archer discovered. For Archer, design may begin with a need, but human creativity is the principle of life as well as design. The notion of unchanging principles is completely alien to Archer. He writes, “Today, we can suspect that all of the laws of nature are arbitrary, or even that there are no laws at all” [Archer, 1989a, p. 76]. Quoting Max Planck to the effect that we have no right to assume that physical laws either exist or will continue in their current manner in the future, he observes: “With scientists taking this sort of view of science, designers should be unembarrassed at accepting the transience of design” [Ibid., p. 76]. Ongoing change, born of human creativity and embedded in the “creative leap” of design thinking, is the organizing principle for Archer. Considering the transience of design and the probabilistic nature of creativity, he writes: The technique of brainstorming in design is based upon the same principle. Here a group of people, well informed on various aspects of the problem in hand, allow their fantasies to run riot, triggering off in one another’s minds a torrent of ideas which no amount of logic could have

446

Richard Buchanan

produced . . . . This brings us back to our original point — setting up goals and criteria and spotting bright ideas are tasks which cannot be abdicated. [Ibid., p. 76]. Considering the implication of creative action, he observes the two kinds of creativity that distinguish art schools from technical colleges: “The creative phase of design has been most systematically developed in Britain in the design departments of art schools. The development phase is the great strength of the technical college” [Ibid., p. 78]. He also remarks, “Engineers may be weak in the systematic construction of the brief and in searching for original design ideas, but they are strong in the techniques of the development of detail” [Ibid., p. 79]. In the span of difference between Alexander and Archer, two other kinds of principles emerged from the design methods movement. The distinguishing feature of these principles, in contrast to the former, is the idea of wholeness and functioning as the cause and explanation of design activity. For Alexander and Archer, the causes of design lie outside of the art or science of design. They lie in the active underlying tendencies of people that require conflict resolution through design or they lie in the chance creative actions of human beings that enable humans to respond to situations of need. In each case, the tendencies or the acts are parts that serve as the cause and as the principled explanation of the whole of a design solution. Alternatively, for Rittel and Jones, the proper functioning of design is the goal and explanation of the development of design and design methodology. As we will see, however, there is a fundamental and principled difference between Rittel and Jones on the issue of what causes the proper functioning of the art or science of design. Does design function well when it is fully developed in its wholeness as a discipline or does it function well when the discipline of design is seen and understood in a whole that transcends the discipline of design, a whole that orders all of life?

Figure 2. Kinds of principles

Thinking about Design: An Historical Perspective

447

Rittel did not find the enterprise of the design methods movement to be a failure because for him it was part of the organic evolution of a field, unfolding much as the design for a product unfolds. The early phase was a testing of ideas that characterized the first generation of efforts to understand design methods. Those ideas were weak or ineffective, but they only set the stage for a later phase of second-generation design methods that came closer to the essential form of design methodology, a form grounded in “argumentation.” He writes, “The second generation deals with difficulties underlying what was taken as input for the methods of the first generation. For example, to set up a measure of performance or an effectiveness function is a focus in the second generation, while in the first generation that was considered an almost trivial task” [Rittel, 1984, p. 322]. This is the case, too, with the issue of participation. “The nightmare of the first generation, implementation difficulties, should disappear or at least be minimized in the second generation . . .” [Ibid., p. 322]. Rittel shares with Archer an interest in uncovering the discipline of design. However, for Rittel this discipline is more than a discipline of method grounded on creative action. Rather, it is a discipline grounded on human nature and human ability — grounded, that is, not on arbitrary foundations but on the natural capability of the human mind to reason and support the human community in its ongoing development. Rittel’s strategy of inquiry is not to organize the act of design. The strategy is to reveal how designers reason and share their reasoning with others through words, symbols, and artifacts. The focus for Rittel is on mental activities and operations involved in argumentation, not on action itself. In turn, the designer’s reasoning is grounded on his or her worldview, which in turn is grounded on the social context where alternative worldviews interact and compete through the ongoing process of argumentation. In short, there is for Rittel a principle of reflexivity in the circle from mental activity to argument and social process and back again. It is the wholeness of the functioning of design that constitutes the principle. And it is on this basis that Rittel forms the idea of a science of design. The science of design has three tasks. First, to further develop the theories of design, to learn more about the reasoning of designers. Secondly, it should pursue empirical inquiries into how plans come about, and what the effects of plans are in comparison with what they intended. Finally, on this basis, it should look for tools to support designers in their work. The human mind is fallible. Methods should be sought to amplify its abilities, even if its [sic] only to keep us from falling prone to our idiosyncracies [sic]. [Rittel, 1988, p. 7] While Archer’s focus was on the principle of creative action as the basis for developing a discipline of design, the focus of Nigel Cross’s work was on the natural principle of human “design ability.” While they employed different strategies of inquiry in their work — instead of a rhetorical strategy of inquiry, Cross’s strategy was cognitive inquiry into the functioning of the mind — both Cross and Rittel

448

Richard Buchanan

saw a natural ground for design ability. Indeed, for Cross, design ability is “a multifaceted cognitive skill, possessed in some degree by everyone.” He goes on to argue that design ability is, in some sense, a natural form of intelligence. “I believe that there are particular, ‘designerly’ ways of knowing, thinking, and acting. In fact, it seems possible to make a reasonable claim that design ability is a form of natural intelligence, of the kind that psychologist Howard Gardner has identified” [Cross, 1995, pp. 115–116]. In addition to pursuing his own studies of design ability, Cross also worked to stabilize a discipline of design science through such vehicles as the journal Design Studies and the book Developments in Design Methodology, which gathered many of the initial papers of the design methods movement. These efforts have helped to sustain one direction of new work growing out of the design methods movement, bringing together a group of researchers who do not necessarily share the same philosophic perspective on design but who do share a loose affiliation in the strategy of a science of design and the principles of such a discipline. Rittel, Archer, Cross and others who pursued the strategy of inquiry — whether rhetorical inquiry or productive science — found the beginning and end of design in the approach taken to problems and in the designer’s understanding of the nature of the problems to be solved. For Rittel, the rhetorical strategy leads to the characterization of “wicked” problems that are essentially indeterminate in nature. For Archer, Cross and others, the design problem may be characterized either as “well defined” or “ill-defined,” and the task of design is to discover and implement a solution. In “Discovering Design Ability,” Cross summarizes this perspective: Design ability is founded on the resolution of ill-defined problems by adopting a solution-focusing strategy and productive or appositional styles of thinking. However, the design approach is not necessarily limited to ill-defined problems. Thomas and Carroll conducted a number of experiments and protocol studies of designing and concluded that a fundamental aspect is the nature of the approach taken to problems, rather than the nature of the problems themselves: ‘Design is a type of problem solving in which the problem solver views the problem or acts as though there is some ill-definedness in the goals, initial considerations or allowable transformations’. [Ibid., pp. 110–1] In contrast to Rittel and others who found the principles of design to lie in design ability, the functioning of the discipline of design, the strategy of approaching design problems, or the worldview of the designer, John Christopher Jones discovered a different kind of principle: a single comprehensive principle of wholeness that is only dimly perceived by designers and others in the human community and that is best approached through intuition, chance, and poetic insight, recognizing the interconnection of more factors than reason can grasp. Nonetheless, his career began in much the same place as those who sought to develop functionality in the discipline of design; he pursued what he regarded as a sound rational ap-

Thinking about Design: An Historical Perspective

449

proach. Later, writing in the poetic and Delphic style that came to characterize his work beginning in the late 1970s, he tells the story of his gradual discovery of the comprehensive principle that transcends other principles and orders life in its own logic, different from the narrow logic of human beings [Jones, 1984, p. 330]. His earlier book, Design Methods, sought to relate all the design methods to each other and to experience. Along the way, however, he found limitations in the principles that grounded such work, and he found a fundamental limitation in design thinking around the division between intuition and rationality — where rationality fails to place design solutions in a wide enough context of experience and life [Ibid., p. 331]. Like others at the time, Jones was interested in the nature of design problems. However, unlike those who focused on wicked problems or ill-defined problems, he focused on the instability of problems and the principled significance of uncertainty. He takes care to explain the significance of instability and its difference from the certainties sought by others whom he calls rationalists — namely, that instability was a sign of widening the problem to the full context that must be addressed by design. Attempting to reduce a problem to any smaller whole, even though such attempts may appear rational, damages the opportunity for a satisfactory solution. Where is the essence of the subject? For me the word in the index with the most sub-entries to it Is ‘Instability of Design Problems’ Which has about ten entries The whole problem becomes more unstable as you widen it As you take more and more of life to be part of the problem you don’t get a more stable problem you get a less stable problem. And this I think is not what the rationalists like. I think that people who approach this subject because it seems rational are those who like certainty in life. If you wish for certainty you might as well leave this subject alone Because design is to do with uncertainty As far as I can see But a lot of people who do wish for certainty do dabble in it And I fear they’re wrecking the subject. [Ibid., p. 332] It is not that Jones opposed reason in design. Rather, he saw its limitations in apprehending the comprehensive principle of order in life. Widening the design problem, rather than narrowing it, was the strategy that he explored first through rhetoric and later through poetry and the arts, including music. This led him to explore design in a different direction than Rittel and those who sought a science of design. Jones focused on design as a social art, linked to the work of experimental artists and to the idea of chance [Ibid., p. 335]. To complete the circle from parts to wholeness, there is both similarity and contrast between Jones and Alexander on the nature of principles. Both men dis-

450

Richard Buchanan

covered principles that they regarded as unchanging and, in a sense, eternal or timeless. However, they differed on the issue of the primacy of the parts or the whole. While Alexander’s principle resided in the parts — the many underlying active tendencies of human beings — Jones’s principle was a single unifying idea, a comprehensive principle that transcends human reason and that can only be approached through chance and inspiration in continuously playing out the interrelations, integrations, and balances of all things in human experience. The contrast with Archer is also apparent, since Archer’s creative acts are arbitrary and diverse while Jones’s “No full stop” is an ongoing dynamic process of the whole. The different strategies of inquiry and the different kinds of principles discovered in the design methods movement help to explain why the movement lost coherence and fractured in so many different directions by the 1970s and 1980s. However, it is important to note that a residue of the movement remained in tact, gathered around the effort to strengthen the foundations of design as a discipline, following the type of reflexive principles articulated by Rittel, Cross, and others. Furthermore, there were continuing participants who wanted to follow the strategies of inquiry that would lead to some form of a science of design — whether an operational science, or a productive science, or a form of design science. These two groups shared enough common ground that they persisted and gradually strengthened as a community through the Design Research Society, based in the United Kingdom, and through other venues of discussion and publication. Indeed, Bruce Archer, perhaps the most persistent optimist of the movement, combined his criticisms of the earlier work with an argument that the design methods movement had not died. “Design methodology is alive and well, living under the name of design research” [Archer, 1984b, p. 347]. In his view, methodology resurfaced in the design community in the many forms of research: “design history, design philosophy, design criticism, design epistemology, design modeling, design measurement, design management, and design education” [Ibid., p. 349]. The design methods movement of the 1960s and 1970s occupies an important place in the development of our understanding of design methodology in the twentieth century, with implications for our understanding of design research. However, just as the “scientific operationalism” of HfG Ulm was not unique to that institution, so too the development of design methodology and the emerging enterprise of design research owe less to the design methods movement than Archer suggested. The investigation of design was broadening around the world, and for reasons that had little to do with the design methods movement. Growing interest in design as a field of emergent new practices, a factor in economic development, a subject of historical inquiry, a partner in technological development, and an intellectual and philosophical problem helped a larger community to emerge in the United States and other parts of the world. Furthermore, there was new interest in understanding the interdisciplinary relationships among design, engineering, and management. One of the most interesting developments of the 1980s was the reemergence of the strategy of Inquiry, taking the form of reflective practice and organizational

Thinking about Design: An Historical Perspective

451

learning [Sch¨ on, 1983]. Drawing heavily on Dewey’s theory of inquiry, Donald A. Sch¨ on offered a strong, popular critique of what he called “technical rationality” in the work of Herbert Simon and many others who sought to apply scientific theory and technique to professional practice. In an influential account of how professionals think in action, Sch¨ on focused on design as a “reflective conversation with the situation,” illustrating the gradual give and take of problem formulation, experiment, and reformulation. While this approach does, indeed, reflect some of the central themes of Dewey’s theory of inquiry, a careful reading of The Reflective Practitioner suggests that Sch¨ on has interpreted the strategy of Inquiry as a kind of skeptical, questioning dialectic, applicable not only to design but to all forms of professional practice. Or, seen from a different perspective, he has broadened design from its narrow application and made it the model of all professional practice — much as Simon, too, sought to extend design to all professional life, but on quite different grounds. Sch¨ on’s work is representative of a significant intellectual trend in the closing two decades of the twentieth century. It was a trend toward rhetoric and dialectic in the study and practice of design. Indeed, it is representative of a loose affiliation of scholars, researchers and design practitioners held together by broad themes such as communication, invention, management, and the social and behavioral sciences. This affiliation had no formal or informal name, but it could be characterized generally as the “humanistic” camp in design. In contrast to this camp, another loosely affiliated group of designers and design researchers continued to pursue the strategy of design science, whether in engineering or in other branches of design. Design methodology was pursued within this group in various forms, including “general design science,” “systematic engineering,” and “engineering science,” but also in various forms of cognitive psychology and quantitative analysis directed toward graphic and industrial design. Despite the differences between these two groups and the gulf of understanding that sometimes seems to divide them, their collective work has served to raise the dignity of design in modern culture. Considering that at the beginning of the twentieth century design was considered merely a minor craft activity on the margins of culture, it is remarkable to see the attention it receives from a variety of quarters at the beginning of the twenty-first century.

BIBLIOGRAPHY [Alexander, 1964] C. Alexander. Notes on the Synthesis of Form. Cambridge: Harvard University Press, 1964. [Alexander et al., 1977] C. Alexander, S. Ishikawa and M. Silverstein. A Pattern Language: Towns, Buildings, Construction. New York: Oxford University Press, 1977. [Alexander and Poyner, 1984] C. Alexander and B. Poyner. The Atoms of Environmental Structure. In, ed. N. Cross, Developments in Design Methodology. New York: John Wiley & Sons, 1984. [Archer, 1984a] L. B. Archer. Systematic Method for Designers. In, ed. N. Cross, Developments in Design Methodology. New York: John Wiley & Sons, 1984.

452

Richard Buchanan

[Archer, 1984b] L.B. Archer. Whatever Became of Design Methodology? In, ed. N. Cross, Developments in Design Methodology. New York: John Wiley & Sons, 1984. [Buchanan, 1995] R. Buchanan. Rhetoric, Humanism, and Design. In, ed. by R. Buchanan and V. Margolin, Discovering Design: Explorations in Design Studies. Chicago: University of Chicago Press, 1995. [Cross, 1984] N. Cross, ed. Developments in Design Methodology. New York: John Wiley and Sons Ltd, 1984. [Cross, 1995] N. Cross. Discovering Design Ability. In, ed. R. Buchanan and V. Margolin, Discovering Design: Explorations in Design Studies. Chicago: University of Chicago Press, 1995. [Dewey, 1958] J. Dewey. Art as Experience. New York: Capricorn, 1958. [Dewey, 1964] J. Dewey. Logic: The Theory of Inquiry. New York: Holt, Rinehart and Winston, 1964. [Findeli, 1991] A. Findeli. Design Education and Industry: The Laborious Beginnings of the Institute of Design in Chicago. Journal of Design History vol. 4, no. 2, 1991, pp. 97-113. [Frampton, 1973] K. Frampton. Apropos Ulm: Curriculum and Critical Theory. Oppositions, No. 3, 1973. [Gropius, 1965] W. Gropius. The New Architecture and the Bauhaus. Cambridge: MIT Press, 1965. [Gropius, 1970] W. Gropius. Scope of Total Architecture. New York: Collier Books, 1970. [Gugelot, 1989] H. Gugelot. Industrial Design in Practice. In, ed. by Kirti Trivedi, Readings from Ulm: Selected Articles from the Journal of the Ulm School of Design (Hochschule fur Gestaltung, Ulm). Bombay: Industrial Design Centre, 1989. [Jaspers, 1951] K. Jaspers. The Way to Wisdom: An Introduction to Philosophy. Trans. R. Mannheim New Haven: Yale University Press, 1951. [Jones, 1984] J. C. Jones. A Method of Systematic Design. In, ed. N. Cross, Developments in Design Methodology. New York: John Wiley and Sons Ltd, 1984, pp. 9-31. [Jones, 1984] J. C. Jones. How My Thoughts About Design Methodology Have Changed During the Years. In, ed. N. Cross, Developments in Design Methodology. New York: John Wiley and Sons, 1984. [Jones, 1992] J. C. Jones. Design Methods. New York: John Wiley, 1992. [Kostelanetz, 1970] R. Kostelanetz, ed. Moholy-Nagy: An Anthology. New York: Da Capo Press, 1970, pp. 81-90. [Kunz and Rittel, 1970] W. Kunz and H. W. J. Rittel. Issues as Elements of Information Systems. In Working Paper No. 131. Berkeley: Institute of Urban and Regional Development, University of California, 1970. [Maldonado, 1989] T. Maldonado. The Emergent World: A Challenge to Architectural and Industrial Design Training. In, ed. by Kirti Trivedi, Readings from Ulm: Selected Articles from the Journal of the Ulm School of Design (Hochschule fur Gestaltung, Ulm). Bombay: Industrial Design Centre, 1989. [Maldonado, 1991] T. Maldonado. Looking Back at Ulm. In H. Lindinger, Ulm Design: The Morality of Objects. Cambridge: MIT Press, 1991. [McKeon, 1998] R. McKeon. Philosophy and the Development of Scientific Methods. Selected Writings of Richard McKeon, Volume 1, Philosophy, Science, and Culture, ed. by Zahava K. McKeon and William G. Swenson. Chicago: University of Chicago Press, 1998. [Meggs, 1983] P. B. Meggs. A History of Graphic Design. New York: Van Nostrand Reinhold, 1983. [Mitcham, 1994] C. Mitcham. Thinking Through Technology: The Path between Engineering and Philosophy. Chicago: University of Chicago Press, 1994. [Moholy-Nagy, 1947] L. Moholy-Nagy, Vision in Motion. Chicago: Theobald, 1947. [Morris, 1914] W. Morris. Art and Its Producers. In Collected Works of William Morris, Vol. XXII, London, 1914. [Rittel, 1972] H. W. J. Rittel. On the Planning Crisis: Systems Analysis of the ‘First and Second Generations’. Bedrifts Økonomen, No. 8, October. 1972, pp. 390-396. [Rittel, 1984] H. W. J. Rittel. Second-generation Design Methods. In, ed. N. Cross, Developments in Design Methodology. New York: John Wiley & Sons, 1984. [Rittel, 1988] H. W. J. Rittel. The Reasoning of Designers. Arbeitspapier A-88-4 (Stuttgart: Institut f¨ ur Grundlagen der Planung). Universit¨ at Stuttgart, 1988.

Thinking about Design: An Historical Perspective

453

[Rittel and Webber, 1969] H. W. J. Rittel and M. M. Webber. Dilemmas in a General Theory of Planning. Panel on Policy Sciences, American Association for the Advancement of Science, 4, 1969. [Sch¨ on, 1983] D. A. Sch¨ on. The Reflective Practitioner: How Professionals Think in Action. New York: Basic Books, 1983. [Simon, 1976] H. A. Simon. Administrative Behavior: A Study of Decisioin-Making Processes in Administrative Organization. New York: The Free Press, 1976. [Simon, 2001] H. A. Simon. The Sciences of the Artificial. 3rd Edition, Cambridge: MIT Press, 2001.

This page intentionally left blank

TYPOLOGIES OF DESIGN PRACTICE Kees Dorst and Kees van Overveld

1

INTRODUCTION

Design is a human activity in which we create plans for the creation of artifacts that aim to have value for a prospective user of the artifact, to assist the user in his/her effort to attain certain goals. These goals can be purely functional, or they can encompass a broad array of cultural and social aims. More often than not, these three kinds of goals go together within one problem situation and they result in contradictory requirements to the design. The creative exploration and negotiation of this broad terrain is the expertise of the designer [Cross, 1990; 2006]. The design project takes place within a volatile real-world context, where the design situation may change any moment (technological advances, policy changes within the organisation, changes in the market) and system borders are hard to distinguish. The output of the design project consists of a ‘design’. That is, a description of the form, function and working of the designed artefact, a description of the interface between this artefact with the outside world, and a ‘use plan’, outlining how the artefact is meant to be used, and within which context it is meant to operate [Houkes, 2002]. The work of the designer also includes the creation of a justification to accompany the design. This is crucial because design is never done in isolation: designers create blueprints for artefacts that other people manufacture, sell and perform maintenance upon, and that aim to enhance the wellbeing of the most important ‘other’ of all, the prospective user. Design projects vary tremendously in purpose, methodology and domain. The differences between, say, designing a household commodity, the organisational structure of a new business, or a symphony are larger than their similarities. Various design sub-disciplines each have developed their own approaches, their own terminologies and their own sets of criteria, stemming from widely different views on what defines ‘quality’ in a design domain. Also within a single design project, there is a large variation of issues. For instance, decisions relating to the mechanical structure of a household implement relate to its manufacturing, which relates to the cost price, which relates to the intended market, which relates to the advertisement strategy. The strings of decisions a designer takes are heavily intertwined. The very complexity, creativity and open-endedness that makes design a fascinating occupation for some, also is highly problematic when we try to observe, Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

456

Kees Dorst and Kees van Overveld

describe and make sense of design. The participants in any discussion on design are bound to have very different conceptions of ‘design’ in their heads — this often thwarts attempts toward a substantial in-depth discussion of the subject. In this paper we will seek to unravel some of this Gordian Knot by exploring the possibility of creating a typology of design practice. If we could agree on such a shared typology, that would clear the way for a strong and clear nomenclature around design. This is not straightforward: designing is a complicated set of activities, taking place in many professions. Within this field of design, we find a cluster of actually quite different professions that traditionally have always defined themselves by the material they used (‘textile design’) by the medium (‘sound’) or by the nature of the designed object (‘mass produced consumer product’). Yet these distinctions are superficial in that they do not touch the nature of the design activity. One can think of many other criteria — besides professional categories — to discern types of design practice. In this paper, we set out to propose different typologies of design practice that are based on deeper, underlying variables that we will argue directly impact the nature of the design activity. Yet a typology of design practices should not only be based on fundamental differences between the types of design activities, but also be close enough to the reality of design in practice to be based on the salient aspects of a design case that is happening in the world. We are searching for a descriptive framework and typology for design that is both fundamental, valid and relevant. What would be the elements of such a descriptive framework for design activity? To describe design activity in its full (real-world) complexity one would need to describe the object of this activity (in this case, the design problem and the emerging design solution), the actor (the designer or the design team/designing organization), the context in which the activity takes place (as far as its impacts upon the activity are concerned) and the structure and dynamics of the complex of activities that is being studied (‘the design process’ ) [Dorst, 1997; Reymen, 2001]. When we look at the design methods and tools that are being developed within the design research community, we see that three of these four ‘aspects of design activity’ are often ignored within the descriptive framework that implicitly underlies our thinking on design. The overwhelming majority of descriptive and prescriptive work in design research focuses on the design process, to the exclusion of everything else. Therefore the design methods and tools that are being developed inevitably focus on enhancing the efficiency and effectiveness of design processes. The models of design processes that have been developed over the years have proven to be very powerful tools in the development of design practice and design education. Yet the process is just one aspect of the total design activity, and this one-sided approach to design research may have hampered attaining deeper understanding of the nature of design. The design process as it can be observed in practice (say, in an empirical study) could easily be the consequence of the properties of the design problem, the specific, even idiosyncratic approach the designer takes or the nature of the context in which the design activity has

Typologies of Design Practice

457

occurred. Thus to arrive at a deeper understanding of design activities, we need to consider more than just the design process. In this paper we take a broader approach to design research and we will consider what is known about design problems, the designer, the structure of design-related thinking processes and the way various design contexts impact upon activities in order to develop a typology of ‘design practices’. First, in Section 2, we will have to deal with the notion of ‘typology’ itself, to frame our undertaking. Then we will set out (in Section 3) to investigate how the current knowledge on design problems within design research could inform the creation of a typology of design practices. Then (Section 4) we will explore the possibility to create a more structural model of design thinking by proposing a notational framework that is focussed on key links that are made within a design (and thus within a design activity). We will critically discuss the possibility to derive insights into the nature of design problems from such a notational framework, and the repercussions of such a framework for a typology of design activities. In Section 5 we will explore how the properties of the designer could influence design activities, by delving into the literature on design expertise. Finally, in Section 6 we will again look at the design activity, but then from the context in which the design problem arises (a technical advance or a change in the market environment). In section 7 we will take stock and devise a way forward.

2

ON TYPOLOGIES

The notion of a ‘typology’, although it has well-defined meanings in areas such as archaeology, linguistics, theology, sociology, and various other fields, doesn’t seem to occur as a canonical term in the field of design research. Generic (that is, non-discipline specific) definitions of the term ‘typology’ appear to centre around ‘the study of systematic, complete and unique classification of types that have characteistics or traits in common’ but such a definition gives little clarifications as to the purpose of a typology. Still, the purpose of a typology is essential to decide which characteristics or common traits should be taken into account. For instance, in the animal kingdom, Linnaeus’ taxonomy counts as a prototypical example. It is a hierarchical typology, consisting of classes that have subclasses, all members from a subclass being member of the parent class by definition. The purpose that Linneaus had in mind was mainly a religious one [Soulsby, 1933], as he wrote in the preface to a late edition of Systema Naturae: Creationis telluris est gloria Dei ex opere Naturae per Hominem solum (‘The Systems of Nature: The Earth’s creation is the glory of God, as seen from the works of Nature by Man alone’): the study of nature would reveal the Divine Order of God’s creation, and it was the naturalist’s task to construct a ‘natural classification’ that would reveal this Order in the universe. Linnaeus’ system has limited use, for instance, for a zoo-keeper who wants to decide which animals to exhibit where in the park, to a veterinary doctor to decide if a particular medicine will work for a particular

458

Kees Dorst and Kees van Overveld

species in case of a particular ailment, or to staff of the World Wildlife Fund to decide which endangered species to protect against extinction. We adopt a very much pragmatic view to typologies: we regard a typology as a means to an end, and different ends require different means; if the end is not clearly defined, it cannot be decided whether the means is appropriate.1 Merely to ‘gain insight’, or to ‘study the relations between one design practice and another’ are too vague ends to distinguish proper typologies from improper ones. So, when proposing a typology in any domain, • we should agree on the purpose of the typology first (who is going to use the typology, and for what reason?); furthermore, • we should define the domain we want to be covered by the typology (in Linnaeus’ case: the question precisely what constitutes a species of living creatures is, as of today, subject to lively debate in the discipline of biology [Hey, 2006]; • regarding completeness and uniqueness, we should ask ourselves whether the domain of the typology forms a closed or open collection: as we will see in a while, the notion of ‘completeness’ in an open domain is rather problematic. ‘Uniqueness’ regards the degree to which two inhabitants of the typology’s domain can be distinguished — and this again requires special care in the case of an open domain; • and finally we should decide what structuring device we will adopt (in Linnaeus’ case, the typology is a hierarchical structure; in this paper we will encounter orthogonal structures (i.e., tables in a database) or network structures; at the end of our exploration, we will see that we might need an even richer structuring device. If Linnaeus would have missed the distinction between reptiles and amphibians, present-day zoos would perhaps not have looked that much different. If, however, a zoo-keeper misses the distinction between tropical animals and arctic animals, or between predators and prey, mortality in zoos would be unacceptably high. So a zoo-keeper, in performing his job, clearly benefits from typologies of the animal kingdom that contain terms such as preferred climate and feeding behaviour. In the world of design, relevant stakeholders are designers, design methodologists, and board members of design studios and industrial design organizations, rather than zoo-keepers. Broadening the view, we can also include customers (purchasing designed artefacts), venture capitalists (funding design processes), teachers (educating prospect designers), and environmentalists (keeping a keen eye on, e.g., sustainability issues of designed artefacts), and perhaps others. The intentions of 1 A quote from Lewis Caroll’s Alice in Wonderland illustrates this: ‘Cheshire Puss,’ she began, [...]‘Would you tell me, please, which way I ought to go from here?’ ‘That depends a good deal on where you want to get to,’ said the Cat. ‘I don’t much care where—’ said Alice. ‘Then it doesn’t matter which way you go,’ said the Cat.

Typologies of Design Practice

459

all such stakeholders can be expressed in terms of variables that should preferably have minimal or maximal values, so that they can be minimized or maximized (or ‘optimized’ in general). In our notational framework, to be presented below, such variables will be called ‘category-II variables’. For instance, if we consider the construction of a typology on the animal kingdom, with a zoo-keeper as primary stakeholder, we could say that ‘mortality in the zoo’ is a category-II variable that should be minimized, and the typology-under-construction should be arranged such that it helps doing this, for instance by distinguishing tropical and artic animals, and perhaps associating these two types of animals with their preferred environmental conditions. For the typology-of-design-practices-to-be-constructed, we can now mention some of these ‘purpose’ variables; a ‘good’ typology should help increasing the values of these variables: • the efficiency of the design process (with possible operationalizations such as the percentage of design decisions that had to be backtracked underway, or the re-usability of previous design processes): a ‘good’ typology could help improve the efficiency of design processes by minimizing the amount of backtracking and / or maximizing the amount of re-usable design decisions; • the manageability of design processes (with possible operationalizations such as the ratio between core activities (=taking design decisions) and overheadactivities (=deliberating about which design decisions should be taken): a ‘good’ typology could help improve the manageability of design processes by reducing the amount of overhead; • the teachability of design (with possible operationalizations such as the rate of successfully graduating design students and failing candidates, or the interteacher variability): a ‘good’ typology could help improve the teachability of design processes, by increasing the rate of successfully graduating students, and / or decreasing the inter-teacher variability. Of course, this is not an exhaustive list of purposes — indeed, all of these purposes are the ‘traditional’ ones that have been elaborated before in Design Methodology, all relating to the process of design. Nevertheless, as we will see later, the traditional structures for building typologies (hierarchies and orthogonal structures) are of limited use for supporting these purposes. The domain of a typology can be open or closed. When Linnaeus contrived his ‘Natural System’, he believed that — and lived under the assumption that — the collection of biological species was God-given: it was fixed at the date of creation and never changed since. It was a closed system. Setting up typologies for closed systems is relatively easy (depending, of course, on the purpose of the typology). A naive approach consists of linearly ordered enumeration. For example, a dictionary, labelling words as substantives, verbs, adjectives, etc., is a straightforward

460

Kees Dorst and Kees van Overveld

typology on (the words of) a natural language; the ordering is given by some alphabet2 (which is itself, in good approximation, a linearly ordered enumeration of speech sounds!). To do justice to the intuitive notion of inheritance, that seems to exist in so many domains, typologies are often organized as hierarchies. The standard form of hierarchy is the single-rooted or single-parent hierarchy, that is: every item has at most one so-called parent item. For instance, a drosophilidae (fruitfly) is a family of the suborder of brachycera is a suborder of diptera is an order of pterygota is a subclass of insecta is a class of hexapoda is a subdivision (subphylum) of arthropoda is a division (phylum) of animalia is a kingdom of eukaryota. If x is an y, then, y is called the parent of x. According to the hierarchical typology, all defining properties of y are found in x, and x has some more properties that are not defined in y. Indeed, y is more abstract than x: formally, it contains only a smaller number (a ‘bundle’) of predicates, or attributes-with-values than x. For instance: arthropoda has as an attribute ‘number of legs’, with arbitrary (even) value; in hexapoda, the value is 6, and this is also true of all hierarchical ranks below hexapoda. Hierarchies can be constructed on closed domains, as in Linnaeus’ original intention; they turn out to apply to open or extendible domains as well, as was found by later generations of biologists. Indeed, it is now believed that there are between 50 and 100 million animal species; only about 1 1/2 million have been named [Rosenzweig, 1996]; in Linnaeus days, the number of named species was only several thousands. Still, most biologists believe that the hierarchical naming structure won’t require any major adjustments to accommodate this huge increase of items-yet-to-be classified. The fact that (standard, single-parent) hierarchies can accommodate with open or extendible domains does not mean that they are suitable to classify design practices. Indeed, different purposes require different structures. In the animal example: the zoo keeper requires a hierarchy where some abstract class has two sub-classes ‘tropical’ and ‘arctic’; the veterinary doctor desires a hierarchy where an abstract class has two sub-classes ‘sensitive to treatment X’, and ‘not sensitive to treatment X’. A single-parent hierarchy, as it turns out in general, can not serve multiple purposes. In particular when, like in the case of typologies for design processes, the collection of category-II variables is not precisely known, the choice for a single-parent hierarchy is hopeless. A first partial remedy is to allow multiple-parent hierarchies. A tiger is both a mammal (so the item ‘mammal’ is a parent item for ‘tiger’) and a tropical animal (so the item ‘tropical animal’ is also a parent item for ‘tiger’). In a multiple-parent hierarchy, both types of relationships can be hosted simultaneously. Multipleparent hierarchies have a number of technical disadvantages, however; since they can always be represented in terms of the more generic orthogonal structures (such as (database) tables), we prefer these as a next candidate for a typology. 2 The words in the Chinese language, and in fact any language that is non-phonetic in nature, cannot be linearly ordered according to some alphabet. Instead, more sophisticated ordering systems (based on the number of brushstrokes, the use of certain elements, etc.) need to be used.

Typologies of Design Practice

461

In an orthogonal structure we have two or more independent attributes (such as ‘preferred climate’ with values ‘tropic’ and ‘arctic’, and ‘feeding habit’ with values ‘carnivorous’ and ‘herbivorous’). The attributes ‘preferred climate’ and ‘feeding habit’ are called orthogonal, since all four combinations (‘tropic’, ‘carnivorous’), (‘tropic’, ‘herbivorous’), (‘arctic’, ‘carnivorous’), and (‘arctic’, ‘herbivorous’) occur in the domain of the typology. In Section 4, we give a more precise treatment of tables using the terms concepts, attributes and values. The essence, however, is that a number of attributes is defined beforehand, and that items are classified in terms of these attributes. For an open or extendible domain, it is impossible to prove that any given set of attributes will be sufficient to fully distinguish the inhabitants of the typology’s domain in a relevant way, where relevance is determined by the goal of the typology. Indeed, for any set of attributes, each attribute with a given set of values, it is always possible that at some future point in time a new item is found that is indistinguishable from an existing item using these attributes only. Indeed, suppose that, in the case of the animal kingdom, skin texture would not be part of the set of attributes: then the distinction between a zebra and a horse might be next to impossible. 3 THE PROBLEM OF DESIGN PROBLEMS Now we will set out to investigate how the current knowledge on design problems within design research could inform the creation of a typology of design practices. This is a logical starting point for our exploration, for two reasons: • anybody familiar with design will intuitively say that there are many kinds of design problems; • designers do react differently to design situations that they apparently perceive as being of different nature. Yet to date there is no well developed and explicit typology of design problems that has been accepted widely within the design professions. Yet, ‘design problem’ is one of the central notions in doing and describing design. Indeed, design is often cast in the form of a reasoning process, a ‘rational problem solving’ process, going from a design problem, where the needs of the users are described in terms of the ‘function’ of the artefact, towards ‘form’, in the guise of statements about the shape and materiality of the artefact, and plans for its production and marketing. The rational problem solving literature that arose in the 1960s and 1970s in the field of Artificial Intelligence has had a profound impact on our thinking about design, and it is still shaping the way design is often discussed today. The introduction of these theories in Design Methodology, in the beginning of the 1970s, helped systemize the then existing models and methods of design, and helped link these to models of problem solving in other fields [Cross, 1984]. There were high hopes that the very nature of design could be captured in a description that was

462

Kees Dorst and Kees van Overveld

based upon considering designing as the solving of ill-structured problems [Simon, 1967; 1992]. Although there have been many developments since then, the original work on problem solving and the nature of ill-structured problems, written by Herbert Simon, still looms large over the field of design research. Design models and methods have been developed within this paradigm; the conceptual framework of rational problem solving has become the normal ‘language’ of thinking and talking about design. There have however, also been fundamental critiques of Simon’s problem solving approach and its applicability to the field of design, and many of the original statements in the problem solving theory that deal with design have since been readdressed and refined. In describing the core of Simon’s conceptual framework, we will now concentrate on the classic 1973 paper ‘The structure of ill-structured problems’ ([Simon, 1973], quotes used in this section are taken from that paper, unless otherwise stated). Within Simon’s theory, the issue of the solution of design problems takes the stage as an example of a wider category of problems, to be called ill-structured problems. In his paper ‘The structure of ill-structured problems’, Simon sets out to explore the relation between ill-structured problems and well-structured problems. He starts with the remark that many kinds of problems that are often treated as well-structured are probably better regarded as ill-structured. Even the limited problems (‘limited’ in the sense of taking place in an enclosed and well-defined world) that are used as standard examples in problem solving and AI literature, like chess playing, display elements of ill-structuredness upon closer scrutiny: even if we regard chess playing as a well-structured problem in the small, by most criteria it must be regarded as an ill-structured problem in the large (over the course of the game). This opens up the way for suspecting that the ill-structuredness of a problem may not be an a priori property of the problem itself, but is linked to the capabilities of the problem solver. In this way, the subject that does the problem solving actually influences the very nature of the problem. Nevertheless, Simon maintains that the problem solving theory that is based upon the solution of well-structured problems should serve as the basis for all problem solving. There is a basic assumption here that even though well-structured problems as such do not exist in the real world, the construction of well-structured versions of ill-structured problems is the way to solve an ill-structured problem. Simon then illustrates the solving of ill-structured problems by taking an example from design. The example deals with the designing of a house, and concentrates on the technical problem of designing the layout of the house. In this case, the structuring actions that turn the ill-structured design problem into a well-structured problem, are done by the architect: Additional specification will be obtained from the dialogue between architect and client, but the totality of that dialogue will still leave the design goals quite incompletely specified. The more distinguished

Typologies of Design Practice

463

the architect, the less expectation that the client should provide the constraints. This means that in a multi-step problem solving process, each problem solver will get the chance to pile interpretation upon interpretation and thus end up taking the problem solving processes in completely different directions. The use of memory and subjective interpretation thus becomes a major influence on the problem solving behaviour of designers. If we take this seriously, then this undermines the very idea of having one, knowable problem at the start of the problem solving process, because the detailed formulation of the problem will vary from designer to designer. But Simon misses this point: . . . the architect will find himself working on a problem which, perhaps beginning in an ill-structured state, soon converts itself through evocation from memory into a well-structured problem, which suggests that there is such a thing as an a priori given ‘design problem’ that, although it undergoes some mild reformulations, essentially stays the same throughout the design process. He has been criticized for this standpoint, and later concedes that: . . . there is merit to the claim that much problem solving effort is directed at structuring problems, and only a fraction of it at solving problems once they are structured. Simon then introduced an unspecified ‘noticing- and-evoking mechanism’, the existence of an ‘indexed memory’, and placed the design process in an ‘effective problem space’: . . . the effective problem space will undergo continuing change throughout the course of the problem solving activity, moving from one subspace to another of the large space defined by the contents of the long-term memory. Despite these theoretical shortcomings, the rational problem solving paradigm, that was largely inspired by the work of Simon, has become a powerful tool for the modelling of designs. Within design research, the emphasis on the rational problem solving is still overwhelming. This is hardly surprising, because the models of design processes that have been developed on the basis of this paradigm have proven to be a very powerful tool in the development of design practice and design education [Roozenburg, 1991; Cross, 1992; Ulrich, 1995; VDI, 1985] . Having some measure of control over their design process has empowered designers and design students to tackle complex problems that otherwise would have been out of reach for most of them. However, as we have argued, the rational problem solving approach to design does have some inherent and fundamental difficulties. They persist despite a huge development within the rational problem solving approach. We can observe them

464

Kees Dorst and Kees van Overveld

still in the contemporary design research that is based on this approach, for example the substantial body of work on the ‘Function-Behaviour-Structure’ (‘FBS’) model of design that has been developed in the research group at the Key Centre for Design Computing and Cognition, under the supervision of John Gero [1990a; 1992; 2002]. For instance, the FBS model ascribes an equally large role to the use of ‘design prototypes’ in determining the ‘framing’ of the design problem [Gero, 1990b], as Simon does to the ‘memory’ and ‘experience’ that a problem solver needs to transform an ill-structured problem into a well-structured one. This large role for ‘experience’ and ‘prototypes’ leads to grave methodological difficulties. Because of the very open-ended way in which the use of ‘design prototypes’ is described in the FBS model, they potentially make up a vital part of the design process, actually bypassing the design process that is modelled in the core FBS model [Vermaas, 2007]. The neat and clear design process model just looses most of its value if it is to be preceded by an undefined, very messy and overwhelmingly influential step called ‘the adoption of a prototype’. Dorst and Cross [2001] have attempted a closer description of problem solving of ill-structured problems by using an empirical study to analyse and describe the design process in practice, as a co-evolution of the design problem and the design solution. Based on their empirical study, they observe that the creation of solutions to ill-structured design problems seems to be a very gradual process, an evolution. Their analysis shows that creative design is not a matter of first fixing the problem (through objective analysis or the imposition of a frame) and then searching for a satisfactory solution concept. Creative design seems more to be a matter of developing and refining together both the formulation of a problem and ideas for a solution, with constant iteration of analysis, synthesis and evaluation processes between the two notional design ‘spaces’ — problem space and solution space. In creative design, the designer is seeking to generate a matching problemsolution pair, through a ‘co-evolution’ of the problem and the solution. Creative design involves a period of exploration in which problem and solution spaces are evolving and are unstable until (temporarily) fixed by an emergent bridge, which identifies a problem-solution pairing. The description of design as the co-evolution of problem and solution leads to the uneasy conclusion that in describing design, we cannot presuppose that there is something like a set ‘design problem’ at any point in the design process. This leads to some very pertinent methodological questions. Can we still describe design in terms of problem solving theories if we may have to abandon the idea that the ‘design problem’ can be identified at all? What, then is the meaning of saying that design is a process running from ‘a problem’ to ‘a solution’ ? We can probably stick to the problem solving theory of design only if we abandon the idea that there is definable problem at the start of the design process, and postulate that it will be constructed later on. This then begs the question how this problem is constructed, and whether this process of ‘problem construction’ can be modelled at all. And if this process of problem construction could be modelled, one could ask whether modelling should be done within the rational problem solving paradigm, or outside.

Typologies of Design Practice

465

We will now proceed by introducing two different approaches that have been taken in addressing this problem. First we will consider the work of Dreyfus and Suchman [Dreyfus, 1992; 2002 and Suchman, 1987], modelling design problems as situated problems, and then we will look at Hatchuel’s ideas on ‘extended rationality’. The description of design as a situated activity involves two important shifts in standpoint. (1) The first fundamental choice that is associated with situated problem solving is that we are first and foremost interested in what design problems are to the designer, seen through the eyes of the designer, in the design situation. This means that we concentrate on the ‘local’ design problem that a designer faces, and ignore the ‘overall’ design problem as something of an abstraction. So we will also have to confront the vagueness (i.e. lack of overview) and subjectivity that is involved in local design actions and decisions. Seen from this perspective, ‘the design problem’ as such does not really exist as an objective entity in the world. There is an amalgamate of different problems that centres around the basic challenge that is described in a design brief. This amalgamate of problems is partly there to be discovered by the designer in the design process, and part of it has to be made by the designer. The process of ‘approaching a design problem’ or ‘dealing with a problematic situation’ becomes the vital clue to understanding what design problems are. This is the second fundamental shift in standpoint (2): for much of the design project the problem solving steps can be quite logical, routine and implicit, without a real choice for the designer. Dreyfus holds that problematic situations are the results of a ‘breakdown’ in this normal, fluent problem solving behaviour. These ‘breakdowns’ then are the moments of real choice [Sch¨on, 1983]. It thus becomes very important to distinguish and describe the nature of these breakdowns, the critical situations in design [Frankenberger, 1996]. Note that the definition of ‘design problem’ has thus been narrowed, and limited to the situations where the routine problem solving has broken down, and that we have to shift our focus from a typology of ‘design problems’ towards the development of a typology of ‘design situations’. If we can be convinced by Dreyfus and others that there is never a (complete) representation of the design problem in the head of the designer, then the only thing left for us to study is the ‘local’ network of links that a designer considers while tackling a design problem in the design situation. The subjective nature of this local network of problems means that we need a model of how designers approach a problematic situation. Hatchuel [2002] analyses the work of Simon on design in its original context, as part of Simon’s bigger project on the development of a theory about ‘bounded rationality’. The aim of this project was to ‘explain human behaviour by simple and constrained, yet informed, decision rules’. The bounded-rationality project spans Simon’s work in economics, Artificial Intelligence and design. In developing this perspective, Simon sees a strong theory of design as crucial. The possibility to develop a strong theory on design (the archetypal ‘science of the artificial’) within this general framework of bounded rationality, serves as a litmus test for the bigger project itself. Simon’s drive to achieve this leads to poetic statements:

466

Kees Dorst and Kees van Overveld

The proper study of mankind is said to be man[. . . ] If I have made my case, then we can conclude that, in larger part, the proper study of mankind is the science of design, not only as the professional component of a technical education but as a core discipline for every liberally educated person. [Simon, 1967, p. 159] Hatchuel argues that Simon is over-eager in his efforts to incorporate design within the general bounded-rationality problem solving theory. Hatchuel illustrates the distinction that he thinks needs to be made between design and problem solving by an example, in which two problem situations are compared: he pictures a group of friends coming together on a Saturday night, the one problem situation is that they are ‘looking for a good movie in town’, the other problem situation is that they set out to ‘have a party’. The first situation is considered to be ‘problem solving’, the second situation is, in Hatchuel’s terms, a real design project. Hatchuel argues that there are three important differences between these situations: • The first difference is that the design situation includes the (unexpected) expansion of the initial concepts in which the situation is initially framed (‘a party’). This makes the solution process a project, instead of a problem. There is no dominant design for what a party should be, so imagination needs to be applied at this very fundamental level. • A second difference is that the design situation requires the design and use of ‘learning devices’ in order to get to a solution. These ‘learning devices’ are sub-processes that help ‘learn about what has to be learned or should be learned’. They include experiments and simulation techniques. • Thirdly, in designing the understanding and designing of the social interactions is part of the design process itself. The group of friends needs to develop a way of reaching a solution, that cannot be supposed to exist before the design situation arises. This point comes very close to the work of Louis Bucciarelli, where he claims that “design is fundamentally a social process” [Bucciarelli, 1994]. • We can add a fourth point here, stating that design problems are open problems, and the choice-problems are closed. The construction of a closed problem within an open problem arena is not a deductive step, but a creative act in itself. From this comparison we can conclude that design undoubtedly includes stretches of ill-structured problem solving, but that it also contains other processes. For Hatchuel design includes problem solving, but it cannot be reduced to problem solving. He states that any model or description method that tries to reduce design to problem solving is bound to miss important aspects of the design activity.

Typologies of Design Practice

467

Concluding remarks on design and problem solving All of this means that the very notion of ‘design problem’ becomes extremely problematic. If the ‘design problem’ in general is not knowable at any specific point in the design process (Dreyfus), if it is evolving in the design process, and if the connotations of the very concepts that are used to describe a ‘design problem’ are shifting as a part of the design effort (Hatchuel), then we need to radically reconsider our use of the term ‘design problem’ as a basis for a typology of design practice. Apparently, subjective elements that are part and parcel of the designer, and the design context also have a strong, and in many cases probably overriding influence on design practice. This has far-reaching consequences for our quest to create a typology of design practice: the attempt to find a foothold (a key organising principle that would allow a typology to be constructed) in the area of design problems will always be flawed. Even a very interesting recent proposal for a model of design problems, that uses three dimensions of design constraints: (1) the ‘generators’ (designer, client, user, legislator), (2) the ‘source’ of design constraints’ (internal or external) and (3) the ‘nature’ of design constraints (radical, practical, formal or symbolic) to typify design problems [Lawson, 2006], is based on the assumptions about the nature of design problem solving that we have had to qualify in this section. We have to conclude that while the problem-perspective is instructive, it is very difficult to arrive at a solid basis for a typology of design practices in this way.

4 PATTERNS IN DESIGN THINKING

4.1

A notational framework

Despite the shortcomings of the ‘problem-perspective’, we do know that designers must respond to an initial ‘problem’ or ‘problematic situation’ in a fairly coordinated way. Maybe we can get to resolution of the question of what the basis of a typology of design practices should be by getting closer to the structure of various design activities, in particular the thinking processes that designers go through during their work. In this Section we will explore what the necessary elements of such a representational structure (‘notational framework’) for design activities would be. We will do this by creating an extensive framework to formally describe design activities. The creation of this framework is a thought experiment, through which we explore whether this is possible at all — the resulting framework should be seen as an example, created for the purpose of achieving a ‘proof of concept’. We do realise that this may be just one of several different formal frameworks that could be used to describe design activities. In this particular framework, the intentionality of the act of designing is the main starting point. A design process is seen as a sequence of actions that lead to an intended goal; the formulation of this goal in terms of operational variables is a prominent feature of our proposal. Our framework also

468

Kees Dorst and Kees van Overveld

aims to do justice to the fact that in design situations we must be able to deal with openness: we must be able to introduce new variables if and when we need. The framework has been set up with the following desired features (‘desiderata’) in mind. Desired properties for a notational framework a. Dreyfus identifies the notion of ‘breakdowns’ where crucial design choices take place. In this abstraction, a design process can be seen as consisting of discrete ‘chunks’ of work, each centring around a single design choice. A ‘choice’ can be seen as a combination of (1) some variable for which the value is to be chosen; (2) a number of alternative values from which we can choose one, and (3) some rationale that justifies why one value is better than another. Our framework shall be such that these three ingredients are well-represented. They will occur in the form of (1) variables (in category I3 ); (2) the types of such variables, and (3) the indirect functional connection of such variables to objectives (see also point e below). b. Dreyfus also observes the ‘network of links’ that a designer considers. ‘Links’ in this context can be thought to refer to causal connections leading from design decisions to their ultimate effect in terms of stakeholder’s happiness (see point e below). To do justice to this view, the notational framework will explicitly represent these links. We do so in terms of a graph; the actual construction of this graph forms a rational reconstruction of the design process. By a rational reconstruction of some process, we mean a consistent description of this process where effort is put in making essential, but implicit or unconscious steps visible, and providing plausible justifications for such steps — even if, during the initial process, these justifications may not have been given. Such a description is usually made a posteriori, after completion of the process; the notational framework, however, is also well-suited to be used to document the process ‘on-the-fly’. c. Hatchuel draws our attention to the role of the design context. Two versions of the same problem, each with a different context, can lead to dramatically different design processes. In fact, part of the creative component of a design process consists of exploring, delineating, and perhaps challenging the context. To pay respect to the crucial role of context, the notational framework shall explicitly address the difference between (knowledge about) the artifact’s internals and its context. For this reason, we will later introduce the distinction between variables in categories I and III. d. Hatchuel also raises the issue of learning, and openness to the expansion of design knowledge during the design process. Every non-trivial design process is an exploration that leads the designer into terra incognita. Tomorrow’s knowledge will supersede what we know today. Therefore, the notational framework should cater for stepwise refinement, that is: replacing global knowledge by more detailed 3 Jargon terminology such as ‘categories’, ‘types’, et cetera, will all be explained in more depth in section 4.2.

Typologies of Design Practice

469

versions.4 As we will see shortly, two mechanisms are provided in the notational framework: one is compound types, the other one amounts to promoting variables from category III (‘constants’) to category IV (variables that are functionally dependent onto other variables). e. Hatchuel further mentions that every design process eventually is a social process that deals with human desires and the satisfaction thereof. This forms a sharp distinction to the idea that designers optimize for ‘objective’ quantities such as safety, robustness, or performance. The bottom line of Hatchuel’s insight is, that we instead (or at least also) should optimize for the perceived safety, the perceived robustness, or the perceived performance. There is presumably some causal connection between performance and perceived performance, and this connection, despite its subjective nature, should also be part of the designer’s concern. Arguing about the working of the (social or psychological) phenomena that connect performance and perceived performance is inseparable part of the design process, and therefore any representation of a design process should potentially accommodate these mechanisms. In the notational framework as we propose it, these ‘hidden’ subjective mechanisms can be conveniently expressed. f. The discussion with regard to ‘openness’ vs. ’closedness’ (as in rational problem solving) of design situations has drawn much attention in the literature. In the notational framework, this discussion is partly reflected in the varieties of types (finite, infinite, closed or open), in particular of category-I variables. The framework helps to subtly distinguish two types of openness: apart from the distinction between open and closed types, we can distinguish adding attributes vs. expanding the extent of an attribute.

4.2

A Notational Framework to annotate Design Activities

The main characteristics of the framework are: • The properties of the designed artifact, as well as other relevant properties we encounter in describing the design process, are given by attributes. Attributes, however, are not fixed. Although some generic attributes exist (e.g., price, physical dimensions, and energy resources will occur in the majority of design processes), most attributes will occur whenever the design process asks for them. They may differ in dependency on the domains involved. • It is a dynamical model, in the sense that it represents the time evolution of a design process. • It is a formal model, with a minimal amount of dedicated semantics built in. On the one hand, this is just enough to allow formal support (as described in [Ivashkov, 2004]), including consistency checking and automated 4 This replacement is not guaranteed to be monotonic. In situations where new knowledge conflicts with the status quo, more advanced forms of formalisation are required. The present form of our notational framework has no provisions for non-monotonic reasoning.

470

Kees Dorst and Kees van Overveld

optimization; on the other hand, it is sufficiently open to serve independent of the application domain. The notational framework consists of two kinds of ingredients: • notions from data modelling (concepts, attributes, values), where variables (=attributes of concepts) are labelled with a category-index according to their role in the reconstructed design process (design variables, context variables, objective variables, and intermediate variables). These notions allow the description of hierarchies (e.g., for distinguishing systems-level and component-level design), orthogonal structures (e.g., product families) and functional dependencies (e.g., how objective5 variables depend via intermediate variables on design variables). They will be helpful to explain the difference between open and closed problems, and they serve to argue about the social mechanisms that underlay the perceived quality of designed artefacts. • notions from process modelling (sequential and parallel steps, iteration, alternatives, refinement, and scoping — each of them being axiomatically definable). These notions serve to describe the dynamics of design processes — both at microscopic scale (the incremental knowledge build-up during a single design process) or macroscopic scale (the interpretation of market and technology trends as far as design is concerned). With the notational framework we will attempt to match empiric observables to model constructs, establishing a connection between real-life design and its rational reconstruction. Our notational framework contains the following terms: Concept — any entity in design discourse. An entity can be an object (‘my desk lamp’, ‘the stakeholder’), or a class of objects (‘brass desk lamps’, ‘the stakeholders involved in this project’). Concepts can both be material (‘the desk lamp we are designing’) or immaterial (‘the intended usage of this type of desk lamps’); they can be existing entities (‘the customer’s writing desk’) or future entities (‘the desk lamp that will fit on the customer’s writing desk’). In various design contexts, various sets of concepts will occur. However, in any design context it is assumed that at least two concepts invariably exist: the artefact to be designed (ATBD) and the stakeholder(s) — the ATBD is intended to satisfy the needs of the stakeholder(s). A concept contains information, and it is necessary to access or to set the information in the concept. To this aim, we introduce the notion of attribute. Attribute — any function that maps a concept onto a value or values (see below) with the intention to express (address, specify, select, . . . ) an atomic unit of information of that concept. For instance ‘colour’, which maps all coloured things onto the set of colours. 5 The word ‘objective’ here is a substantive (‘objective’ as ‘goal’, ‘aim’), not an adjective (as opposed to ‘subjective’).

Typologies of Design Practice

471

Value — an element of the range of an attribute. ‘Red’ is a value of the attribute ‘colour’. The range of all values for an attribute is also called6 the type of that attribute. For instance, the type of ‘distance’ equals all non-negative real numbers; the type of ‘usage’ is the set of all possible usages (e.g., of an artefact). We observe that the mapping of a concept to a value can take place in various distinct ways: (i)

Choice or decision: a designer decides that the colour of the lamp shade to be designed shall be yellow, not red;

(ii) Consequence or effect of (sets of) decisions: as a result of the design decisions, this device will eventually weigh 350 grams; (iii) Observations, measurements or estimates: the density of brass is . . . gr/dm3. The difference between these various associations between attributes (of concepts) and values will be crucial for our understanding of the various roles attributes can play in a design process, as will become clear in the next section. Also notice that some types may be so called compound types, that is: sets of elements that are concepts in their own right. For instance: for an attribute ‘usage’, applied to the concept ‘deskLamp’, we may find the value ‘illuminatingTheTable’. The value ‘illuminatingTheTable’, however, is a concept in its own right in the sense that it contains further information that may be relevant for the design process (such as the colour of the light shining on the table, the area of the illuminated spot, and others). Values in a simple (non-compound) type are not concepts, they are just simple values. ‘Red’ or ‘3’ or ‘true’ are simple values. It is quite common, however, that the type of a new attribute (that is, an attribute that is newly introduced in a design discourse) first is simple, and that only at a later stage this type is converted to compound. For instance, ‘red’ may, at a later stage in the discourse, have attributes such as its brightness, its saturation or its glossiness. The sequential refinement of simple values that become promoted to concepts in their own right happens often with designs where an ATBD is assembled from components taken from a catalogue. For some types, we know immediately that they will soon be compound types (such as ‘shape’), but even then we can treat them at first as simple types (say, {round, rectangular, triangular}) which eliminates the necessity to immediately specify further attributes. This is sometimes called ‘separation of concern’ or ‘deliberate vagueness’ [Cross, 1990]: we deliberately choose the level of concreteness at which we want to discuss a concept. Apart from the distinction between compound and simple, there is the distinction between what will be called finite and infinite types and closed and open types. This distinction is necessary, among other things, in the light of the debate about open and closed design, as explained earlier in this section (see item f from the ‘desired properties’). The following holds: 6 The use of the word ‘type’ here should not be confused with its use in the phrase ‘the type of design situation’.

472

Kees Dorst and Kees van Overveld

• A finite type has a finite number of values (for instance, the number of different members of a list of components with given functionality taken from some catalogue). A compound type is finite if all its attributes are finite. • An infinite type has a transfinite number of values (for instance, an integer, rational or fixed point number. The mathematically relevant distinction ‘countable’ vs. ‘non-countable’ is less essential in most design contexts). A compound type is infinite if at least one of its members is infinite. • A closed type is finite, or if it is not finite, there is an algorithm7 to generate all its values. For instance, an integer number is a closed type — but also colours and strings of finite length are closed types. A compound type is closed it all its members are closed. • For an open type, we cannot give an algorithm to give all values. For instance, the type of the attribute ‘shape’ (other than ‘one element from a given finite list of shapes’) is an open type. It is impossible to give an algorithm to systematically produce ‘all’ shapes.8 A compound type is open if at least one of its members is open. Signature of a concept — for a concept C, the signature is the set of attributes that apply to C and that have been identified thus far during the design project. Notice: capturing all information in a concept would require an infinite signature. Signatures tend to grow over the lifetime of a concept. The notion of signature is required to deal with item d from the ‘desired properties’. Extent of an attribute — for an attribute A, the extent is the set of concepts to which A may be applied. Notice: adding a new concept to the extent of a given attribute, such as ‘usage’ is often considered to be a creative act. This form of ‘inventing’ is characteristic of much of innovative design: it usually can be seen as extending the extent of some ‘usage’-like attribute.). For example, the extent of the purpose ‘is useful for individual motorized transport at moderate speeds on pathways’, which traditionally contains concepts such as motorcycle or scooter, was recently enhanced by the development of the Segway. This extension of the 7 It is problematic to generate real numbers by an algorithm. That means that compound types that contain real numbers as attributes also should be considered as open types. Real numbers, however, are mathematical abstractions. In any practical situation the difference between a real number and a fixed point number (that is, a rational number with a given maximal number of decimals) can not be assessed. Fixed point numbers can trivially be generated by an algorithm: they can be enumerated. Therefore we ignore the subtle problems related to real numbers. 8 It is possible, however, to parameterize a (large) collection of shapes in terms of a finite or at least closed set of parameters. In this way, design problems that are fundamentally nonalgorithmically solvable can be forced into ones that can be tracked — sacrificing an open set of residual shapes that cannot be produced. An example is the use of spline curves and spline surfaces (a class of mathematically defined parametric objects) to represent shapes in computeraided geometric design.

Typologies of Design Practice

473

concept (or ‘the displacements of concepts’ as described in [Sch¨on, 1994]) is far from trivial. Enlarging the extent of one or more attributes is most often considered as a creative act; a different class of creative acts is the extension of the type of an attribute. E.g., adding an element to the type of ‘shape’. Also, adding the value ‘one’ to the type of the attribute ‘number of axes’ of the concept ‘motorized vehicle’ is another way to interpret the invention of the Segway. We will revisit various types of creative acts when we discuss ‘revolutionary’ design as opposed to ‘normal’ or ‘incremental’ design. Similar as the notion of signature, the notion of extent is required to deal with item d from the ‘desired properties’. Variable — a variable, finally, is an attribute, applied to a concept. For the purpose of modelling design processes, we distinguish 4 categories of variables. The first 3 categories (I, II, and III) were already introduced above when we explained the notion of ‘value’, as distinct ways to map a value to a concept ((i),(ii), and (iii)); the fourth one (IV) is a direct consequence from items b, d and e from the ‘desired properties’. • Category I: design variables or choice variables — these encompass the design space. All design variables are attributes of the ATBD — but not all attributes of the ATBD are in category I.9 Given a signature of the ATBD, category I theoretically spans the set of all possible design alternatives. The design space is a one-to-one mapping of the Cartesian product10 of the types of all category-I variable to the space of possible ATBD’s that can be characterized by these category-I variables. Indeed: every point in this Cartesian product corresponds uniquely to a single set of choices for all choice variables, and therefore it corresponds to a single unique design. This closely relates to the issue of open or closed design situations. Indeed, if the signature of the ATBD is known, and all category-I variables have finite types, the design problem in principle amounts to selecting the ‘best’ point11 in the design space — which in theory could be realized by exhaustive enumeration. Things are very different, however, if design either includes extending the signature of the ATBD (adding attributes to the ATBD or to one of the sub-concepts that are values of one of the compound attributes of the ATBD), or if some of the category-I variables has an open type. In design debates the distinction between the cases where the ATBD has or has not a given signature, and the cases where category-I variables have open or closed types is not always made — and much confusion results.

9 Shortly, we will see that many of the attributes of the ATBD will be in category IV, for example. 10 The Cartesian product of a sequence of sets is the set of all possible tuples where each tuple is a sequence of elements, one element taken from each of the sets. 11 When we discuss category II, we will articulate the subtleties related to ‘best’.

474

Kees Dorst and Kees van Overveld

• Category II: objective variables, goal-related variables or satisfaction variables — these encompass the objective space.12 All objective variables are attributes of the stakeholder(s), namely those attributes that represent his or her degree of satisfaction (for instance: price, speed, safety, effectiveness, . . . ). Variables in category I should be chosen such that values of variables in category II are, in some sense, optimal, in order to reflect the utmost subjective satisfaction of the stakeholder(s). Again, in the case of given signature of the ATBD and closed-type category-I variables, this reduces the design problem to an optimization problem — assuming that we are able to assess the values of the category-II variables as functions of the relevant category-I variables. It is important that category-II variables represent subjective variables only. Design should not be optimized with respect to objective quality indicators: it should attempt to maximise the perceived representatives of such quantities. This is the essence of part of Hatchuel’s argumentation, and it occurs in item e of the ‘desired properties’. In all cases, the values of (subjective) category-II variables derive from inter-subjective variables of the ATBD by means of some biological, psychological and / or sociopsychological process.13 This is true for physical variables, but it is even more prominently the case for economical variables (the cost price of some commodity is perceived as ‘expensive’ or ‘cheap’ under the influence of a host of contextual parameters, other than the actual price in Euros), functional, aesthetical or prestige-related variables. A second issue in dealing with the category-II variables lies in the fact that we usually have several of them. Therefore, ‘optimization’ always should be read as ‘multi-objective optimization’.14 • Category III: context variables (corresponding to (iii) above) — these encompass all variables for which the designer cannot make decisions, but that should be taken into account in order to predict or understand the values of variables in category II. Examples are physical or economical or demographical constants, legislature and variables taken from design requirements. • Category IV: Finally, there is a further class of variables, to be called auxiliary variables, which encompass all variables for which the designer cannot make decisions — although the stakeholder(s)’ satisfaction may depend indirectly on auxiliary variables. 12 The word ‘objective’ here is quite unfortunate. As we will see, the objective space is spanned by variables that represent subjective notions only — in line with item e of the ‘desired properties’. 13 The only exception would be if the category-I variables would also be subjective — more in particular: if they would exist in the same mind as the category-II variables. This would relate to a particular design process, namely the case where one designs one’s own daydream. Although this may seem far-fetched, it relates to the practical case where market strategies are designed that mainly focus on re-programming the norms, values and appreciation metrics of the prospect customer. 14 For the mathematical technicalities related to Multi-objective optimization, we refer to [Ivashkov, 2004] and to the article by Kroes, Franssen & Bucciarelli about rationality in engineering design in this Volume.

Typologies of Design Practice

475

Now we can envisage a model that depicts dependencies among variables of different categories. Variables in categories I and III are purely independent; variables in category II are purely dependent, and variables in category IV are both dependent and independent. The dependencies form a directed, acyclic graph. Among other things, this graph can be used to support reasoning about design rationales and to help clarifying assumptions in the design process [Ivashkov, 2004]. With the above notions of concepts, attributes, and values and the four categories we can represent a design-related thinking process (or at least the knowledge contents as it is built up during design discourse) as a process where a graph as in Figure 1 is constructed. In the idealized case, the construction proceeds starting from category-II variables, working from right to left through the graph. A construction step can be one of: • expressing the dependency of an existing variable (category-II or categoryIV) in terms of other variables. These other variables can either exist or they can be new. • At any point, a functional dependency can be refined or altered. This usually amounts to a refinement of the design-related knowledge: either a detailing step in which a simple variable (say, ‘shape’) is replaced by a compound variable; or a step where a variable that upon introduction ended up in category-III, but is now moved to category-IV. • Simplification of the model: that is, the recognition that a particular functional dependency has so little impact on the values of category-II variables that we can do without — for instance moving a category-IV variable to category-III. The construction of the graph is complete if all category-II variables are completely expressed in terms of category-IV variables; and all category-IV variables are completely expressed in terms of other category-IV variables or category-I or category-III variables.15 ‘Complete’ does not mean that the model is ‘correct’ or ‘accurate’ — it merely means that we have constructed a (first) approximation to a consistent account of all the causal mechanisms that connect the design decisions in category-I to their intended effects in category-II. Once we have a complete model we can investigate its sensitivity for certain modifications, we can extend it where we feel that important causal connections have not been done justice, or we can try to optimize it by tuning the category-I variables. The above process can be interpreted as a rational reconstruction of design activity. In this reconstruction, we find ample opportunity to introduce those concepts and variables to express the unique features of the design situation at 15 The case where a category-II variable directly depends on a category-I variable only could occur if the designer and the stakeholder are one and the same person (that is, the one who takes the decision is the one who benefits from this decision). Indeed: category-II variables represent subjective satisfaction. In settings of industrial design practice this is usually not the case.

476

Kees Dorst and Kees van Overveld

•

II: objective variables

•

I: design variables •

•

•

• • •

• •

• IV: auxiliary variables

•

III: context variables Figure 1. The four categories and their dependency relations hand. Apart from the improved clarity, the opportunity to verify the consistency of the design, and the option to perform automated optimization for some of the category-I variables, we can therefore use this rational reconstruction to distinct the present design situation from others — that is, to categorize design situations.

4.3 Patterns in Design Processes We proposed a notational framework as a rational reconstruction of design activities. We have effectively created a ‘toolbox’ that will help trace the precise steps that are happening in design thinking, as a first step in establishing patterns in design activity. Variables have been introduced as the attributes of concepts, and the rational reconstruction of design activities can be seen as building a graph of dependencies among various variables, leading from design decisions to the subjectively perceived happiness in the stakeholders. This notational framework is not in itself a typology of design practices. Yet it provides a basis for assessing features of the things that are classified, thus providing a basis for such a typology. Once design processes have been documented and reconstructed in terms of the notational framework, we can start studying the resulting graphs — which are formal structures instead of anecdotic episodes of human experience. From the theory of graphs [Biggs, 1986], much is known about patterns in graphs. For instance, the interconnectivity between various categoryIV variables tells us how side-effects of decisions propagate through the design process. Trade-offs between different category-II variables can be analysed from the topological structure of the graph. A modular design reveals itself in weakly connected sub-graphs, et cetera. At the end of the rational reconstruction, we have a graph according to Figure 1; furthermore we may have an chronological account of the subsequent additions of

Typologies of Design Practice

477

nodes (variables) and edges (dependencies) in the graph. This chronological account also allows the identification of very distinct patterns in design processes. For instance, a process that strictly proceeds from ‘right to left’ (that is, introducing all the category-II variables first, and next developing them into category-I and category-III variables) is very different from a ‘left-to-right’ process — which may start from a single, more or less serendipitous invention (the non-permanent glue that led to the famous M3 post-its). We summarize: when the notational framework is used to build a rational reconstruction of the design process, this process is cast into the form of a graph (plus its chronology); patterns in the design process can be studied by analysing patterns in the graph. We will return to this point in the concluding section of this chapter, where we give further examples of patterns that may be studied in this manner. 5

TYPES OF DESIGNER: DESIGN EXPERTISE

It is precisely the need to model the kinds of subjective elements in design practice that leads us to the need to create of a framework to describe ‘the designer’. For this we turn to the theoretical work done in other fields on expertise development, and see if we can apply them to characterize the competence development of a designer in the course of her / his career. The assumption we need to make in adopting this approach to design expertise is to say that we expect expert designers to work differently from novices, and that gaining design expertise is not a gradual process — both of these statements can be well-supported by earlier research. Hubert Dreyfus [2002] distinguishes six distinct levels of expertise, corresponding with six ways of perceiving, interpreting, structuring, and solving problems. 0) Na¨ıve. This is an extra level, preceding the ‘novice’ level that is the start of the Dreyfus model. This state is required in a model of design expertise since designlike tasks are not only performed by professionals but also by ordinary people in their everyday life. This na¨ıve state of designing is adequate for everyday use in conventional situations. Many students that enter design schools will display this na¨ıve design behaviour. They do not yet understand that design is a series of activities, and treat it more as a one-off choice from a set of design solutions that they know and want to emulate (‘I want to make something like that’). 1) A novice will consider the objective features of a situation, as they are given by experts, and will follow strict ‘rules’ (as in ‘the accepted rules of the game’ or ‘conventions’ rather than algorithms) to deal with the problem. In this novice stage the students encounter design as a formal process for the first time. To tackle the complexities of design they need to learn a whole series of techniques and methods of representation. 2) For an advanced beginner the situational aspects are important, there is sensitivity to exceptions to the ‘hard’ rules of the novice. Maxims are used for guidance

478

Kees Dorst and Kees van Overveld

through the problem situation. At least one aspect of the increased expertise developed by these students during their education is in terms of their acquisition of schemata or ‘design prototypes’. 3) Competent designers work in a radically different ways. They select the elements in a situation that are relevant, and choose a plan to achieve the goals. Problem solving at this level involves the seeking of opportunities, and of building up expectations. In process terms a competent designer is likely to be able to become the creator of the design situation, through strategic thinking. This is a very empowering ability, in contrast to the earlier levels of expertise in which the designer was basically just reacting to design situations as they might occur. 4) The real expert has many years of experience which allows him/her to recognize high-level patterns in design situations and respond to a specific situation intuitively, and performing the appropriate action, straightaway. There is no problem solving and reasoning that can be distinguished at this level of working. This can actually be a very comfortable level to be functioning on, as a well-respected creative professional within an established field. Yet the expert is quite vulnerable to radical changes in the context of the profession. 5) With the next level, the master, a new uneasiness creeps in. The master sees the standard ways of working that experienced professionals use not as natural but as contingent. A master displays a deeper involvement with the professional field as a whole, dwelling on success and failure. This attitude requires an acute sense of context and openness to subtle cues. The master designer is really a development of the Expert who may have taken their set of guiding principles to a level of innovation such that their own work is seen as representing new knowledge in the field. At this level of performance, designers are producing design ideas that are innovative responses to situations that may have been previously well understood. Such work is published and becomes the new precedent for other designers to study. This could be deemed ‘practice based research’. 6) The visionary consciously strives to extend the domain in which he/she works. The world discloser develops new ways things could be, defines the issues, opens new worlds and creates new domains. A world discloser operates more on the margins of a domain, paying attention to other domains as well, and to anomalies and marginal practices that hold promise for a new vision of the domain. The design world deliberately creates an opportunity for this with design idea competitions, exhibitions, and the publication of professional journals. Most of these levels are intuitively recognizable to anyone involved in design education or design practice. But note that these fundamentally different ways of looking at problematic situations can actually co-exist in a design project. Designers display rule-following behaviour, as well as the interpretation and reflection that characterise higher levels of expertise at work. And in this model, design expertise is described as a set ordering of discrete states, although it is far from clear that individuals would necessarily progress one level at a time. But the levels are

Typologies of Design Practice

479

distinct in that what is required developmentally to move up a level in each case is different, and, most crucially for this paper, that each level comprises its own kind of problem solving and reflection. The notational framework that was introduced in section four can be used to construct detailed analyses of the differences in ways of working at different levels of the expertise model. When a designer moves to a higher level of expertise, the content of the concepts a designers deals with seems to change: the number of category-I and category-III variables under consideration will increase, and the choice of concepts will be different — the higher levels of expertise can be expected to involve more complex concepts, with larger ‘signatures’. In looking at design processes at different levels of expertise, we could hypothesise that overall the order in which category-I and category-II variables are introduced will be different at higher levels of expertise. These and other hypotheses on the general connections between the 6+1 levels of expertise development and the formal framework of section 4 should be established and tested through empirical investigation. One could then embark on a more finegrained exploration of the strategies that individual designers use. For instance, hypotheses such as the following could be the starting point for empirical studies into design strategies: (1) During the design process, various forms of refinement take place. The following forms can be distinguished: (a) variables that enter the design discourse as assumed constants in category III, but are promoted to category IV when mechanisms of causal dependency, involving these variables, are discovered later; (b) enlarging the extent of an attribute, that is, the creative act of proposing a new value for a known attribute; (c) extending the signature of a concept, that is, seeing a new attribute for a known concept, or (d) replacing a simple (= noncompound) type by a compound type. It is quite likely that designers at various levels of expertise can be distinguished in their use of mechanisms (a) . . . (d). (2) A single shot construction of the complete graph from figure 1, starting in category II and working towards categories I and III without any backtracking is in fact quite rare. In practice, designers will jump between extending categories I and II and back, perhaps leaving sections of the graph under construction nonconnected during the process. It can be expected, however that novices and more advanced designers have different strategies in this respect; again, empirical observation could reveal significant differences among the various levels of expertise.’ On a metalevel — the level of though and reflection on practice that drives the development as a designer — the kind of issues that are faced by the designer as a novice (‘How can I use my methods?’) are quite different from that on the advanced beginners level (‘When should I use this particular method/rule of thumb?’). Likewise, the reflection that takes place on the novice-level deals with the rules themselves, the reflection for the advanced beginner centres on the applicability of a rule in a specific design situation. And so on.

480

Kees Dorst and Kees van Overveld

6 TYPES OF DESIGN CONTEXT Now in our fourth approach (the first three being: design problems, the formal modelling of patterns in design thinking, and types of designers) we propose to model and typify design practices focusing on two key forces in the context of a design project, namely: market and technology. In doing this we build upon one of the classics in Innovation Management, the Ansoff Matrix that spells out four different innovation strategies for companies, by mapping the ways in which product and market can be related (see [Ansoff, 1968]). This matrix is now widely used in New Product Development in a slightly reformulated form, to concentrate on the interaction between market and technology as drivers for innovation. The market is the configuration of manufacturers, distributors and suppliers, and the sales channels that connect them. These channels include both material (goods, products) and immaterial (communication, advertisement) streams. Technology is a conglomerate of existing knowledge, expertise and infrastructure related to solving technological problems. Technology can be either publicly available (texts, scientific journals, the Internet) or proprietary (patents, licences, R&D laboratories, university research centres). Technology feeds into a design process, and the products of a design process feed into the market; the market, finally, leads to consumers purchasing, possessing, using, and disposing of designed products. At first sight, the interaction between technology and market may be considered as a simple one-way, causal process, where technological decisions eventually lead to more or less successful market results. In other words: all category-II variables exist in the end-consumers, and all category-I variables exist in the realm of technology. This is too naive, however. The dynamics of market and technology are coupled. ‘Design’-decisions can be taken anywhere, ranging from the development of enabling technology to marketing decisions. There are mutual influences. Developments in technology may follow from market trends, and vice versa. The distinction between market-driven and technology-driven design situations reflects itself, among other things, in terms of our formal framework from section 4, in the attribution of certain variables either in category I or in category III. For instance, market-driven design will assume market parameters (e.g., pricing, volume and advertisement-related variables) to reside in category III, whereas technology-driven design may attribute them to category I. So, again, one of the purposes of the typology of design problems we are constructing is to distinguish these various influences and to enable unravelling the various dependency paths. In particular, we want to distinguish design processes that result from autonomous, internal processes (roughly: the distinguishing feature is the importance of category-I-variables) from those that follow from external drivers (roughly: the distinguishing feature is the importance of category-III-variables). Therefore we need a vocabulary to tell apart these various mechanisms. We will use the terms push and pull [Ulrich, 1995]. The distinction between pushing and pulling, applied to market and technology leads to four types of forces that drive innovation and design:

Typologies of Design Practice

481

market push (on consumers): the market is seen as a moving inert object: even in the absence of external forces it attempts to continue its movement due to internal mechanisms such as competition and latency in supply chains. Indeed: players in the market are assumed to co-exist in free economic competition. Competition is assumed to be a process amongst vendors and suppliers, internal within the market. Next generations’ products need to possess selling features that are novel — even if there is no a priori customer need for such features. For instance, in markets such as fashion, automobiles, mobile phones, et cetera, it is expected that vendors offer a new collection of products every year — simply because they anticipate their competitors doing the same: a mutual belief that sustains itself. technology push (on the market): technology is seen as a moving inert object: even in the absence of external forces it attempts to continue its movement due to mechanisms that are internal to the domain of technological development. Because of the intrinsic link between (applied) technology and (fundamental) scientific research, there is a continuous stream of novel technology options (novel materials, novel devices, novel components, novel manufacturing techniques, ...). It seems like an autonomous process that sooner or later these technological options give rise to new products or systems. For instance, RFID technology has now become so cheap and robust that various companies invest in projects intended to seek applications. market pull (by consumers): the ‘world outside’ consists of consumers. The demand for new products or services can result from existing customer needs. Such customer needs can take various forms, for instance: • primary consumer needs, such as food, shelter, health and comfort (essentially, the Maslow-pyramid, see [Maslow, 1954]); • social phenomena, such as a commonly felt uneasiness with respect to criminal offences asks for products in the range of security and crime prevention (improved restricted access systems, surveillance systems, ...); • an earlier generation technological product has certain shortcomings. For instance, using mobile phones in cars jeopardizes driving safety, which asks for hands-free systems; • hypes and social trends: the merchandizing that accompanies large manifestations or public events (such as international sports tournaments, popular television series, movies, ...) often forms substantial product- and market volumes. technology pull (by the market): the ‘world outside’ consists of the market. Some developments in technological research are triggered by demands from the market. For instance, there is a large amount of fundamental scientific research being done in the development of longer-lasting batteries for mobile applications — sparked by a market-driven proliferation of mobile devices, and the increasing number of energy-consuming functionalities that these devices are decked out with.

482

Kees Dorst and Kees van Overveld

The difference between ‘push’ and ‘pull’ is clarified in the schema of figure 2. The lighter end of each arrow indicates the initiative-side of the force: technology push and market push originate in technology and market, respectively; technology pull and market pull originate in the market and the consumers, respectively. c o n su m e rs

m a rk e t

te c h n o lo g y te c h n o lo g y p u s h

m a rk e t p u sh

te c h n o lo g y p u ll

m a rk e t p u ll

Figure 2. Technology, market and consumers and the pull and push relations between them Some design projects are mainly driven by one of the two market forces, some design projects are mainly driven by one of the technology forces, and there are various combinations. In all, this gives us 9 types of driving forces for design projects, as depicted in table 1. Table 1. Driving forces for design projects technology push

market

push neither push nor pull pull

pull

A D

neither push nor pull B E

G

H

I

C F

Next we give just a couple of examples of what happens in these boxes to explain the strengths and weaknesses of this approach to understanding design practice. • A: Technology push and market push — In this case, there is both an autonomous technological development and a strong competition in the market. For instance, in hi-tech domains, technology push fuels the competition between major players that are each assumed to posses their own research facilities. Research results are proprietary, e.g. in the form of patents or specialized equipment, and these form the critical assets that should bring competitive advantage. For instance, subsequent generations of semi-conductor devices come to the market as a consequence of Moore’s law, a semi-empiric rule that states that the amount of computing or storage capacity, available per dollar, doubles every 18 months. Major semi-conductor vendors rely on their own specialized research labs.

Typologies of Design Practice

483

• B: Market push, technology neutral — This is the case of highly competitive markets, that do not rely on novel technology in order to innovate. A breakthrough can consist of exploiting a new market niche, often by promoting a new usage for an existing product. For instance, long before it became a mass-commodity, the walkman existed as a compact audio recording device, typically to be used in office applications (e.g. as a Dictaphone). The discovery that this device could be used to bring musical entertainment while doing outdoor sports ignited a massive PR campaign by Sony, but hardly any technological modifications were required. • D: Technology push, neutral market - Technology push in a market neutral context occurs if no competitive arena with market players exists (yet). The famous M3 Post-it product was not the result of an attempt to improve M3’s market position in the stationery market. Rather, the haphazard invention of a sticky, but not permanent glue asked for an application. Often, innovations that can be labelled as ”technological serendipity” fall in this category. The nine cells together give us a more detailed framework to consider the classic division between normal and revolutionary design [Vincenti, 1990]. Revolutionary design would be most likely to occur when market AND technology forces have been combined in a creative and novel way (in table 1, those are the cells, A, C, G and I).N ormal design would possibly occur when the design project can be seen as a direct and fairly literal response to either force, resulting in an artefact of comparatively low novelty (in table 1, cells D, E, F and H). A major part of design practices is like that: not necessarily displaying creative leaps that are often associated with design. Designers create value through the artifacts they envisage and realise, as a creative response to an ever changing world. This response can be an active one, in the sense that some designed artefacts actually aim to considerably influence the way we live — or it can be a more passive one, an evolutionary processing of the changes in the world in a next generation of artefacts. Much of the work that designers do is the creation of such small, incremental change as a response to slight changes in the forces of the market or technology. Even companies that present themselves as ‘innovative’ have to watch their resources and they tend to innovate as little as possible, but just enough to stay ahead of the competition. One could argue that the typology of design contexts that is outlined above is of limited use, because the small but important ‘active’ strand of design, that aims to be ‘Revolutionary’ largely escapes the forces of Market and Technology. We would respond that most of design is not of this type, and capturing 90% of design practice is a noble achievement in itself. And even the ‘Revolutionary’ type of design practice will always be in a dialectical relationship with the contextual forces we have described.

484

7

Kees Dorst and Kees van Overveld

CONCLUSION: TOWARDS A MULTI-PARENT TYPOLOGY OF DESIGN PRACTICE

In our quest for developing a typology of design practices we have come up against the sheer complexity of ‘design’ — and hence our undertaking has not been a simple one. We have constructed a rich and diverse picture of the field. But that is not enough. For a typology of design practices we need a way of connecting these frameworks into a persuasive whole. We would want to arrive at a typology that is intellectually solid, acceptable as a valid representation of real-world design and practical for purposes that one might have. Yet after all this work we have to accept that such an integrated typology is not within our reach yet. The four frameworks we have used to describe design, despite connections between them, are based on four different aspects of the design situation: the design problem, the design activity, the designer and the design context. Our quest for clarity has led us to four descriptions of design from different viewpoints, in nearly incommensurable terms. There is no single grand unified description behind these four partial descriptions. For instance: one cannot validly say that a certain type of design context will always be connected to a certain type of problem solving process, a certain subjective approach by the designer, or a specific dynamic in the treatment of category-II design variables. Although there might be connections (co-occurrences) that are much more plausible than others, and the notational framework that was presented in Section 4 can act as a kind of Esperanto, a language on the very detailed and intricate level of steps in design thinking in which the concepts used in the other descriptions can be expressed. We know that there are patterns in design, because some of these patterns have been given names in design practice. Thus we hear designers and design thinkers talk in meta-terms, as we discussed: some design projects are classified as ‘Normal Design’ and some as its apparent opposite, ‘Revolutionary Design’. These metaterms can be loosely described within the frameworks we have presented here; yet it is difficult to pin them down, as they are no scientific terms: in everyday parlance, people do not define what aspect of design is supposed to be ‘normal’ or ‘revolutionary’. Nor is it explained whether these are supposed to be objective or subjective terms. The frameworks we have presented here challenge designers and design thinkers to define what they mean when they use terms like this in dealing with design issues– a key prerequisite for attaining better, more in-depth discussions on design. The framework we have built doesn’t provide us with a typology of design, yet it does provide us with some of the tools we need for constructing such a typology. And as tools, they are related: the four aspects of design that have been the central thread in our thinking can be seen as layers of description that can be used in relation to one another in a process of creating a typology of design. When we start creating a typology of design, it makes sense to use these tools in a particular order: from the general to the particular, and from the ‘from the outside inwards’, as it were: from the context in which design takes place through some

Typologies of Design Practice

485

intermediate steps all the way to the very detailed level of concrete design actions and design thinking. The four different frameworks we have discussed support these steps: the context-based approach (Section 6) provides a characterisation of the framework in which design takes place, and the 9 different types of context will result in different types of design problems (See Section 3). These are to be interpreted by the designer, which is where the designer’s subjectivity and expertise level (Section 5) come into play. Then this will translate into a specific set of issues that need to be resolved in the design situation, leading to design activities that can be modelled explicitly in a notational framework like the one presented in Section 4. Thus what we have arrived at is a broad, 4-part framework for the description of design practices. A typology of design practices can be developed from this by using this fourfold framework as a tool to describe many cases of design practice, and then see what patterns emerge. Those groups (clusters) of related design practices will actually be the types within the typology. From this bottom-up process (from empirical data to patterns, to initial categories, to the establishment of types), the organising principles could then be derived. This bottom-up process is the only way forward: at the end of Section 2 we already intimated that the we would need to create a multi-parent typology for an activity as complicated as design. A single-parent typology could have been be attained in a neat, top-down manner through the logical analysis of a single key underlying principle. The creation of such a multiparent hierarchy for design (here based on four ‘parents’: the design problem, the design (thinking) process, the properties of the designer and the design context) cannot be derived in such a manner. The patterns in design practice that make up the multi-parent typology can only emerge through painstaking description of many cases. That is a formidable task, the enormity of which could in practice be alleviated by the fact that a typology is always teleological, always for-something. That goal will determine which parts of this overall, four-layer framework for typifying design practice are more relevant to the user of the typology, or where these descriptive frameworks need to be extended (e.g. the description of the designer that has been presented in this paper is exceptionally poor: designers are much more complicated beings than just exponents of an expertise level). A prerequisite for this is that the detailed notational framework that we described in Section 4 (or frameworks like it — one can base these on several different principles) should be further developed — such a rigorous and detailed framework for describing design is needed, as an Esperanto that bridges the four descriptive frameworks. This is a great challenge: any detailed description of design is bound to be complex. Upon reflection, this could be the reason such a notational framework has never been fully developed before — it will be too complicated, cumbersome and time-consuming to be directly useful in design practice, and the academic field of Design Studies tends not to stray too far from practical application. Yet such a notational framework is absolutely vital for the creation of

486

Kees Dorst and Kees van Overveld

a typology of design, a typology that will in the end not only be intellectually interesting, but also have important practical applications. The four-layer framework and the typology that can spring from it could be a way to move beyond the unbearable vagueness of the word ‘design’, and serve to make discussions around design more precise, allowing a deeper intellectual engagement with this fascinating and deeply human activity. BIBLIOGRAPHY [Ansoff, 1968] H. I. Ansoff. Corporate Strategy: An Analytic Approach to Business Policy for Growth and Expansion. Penguin Books, 1968. [Biggs et al., 1986] N. Biggs, E. Lloyd, and R. Wilson. Graph Theory 1736-1936. Oxford University Press, 1986. [Bucciarelli, 1994] L. L. Bucciarelli. Designing Engineers. MIT Press, Cambridge MA, 1994. [Carroll, 1865] L. Carroll. Alice’s Adventures in Wonderland. Oxford, UK, 1865. [Cross, 2006] N. G. Cross. Designerly Ways of Knowing. Springer Verlag London, 2006. [Cross, 1984] N. G. Cross. Developments in Design Methodology. Wiley, Chichester, 1984. [Cross, 1990] N. G. Cross. The Nature and Nurture of the Design Ability. Design Studies, 11(3), 127-140, 1990. [Cross, 1992] N. G. Cross. Roozenburg N.F.M. Modelling the Design Process in Engineering and in Architecture. Journal of Engineering Design 3(4), 325-337, 1992. [Dorst and Cross, 2001] C. H. Dorst and N. G. Cross. Creativity in the design process: coevolution of problem-Solution. Design Studies 22, 425-437, 2001. [Dorst, 1997] C. H. Dorst. Describing Design - A Comparison of Paradigms. Thesis TUDelft, 1997. [Dreyfus, 2002] H. L. Dreyfus. Intelligence without representation — Merleau-Ponty’s critique of mental representation. Phenomenology and the Cognitive Sciences, 1, 367–383, 2002. [Dreyfus, 1992] H. L. Dreyfus. What Computers Still Can’t Do. MIT Press, Cambridge MA., 1992. [Frankenberger and Badke-Schaub, 1996] E. Frankenberger and P. Badke-Schaub. Modeling design processes in industry — empirical investigations of design work in practice. In: O. Akin and G. Saglamer, eds., Proceedings of DMD’96, Istanbul, 1996. [Gero and Kannengiesser, 2002] J. S. Gero and U. Kannengiesser. The Situated FunctionBehaviour-Structure Framework. In: Gero, J.S. (Ed.) Artificial Intelligence in Design ‘02 Kluwer, Dordrecht, 89-104, 2002. [Gero et al., 1992] J. S. Gero, K. W. Tham, and H. S. Lee. Behaviour: A Link Between Function and Structure in Design. In: Brown, D.C., M.B. Waldron and H. Yoshikawa (Eds.) Intelligent Computer Aided Design, Elsevier, Amsterdam, 193-225, 1992. [Gero and Rosenman, 1990a] J. S. Gero and M. A. Rosenman. A Conceptual Framework for Knowledge-Based Design Research at Sydney University’s Design Computing Unit. Artificial Intelligence in Engineering, 5 (2), 65-77, 1990. [Gero, 1990b] J. S. Gero. Design Prototypes: A Knowledge Representation Schema for Design. AI Magazine 11 (4), 26-36, 1990. [Hatchuel, 2002] A. Hatchuel. Towards Design Theory and expandable rationality: the unfinished program of Herbert Simon. Journal of Management and Governance, 5, 3-4, 2002. [Hey, 2006] J. Hey. On the failure of modern species concepts. Trends Ecol Evol 21, 447-450, 2006. [Houkes et al., 2002] W. Houkes, P. E. Vermaas, C. H. Dorst, and M. J. de Vries. Design and Use as Plans: An Action-Theoretical Account. Design Studies 23, 303-320, 2002. [Ivashkov, 2004] M. Ivashkov. ACCEL: a Tool Supporting Concept Generation in the Early Design Phase. PhD thesis, ISBN 90-6814-581-9, Eindhoven University Press, Eindhoven, the Netherlands, 2004, urlhttp://alexandria.tue.nl/extra2/200411683.pdf [Lawson, 2006] B. R. Lawson. How Designers Think — The Design Process Demystified. Architectural Press, Oxford, 2006.

Typologies of Design Practice

487

[Maslow, 1987] A. Maslow. Motivation and Personality. NY: Harper, 1954. Second Ed. NY: Harper, 1970. Third Ed. NY: Addison-Wesley, 1987. [Reymen, 2001] I. M. M. J. Reymen. Improving Design Processes Through Structured Reflection: A Domain-independent Approach. PhD Thesis Eindhoven University of Technology, Eindhoven, 2001. [Roozenburg and Cross, 1991] N. F. M. Roozenburg and N. G. Cross. Models of the Design Process — Integrating across the Disciplines. In V. Hubka, ed., Proceedings of ICED 1991, Heurista, Zurich 1991. [Rosenzweig, 1996] M. L. Rosenzweig. Species Diversity In Space and Time. Cambridge University Press, Cambridge, UK., (revised ed.) 1996. [Simon, 1992] H. A. Simon. Sciences of the Artificial. The MIT Press, Cambrigde MA, 1967, 1992. [Sch¨ on and Rein, 1994] D. A. Sch¨ on and M. Rein. Frame Reflection. Basic Books, NY, 1994. [Sch¨ on, 1983] D. A. Sch¨ on. The Reflective Practitioner: How professionals think in action. London: Temple Smith, 1983. [Simon, 1973] H. A. Simon. The structure of ill-structured problems. Artificial Intelligence, 4, 181-201, 1973. [Suchman, 1987] Suchman, L.A. Plans and Situated Actions. Cambridge University Press, Cambridge UK, 1987. [Soulsby, 1933] B. H. Soulsby. A Catalogue of the Works of Linnaeus. London, UK, 1933. [Ulrich and Eppinger, 1995] K. T. Ulrich and S. D. Eppinger. Product Design and Development. McGraw Hill, NY, 1995. [VDI 2221, 1985] VDI 2221. Methodik zum Entwickeln und Konstruieren technischer Systeme und Produkte. VDI-verlag, D¨ usseldorf, 1985. [Vermaas and Dorst, 2007] P. E. Vermaas and C. H. Dorst. On the conceptual framework of John Gero’s FBS model and the prescriptive aims of design methodology. Design Studies 28 (2), 133-157, 2007. [Vincenti, 1990] W. G. Vincenti. What Engineers Know and How They Know It — analytical studies from aeronautical history. Johns Hopkins University Press, Baltimore, 1990.

This page intentionally left blank

TRANSLATING CUSTOMER REQUIREMENTS INTO TECHNICAL SPECIFICATIONS Marc J. de Vries

1

INTRODUCTION

One of the case studies in Walther Vincenti’s book What Engineers Know and How They Know It deals with the establishment of design requirements that are related to pilots’ experiences with aircraft [Vincenti, 1990]. Vincenti describes how aircraft designers in the period 1918-1943 gradually learnt to use pilots’ flying experiences to set design requirements. Vincenti shows that this was by no means an obvious issue for designers at the beginning of that period. What was obvious is that flying qualities of an aircraft were related to the balance between the stability of the aircraft and the extent to which the aircraft yielded to control actions. Stability was seen as a necessity by the designers, because otherwise the pilots constantly had to correct the motions of the aircraft even when flying in a straight line. But too strong a stability would make it too difficult to change the flight direction. This was expressed in a rule of thumb: “stable, but not too stable”. That, of course, was a pretty vague way of stating a design rule. Besides its vagueness, another disadvantage of the rule was that it did not take into account the experience that pilots in the early days of flight had no problems at all flying fairly unstable aircraft. To the contrary: some of those unstable aircraft were seen as among the best by pilots. From Vincenti’s account it can be derived that in the 1918-1943 period a number of learning processes took place. In the first place designers learnt to recognize the pilots’ experiences as a factor that is important for their work. They gradually realized that the theoretical considerations concerning stability versus control did not guarantee satisfaction of pilots. But their problem was that pilots expressed their needs in so vague terms that it was difficult to relate them to the theoretical concepts (stability, control) that the designers used. Most pilots were able to express their opinion about the flying qualities of an aircraft as ‘good’ or ‘normal’, but that was, of course, not very informative for designers. A second learning process was to get to know the influence of aircraft characteristics on the stability and control. This was a matter of experimentation, both in wind tunnels and in real flying. Thus the designers learnt what variables were relevant for stability and control. For example, the influence of the force Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

490

Marc J. de Vries

exerted on a steering stick on the elevator angle and the resulting flying qualities for various designs was studied. Such knowledge enabled them to establish the balance between stability and control as they as designers perceived it to be appropriate. The third learning process was to relate this balance with the pilots’ experiences, vague as they were expressed. Pilots became motivated to contribute to this learning process as they found that in longer flights a lack of stability (about which they had not cared before in short flights) did result in tiredness because of the constant need to control the motions of the aircraft. Evidently it was in their interest that designers would find the right balanced between stability and control. This learning process was a matter of continuous conversations between engineers and pilots. Vincenti describes how this demanded certain persons who served as ‘translators’, who were able to understand what pilots meant and to translate it into terms that engineers could refer to. Thus a process took place that Vincenti characterizes as the transformation of an ill-defined problem with subjective human elements (the pilots’ experiences) into a well-defined objective problem. A similar struggle to relate users’ desires to product characteristics can be found in one of the many research fields in which the Philips Natuurkundig Laboratorium was active: recording of sound and images. This industrial research lab was and still is one of the largest in the electronics industry. Its history shows that it was one of the few that survived history because of its ability to seek a proper balance between a focus on its company’s direct interests and a more independent search for knowledge of phenomena that might lead to improved or new products [De Vries, 2005b]. One of the fields to which the lab has contributed substantially is that of optical recording (of which today’s primary examples are CD and DVD). The struggles that the lab went through in this field are very much concerned with relating customers’ requirements and physical features. The development of optical recording in the Nat.Lab. was not triggered by customers’ wishes (as was usual in the 1947-1972 period of the lab’s history), but by the technical possibilities. A Nat.Lab. researcher had noticed a disc containing optically stored information (small images), which was presented at a technical exhibition in Italy. From that the idea was born to develop a disc with optically stored information for educational purposes. A list of expected customers’ requirements was made up and contained such issues as: capability of storing both images and sound, a two-dimensional medium, random access, and preferably not of a ribbon-type. This list reveals that it is improbable that customers had been consulted as all of these are already technical translations of expected underlying customer concerns. Probably the perceived customer concerns were such issues as: video as the desired content of information, compactness, and durability of the medium. These requirements were seen as compatible with the idea of a disc on which an analogue video or audio signal could be stored in the form of dimples that were read by a laser that had been attached to a movable arm. Philips was not the only company that came up with such technical translations of the expected customers’ concerns, and thus a race started to be the first with a prototype in which all of

Translating Customer Requirements into Technical Specifications

491

this functioned well. In the end the race appeared to have only losers. No company was able to bring out the product with commercial success. Reflection on this failure revealed that important customer requirements had been overlooked, and also that the translation into the chosen physical features had been wrong in a number of cases. Customers appeared to prefer a wide availability of software (c.q. the availability of a wide variety of movies on disc) above the requirements that had been used in the early phases of development. The success of the product that later emerged as a spin-off from this failed innovation, the CD, indicated that customers also were interested in sound without images (i.e. music) when the quality of the sound was better than that of a vinyl record. Also the compactness, which certainly had not been a priority in the development of the video disc (it was as big as a vinyl record) appeared to be valued by customers much more than expected. These are examples of overlooked customer requirements, which might well have been discovered earlier if customers had been consulted. But also it appeared that the translation of customer requirements into product specifications had not been unproblematic. Before the idea of the storage of information in the form of dimples in the surface of the disc was used, several other options had been tried out: small pictures on a disc (to be read by a television camera pick-up tube, electrical charges on a disc, and small holograms on a disc (most of these options had come from the RCA lab in the USA). Apparently it was by far not obvious how the perceived customer requirements could best be addressed technically. In this case the lessons were learnt in a very costly process. The two cases that have been sketched above are from quite different technological areas, but show important similarities. In both cases there is a struggle to become aware of and deal with customer requirements. Both stories show that listening to users was by no means regarded as a necessary component of design work. Also both cases indicate struggles to relate customer requirements to technical realizations. That, too, appeared to be a non-trivial matter. In the following sections the transitions from customers’ requirements to physical features will be analyzed more systematically. First we will see how the role of the customer in technological developments has changed in the second half of the 20th century, and the role Continental philosophy of technology had in that. This change has invoked the need to reflect on how to transform customer requirements into technical specifications. Then it will be shown that the nature of this transition can be described in terms of a transition between languages and between artifact natures. The latter transition entails different types of knowledge related to those natures. Next we will see how in practice methods are used to support the transition from customer requirements to technical specifications. Two of such methods will be analyzed in some more detail. The analysis will show that it is important to take into account the methodological assumptions of each of these methods, and how these can make a difference for various applications (artifacts, materials, software). Then the problem of the customer as an aggregate rather than as a collective, which features in the practice of using such methods, will be discussed. Then we will see the importance of cooperation between disciplines for transform-

492

Marc J. de Vries

ing customer requirements into technical specifications and the consequences this has for the nature of technological sciences. Finally we will critically examine the term ‘translation’ that will be used throughout this chapter. 2 THE CHANGING ROLE OF CUSTOMERS IN PRODUCT DEVELOPMENT The historical case studies in Section 1 illustrate how engineers can struggle with the challenge how to transform the perceived requirements of future clients and users into physical features of the artifact that they are designing. The fact that engineers seek translations from customer requirements into physical features does not necessarily mean that customers themselves have a say in this process. It may well be that the engineers do not feel any need to consult customers because they are convinced that they themselves know what is good for customers, perhaps even better than the customers. Both case studies have illustrated that. They are the experts, while the customers are only laypeople. This attitude is visible in the ‘technology push’ view on technological developments. It was summarized in the slogan of the 1933 World Fair: “Science Invents; Industry Applies; Man Conforms”. This view flourished in the 1950s and 1960s when there was a general feeling that technology could play an essential role in building up a new society in which all the bad memories of WWII would be overruled by the wealth and prosperity that technology would bring. There was economic growth and this offered people ample opportunity to take advantage of the new technologies. This is, for instance, the period of the electrification of the households (both in Europe and in the USA; see [Morton, 2002]). New electrical artifacts invaded the lives of all people and not just the happy (richer) few. In an industrial research laboratory such as the Philips Natuurkundig Laboratorium the dominant idea was that the engine of technological progress is not customer requirements, but science. The focus was not so much on transforming customer requirements into physical features, but transforming natural phenomena (as studied in science) into physical features. This idea, too, was reflected in the World Exhibition slogan. It was also the core of Vannevar Bush’s vision in the report ‘Science: The Endless Frontier” that was published in the USA shortly after the end of WWII. His advice to the president was to stimulate ‘basic’ research as this would almost certainly lead to industrial progress. The great example for this doctrine was the atomic bomb, which was a spin-off from nuclear physics. Not long after that another example came up: the transistor, which was clearly a result of applying solid state physics theory, as earlier efforts to copy the structure of a triode valve into solid state in a purely experimental way had failed. Customer requirements had played no part in this. They only served as a trigger for the development. Searching for the transistor was motivated by the need to have a more compact amplifier for telephone switching devices. Ironically the first applications had nothing to do with telephony: they were in hearing devices, the computer and in (transistor) radios. So the driving force behind the development of the many

Translating Customer Requirements into Technical Specifications

493

new artifacts in the 1950s and 1960s was believed to be not so much customer requirements but rather scientific knowledge. One can question, though, if that is the real picture. The history of the Philips Natuurkundig Laboratorium in that period shows that those cases in which the development of a new artifact was the result of applying knowledge produced by ‘basic’ research in the laboratory were the exceptions rather than the rule. Many ideas coming from the lab were rejected by the Product Divisions (PD’s) of Philips as not commercially feasible. Evidently, there was concern about the opinion of customers. This does, however, not mean that those PDs actively approached customers to get to know their desires. It was the engineers working in the PDs who claimed to know what customers would appreciate. So although their idea was not to transform natural phenomena (as studied in ‘basic’ research) into physical features, but rather (perceived) customer requirements, the customers themselves did not have a say in the transformation process. In this context philosophy of technology started to emerge as a new discipline. The number of artifacts that were introduced in households and in society in general, was so large that a number of Continental philosophers started writing about technology with great concern. Jacques Ellul presented the view that there was no longer just a collection of separate artifacts, but that all these artifacts worked together and got the character of an autonomous system on which society had lost grip. Martin Heidegger warned for the effect that these omnipresent artifacts had on our perception of reality, namely as only a resource. Both Ellul and Heidegger were pessimistic about the possibilities of changing this situation1 . I will not go into the various ways in which these views can be criticized but just mention it as a token of the uneasiness that several Continental philosophers felt about the effects of the omnipresence of artifacts. Their writings were one of the factors that would eventually cause a change in the role of customers in product development. They created an awareness that without a conscious social influence on technological developments, technology would not only be beneficial but could also become a threat. One thing, though, should be mentioned. Historical research has shown that customers did have a substantial influence on the implementation and diffusion of artifacts. Customers, for instance, had a substantial impact on the choices between different types of energy supply (gas versus electricity) in households. This, however, does not mean that customer requirements also had an effect on the development of technology. By the end of the 1960s the situation changed, perhaps not primarily because of the philosophers’ warnings, but because the economic growth came to an end. Also, the first negative large-scale effects of technology on the natural environment became evident. People started questioning the perceived dominance of science and technology in society. It was the time in which Technology Assess1 Much later, Albert Borgmann went in Heidegger’s footsteps and focused on individual artifacts (see [Borgmann, 1984; see also [Verbeek, 2005]). According to him the artifact offers commodities that disengage us from reality. The artifacts serve as intermediaries between us and the lifeworld and thus cause us to loose contact with the natural environment.

494

Marc J. de Vries

ment emerged as an instrument to gain social control over the development of new artifacts by investigating their possible effects on various aspects of social life. This did not work out for all technological developments. Nuclear energy, for instance, was developed in spite of substantial social resistance. But the new attitude towards technology did have an impact on mass consumer product development. The economic stagnation caused consumers to be more critical towards new technological products. Until then customers could afford to buy any new product but now customers were more selective and looked more carefully if the new product really addressed their needs and desires. Such changes also affected the work in industrial research labs. In the Philips Natuurkundig Laboratorium, for example, it created a new relationship between the lab and the PDs. The lab had to take into account the desires of the PDs in their choice of research areas. Also, new mechanisms for co-operation between the lab and PDs were created in order to help the lab get a better understanding for what customers would value. But perhaps the most important change for the topic of this chapter is that also mechanisms were developed to help engineers get an understanding of what customers wanted by getting information from them rather than by working with the engineers’ own estimations of what customers would desire. This development can be seen in the way the concept of quality changed from the 1970s on. Had it been originally intended to be focused on the control of production processes to prevent the sell of malfunctioning artifacts, later it turned into a concept that involved whatever could be done to design and produce artifacts so that it would please the customer, and also it was felt that the customer’s own input was indispensable for that. This perspective on quality was named ‘Total Quality Management’. In the (Continentally oriented) philosophy of technology such changes were reflected in new ideas about the role of society in technological developments [De Vries, 2005a]. Philosopher Andrew Feenberg showed that users could not only influence the diffusion of new artifacts in society, but also their function. He used the example of the Minitel computer system in France to illustrate this [Feenberg, 1999]. Customers hacked the system and redefined the artifact after its introduction. It was meant to be a system for disseminating information by the government, but soon was changed into a system for exchange of information between users by those users themselves. This, however, still does not mean that users were able to influence the design of the artifact. They did redefine its function, but they did not change the physical features of the system. The same limitation can be seen in the view of the social constructivists, who claim that each technical artifact is a social construct, because it is the users rather than the engineers that determine what it is (for). Wybe Bijker came up with the — now classical — example of the bike [Bijker, 1995]. Some users saw it as a ‘macho machine’ that boys could use to show their courage to girls rather than as a transportation means. This, however, does not mean that the bike was designed as a ‘macho machine’ by listening to the customers’ (i.e., the boys’) requirements. Not only the social constructivists, but also the Continentally oriented philosophy of technology does not offer much clues about how customers’ requirements really impact the development of artifacts.

Translating Customer Requirements into Technical Specifications

495

Perhaps this has to do with the fact that nearly all philosophers in that tradition take an outsiders’ perspective. They do not very much investigate the work of the engineers ‘from inside’. In the remaining sections of this chapter I will draw primarily from an analytically oriented stream in the philosophy of technology, in which we find more of that ‘insider perspective’. 3 DIFFERENCES BETWEEN USERS’ AND DESIGNERS’ DESCRIPTIONS OF TECHNICAL ARTIFACTS We have seen historical examples of how customer requirements were translated into physical features, and we have also discussed briefly the changing role of society in the development of artifacts in technology. I will now make a first analysis of why at all there is a transformation from customer requirements into physical features of the artifact. I will argue that two types of transformations are at stake here: translation between languages and a translation between natures of the artifact. One of the complicating factors in the transformation of customer requirements into physical features is that customers and engineers speak different languages. In a Wittgensteinian term, one could say that they use different language games. Wittgenstein has pointed out that in different contexts people can use different rules for using words. This can be recognized in our topic. Engineers use a language game in which a precise definition is important. A list of requirements as set up in a design process contains well-defined terms and to a large extent it is quantitative, which gives it an even more exact appearance. Even when uncertainties are involved they are quantified (for instance, in statistical production error margins). In engineering the often vague way in which a term is ‘defined’ in ordinary life’s language is not acceptable. Engineers do not consider such terms to be valid elements in their language. The general public, though, does not necessarily experience its language as ill defined (this is the predicate that Vincenti uses to characterize the customers’ language; he also used the term ‘translator’ for a person who was able to make connections between users’ and designers’ concerns). In that context other norms are used for determining whether or not a term is well defined or ill defined. Also the terms engineers use are very specific for their profession. Most terms are unknown to non-engineers. Even when engineers use a term that also features in ordinary daily life language, it often has a different meaning. Engineers using the term ‘weight’ do not mean the same as in the general public’s use of the term (they would use the term ‘mass’ in stead of what the public calls ‘weight’). So both the meaning of terms and what is accepted as a valid term can differ between engineers and customers. This difference in language games can easily give rise to misunderstandings and other difficulties in the communication between engineers and customers. Such difficulties can be recognized in the practice of industrial companies when engineering departments communicate with management, marketing and sales departments. This is the basis of many jokes in the well-known comic Dilbert by Scott Adams. The engineer Dilbert often finds

496

Marc J. de Vries

himself misunderstood by the management of the company in which he works. In the comic this makes us smile. In industrial practice, though, it is a matter of real concern. Apart from a language transformation, also a transformation between two natures of the artifact takes place when customer requirements are transformed into physical features. A technical artifact can be described according to two natures: a physical nature and a functional nature [Kroes and Meijers, 2006]. When we consider the physical nature of the artifact we are dealing with its size and shape, its weight, color, smell, its materials, its conductivity, etcetera. When we consider its functional nature we focus on what it can be used for. Customer requirements in most cases refer to the functional nature of an artifact (later on I will show that exceptions exist). In fact this is what customers can speak about most comfortably, as they lack an expertise in many of the aspects of the physical nature. Physical features refer to the physical nature. This is where the engineers have their expertise. What is needed in the design of the artifact is a translation from statements about a desired functional nature into statements about a possible (and preferably also a feasible) physical nature. This is in fact the core of the design challenge.

4 KNOWLEDGE ABOUT THE ARTIFACT’S NATURE In expressing their desires, customers draw from knowledge they have about existing artifacts, that can be referred to when new artifacts are designed (in fact most designs can be regarded as re-designs). As suggested in the previous section, both engineers and users have knowledge about artifacts. This knowledge can be at least of three different types:2 knowledge of the physical nature of the artifact, knowledge of the functional nature, and knowledge of the relationship between the physical and the functional nature [De Vries, 2003]. The last-mentioned type of knowledge, of course, is at least as important for the topic of this chapter as the other two types of knowledge. It is that type of knowledge that enables the engineers to make the transformation from customer requirements (related to the functional nature of the artifact-in-design) to the physical nature (related to the physical features of the artifact). Two sub-types can be distinguished here. The first type can be formulated as follows: person S knows that artifact A’s physical property p (or a combination of properties pi ) makes it suitable for carrying out with A the action Ac that results in the desired change in state of affairs φi → φi+1 . For example: a person knows that the length of a screwdriver makes it suitable for opening the lid of a tin can (note that this was not what the artifact was intended for, but the user can decide to ignore the proper function of the artifact and use it for an accidental function). This is the type of knowledge that the customer uses to identify what an existing artifact can be used for, based on its observed 2 The list here is by no means exhaustive but limited to those types that suffice to show the differences between customers and engineers.

Translating Customer Requirements into Technical Specifications

497

physical nature. The second sub-type can be defined as follows: in order to let action Ac with artifact A result in the desired change in state of affairs φi → φi+1 , A should have physical property p (or a combination of physical properties pi ). For example: a person knows that in order to make a hammer hit nails into a piece of wood, it needs to have a heavy head. This is the type of knowledge that the designer uses to identify physical properties that are needed to make the artifact function as it should. It is important to note that the desired change in state of affairs is not only the realization of the function in a strict sense (for instance: the nail being inserted into the piece of wood). It can also include various desires that are related to this function realization. Safety is a good example of that. Implicitly included in the desired change of state of affairs is that in the new state of affairs I am still healthy. Also being bankrupt because of the high cost of realizing the function can be reckoned as undesirable in the new state of affairs. This means that safety and costs are seen as possible elements in the functional nature of an artifact in my account. This, of course, also holds for various other non-physical properties of the artifact that users can include in their desires. It should also be noted that the three types of knowledge that have been mentioned so far (knowledge of the physical nature of an artifact, knowledge of its functional nature, and knowledge of the relationship between physical and functional nature) do not exhaust all knowledge about the artifact, even not when knowledge of the functional nature is taken in the broad sense as indicated. There is, for instance, also knowledge of how to operate the artifact (both users and designers can have this). I have deliberately selected those types of knowledge that are directly involved in the transformation of customer requirements into physical features of the artifact. Sometimes users have beliefs about the relationship between the physical and the functional nature of an artifact too when asked to express their desires with respect to an artifact-in-design, but this does not mean that they also have knowledge of the third type in my short list above. As an example the following experience can be mentioned. A producer of isolating material that is used in buildings found out accidentally that the same material can be used as a substrate for plants. The engineers wanted to optimize the material for this new application as it had not been designed for this originally, and in order to do that they asked farmers about their requirements for such material. The farmers told them that they wanted the material to be homogeneous, in order to guarantee homogeneous growth of plants over the substrate (this was important because the watering system was equally divided in the space where the plants are grown). The engineers knew that this was not a necessary condition for homogeneous plant growth. But the farmers were so insistent about the homogeneity of the material that the company decided to give in and took homogeneity of the material rather than homogeneous plant growth as a requirement. In this case the farmers’ belief about the relationship between the physical nature of the material (homogeneity) and its functional nature (causing homogeneous plant growth) was false. In general the engineers’ beliefs about the relationship between physical and functional nature will be more reliable than the

498

Marc J. de Vries

users’ beliefs, given their greater expertise of the physical nature. False beliefs about the relationship between the physical and the functional nature also can cause problems in the use of an artifact, in particular when accidental functions are at stake. The proper function can be expected to be in accordance with the physical nature of the artifact because the designer chose that nature based on his or her expertise. But accidental functions may be based on false beliefs of what manipulations the physical nature of an artifact will allow. Even total breakdown of the artifact may be the effect of this. Therefore designers may want to consider possible accidental functions in order to adapt the physical nature so that dangerous situations in cases of accidental use will be prevented.

5

TRANSFORMING USERS’ REQUIREMENTS INTO PHYSICAL FEATURES

Having seen the different types of knowledge engineers (and users) can have of an artifact, we are now ready to examine more closely what can be meant by the terms ‘transformation’ and ‘translation’ that so far we have used as if their meaning was obvious. ‘Transformation’ means that we go from one form to the other. In the case of what this chapter is all about, it means going from one type of information to the other. What designers want to do is go from information about the artifact in terms of customers’ requirements to a different type of information about the artifact, namely about its physical features. But how can designers do that. Perhaps the term ‘translation’ gives more insight into that. Translation is not a one-to-one projection. Usually, a translator has more than one word to choose from as the translation of a word in another language, or more than one grammatical construction that is equivalent to the construction in the original language. Ignoring this fact is what leads to the often clumsy translations that are generated by translation software. This is the same in translating customer requirements into physical features of the artifact. One customer requirement can be related to more than one physical feature. For instance, in the aircraft case with which this chapter started, the flying qualities were related to stability and control. In that case both had to be taken into account, as it appeared to be the balance between the two that determined the flying qualities. Vice versa, one physical feature can be related to more than one customer requirement. As the translation of customer requirements is based on relationships between customer requirements and physical features, knowledge of these relationships plays an important part in this translation. The fact that there is no one-to-one relationship has as a consequence that a back-and-forth process has to take place in order to reach a full understanding of the translation. Suppose a certain customer requirement has been found to be related to a certain physical feature, then it is relevant to investigate if that physical feature is related to other customer requirements as well. Satisfying one customer requirement by certain design decisions with respect to that physical feature may have consequences for the extent to which the other

Translating Customer Requirements into Technical Specifications

499

customer requirement, related to the same physical feature, is satisfied.3 The complexity of relationships makes it difficult for designers to make the translation in their minds only. In particular when artifacts get complex, designers feel a need to use methods that visualize the relationships and thus help to acquire a perspective on the totality of relationships between customer requirements and physical features. Some of these methods will be discussed and analyzed in the next section. 6 TRANSFORMATION METHODS Methods that support the translation of customer requirements into physical features are one kind of a whole set of methods that deal with customers requirements. The focus on customers in industrial business contexts has been developed under the umbrella name of Total Quality Management (TQM). The term ‘quality’ has broadened substantially in the 1970s and 1980s. As for the original meaning of the term ‘quality’ for this broader concept too methods were developed. Thus a whole set of methods emerged, related to different phases of the lifecycle. Many of these methods were concerned with the design phase. This can be told by their names: “design for manufacturing”, “design for assembly”, “design for logistics”, “design for cost”, “design for recycling”, etcetera. Most of these methods deal with customer requirements indirectly. They help the designer to choose a physical nature for the artifact that will be easier to manufacture, easier to assemble, easier in terms of the logistics that are needed for its production, cheaper, or easier to recycle. “Easier” mostly results in “cheaper”, and this advantage can be used to lower the price of the product, which will please the customer. Thus the lower price is an indirect result of the “design for X” methods. Some TQM methods, though, have a more direct relationship with the customers’ interests. Two examples of those are: Quality Function Deployment (QFD) and Value Analysis (VA). I will first give the ‘cookbook’ description of the methods and postpone the methodological critique to the next section. The QFD method reflects the idea of a physical and a functional nature of an artifact. The method is based on a matrix in which customers’ requirements (related to the functional nature of the artifact) are in the rows of the matrix, and the physical features of the artifact (its physical nature) are in the columns.4 The physical 3 In this respect Andrew Feenberg’s proposal for a two-step instrumentalization process (made in his book Questioning Technology) seems to fall short. He suggested that there is primary instrumentalisation in which de-contextualisation takes place and the designer can concentrate on the physical nature of the artifact, after which re-contextualization takes place in the phase of secondary instrumentalization. In the first place this process seems to suggest still a certain technological determinism, because early in the development the social context plays no part. In the second place it lacks the back-and-forth movement that was described above. 4 QFD literature commonly uses the term ‘technical’ features for the content of the columns. Here the term ‘technical’ does not include functional aspects of the artifact, although the word ‘technical’ in itself suggests so. To avoid misunderstandings about this I have used the term ‘physical’ in stead of ‘technical’. The word ‘technical’ is used only in the expression ‘technical specification’, as this is very common language in design and also includes functional aspects.

500

Marc J. de Vries

features are usually quantitative parameters of the artifact. The relations between customer requirements and physical features are expressed in the cells of the matrix. This is usually done on a nominal scale (though expressed in numbers: 9 for a strong relation, 3 for a moderate relation, 1 for a weak relation, and 0 for total independence). By adding the relationship scores for each column a total score can be derived that indicates the importance of the physical characteristic in that column. Often a weight factor for each customer requirement is added to each row and multiplied with the relation scores before adding the scores in a column. This enhances the meaning of the total score: it then shows which physical features have the strongest relations with the most important customer requirements. Those features of course should be given priority when designing the product to make sure that they fit with the customers’ requirements.5

Technical Correlations

Technical Characteristics Weight Factors Customer Requirements

Customer Perceptions (benchmarking) Improvement Rates

Prioritised Requirements Interrelationship Matrix

Technical Targets

Figure 1. A QFD matrix 5 The ‘roof’ on top of the matrix shows the relationships between the physical features. This part of the method indicates where trade-offs are necessary because there are conflicts between requirements concerning physical features being given their optimal value. Trade-offs are not further discussed in this chapter because they are independent of the actual issue we are dealing with, namely the translation of customer requirements into physical specifications.

Translating Customer Requirements into Technical Specifications

501

Value Analysis is focused on trade-offs between functions the customer will value and the cost of realizing those functions in (specific parts of) the artifact. The analysis is based on an existing design of an artifact that is to be redesigned, or a draft design of a new artifact. First an analysis is made of all the functions and sub-functions of the artifact and to each function and sub-function a score is attached that represent the extent to which customers value that particular function or sub-function. Then to each function and sub-function the price of the part in which that function is realized is attached. Finally a comparison between values and prices indicates mismatches: parts with high costs but low values for customers can be deleted or maybe their function can be designed into an already existing part by reshaping it; parts with low costs but high customer appreciation may be improved by, for instance, choosing better and more expensive materials. As in QFD, here too relationships between the physical nature (the presence of certain parts) and the functional nature (the extent to which customers appreciate certain functions) are identified. Another similarity between QFD and VA is the effort to express these relationships in a quantitative way. Of these two methods, QFD is the most encompassing because the value that customers give to each function is represented in the weight factors that are added to the rows. For that reason QFD is chosen for further analysis in the next section. The conclusions resulting from this analysis, though, are relevant for other methods as well and therefore they offer more general insights into the issue of this chapter than only for QFD.indexdesign methodology 7

METHODOLOGICAL REFLECTIONS

In this section a critical analysis of the QFD method will be presented. I will first discuss how analyses became part of design methodology after an earlier period of fairly na¨ıve views on methods. Then I will move on to focus on the assumptions that underlie the QFD method and first show what assumptions the QFD method has in terms of identifying customer requirements, for identifying physical features, and for identifying the relations between customer requirements and physical features, and finally I will show that QFD in fact does not make the actual translation from customer requirements into physical features, but only helps to structure the problem of this translation.

7.1

Moving away from a na¨ıve view on methods in design methodology

In his survey of the history of design methodology, Nigel Cross showed how this discipline originally had a fairly na¨ıve view on design processes [Cross, 1984]. It was assumed that all design processes, irrespective of what was designed, should be executed in three main consecutive phases: analysis, synthesis and evaluation. It was only later that design methodologists found that reality was too complex for such a simple scheme. Both as a descriptive and as a prescriptive representation

502

Marc J. de Vries

of design processes the analysis-synthesis-evaluation scheme appeared to fail. The real process that designers went through was much more complicated and differed substantially between different design areas. In particular differences between mechanical engineering design and architectural design were investigated later on in the history of design methodology. Also in the philosophy of technology differences between the design processes in different areas were investigated. Sarlemijn, for example, described three different types of technologies of which the development has quite different features [Sarlemijn, 1993]. Experience-based technologies in his terminology are technologies in which designers’ (and users’) experience with previous designs is the basis for the design work, more so than fundamental theories that describe the natural phenomena on which the functioning of the product is based. Such theories feature in what he called macro-technologies, at least the classical theories of mechanics, thermodynamics and electromagnetism. The third type of technology in Sarlemijn’s taxonomy is that of microtechnologies in which fundamental theories of microparticles and microstructures are a necessary prerequisite for the development of artifacts. By analyzing a number of case studies (for instance, the transistor and the Philips Plumbicon) Sarlemijn showed that customers’ requirements can well be taken into account in experience-based technologies (customers have a fairly good idea of what the product can do) and that such requirements are more difficult to deal with in early phases of microtechnological developments as in those cases understanding of the phenomena requires priority [Sarlemijn, 1993]. It is remarkable that in spite of all the lessons that have been learnt in design methodology, the way TQM methods are used in practice often is very much in line with the early days of design methodology. Again, strong claims of generality are made. TQM literature does not pay attention to differences between different design areas, for instance differences with respect to the extent to which it is possible to deal with customer requirements in early phases of the design process. This suggests that such methods as QFD and VA apply to any design problem in a similar way. Such a na¨ıve approach has resulted in disappointing experiences in business companies in which those methods were used. The analysis of such experiences has added to the awareness that it is necessary to make careful methodological reflections on the methods before applying them to a certain situation. Design methodology and philosophy of technology can have practical implications here. Methods are based on assumptions. Methodological analyses can make clear what those assumptions are and the extent to which they are fulfilled in certain situations. This is why methodological reflection can contribute to the successful applications of methods. This is the case, too, for TQM methods. Usually those methods are described in terms of a prescriptive sequence of actions. Several of those actions require knowledge in order to be carried out successfully. The ‘cookbook’ description assumes that this knowledge is present, but does not state that explicitly. This is where problems will occur: the agents carrying out the actions lack knowledge but are not aware that this causes mistakes in the application of the method. This results in faulty outcomes of the method. There is, though, no

Translating Customer Requirements into Technical Specifications

503

‘mechanism’ that warns the agent when outcomes are based on the input of false beliefs when carrying out the steps in the method. This can be illustrated in the case of QFD.

7.2

Assumptions concerning the identification of customer requirements, physical features, and the relations between them

An important assumption of QFD is that there are customers who can explicate their desires. This assumption is crucial in what Sarlemijn has called the ‘American’ approach to QFD, in which the direction in which the product development goes is determined by customers’ requirements. Sarlemijn contrasts that approach with what he calls the ‘Japanese’ approach to QFD, in which the company sets out the directions for product development on the basis of its own strategic considerations, but consults the customer when it comes to further details [Sarlemijn, 1995]. For instance, when Honda decided to upgrade their successful Civic car model to the high end market this was not done because customers had asked for that, but because the company itself had identified this as a promising market opportunity. It is only within the context of this decision that the requirements of customers in this new market segment were investigated and used for getting specific information about how to develop the Civic into a more luxurious and expensive car. Of course Honda most probably had indications that customers would appreciate a Civic model in a more expensive mode. The Japanese approach thus takes into account the dynamic relationship between company and customers more than the American approach does. In the American approach there is only oneway communication: the customer gives information and the designer uses it. In the Japanese approach designers use an assumed customer wish, then inform the customer about the possibilities of a more expensive Civic model, and then ask the informed customer about his or her desires. Asking customers for their requirements assumes that they are able to express those. Kano has argued that only one of three categories of desires will come out, namely those desires that customers are aware of and will use to evaluate the new product. Kano call’s those qualities linear because the better these are realized the higher the customer’s appreciation. According to Kano there are two other categories of qualities, namely those qualities that the customer does not mention because he takes them for granted. Such qualities are not linear, because if they are realized this does not result in more appreciation, while if they fail to be realized the customer will strongly disapprove. The third category of qualities in Kano’s view are those qualities of which the customer has no imagination and therefore will not miss them when not realized, while if the company is able to realized them the customer will be happily pleased. These can be all sorts of gadgets and add-ons that the more talented designers can come up with. Before the assumption that customers can express their desires comes the assumption that it is possible to identify who should be regarded to be the customers. This can be problematic. An example is the development of coaches for bringing

504

Marc J. de Vries

tourists to their destinations. Companies producing those coaches are faced with the problem of a manifold customer. Prima facie it seems that the tourists are the primary customers. But this is a simplification of a more complicated situation. The tourists do not determine which company’s coach travel organizations will buy. At best such companies try to get to know what their clients, the tourists, value and take that into account when ordering the coaches. But they will also take into account other party’s interests. The drivers, for instance, will probably also get a say in this, as well as the garage that will maintain the coach. It is not difficult to imagine that there will be conflicts between the requirements generated by different clients. The tourists, for instance, will appreciate luxurious seats, while the travel company will be more interested in the price of those seats. This highly complicates the step of QFD in which one list of customer requirements is to be generated for the rows of the matrix. In a way the choice of whose requirements will be chosen for prioritizing is an aspect of what Langdon Winner has called the ‘politics of artifacts’ [Winner, 1986], because this choice will at least partially determine who will benefit from the product mostly. It raises the broader issue of who decides about priorities in customer requirements. Is it the designers? Or is it the company management? Or is it the customer? Such crucial decisions are not supported by the method. The QFD method also assumes that it is possible to come up with one list of physical features (the columns of the matrix) that can be related to the various customers’ requirements (the rows). Particularly in the case of what Sarlemijn called microtechnologies, when the phenomena underlying the functioning of the artifact are not yet fully understood, this can be a serious problem. The development of nanodevices is a striking example of that. The exact quantum behavior of those devices is certainly not fully understood yet. Then there is the crucial step of filling in the cells of the matrix. In fact this is exactly the step in which the transition from customer requirements into physical features takes place. The method assumes that there is knowledge of the relationships between the customers’ requirements (the functional nature of the artifact) and the physical features (the physical nature), because that is precisely the type of knowledge that is at stake here. That knowledge is not logically derivable from knowledge of the physical nature, neither from knowledge of the functional nature. It is really a separate type of knowledge, which usually is gained in practice by using existing artifacts. In cases of artifacts that are relatively new his knowledge may well be absent and sophisticated guesses is probably the best input one will have in that situation. This has the risk, though, that false beliefs are used as the input for the relation scores in the cells of the matrix.

7.3

Does QFD actually make the translation?

The observations described above bring us to an important point for the topic of this chapter: what is the actual contribution of the TQM-methods to the desired transition from customer requirements to physical features? We can see that

Translating Customer Requirements into Technical Specifications

505

none of the methods that have been described includes a mechanism to make the transition from customer requirements into physical features. This becomes very clear in QFD: the knowledge that is needed to fill in the relation values in the cells of the matrix can not be derived from any information that has been inserted into the matrix already, neither from any information that was already in the matrix beforehand. The only thing QFD and similar methods do is to make the problem of the transition explicit and structure it. QFD, for instance, helps to get insights into what the consequences of certain choices in physical feature are for the extent to which customer requirements are met. This function is supported in QFD software, by means of which consequences of alterations in the numbers in any part of the matrix can be calculated by the software. But this structuring function of the method only works if we have this knowledge of the relationship between physical and functional nature, and that knowledge does not come out from the method. So in fact the contribution of the often highly praised methods in fact is quite limited and does not solve the problem that is at the very heart of the design process, namely the proper transformation of customer requirements into physical features. Once the users of the methods have become aware of certain assumptions not being fulfilled and of the limitations of the method (in that it does not solve the problem of the transition from customer requirements but merely structures it), they can still exploit the method for a useful purpose. In the case of QFD, for instance, the identification of what knowledge of the relationship between the physical and functional natures of the artifact is lacking can be used to set up an R&D project in which such knowledge is acquired. For such projects inquiry among customers and scenario methods can be used [Courage and Baxter, 2005]. Users can be asked to express their experiences with existing products or to respond to prototypes or mockups of new products. Thus the effect of changes in the design on the customers’ satisfaction can be assessed. In scenarios the effect of different design decisions on the customers’ satisfaction can be explored in a more theoretical way (by reasoning rather than by observing customers’ responses). One could speak here of ‘pre-QFD’ studies, as Sarlemijn has suggested. Increase in the knowledge of relationships between the physical and functional nature may also provide input for new creative solutions for fulfilling customers’ requirements by physical features. New translations of requirements into features can be brought forward by means of which certain conflicts between the realizations of desirable physical features can be solved. Such conflicts are identified in the ‘roof’ of the QFD-matrix (see figure 1). A simple example may illustrate this. When designing a new flat iron, designers are faced with a conflict between ease of manipulating the iron and exerting sufficient pressing force on to the cloth when both requirements are translated in terms of the weight of the iron (one requirement then requires a low weight, while the other requires a high weight). The conflict can be solved by translating the pressing requirement not by weight but, for instance, by a vibrating movement of the iron’s sole. Thus variations in translations can help to solve conflicts between requirements. The use of these methods as structuring

506

Marc J. de Vries

devices, however, only works if the artifact is not too complex. If, for instance, the QFD-matrix consists of many rows and columns, the designers loose insight into the overall pattern in the relationships, which forces them to a total trust in the mathematical manipulations in the method. Such trust, however, is unjustified, given the fundamental methodological questions that can be raised with respect to the use of numbers in the method. Different types of measurement scales (nominal, ordinal and interval) are mixed up in the method,6 which makes the calculations highly questionable. 8 MATERIALS AND SOFTWARE VERSUS ARTIFACTS AS APPLICATION FIELDS QFD and similar methods have been developed as methods for designing artifacts. But in the course of time the applications became manifold. As new types of applications were tried out, it became evident that the suggestion so often made in the methods’ handbooks as if they could be used irrespective of what was to be designed, was wrong. One of the applications in which it became clear that it deviated from application to artifacts is that of the design of new materials. An example of this was the use of QFD to design a new material for car bumpers and dashboards. A chemical company in the Netherlands made an effort to implement QFD for this purpose and soon found out how materials ‘behaved’ differently from artifacts in the QFD matrix. As remarked previously one of the options for QFD is to ‘play around’ with the translation of customer requirements into physical features by changing the columns (in which the physical features have been written) to see if a different translation of a customer requirement can help solve conflicts between physical features. This appeared to be problematic in the case of materials. The columns in the matrix in that case are not artifact properties but material properties. For artifacts it is not difficult to change one feature without affecting the others. One can add new parts to an artifact that do not cause dramatic changes in the overall design, and thus leaves many other physical features unaffected. In the case of a material, though, this is much more difficult. One can not easily change, for instance, the hardness of the material without affecting the other properties. Doing this would require making changes in the composition of the material and then ‘automatically’ numerous other material properties change also. Perhaps the best way to change material properties without affecting all the other properties is by working with additives that only change 6 This

problem is not unique for QFD but features in other methods for investigating people’s opinions too. Such opinions are usually measured in a nominal scale, for instance by asking respondents to indicate if they strongly agree, just agree, disagree or strongly disagree with a given statement. But then usually these nominal representations of the person’s opinion are transformed into an interval scale by assigning a score of 1, 2, 3 and 4 to the four respective nominal expressions, while in fact it is unknown if the distance between the consecutive alternatives is indeed equal. The same is done in QFD. Relations between customer requirements and physical features have not been measured and can only be estimated in nominal terms, yet they are filled in as scores on an interval scale.

Translating Customer Requirements into Technical Specifications

507

specific properties. But often the options for doing that are limited. Although this experience was related to the use of QFD, it also holds for similar methods. For VA it is already difficult to imagine how at all this method could apply to materials, as there are no parts in the material, while the method is based on the assumption that there are. In that respect this observation reveals an interesting aspect of design that is seldom addressed in design methodological literature, namely that the design of a material is quite distinct from the design of artifact in that the physical nature of the material is more homogeneous than that of artifacts. This limits the freedom of designers to ‘play with’ relationships between the physical nature and the functional nature substantially. Another application area where QFD-type methods need careful methodological consideration is that of software. Here the ‘physical nature’ of the artifact (if at all one is prepared to call a computer program an artifact given its non-material character) consists largely of (sequences of) symbols and perhaps time periods needed for the execution of parts of the program. The latter type of physical feature suggests that software is to be regarded as a process rather than as a product. That may have large consequences for the application of QFD and similar methods. There is, unfortunately, not much literature available which reflects on those consequences. Yet such methods do seem to be applicable to software. For QFD there is already a standardized term for that: Software QFD or SQFD [Haag, Raja and Schkade, 1996].

9 THE PROBLEM OF CUSTOMER AGGREGATES Handbooks of TQM methods that aim at transforming customer requirements into physical features (e.g. [King, 1989] for QFD) are fairly naive in prescribing the user of the method to ‘go to the customer and ask for his/her wishes’. Of course it is not stated as bluntly as that, but the handbooks do tend to overlook the problem of the fact that the customer is not a single agent, but an aggregate of agents, who can hardly be expected to share the same complete set of requirements. This holds in particular when the customers are not only users, as we have seen in the coach development example. This experience with QFD reveals another interesting philosophical issue with respect to the transformation of customer requirements into physical specifications. The customers’ desires to be used as an input for the design process are not single agents’ desires, neither is there a collective customer desire. For many product developments there is no collective of customers that designers can address. The customer market usually is an aggregate of individual customers.7 This problem has consequences for the use of QFD and similar methods. Because there can only be one set of customer requirements in the rows of the QFD-matrix, one has to pretend that there is a set of collective customer requirements, even though this is usually not the case. 7 There are cases in which this problem does not occur, namely when a product is developed for a company, e.g. in the case of ship building. Then the customer is a collective, not an aggregate.

508

Marc J. de Vries

Van de Poel has shown the impossibility of translating individual into collective customer demands without violating one or more reasonable conditions due to Arrow’s Impossibility Theorem [Van de Poel, 2007].8 In some cases it can be justified to pretend that there is a collective of customers, because sometimes customers do influence each other’s behavior and then show strong group awareness, as we can learn from e.g. the sudden popularity of new gadgets among young people. Not being part of a group can put people in an uncomfortable, socially isolated position. Still, customers do not entirely commit themselves to the obligation of buying a product when other customers in the same market will do that. In many cases the best designers can hope for is a sort of ‘weak obligation’ with the customers to do so. But as there is no common decision to commit to such an obligation when the product is introduced into the market, there will always be uncertainty about whether or not such a situation of ‘weak obligation’ in the above mentioned sense is present. The strategy to work towards creating this ‘critical mass’ of buyers in fact is what the Strategic Niche Management approach aims at [Kemp, Schot and Hoogma, 1998]. The basic assumption in this approach is that a product has more chance of reaching a certain market when first it has found footage in a specific part of that market, that is, in a niche. Once the value of the product for customers has been proven in the niche, there is a fair chance that it will also reach the whole market of which the niche is a part. The example of the CD has shown that innovations sometimes take place without a niche first being ‘converted’ but in many other cases it seems that innovations did take place via a niche. In the niche it is easier to investigate the customers’ needs because it is a specific group of customers in which the variety of desires can be expected to be less broad than in the whole market (or at least one can better identify which specific desires have a high importance to the customers in that niche). This is what is called a ‘market niche’. A second type of niche is a ‘technology niche’. In such a niche the previously mentioned lack of obligation can be solved by creating obligations, for instance by subsidizing certain new environmentally-friendly artifacts by the government. The attractiveness of the new product to customers is not intrinsic in the artifact, i.e. based on its fit with customer needs (although this has to be there to a certain extent also), but it lies outside the artifact.

10

THE MULTI-DISCIPLINARITY OF ENGINEERING SCIENCES

A final issue that needs to be addressed are the consequences of the considerations presented in this chapter for the nature of engineering sciences. Probably the most obvious conclusion we can draw is that engineering sciences need to be 8 These conditions are: collective rationality (the collective preference ordering must be complete and transitive), unrestricted domains (no restrictions to how an individual orders the alternatives), the Pareto principle (if everyone prefers A over B, then the collective ought to prefer A over B), independence of irrelevant alternatives (the ordering of two alternatives must not be influenced by the inclusion or exclusion of an alternative), and absence of a dictator (an individual whose preferences determine the collective preference).

Translating Customer Requirements into Technical Specifications

509

multidisciplinary or interdisciplinary (depending on how one defines these terms) if they are supposed to include knowledge of how to transform customer requirements into physical specifications. For this translation, as we saw, knowledge of the physical nature of artifacts, knowledge of their functional nature, and knowledge of the relationship between the physical and the functional nature are needed. Those different types of knowledge come from different disciplines. Knowledge of the physical nature may be drawn from various engineering sciences and natural sciences such as physics, chemistry, biology, and from mathematics (the latter because it also concerns size and shapes). Knowledge of the functional nature (including knowledge of customer requirements) may draw from psychology, sociology, anthropology, and other social sciences. Knowledge of the relationship between the physical and the functional nature is primarily something we expect to be developed in design practice. Development of that knowledge can be expected to draw on the other two types of knowledge. This is why design practice can be expected to be an activity in which knowledge of different disciplines is needed. The way knowledge of different disciplines is combined can vary between different practices. Margaret Boden has come up with a taxonomy of levels of interdisciplinarity that is often referred to in the literature [Boden, 1997]. She defines six levels of interdisciplinarity, ranging from encyclopedic (in which case people from different disciplines at best know of each others’ presence but do not co-operate) to integrated (in which case there is an intensive co-operation of people from different disciplines working on common problems and using common concepts and methods). The way these levels are defined is purely in social terms. It represents the extent to which people from different disciplines co-operate. That taxonomy can also be used to investigate design practice. As in Boden’s taxonomy, design practices show the same sort of range up to levels of intensive co-operation between designers/engineers and marketing people. For QFD and methods like QFD that means that in the process there are frequent sessions of people from different departments in which they discuss, for instance, the scores of relationships between customer requirements and physical features in the cells of the QFD-matrix. The content of the QFD-matrix in such cases really can become a shared belief for all participants in those sessions. This sociological approach to interdisciplinarity is certainly an important one. In Vincenti’s case study on the design of aircraft, he described how the crucial step in the process of learning how to relate pilots’ experiences to the physical features of the aircraft (stability, controllability) took place by intensive contact between pilots and engineers. As social epistemology points out, knowledge is embedded in people, and social interactions between people can have a great impact on the way knowledge can be made explicit and available for practical purposes. The integration of knowledge from different disciplines faces several barriers, as practice has shown. The difference in language games between users and designers is mirrored by the difference in language games used by marketers and engineers in an industrial company. This difference in language can be accompanied by

510

Marc J. de Vries

a mutual suspicion about the relevance of the other party’s contribution to the industrial process. In that respect C.P. Snow’s ‘Two Culture’ story still holds today. The difference in language can also be related to what seems to be a certain ‘incommensurability’ of disciplines. Sometimes the same terms are used in different disciplines, but with a different connotation. Formally the concept of correlation has the same (mathematical) definition in both natural sciences and social sciences. But when a social scientist finds a correlation of .40 and concludes from that that there is a significant relatedness, the natural scientists will immediately raise their eyebrows. Similar problems will occur when marketers and engineers try to work together but get frustrated because concepts they both use appear to have different connotations for both parties. To overcome such problems an intermediary could serve a useful role. For educating such intermediaries special programs have been set up at universities and colleges worldwide. Usually these programs offer a curriculum that combines engineering courses and social science courses. Integration of the two types of disciplinary knowledge takes place in project work. Practice, here too, has indicated that it is not easy to realize integration of knowledge from different disciplines. Yet, a new field of scientific study has emerged. In that new ‘inter-discipline’ entirely new concepts and theories are developed that are not found in the original contributing disciplines (such as: socio-technical landscapes, and technological regimes and trajectories). Such new concepts may not be directly operational for transforming customer requirements into physical features, but they do offer a broader understanding of how new technologies and products need to be seen as one totality with social developments. Customers are not only individuals but they are also part of these social developments. Therefore a broader understanding of what happens in the social context of technology is an important input for the transformation process in which the desires of this socially embedded customer are to be transformed into the physical features of a product. 11 AMBIGUITY OF THE TERM ‘TRANSLATION’ Throughout this chapter I have used the term ‘translation’ as if it were unproblematic. But it is not. As we have seen in the critical analysis of the QFD method (in section 7), QFD claims to enable designers to make the translation from customer requirements into physical features, but in fact it only structures this problem. The actual translation remains an unclear step in the process. Somehow the designer must already have ideas about this before the QFD method is applied; otherwise it is not possible to identify relevant physical features and their relations to customer requirements. The method itself does not generate these features. In fact, there is no method or analysis that really accounts for this translation. It requires creativeness and know-how of designers. For that reason the term ‘translation’ is ambiguous. On the one hand, it seems appropriate in that it relates to a real aspect of the design process: the step that entails relating two different sets of terms (customers’ requirements and physical features). But the term ‘translation’ is not

Translating Customer Requirements into Technical Specifications

511

commonly associated with creativeness. The role of the translator of a book is seen as fairly uncreative, as the translator is not expected to add new thought to the text. The ‘translation’ of customer requirements into physical features, however, does demand the input of new ideas into the process. Our understanding of how this happens is still very limited. There is no philosophical account available yet that offers a good understanding of this. We have seen how various philosophical concepts help to understand the nature of the problem of the ‘translation’, but how the ‘translation’ takes place is still a fairly mysterious process. Evidently, there is still room for more philosophical reflections here. ACKNOWLEDGEMENTS I want to thank Peter Kroes, Anthonie Meijers, Krist Vaesen and Sjoerd Zwart for their comments on earlier versions of this text. BIBLIOGRAPHY [Bijker, 1995] W. E. Bijker. Of Bicycles, Bakelite and Bulbs: Toward a Theory of Sociotechnical Change. Cambridge University Press, 1995. [Boden, 1997] M. Boden. What is interdisciplinarity? In: Cunningham, R., ed., Interdisciplinarity and the Organisation of Knowledge in Europe (Euroscientia conference), pp. 13-26. Office for the Official Publications of the European Communities, 1997. [Borgmann, 1984] A. Borgmann. Technology and the Character of Contemporary Life. University of Chicago Press, 1984. [Courage and Baxter, 2005] C. Courage and K. Baxter. Understanding Your Users. Elsevier, 2005. [Cross, 1984] N. G. Cross. Developments in design methodology. Wiley, 1984. [Feenberg, 1999] A. Feenberg. Questioning Technology. Routledge, 1999. [Haag et al., 1996] S. Haag, M. K. Raja, and L. L. Schkade. Quality Function Deployment usage in software development. Communications of the ACM, 39(1), pp. 41-49, 1996. [Kemp et al., 1998] R. Kemp, J. Schot, and R. Hoogma. Regime shifts to sustainability through processes of niche formation: the approach of strategic niche management. Technology Analysis & Strategic Management, 10(2), pp. 175-198, 1998. [King, 1989] B. King. Better designs in half the time. Implementing Quality Function Deployment in America. GOAL/QPC. 1989. [Kroes and Meijers, 2006] P. A. Kroes and A. W. M. Meijers. Introduction. The dual nature of technical artefacts. Studies in the History and Philosophy of Science, 37, pp. 1-4, 2006. [Morton, 2002] D. L. Morton. Reviewing the history of electric power and electrification. Endeavour, 26(2), pp. 60-63, 2002. [van de Poel, 2007] I. van de Poel. Methodological problems in QFD and directions for future development. Research in Engineering Design, 18, pp. 21-36, 2007. [Sarlemijn, 1993] A. Sarlemijn. Designs are cultural alloys. STeMPJE in design methodology. In: Vries, M.J. de, Cross, N. and Grant, D.P., eds., Design Methodology and Relationships with Science. pp. 191-248. Kluwer Academic Publishers, 1993. [Sarlemijn, 1995] A. Sarlemijn. Kwaliteitsfunctie-ontwikkeling; kiezen voor een Japanse of een Amerikaanse aanpak?. In Sarlemijn, A. and Boddendijk, H.G., eds., Produkten op maat. QFD als gids bij produktontwikkeling. Boom, pp. 21-54, 1995. [Snow, 1959] C. P. Snow. The Two Cultures and the Scientific Revolution. MacMillan, 1959. [Verbeek, 2005] P.-P. Verbeek. What Things Do. Philosophical Reflections on Technology, Agency, and Design. The Pennsylvania State University Press, 2005. [Vincenti, 1990] W. G. Vincenti. What Engineers Know and How They Know It. Johns Hopkins University Press, 1990.

512

Marc J. de Vries

[de Vries, 2003] M. J. de Vries. The Nature of Technological Knowledge. Extending Empirically Informed Studies into What Engineers Know. Techn´ e 6(3), pp. 1-21, 2003. [de Vries, 2005a] M. J. de Vries. Teaching About Technology. An Introduction to Philosophy of Technology for Non-philosophers. Springer, 2005. [de Vries, 2005b] M. J. de Vries. 80 Years of Research at the Philips Natuurkundig Laboratorium. The Role of the Nat.Lab. at Philips. Amsterdam University Press, 2005. [Winner, 1986] L. Winner. The Whale and the Reactor: A Search for Limits in an Age of High Technology. University of Chicago Press, 1986. [Wittgenstein, 1953] L. Wittgenstein. Philosophical Investigations, ed. G.E.M. Anscombe and R. Rhees, transl. G.E.M. Anscombe. Blackwell Publishers, 1953.

FOUNDATIONAL ISSUES OF ENGINEERING DESIGN Peter Kroes

1

INTRODUCTION

The homo faber notion stresses the tool-making aspect of human agency.1 Such tools are technical artefacts, they are the material outcomes of human productive activity (or work). Due to their increasing complexity, the production of technical artefacts has evolved; there has been a shift from experience-based crafts to professions where use is made of highly specialised scientific and technological knowledge. With professionalisation has come division of labour bringing to the fore the different kinds of activities that often remain implicit and inextricably intertwined in the crafts. One of the main divisions of labour follows the dividing line between the mental and physical activities involved in making technical artefacts, between conceiving a technical artefact and actually making or producing one [Dym, 1994, p. 15; Pahl and Beitz, 1996].2 The conceiving side is termed designing and it is done by specialised professionals, namely designers. The result of any designer’s work is a design of a technical artefact. A design, roughly, is a plan or a description (which may include drawings) of a technical artefact. As such, a design is not a technical artefact in itself but merely a representation thereof. A design may include a plan or a description of how to make the artefact in question and may go on to function as a blueprint for its physical realization, that is, for the actual manipulation of matter so that it results in a particular kind of material object.3 It is in this making phase that the production or manufacturing engineers and the production facilities come into play. The designing aspect of making technical artefacts appears to be of particular importance in cases where making involves more than simply reproducing an existing kind of technical artefact. Copying the design of a technical artefact is not really designing. 1 For

an interesting discussion on the notion of homo faber, see [Arendt, 1958]. instance, Pahl and Beitz [1996, p. 1] remark that “The mental creation of a new product is the task of the design or development engineers, whereas its physical realization is the responsibility of manufacturing engineers.” 3 The notion of making a technical artefact is ambiguous; it can refer to the intentional creation (designing) of an artefact with a particular function or to its actual physical/causal production (for instance involving workers in a production facility who may not know what they are producing); for an interesting discussion on these two interpretations, see [Thomasson, 2007]; here we concentrate on making technical artefacts in the first sense. 2 For

Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

514

Peter Kroes

Making designs or plans for new technical artefacts points to a feature of human agency that extends far beyond the domain of material production. We make all kinds of plans in the sense of considered series of actions, which may or may not involve the use or making of technical artefacts. The homo faber idea presupposes that human beings are planning agents, agents with the ability to form and execute plans [Bratman, 1987]. According to Bratman we are planning agents because we are, at least to some extent, rational agents in that we reflect on the outcomes of possible courses of actions, in other words, on the execution of possible plans. On top of that we need to plan our actions in order to coordinate our own actions in relation to the different goals we pursue simultaneously while coordinating our actions with those of others. The making of technical artefacts presupposes such planning capacity not only with regard to the production of an actual artefact on the basis of a plan or design but also with regard to the planning how to use the technical artefact in question. In Section 7 we will argue in more detail that engineering design may be interpreted as an activity of making ‘use plans’. In order to avoid misunderstanding it is necessary to qualify the above-mentioned sharp distinction between the designing and the producing of technical artefacts in terms of mental and physical activities. Especially when it comes to mass produced products, the designing of a technical artefact may also involve actually making a prototype. The function of such a prototype is to test and evaluate the proposed design before it goes into mass production. In much engineering design practice demonstrating that a proposed design ‘works’ by building a prototype is seen as an integral part of the actual designing phase. In such cases, the actual mass production of the technical artefact remains external to the design task but the designing part itself is no longer a purely mental activity since it involves building and running experimental tests on prototypes. Still, the outcome of the design phase, insofar as it is a plan for a technical artefact, is a mental product. In some design practice prototype creation may be virtually impossible (for instance, when designing a new harbour). Even then drawing a sharp distinction between design and production may be problematic because during implementation it may well be necessary to redesign part of the original design so that the design activity actually extends into the production phase. So, while conceptually we may distinguish between the mental and physical creation of technical artefacts it may, in practice, be difficult to separate them. In the following discussion we will restrict ourselves to the design aspect of creating or making technical artefacts. This is generally considered to be the most interesting aspect because it is assumed to require a degree of intelligence and creativity, whereas the actual production side simply involves executing a plan (the design) whilst tapping the appropriate physical and organisational skills and making use of production facilities. This attitude towards the design and production aspects of technical artefacts may reflect the rather pervasive difference in the way science and technology have been appraised as part of Western thinking since Greek antiquity. We will not question this attitude here but will merely

Foundational Issues of Engineering Design

515

reiterate our observation to the effect that in actually making technical artefacts both aspects may be intertwined in inextricable ways. We will focus on the engineering aspects of designing and not so much on the aesthetic aspects. Clearly, in some engineering design practice, such as in industrial and architectural design, aesthetic considerations often play a prominent part. There the notions of designing and design are often primarily associated with the aesthetic qualities of designed objects and aesthetic criteria may figure prominently in the criteria used to evaluate proposed designs. Such engineering design practice comes very close to art. In many branches of engineering design, though, aesthetic considerations only play a minor role, if at all. There it is the design and development of technical artefacts that can fulfil practical functions that takes centre stage. In those branches, proposed design solutions are primarily evaluated on the basis of criteria such as effectiveness, efficiency, costs and durability rather than on aesthetic criteria. It is this kind of engineering design, in which the solution to technical problems has a major role, that we are primarily interested in here. Two questions relating to engineering design will concern us: (1) What kind of activity is engineering design? and (2) What is a technical design? The first question in which the notion of design is verbal is about the process of designing; while the second is about design in the nominal sense and so concerns the outcome of a design process. Before we can turn to these questions there is yet another preliminary issue that has to be addressed. Just a brief look at the range of engineering design disciplines and at the divergence in the kinds of things designed by engineers is sufficient to make one question whether it is indeed sensible to endeavour to generally characterise engineering design either as a process or as a product. There are dozens of different engineering disciplines (mechanical, electrical, civil, chemical, agricultural, bio(medical), material, mining, computer etc. ) that design myriad products ranging from mass produced computers to purpose-built oil platforms, from telephones to high rise buildings, from components to complex systems, from micro-organisms to software, from multi-purpose materials to highly specialized single-purpose devices etc. Accordingly there is also great variety in engineering design practice, not only as far as the required competences and skills of design engineers goes but also in the composition of the design teams. Some design projects may be carried out by a single designer while others require large, multidisciplinary teams of design engineers. There is also much variety in the kinds of design problems that need to be solved. Vincenti [1990], for instance, distinguishes between normal and radical design problems and between design tasks that are high and low in the design hierarchy. Does such variety reflect any unity? Is it possible to generally characterise engineering design processes and to pinpoint domain-independent principles and procedures for engineering design? This has been a topic of controversy (see, for instance, [Reymen, 2001]). Naturally much depends on the level of abstraction chosen. It is easy to cite very general problem solving strategies (e.g. analyse,

516

Peter Kroes

synthesise and evaluate) to characterise engineering design but in that way much if not all of what makes engineering design a specific kind of problem solving is lost. Conversely, if we zoom in too much it becomes difficult to recognise the common elements in different design practices. One factor driving the search in engineering practice for systematic, domain-independent design principles is the ever-growing complexity of the objects of design. In recent decades this has led to the emergence of new fields such as system design and design methodology which study the principles and procedures of engineering design in order to rationalise and improve design practice [Sage, 1992; Cross, 1994 (1989); Hubka and Eder, 1996; Pahl and Beitz, 1996; Dym and Little, 2000]. Within these fields, the analyses of and proposals for engineering design methods are often domain-independent. In the following section we will take as our starting point some general, domainindependent features of engineering design as proposed by engineers themselves. 2

ENGINEERING DESIGN AND SCIENCE

Modern engineering design is a science-based activity but that does not make it a branch of applied science. Indeed, the solving of design problems is taken to be something very different from the solving of scientific problems. Designing is even considered by some to be the salient feature of technology that distinguishes it from science [Mitcham, 1994, p. 220]. We will discuss two features of engineering design that make it an intrinsically different activity from scientific research. The first concerns the decisional nature of engineering design, the second the wide variety of constraints laid down for designs. As our starting point we take the following general characterisation of engineering design by the Accreditation Board for Engineering and Technology (ABET); it states that engineering design:4 is the process of devising a system, component, or process to meet desired needs. It is a decision-making process (often iterative), in which the basic science and mathematics and engineering sciences are applied to convert resources optimally to meet a stated objective. Among the fundamental elements of the design process are the establishment of objectives and criteria, synthesis, analysis, construction, testing and evaluation. Although not explicitly stated, the systems, components and processes devised are assumed to be of a material nature; the design of (part of) an organisation or institution is not considered to be the domain of engineering design proper.5 4 http://www.me.unlv.edu/Undergraduate/coursenotes/meg497/ABETdefinition.htm; accessed November 14, 2006. Note the prevalence of the technology-is-applied-science idea in this conception of engineering design: the application of scientific knowledge to engineering design is explicitly mentioned. 5 According to Simon [1969/1996, p. 111], however, the intellectual activity of designing material artefacts is not fundamentally different from the designing of organisational structures or procedures.

Foundational Issues of Engineering Design

517

The stated objective is laid down in what is usually termed a list of specifications. The list is derived from the function or functional requirements that the thing to be designed (i.e. the system, component or process) is expected to satisfy. These functional requirements are related, in turn, to certain human ends (or needs). If the designed artefact meets all the specifications, it is deemed suitable to realize the desired function. Whether that indeed turns out to be the case depends very much on whether the list of specifications adequately meets the functional requirements. If it does and if the reasoning from end to function has also been performed adequately, then the designed artefact may be expected to be a reliable means to the specified end. A striking feature of the ABET definition given above is that it characterises engineering design as a decision-making process and not simply, as is so commonly done, as a problem solving process. Characterising engineering design as mere problem-solving may indeed be misleading because of the dominant view that problem-solving involves finding or discovering the ‘right’ solution, the solution which, in principle, is uniquely determined by the problem structure. This discovery-picture has been traditionally associated with the kind of problemsolving that takes place in science or mathematics but it does not correspond to the kind of problem-solving peculiar to engineering design. Design problems are often ill-structured [Simon, 1984] or wicked problems [Rittel and Webber, 1984], which for instance means that there may be no definite formulation of the design problem itself, insufficient criteria to evaluate proposed solutions and no clear idea of the solution space. As indicated by the ABET definition, engineering design is partly all about clearly establishing objectives and criteria by which to judge the proposed alternatives. Decisions thus have to be made that are to a large extent underdetermined by the problem formulation. Such decisions may have a significant effect on the aim and the outcome of the design project. But even if there is a clear and unambiguous formulation of the objective in the form of functional requirements together with a list of specifications, decisions have to be made concerning the promising options to work on. It may turn out that some of the specifications conflict in which case decisions about trade-offs have to be made. Alternatively, given the state of the art technology or the available resources it may not be possible to come up with a solution that satisfies all specifications. One then has to decide how the list is to be adjusted. When there is a set of alternative solutions satisfying all the specifications, there is no guarantee that one particular solution can be embraced as the best or optimal solution for rather fundamental reasons linked to multiple criteria evaluations [Franssen, 2005]. In such cases the lack of rational procedures leading to the determining of the best option means that decisions again have to be taken about which option to choose. These various kinds of decisions are all part and parcel of engineering design practice. The actual decisions taken may have far-reaching consequences for the outcome as they will shape the artefact that is being designed. This ‘decisional’ nature of engineering design reflects the idea that engineering design is much more a process of invention than a process of discovery. It is about the creation of new

518

Peter Kroes

objects, not the discovery of what already exists. So the decisional nature of engineering design is not so much to be interpreted as the making of choices between pre-given, existing options but as the creating of various design options leading to the final technical artefact by making decisions that fix their properties. Note that not all the problems that have to be solved in engineering design necessarily involve the kind of decisions discussed above; for instance, part of a design process may involve ‘finding out’ the maximum load a proposed construction can bear. Such problems are the domain of technological (engineering) research; they are not the kind of ‘creative’ decisions that have to be made when solving design problems.6 The second important feature of engineering design to be discussed here is the variety which exists in the kinds of constraints that design engineers have to deal with when designing. According to the ABET engineering students have to learn to solve engineering design problems under a (ibidem) “variety of realistic constraints, such as economic factors, safety, reliability, aesthetics, ethics and social impact.” This variety of constraints is reflected in the list of specifications, which means that a variety of factors (economic, safety, reliability, ethical etc.) determines the design problem and the ultimate shape of the object of design. With scientific problems such constraints are virtually absent; because of this scientific and engineering design problems are essentially different.7 Both domains are governed by different kinds of values, norms and success criteria. According to the ‘ivory tower’ model, science as a cognitive activity is ideally guided by epistemic values only, such as truth, empirical adequacy, simplicity and explanatory power. In practice this view of science may be inadequate but it nevertheless highlights an important feature of scientific research, namely that from a cognitive point of view the results ought to be basically independent of the wider social context. More or less the same may be said of engineering research or ‘applied science’: once the technologically interesting topics have been chosen, the research will proceed according to the same values as those abided by in science. Research, whether scientific or technological, is mainly guided by the values and norms of theoretical rationality that deal with the issue of what to believe. In contrast to science, the wider social context is of paramount importance to engineering design, since it is embedded within a broader framework of product creation processes. As part of these processes, problem solving in engineering design is subject to other kinds of values, norms and success criteria. Proposed so6 This is not to say that science is not a creative enterprise. In science the creative aspect is traditionally considered to reside primarily in the activity of representing some pre-existing world, not in creating that world. This traditional view of science has come under attack from social constructivist quarters (see, for instance, [Barnes et al., 1996]). Hacking [1983] has also challenged this view by claiming that in experiments physical phenomena are created. For a criticism of this view, see [Kroes, 2003]. 7 Of course, just as in engineering design, scientific research as a human enterprise is subject to all kinds of constraints, for instance, constraints deriving from the limited resources, the risks associated with performing experiments, the possible social consequences etc. In this respect, there is not much difference between engineering design and scientific research.

Foundational Issues of Engineering Design

519

lutions are evaluated in terms of pragmatic criteria such as effectiveness, efficiency, feasibility, costs and safety. Indeed, engineering design is guided by the demands of practical rationality, with issues relating to the course of action to take in order to achieve given ends. These kinds of actions and ends are always embedded in broader, value-laden social contexts that impose their own constraints on viable solutions. These constraints may change in the course of time; for instance, in recent decades sustainability has emerged as a new important constraint on engineering design. The following domain-independent definition of engineering design by Dym [1994, p. 17] suggests that engineering design is subject to two different kinds of success criteria: Engineering design is the systematic, intelligent generation and evaluation of specifications for artifacts whose form and function achieve stated objectives and satisfy specified constraints. One set of criteria for evaluating proposed design solutions is the list of specifications that has to be ‘achieved’ and is derived from the stated objective and related function. The other set consists of constraints that have to be ‘satisfied’. Dym remarks that one may question whether a clear distinction between the set of constraints and the list of specifications can be made. For instance, the condition that a car engine has to satisfy a legal standard in conjunction with pollution may be taken as an element of the list of specifications but it may also be seen as a constraint. Whether the distinction is meaningful or not, it is the variety in kinds of constraints/specifications imposed on design solutions that is important. Because of this variety conflicts between constraints/specifications may easily arise (for instance, between safety (more mass) and sustainability (less mass) requirements for cars). This is the reason why finding ‘clever’ trade-offs between conflicting specifications/constraints plays such a prominent role in engineering design practice. These two features of problem solving in engineering design, its decisional nature and the wide variety of constraints, are what make it an activity that is very much distinct from the kind of problem solving that goes on in science. Although modern technology is science based, problem solving in engineering design is not a type of ‘applied science’. Any characterisation of engineering design as an activity deriving from science neglects or downplays the importance of the features sketched above. Below, in Section 5, we will discuss yet another feature that distinguishes design from scientific research. It concerns the kinds of reasoning employed by scientists and design engineers in their problem solving. Virtually all analyses of scientific reasoning amount to variations of inductive, (hypothetical-)deductive or abductive forms of reasoning. In designing, a different kind of reasoning takes centre stage, namely means-end reasoning. For a further explanation of the nature of engineering design we first turn, however, to the work of Simon who in his classic The sciences of the artificial [1996 (1969)] presented an analysis of technical artefacts and of designing that nicely indicates what engineering design is all about.

520

Peter Kroes

3 THE NATURE OF ENGINEERING DESIGN According to Simon engineering design deals with the synthesis of artificial things and engineers, in particular designers, are [1969/1996, p. 4-5] “concerned with how things ought to be — how they ought to be in order to attain goals, and to function.” Instead of taking the world for what it is (as in science) engineering design seeks to change the world to meet given needs, desires or goals. More specifically, engineering design contributes to the development of the material means that people may use to achieve their goals. These material means, technical artefacts, have a function; when functioning and used properly they are supposed to bring about effects that are conducive to achieving the ends associated with their function. The normative character of what engineers have to deal with is reflected in normative statements about technical artefacts.8 A malfunctioning television set or a bad screwdriver are not the things they ought to be in the sense that they do not show the kind of behaviour (when used properly) they ought to show. Simon analyses the function or purpose of a technical artefact in the following way [1969/1996, p. 5]: Let us look a little more closely at the functional or purposeful aspect of artificial things. Fulfilment of purpose or adaptation to a goal involves a relation among three terms: the purpose or goal, the character of the artifact, and the environment in which the artifact performs. For instance, the purpose of a clock is to tell time and the clock’s character has to do with its physical makeup (gears, springs etc. in the case of a mechanical clock). Finally, the environment is important because not every kind of clock is useful in every environment; sundials can only perform their function in sunny climates. Simon’s analysis of artefacts is represented in a schematic way in Figure 1.9 According to Simon the environment of an artefact is very important because it is what moulds the artefact. He considers the artefact to be a kind of [1969/1996, p. 6] ‘interface’ between “an ‘inner’ environment, the substance and organization of the artefact itself, and an ‘outer’ environment, the surroundings in which it operates.” The inner environment of the artefact, its character, is shaped in such a way that it realises the goals set in the outer environment. Therefore, the science of the artificial must focus on this interface, since the [1969/1996, p. 113] “artificial world is centred precisely on this interface between the inner and outer environments; it is concerned with attaining goals by adapting the former to the latter.”10 The task of engineering design is to come up with descriptions of 8 For an analysis of the normativity of technical artefacts, see Franssen’s contribution to this Volume. 9 The arrows stand for conceptual implication: the notion of an artefact conceptually implies the notion of a character, a goal or purpose and an environment. 10 This remark suggests that there is just a one-way influence from the outer to the inner environment. The design of technical artefacts, however, may also be a matter of adapting the outer to the inner environment (for instance, by adapting the behaviour of prospective users through training).

Foundational Issues of Engineering Design

521

Goal/Purpose

Artefact

Character

Environment

Figure 1. Simon’s interpretation of a technical artefact technical artefacts for which the inner environment is appropriate or adapted to the outer environment.11 What makes Simon’s analysis of technical artefacts and engineering design so interesting is that it draws attention to the tensions between the inner and outer environments of technical artefacts, the tension between what artefacts do or are capable of doing, and what they are expected to do within some context of human action (i.e. the ‘rich’ outer environment that imposes so many constraints). It is this tension which, according to Staudenmaier [1985, p. 103], is the defining nature of technology. Indeed, engineering design is about filling in the “substance and organisation” of the inner environment so that it meets all the requirements or constraints imposed from the outer environment. In so doing, engineers have to take into account what is physically and technologically possible. It is the tension between the set of physical and technological constraints that apply to the contents of the inner environment and the set of constraints that derive from the outer environment (contextual constraints: functional specifications and other requirements) that defines the core of engineering design (see Figure 2) (for more details, see [Kroes, 1996]). This tension is one of the main driving forces behind the development of technical artefacts (see, for instance, Petroski’s [1992] principle of ‘form follows failure’). Obviously apart from these ‘market pull’ factors, advances in technology may also drive the development of technical artefacts (the ‘technology push’). This interface character of technical artefacts explains the difficulties engineers have when disambiguating and fixing the meaning of the notion of function, especially in relation to notions of behaviour and purpose. Rosenman and Gero 11 It is the distinction between inner and outer environment that also lies at the basis of Hubka and Eder’s [1996, p. 108-114] theory of the properties of technical systems (technical artefacts).

522

Peter Kroes

Physical and Technical Constraints

Inner Environment

Design of Technical Artefacts

Contextual Constraints

Outer Environment

Figure 2. Design and the tension between inner and outer environment

[1994, p. 199], for instance, remark that engineering design involves concepts from “both the human sociocultural environment and the physical environment”; this means, in Simon’s terminology, concepts from the outer and inner environment. Within engineering practice, functions are usually represented in terms of input-output relations. However, there is no consensus on whether functions correspond to properties of the physical system making up the inner environment of the technical artefact or to properties of the outer environment in which the artefact is embedded. The former interpretation links it to physical properties (capacities) of the technical artefact, the latter to the ends pursued in the outer environment. Pahl and Beitz [1996, p. 31] apply “the term function to the general input/output relationship of a system whose purpose is to perform a task” whereas Hubka and Eder [1996] interpret the notion of function in terms of the internal processes taking place within a technical system. According to Roozenburg and Eekels [1995, p. 96] “the function of a system is the intended transformation of inputs into outputs.” Sometimes a distinction is made between two different kinds of function, one referring to actual behaviour, the other to intended behaviour (see Chandrasekaran [2005] for various definitions of functions within an engineering context; Chandrasekaran and Josephson [2000] distinguish between environmentcentric and device-centric views on function, which roughly correspond to functions as seen from the outer and inner environments). This ambiguity surrounding the notion of function appears to be closely connected to the fact that technical artefacts act as an interface between a social/intentional outer environment and a physical inner environment. Simon’s distinction between inner and outer environment points to two different ways of contemplating technical artefacts. Looked at from the outer environment, which is typically the user’s perspective, the technical artefact presents itself primarily, whatever its inner environment, as a means for achieving a goal or end. From this means perspective the artefact is characterised primarily in a functional way; the inner environment remains a black box. If we look at it from the inner environment and forget that the object is the result of a process of adaptation to the outer environment then the artefact presents itself merely as a physical system; from this perspective, the goal that it fulfils in the environment remains a black

Foundational Issues of Engineering Design

523

box. This is typically the way a physicist would analyse a technical artefact. As Simon [1969/1996, p. 7] remarks: “Given an airplane, or given a bird, we can analyse them by the methods of natural science without any particular attention to purpose or adaptation, without reference to the interface between what I have called the inner and outer environments.” Different kinds of descriptions of technical artefacts are associated with the inner and outer environments or with the physical and functional descriptions. If we take engineering design to be a process of devising an artefact that is adapted to some specific environment, then it starts from the outer and proceeds to the inner environment. So, engineering design may be interpreted as a process in which a transition is made from a function to a physical structure.

4

ENGINEERING DESIGN: FROM FUNCTION TO STRUCTURE

From the point of view of the object of design, an engineering design process does indeed start with a description of the anticipated behaviour of that object, (i.e. its function), and end with the description of a physical structure that realises that expected behaviour.12 In engineering design, therefore, a functional description of an object has to be translated into a structural description. The latter description specifies, for every relevant detail, the geometrical and all the other physical and chemical properties of the technical artefact. What is relevant in the description is primarily determined by the function of the technical artefact. Anything that has a direct bearing on the performance of the function is relevant. Let us first have a closer look at these two different ways of describing. In a functional description the object of design is represented as a ‘black box’, which is the means that transforms a certain input into a desired output. Quite how this input is transformed into the output is left in the dark. It is precisely the aim of the design process to come up with a viable proposal for how this may actually be achieved. Depending on the case in hand, the input and output may be described either in a more qualitative or or a quantitative way. For instance, in the ASM Handbook Materials Selection and Design [1997, p. 22 ff] the function of a fingernail clipper is described as serving to “remove excess length on fingernails” and the inputs and outputs of this function are specified in terms of energy, material and signal flows (see Table 1; also [Otto and Wood, 2001, Ch. 5]). In certain cases, the relation between input and output may be partly represented by a mathematical function, for instance the relation between the input and output signal of an electronic circuit. Note, however, that in these cases the mathematical function is not to be confused with the technical function of the electronic circuit. The mathematical function can only partly represent the technical function because it does not include the normative aspect of the technical function (that is, that the device ought to, or is intended to realise the specific mathematical relation between input and output signal). 12 See

also [Dym and Little, 2000, p. 113].

524

Peter Kroes

Table 1. Input and output of fingernail clippers

Energy Material Signal

Input

Output

Finger force Kinetic (hand motion) Fingernail Hands, Debris Long nail, hang nail, rough nail

Sound Sound Kinetic energy in nail Cut nail, Hands Debris Good appearing nail

A purely functional description is opaque with regard to the structure of the object that realizes the function. This is a direct consequence of the fact that a functional description is the result of viewing the object from a means-end perspective. What is of primary importance from this perspective is that an object, irrespective of its specific constitution, can be used as an effective and efficient means to a certain end. In other words, this is — as already mentioned — typically the perspective of the user of a technical artefact. When we turn our attention to structural descriptions of technical artefacts, the situation is the reverse: just as a functional description is opaque with regard to the structure that realizes the function described, so a structural description of an object is opaque as regards the function performed by that object. Just to illustrate this, imagine that the fingernail clipper of Figure 3 drops out of an aeroplane and lands at the feet of the chief engineer of an as yet undiscovered tribe in the Amazon forest. She carefully studies the structural properties of this object that is totally unknown to her and its behaviour under various circumstances before coming up with all kinds of input-output schemes, that is, with all kinds of possible functions. If she is lucky, the input-output scheme of Table 1 will be among these functions. But even then, how is she to determine which one of the many inputoutput schemes is the intended one corresponding to the function of this object? In principle this appears impossible on the basis of a structural description of the object; even the most detailed description of all structural properties of the object and of its physical/chemical behaviour will not enable the engineer to deduce that it is a fingernail clipper.13 Thus, from a functional point of view, a structural description is also a black box description. The situation is in fact symmetric from the point of view that each mode of description black boxes the other. Purely structural descriptions of technical artefacts are typically of interest in the context of the production of technical artefacts. Such descriptions provide all 13 Such situations may occur in archaeology but also in ‘reverse engineering’, in which technical products are taken apart in order to analyse their working and the functions of their parts. Sometimes it may not be possible to reconstruct the function of a part on the basis of its physical properties and its place in the whole artefact.

Foundational Issues of Engineering Design

525

Figure 3. A ‘physical object’ with input-output relations

the information that is necessary for the making of a technical artefact; in principle it is not necessary to know the function of the object being made. Examples of purely structural descriptions of technical artefacts underscoring their importance for engineering practice may be found in ISO norms for standardised technical objects. In order to ensure that such standardised technical objects can in practice be replaced by any other item of the same kind, these norms minutely register every relevant structural detail. Structural and functional descriptions of technical artefacts correspond to descriptions from the physical and design stance as defined by Dennett [1987, p. 15 ff]. Nevertheless, the role of such stances in engineering practice differs from that attributed to them by Dennett. His distinction between the physical and design stance is of a methodological nature; it concerns the best strategy for predicting the behaviour of systems. Depending on the complexity of the system under consideration, the behaviour of a system may best be predicted from a physical or a design stance (here we do not consider the intentional stance). Descriptions from the physical and design point of view always provide different ways of dealing with the same problem of predicting behaviour. Within engineering design practice, the structural (physical stance) and functional (design stance) descriptions have different methodological roles to play. In this respect it is not the prediction of the behaviour of an already existing system that is at stake, but rather the design of a system that is expected to exhibit certain behaviour. The description of an object qua technical artefact, that is, as a physical object with a function, requires a structural and a functional description, irrespective of the complexity of the relevant system. The difference regarding the methodological role of these two types of description clearly comes to the fore with simple technical artefacts. It is often possible to apply either the physical or the design stance to predict their behaviour. In such cases, the structural and functional descriptions become alternative descriptions for predicting behaviour. From an engineering point of

526

Peter Kroes

view, however, they are not alternative descriptions but they do complement each other and each is indispensable when it comes to describing the object involved as a technical artefact. The above view of engineering design as a translation of a functional description into a structural description of the object of design, raises interesting questions on how engineers do such translating and the kind of relations that exist between the functional and the structural descriptions of an object. As regards the last point, it should be noted here that function does not uniquely determine structure, since different objects may perform the same function and that the reverse is also true because one and the same object may perform different functions. If it is assumed that the functional description of an object implies normative claims (about what it ought to do), then from a logical point of view the deduction of a functional description from a structural description is problematic. Such a deduction would be vulnerable to some form of naturalistic fallacy objection from adherents to the is-ought dichotomy. Furthermore, just as it is generally considered problematic to derive ‘ought’ from ‘is’, so it is considered problematic to do the reverse, that is, to deduce the actual physical properties of an object (‘what it is’) solely from a knowledge of its function (‘what it ought to do’). There therefore appears to be a logical gap between the functional and structural descriptions of an object. Nevertheless, the function and structure of technical artefacts are taken to be intimately related, not only in the sense that the physical structure realizes (or is supposed to realize) the function, but also in the sense that apparently designers are able to reason successfully from functional to structural descriptions or vice versa (for instance when they justify a proposed design by explaining why it realises the required function).14

5 MEANS-END REASONING What kind of reasoning and knowledge is involved in translating a function into a structure? In practice designers make use of methods like constructing morphological charts or function-means trees to go from function to structure; these charts or trees give a graphical representation of functions (and sub-functions) and the various known ways of realising them [Cross, 1989/1994, p. 106 ff; Dym and Little, 2000, p. 116, 146 ff]. They present in a condensed way the available alternatives for filling in the black boxes corresponding to the (sub)functions. As mnemonic devices they do not give any clues about what kind of reasoning may lead from a function to a structure. In solving this translation problem, ‘means-end’ reasoning appears to be of paramount importance, since the design process is all about finding or constructing the appropriate means for achieving certain ends. In spite of its importance to engineering practice and daily life in general, the formal (logical) analysis of means-end relations and reasoning has received relatively little atten14 For a discussion on the ‘coherence’ of structure and function of technical artefacts, see [Kroes, 2006].

Foundational Issues of Engineering Design

527

tion [von Wright, 1963; 1972; Segerberg, 1992]. Recently, research into artificial intelligence has triggered more interest in this kind of reasoning [Pollock, 2002]. Within engineering design practice there is also a great interest in the formal analysis of functional reasoning, a phenomenon which appears to be closely related to means-end reasoning, because of attempts to develop formal tools to represent the objects of design and supporting functional reasoning about these objects [Dym, 1994; Chittaro and Kumar, 1998].15 Means-end reasoning may be seen as a form of practical inference about what needs to be done to achieve an end. In that respect actions are taken to be means to certain ends (states of affairs in the world) [Hughes et al., 2007]. From a technological point of view, however, objects may also be viewed as means to action ends (for instance, a knife is a means to cutting bread, a pencil a means to writing). A formal analysis of means-end relations and reasoning in which objects and not actions are means has still to be made. In his seminal paper on practical inferences von Wright [1963, p. 161] analyses the following kind of reasoning: One wants to attain x. Unless y is done, x will not be attained. Therefore y must be done. Von Wright calls x the end and y, which is an action, a means to that end. This type of argument concerns necessary means to ends and the conclusion, which states an action, expresses a practical necessity. This practical necessity is derived from the statement of an end and from a conditional statement based on the causal structure of the world. The question one has to ask is whether such arguments are logically conclusive. Prima facie this seems not to be the case since the premises consist of descriptive statements and the conclusion remains prescriptive. But for von Wright this is not a convincing argument against logical conclusiveness. Two features of these types of arguments are of particular interest from the point of view of engineering design. The first has to do with the fact that with these kinds of practical inferences a transition is made from descriptive to prescriptive statements. This strongly supports our claim that this kind of reasoning may play an important role in engineering design. We have already noted that reasoning from function to structure and vice versa is problematic because of the alleged logical gap between functional and structural descriptions of technical artefacts, a logical gap stemming from the is-ought dichotomy. But here we come across a form of reasoning that appears to defy this dichotomy and may therefore open up possibilities for reasoning from function to structure and vice versa. There are still, however, many problems to be solved. The practical kind of inference studied by von Wright concerns practical necessity in relation to human actions. But function statements about technical artefacts are connected to statement about what objects, when considered as means, ought to do and not about what human 15 For

a discussion on means-ends reasoning, see Hughes’s contribution to this Volume.

528

Peter Kroes

agents ought to do. It is not clear how such statements about objects are related to ‘ought to do’ statements about human beings and whether it is possible to construct practical inferences for objects analogous to the one above about human actions. Furthermore, engineering design is primarily about reasoning from function to structure, that is, reasoning from the normative to the descriptive, whereas in the example given above concerning practical inference, the reasoning proceeds in the opposite direction. Again, whether practical inferences from the normative to the descriptive may be constructed remains open to discussion. Finally, given the multiple realisability of technical functions, practical necessity seems too much to ask of the conclusions of means-end reasoning that goes on in engineering design. The second feature concerns the second premiss, the content of which is a conditional relation between the means and the end, and which, as von Wright remarks, is based on a causal relationship. Not surprisingly this closely links means-end reasoning to the causal structure of the world. If we know that event A causes event B,16 then we may realise the occurrence of event B by bringing about event A, if this is technologically possible and if there are no interfering circumstances. So the action of bringing about event A may be considered to be a means for the occurrence of event B, the end. The causal relationship in itself does not imply practical necessity, that is to say, event A is not a necessary means for event B. For that to happen a much stronger conditional statement is required, namely that the bringing about of the occurrence of A is the only practically feasible course of action for bringing about event B. The intimate relation which exists between means-end reasoning and causal relations explains why scientific knowledge plays such a dominant role in modern design practice. This leads to the question of the kind of knowledge used to solve design problems. As has been argued, it would be misleading to assert that engineering design is simply the application of scientific knowledge (or knowledge produced by the engineering sciences). A knowledge of natural phenomena is certainly not all that is needed to solve design problems. Hubka and Eder [1996] have attempted to develop a design science, which they take to be a system of logically related knowledge about designing and for designing. In their enumeration of the various kinds of knowledge needed for engineering design, knowledge from the engineering sciences is just one item in a long list [1996, p. 72]. In a similar vein, Dym and Little [2000, p. 22-3] remark that the majority of the many questions that have to be posed when designing a relatively simple object such as a safe ladder cannot be answered by applying the mathematical models of physics. According to Vincenti [1990, Ch. 7] the anatomy of engineering design knowledge includes at least six different categories of knowledge, some of which do not derive from scientific knowledge at all, such as the ‘know how’ acquired on the shop-floor. All these various kinds of knowledge are important for turning a functional description of the object to be designed into a structural description. Ryle’s [1984] distinction between ‘knowing that’ and ‘knowing how’ may be of particular relevance when 16 More

precisely, tokens of event type A cause tokens of event type B etc.

Foundational Issues of Engineering Design

529

analysing the kinds of knowledge used to solve design problems, because of the intimate relationship between designing and knowing how to make or do things. To conclude, when compared to science the kind of problem solving prevalent in engineering design not only appears to employ distinctive forms of reasoning but also distinctive forms of knowledge. Up until now, the nature of design knowledge, or more generally technological knowledge, has not received much attention in epistemology.17 This is even more true of the formal analysis of means-end reasoning in logic. 6

PHASE MODELS IN ENGINEERING DESIGN

We now briefly move to a topic that has received much more attention, especially from design methodologists, namely the matter of the modelling of design processes in terms of rationally prescribed steps or phases and the development of design tools.18 These phase-models and design tools are supposed to contribute to the improvement of actual design processes. Most of the models are more or less detailed variations on the basic analysis-synthesis-evaluation cycle. As long as designing remains an activity performed by one single individual, these phases will be mainly relevant from a conceptual point of view. As soon as designing becomes a matter of teamwork, which tends to be the situation in modern industry where complex and large systems are dealt with, the phasing of the design process becomes an important management tool for organising, controlling and steering the process of product development. One matter that hampers discussions on the usefulness of implementing such phase diagrams in engineering practice is the criteria for evaluating and measuring the success of the outcome of an engineering design process. From a strict engineering point of view, the simplest success criterion is to meet the list of specifications while satisfying the given constraints. This assumes that the list of specifications is immutably fixed at the beginning of the design process, which is not often the case. Because of problems encountered on the way, they may have to be adjusted during the design process. Moreover, as was remarked before, decisions about the performance criteria to use and the development of methods for measuring such performance criteria are often an integral part of the design process. On top of this, various participants in the design process may evaluate the outcome in different ways. In spite of these difficulties, design methodologists claim that the implementation of systematic approaches to design improves the design process (see, for instance, [Pahl and Beitz, 1996, p. 499-501]).

17 For a discussion on the nature of technological knowledge, see Houkes’s contribution to this Volume. 18 A more detailed account and discussion of these phase models may be found in the chapter on Rationality in design of this volume. See, for instance, also [Cross, 1989/1994, Ch. 2].

530

Peter Kroes

7

DESIGNING PLANS RATHER THAN MATERIAL OBJECTS

So far we have analysed the nature of engineering design mainly from the point of view of the object of design. Our overall perspective has been the making of a (new) technical artefact and we have particularly analysed how a technical artefact, as an object of design, is described at the beginning and at the end of the design process. This object-oriented view on engineering design is rather dominant among engineers. It is true that the usual characterisation of the outcome of a design process stresses that it is a production plan and not a real material object but that is simply a consequence of the prevailing division of labour. The design phase is followed by the production phase which results in the real, material technical artefact to which the designing was geared. If one changes perspective, though, from the technical artefact making to the using side one sees that this cannot be the whole story behind engineering design. From the perspective of the user it is not the making of a technical artefact that matters but rather how to implement it in order to realise his or her goals. To that end it is not a fabrication plan that the user needs but an instruction manual which describes how the artefact can be implemented. An instruction manual or use plan is needed to make the function ‘accessible’ to the user. A technical artefact without a manual or a use plan is in principle of no practical use.19 Thus, from the point of view of the user it is not the production plan for the material technical artefact that matters, but rather its manual or use plan. It may well be the case that when designers characterise the outcome of design processes as production plans for technical artefacts, they implicitly take the manual to be part of the technical artefact. It is, however, important to make its role explicit because it enables attention to shift from material objects to actions and plans in which objects have a role. From an action-oriented view, engineering design is about making use plans concerning how goals may be realised with the help of technical artefacts. Technical artefacts may be said to be embedded within such use plans. Following this line of reasoning, Houkes and Vermaas et al. [2002; 2004] have developed an action-theoretical account of the designing and using of technical artefacts. In it they reconstruct the design and use of technical artefacts in terms of plans, intentions and practical reasoning. They take plans to be goal-directed series of considered actions and they see a use plan for an object as a series of actions involving the manipulation of the object in order to achieve the goal of the plan. They divide the design process into two different activities, namely use plan design and artefact design. Each of these activities is reconstructed in terms of plans and the plan for artefact design is embedded in the plan for use plan design. In their account, the interaction between designers and users does not simply involve the transfer of a technical artefact but also, and primarily, the communication of a use plan [Vermaas and Houkes, 2006, p. 7]. 19 The most common form of a use plan is a written manual; simple technical artefacts often come without a manual as the use plan is presumed to be known to the user and so remains implicit.

Foundational Issues of Engineering Design

531

An attractive feature of this action-theoretical interpretation of engineering design is the central role it attributes to practical rationality/reasoning. If plans are the outcome of engineering design then these plans, irrespective of whether they involve the manipulation of objects, have to satisfy the demands of practical rationality. This applies to the use plan but also to the plan for artefact design which is embedded in the use plan. Bratman [1987, p. 31] discusses two of the demands placed on plans. The first concerns consistency constraints; plans should be internally consistent (they should not include incompatible goals) and consistent with the beliefs of the agent who executes them. Plans should furthermore be means-end coherent which requires that they be broken down into preliminary steps, sub plans and means so that in the eyes of the agent they may be successfully executed. According to Houkes et al. [2002, p. 320] these demands of practical rationality may lead to norms for good and bad design and use. They question, however, whether such demands on plans exhaust the norms operative in engineering design and artefact use. They note that their approach to engineering design and artefact use has an intellectual bias: in line with what was posited in the introduction, the actual executing of plans is not considered to be an interesting topic in its own right. This leads to an interpretation of the demands placed on practical rationality that relates primarily to rational deliberation, a situation which also appears to be the case regarding the demands that Bratman imposes on plans. Actually making things or executing plans may impose additional demands. For instance, it is not clear whether or to what extent the notion of means-end coherence can independently account for the important role of the norm of efficiency in engineering design. This action-theoretical approach to engineering design analyses the nature of designing and its output from the point of view of what Simon calls the outer environment. It takes as its precept practices of intentional human action in which technical artefacts are used to realise ends. Without recourse to this context of human action it is impossible to adequately characterise engineering design and technical artefacts. Unfortunately this approach engenders the same problem of how to translate a function into a structure as that encountered in the objectoriented approach to engineering design. From an action-theoretical point of view, technical artefacts provide ways of achieving certain goals; but how can we move from an ‘outer environment’ description of artefact x in terms of what it is for (x is for y-ing) to an ‘inner environment’ description that specifies the physical make-up of x? How do engineers manage to jump back, so to speak, over the ‘for-operator’ to the object itself? Whether we examine engineering design from an object-oriented angle or from an action-oriented angle the problem remains. 8 A TECHNICAL DESIGN So far the notion of design, when used nominally, has referred mainly to the outcome of a given design process. From the point of view of the product creation process, this outcome is usually taken to be a production plan for objects that

532

Peter Kroes

are still virtual. This is not the noun-type notion of design we are interested in here. When referring to a car design, for instance, what is meant is not usually its production plan but something that has more to do with the properties of the car itself, irrespective of whether that car actually exists or how (if it indeed exists) it was actually produced. It is not easy to grasp what this ‘something’ is. Whatever it is, the design of the car remains an important facet since it more or less determines the accompanying structural and functional properties. It even becomes a defining feature of the car in the sense that its design makes the technical artefact an artefact of a particular kind, namely the kind ‘car’. In order to come to terms with what, in this sense, a design actually is consider the design of the Newcomen steam engine, which is represented graphically in Figure 4. The main function of this kind of steam engine was to power pumps to drain mines, and this was achieved by producing a reciprocating motion in the great beam, which was activated by the motion of the piston etc. The drawing not only provides information about structural features of the design but it also presents part of the form and the layout or organisation of the various parts of the Newcomen engine. Even if we were to add all the relevant structural information to this drawing we would still not end up with a complete representation of the design of the engine. As a designed object, the Newcomen engine has a purpose but that purpose is not contained in the structural representation of the design. For a representation of this aspect of the design of the Newcomen engine it is necessary to add information about its overall function, the functions of its parts, means-end relations and how the machine operates. In order to highlight the purposeful character of a design, a representation of a design thus has to include information about its structural and functional features. Note that, by contrast, a design as a production plan for a still virtual technical artefact does not necessarily include information about the functional properties of the artefact and all its parts. According to Dym and Little [2000, p. 10] a production plan has to be such that “the fabrication specifications must, on their own, make it possible for someone totally unconnected to the designer or the design process to make or fabricate what the designer intended in such a way that it performs just as the designer intended.” It is sufficient for the production plans to contain a purely structural description of the technical artefact. In principle it is not necessary to include a functional description of the artefact since a functional description does not specify in a “clear, unambiguous, complete, and transparent” (ibid.) way the physical properties of the object to be produced. So the notion of a design seen as a production plan is clearly different from the notion of a design that is central to determining the category to which a technical artefact belongs. Anything that is called a design in the sense intended here may vary greatly in engineering practice in terms of level of detail and can be anything from a rough sketch, as displayed in Leonardo Da Vinci’s drawings of machines, to a complete description of every minute detail of a prospective or existing artefact. We will assume that a representation of a design of a technical artefact has to be a combined description of all of its relevant physical and functional properties (relevant in the

Foundational Issues of Engineering Design

Figure 4. The design of a Newcomen engine

533

534

Peter Kroes

light of the performance of its overall function). A functional description represents only half of the design of a technical artefact since different physical structures may realise the same function and different physical realisations imply different designs of the artefact. The same is true of structural descriptions of a technical artefact: one and the same physical object may perform different functions on the basis of different designs (reflected in different structural and functional decompositions of the same object). Neither the functional nor the structural descriptions on their own completely capture the design of a technical artefact; the functional design omits the structural side while the structural design lacks the functional design properties. This just goes to show that when describing technical artefacts both the structural and the functional properties are indispensable in engineering practice [Kroes, 2006, p. 139]. A main difficulty when further clarifying the notion of a technical design lies in its association with the notions of purpose and function. Artefacts based on a technical design are said to have a purpose and this purpose is conferred on them by their design. Indeed, the notion of a design has strong teleological connotations in that a designed object (i.e. an object based on a design) has a specific property of ‘for-ness’: it is for doing something or for being something.20 This teleological character of designs may be captured by characterising them as some type of plan since plans are associated with purposes and goals. In this context, however, a plan is not a considered series of actions. As technical artefacts do not execute plans that would not make any sense. A plan may rather be taken to be something like a ‘purposeful or teleological arrangement or organisation’ of physical objects showing the adjustment of means to an end. But how is this to be interpreted? One way to interpret the purposeful nature of a design (or of an object based on a design) is by tracing a design, like a kind of plan, back to its origin. A plan is a mental construct that has its origin in the mind of the designer. It may be taken to inherit its purposeful nature from its designer. This line of reasoning is used in arguments from design. In its most famous form, it is an argument for the existence of God. The purposefulness (together with other features) of certain natural systems, in particular of biological organisms and their parts, is taken to be proof that they are designed objects, a fact which is then used as an argument for the existence of a supernatural intelligent designer [Ratzsch, 2005; Russell, 2005]. According to Ratzsch [2005, p. 2] arguments from design are rather unproblematic in the case of technical artefacts or, more generally, in the case of things that “nature could not or would not produce.” He claims for instance that for a DVD-player the conclusion that it was designed by human beings is “nearly inescapable”.21 A similar claim was made more than two hundred years ago by Paley with regard to a watch; he stated that when we examine a watch, what we see are [Paley et al., 2006, p. 14]:

20 For an analysis of the notion of teleology in relation to technical artefacts see, for instance, [McLaughlin, 2001; Perlman, 2004]. 21 Feh´ er [1993] presents an interesting thought experiment that puts this claim to the test.

Foundational Issues of Engineering Design

535

contrivance, design; an end, a purpose; means for the end, adaptation to the purpose. And the question, which irresistibly presses upon our thoughts, is, whence this contrivance and design. The thing required is the intending mind, the adapting hand, the intelligence by which that hand was directed. In this way, the purposefulness of a technical design (and of a technical artefact based on that same design) may be directly related to, and considered to be derived from, the intentionality of a human designer. Still, this does not lead to a clearer picture of what a design, as a defining feature of a prospective or real artefact, is. Things become even more complicated when a technical artefact, as a designed object, is taken to be the ‘embodiment’ or ‘material realisation’ of a design. What does it mean for a physical object to embody a design, a mental plan, and to what extent does it inherit the purposefulness of a design? This notion of technical artefacts appears to turn them into objects made up of a mental (i.e. intentional) and a physical side; they become, so to speak, creations of mind and matter. The physical features of technical artefacts are necessary when accounting for their causal efficacy and their intentional features account for their purposefulness (i.e. their functions). Insofar as they are products of the mind they inherit the teleological nature of the intentional action of the designer. In line with this thread of argument, technical artefacts are objects with a dual nature; they have physical features as well as intentional features.22 To conclude, from a conceptual point of view no clear analysis of the notion of a design of a technical artefact has yet been provided. In engineering practice these conceptual problems do not appear to be very important. In fact, one can search almost in vain in engineering handbooks for an elaborate analysis of what a design, in the sense intended here, incorporates.23 From a pragmatic point of view, what is much more important is how designs of technical artefacts are unambiguously represented. The growing complexity of modern technical artefacts and the use of computers in supporting solutions to engineering design problems have increased the need for more formal, unambiguous representations of designs. Such representations are vital to the development of engineering data management systems for computer aided design (CAD). It is especially the formal representation of functions that proves to be problematic [Dym, 1994]. Much work is currently being done on developing taxonomies of functional primitives (a field sometimes referred to as ‘functional modelling’), on functional representation and functional reasoning in AI-quarters, the aim being to support engineers in their solving of design problems and in the accurate representation of designs. 22 For a discussion on the dual nature of technical artefacts, see the special issue “The dual nature of technical artefacts” in Studies in History and Philosophy of Science, vol. 37, no. 1. 23 When used as a noun, the notion of design usually refers to a fabrication plan. Hubka and Eder [1996, p. ix] mention the interpretation of a design as the outward appearance and pattern of artefacts; this interpretation is not particularly of relevance to the present discussion.

536

Peter Kroes

9 COMPLEXITY AND THE TRADITIONAL DESIGN PARADIGM In this final section I will draw attention to a new field of engineering that has emerged in recent decades, namely systems engineering. It focuses on the design, development, maintenance and control of complex, large scale technological systems (see, for instance, [Sage, 1992]). It is interesting to take a closer look at this field because it is where engineers meet the limits of the applicability of what I will call their traditional design paradigm. Let us first take a closer look at this design paradigm. It is made up of three assumptions about the kind of technical artefacts that are designed. This category is exemplified by stand alone consumer products. Most of the examples used so far fall into this category. These technical artefacts may be used by individuals or by groups, more or less in isolation of their wider technological and social context. What is required for the proper performance of their function is a technical artefact that does not malfunction and is properly implemented. To phrase it in Simon’s terminology, the inner and outer environments of the technical artefact have to behave as they ought to. This brings us to the first important feature of the traditional design paradigm, namely the assumption that it is possible to clearly separate the object of design from its environment. In his analysis of engineering design Simon, for instance, simply assumes that this does not give rise to any problems. The second feature concerns an assumption about the nature of the object or system to be designed or, more to the point, the content of the inner environment. Traditional engineering concerns itself with the designing of the hardware (the manual is more or less taken for granted). What is designed is a material technical object. The final feature of the traditional design paradigm is that it is assumed that the behaviour of the systems designed can be fully controlled by controlling the behaviour of its parts, at least when the designed system is used under conditions specified within the design specifications. Given that the artefact is made up of physical parts, this control amounts to the control of the behaviour of these physical parts through a set of control parameters. These three assumptions about the objects of design, which are not independent of each other, together characterise the traditional design paradigm. Certain features of the kind of systems designed within the field of systems engineering appear to undermine the applicability of the traditional design paradigm in this field. Systems engineering arose in response to the ever more complex systems designed and developed by engineers. This development not only challenged engineers in relation to the designing of such complex systems but it also presented questions concerning the designing and organising of the engineering design process allied to such complex systems [Ottens et al., 2006]. Here we concentrate on two features of the kinds of systems designed that pose questions in conjunction with the applicability of the traditional design paradigm. The first feature concerns the socio-technical nature of the systems designed, the second the possibility of emergent behaviour in complex systems.

Foundational Issues of Engineering Design

537

One of the types of systems studied and designed within systems engineering is large-scale infrastructural systems, such as electric power supply systems or public transport systems. The behaviour of these systems is significantly affected by their technical elements but the functioning of the systems as a whole depends as much on the functioning of these technical components as on the functioning of social infrastructures (legal systems, billing systems, insurance systems etc.) and it depends on the behaviour of human actors. From an engineering point of view this draws attention to the issue of whether the social infrastructure is to be regarded as part of the outer environment and modelled as a series of constraints for the design of technical systems or simply taken as part of the system to be designed. An important argument in favour of including these social elements within the system is that the technological and social infrastructures have to be attuned to each other if such systems are to operate successfully. If social elements are included within the system, as is often advocated, then the implication is that systems engineering has to deal with socio-technical systems. These are hybrid systems consisting of elements of various kinds, such as natural objects, technical artefacts, human actors and social entities like organisations and the rules and laws governing the behaviour of human actors and social entities.24 The traditional design paradigm no longer seems to be a suitable basic framework for the design and control of socio-technical systems. To begin with, there is the problem of where to draw the line between the system under consideration and its environment. This is a conceptual problem that systems engineering inherits from systems theory [Kroes et al., 2006]. If the function of a system is taken to be that which gives the system cohesion, then it is rather obvious that all elements relevant to the functioning of the system should be included. So human agents and social institutions would have to become integral parts of the infrastructure systems alluded to above. But how is the function of, for instance, an electric power supply system to be defined? Different actors may have different views on this and may therefore have different opinions on what constitutes part of the system and what belongs to its environment. The socio-technical nature of the systems designed also means that the nature of the system to be designed changes. The inner environment will no longer consist of only material objects. The design of these systems not only involves the design of technical but also of social infrastructures from the point of view that they are tailor-made to match each other. Finally, the idea that these systems can be completely designed and controlled has to be abandoned. The behaviour of human agents and social institutions cannot be controlled in the way that the behaviour of technological systems can be controlled. In the traditional design paradigm it is assumed that there is a vantage point outside the designed system from which design and control is overseen. That is not the case with socio-technical systems in which various actors, with their own interpretations of the function of the system and their role in realising it, set out 24 Within the field of STS these systems are often referred to as heterogeneous systems; see, for instance, [Bijker et al., 1987].

538

Peter Kroes

to change or re-design parts of the system from within. For this reason even the notion of designing socio-technical systems becomes problematic. The second feature of complex systems that threatens the applicability of the traditional design paradigm lies in the possible occurrence of emergent phenomena. In recent times, emergent phenomena in complex technological systems have become quite a topic of debate in engineering circles [Buchli and Santini, 2005; Deguet et al., 2005; Johnson, date unknown]. The science and engineering of complex systems are turning into fields in their own right in which emergent phenomena are widely coming to be seen as a defining feature of complexity.25 Complex systems may exhibit non-linear, chaotic behaviour that results in processes of selforganisation and in emergent systemic properties like adaptivity, robustness and self-repair [Bertuglia and Vaio, 2005]. From an engineering point of view such properties may be desirable but the drawback is that their occurrence may be unexpected and unpredictable. That makes it difficult to control such features. The desire to control emergent phenomena in complex systems is driven on the one hand by the fact that they may be dangerous (blackouts in electric power supply systems are often claimed to be such emergent features) and on the other hand by the fact that they may contribute to some desired property of complex technological systems (e.g. complex adaptive systems may be more robust in relation to changing conditions in the environment).26 Whether blackouts in electric power supply systems are genuine examples of emergent phenomena or whether other real examples can be given, remains to be decided. Assuming, however, that emergent phenomena may occur in complex technical systems, they do pose a real challenge to the traditional design paradigm. This challenge is not related to the first and second features of this paradigm. Emergent behaviour may occur in systems where it is not problematic to establish where the boundary with the environment lies and where there is not necessarily evidence of ‘hybrid’ systems (although the socio-technical systems discussed above may prove to be a promising class of systems exhibiting emergent behaviour). It is the third feature, the assumption about the control of the behaviour of the system that has to be renounced with regard to emergent behaviour. The emergent behaviour of a system cannot, more or less by definition, be reduced to the behaviour of the constituent parts of the system. This means that techniques like functional decomposition cannot be applied to functional properties of systems that are based on emergent phenomena. It also implies that the behaviour of the system as a whole cannot be completely controlled by controlling the behaviour of its parts. So, emergence and control do not go hand in hand. According to Buchli and Santini [2005, p. 3] “there is a trade-off between self-organization [and 25 See, for instance, the pre-proceedings of the Paris conference (14-18 November 2005) of the European Complex Systems Society, ECCS’05 (http://complexite.free.fr/ECCS/ ); this conference hosted satellite workshops on topics such as Engineering with Complexity and Emergence and Embracing Complexity in Design. 26 Kasser and Palmer [2005] distinguish between three types of emergent properties namely undesired, serendipitous and desired; serendipitous features are described as “beneficial and desired once discovered but not part of the original specifications”.

Foundational Issues of Engineering Design

539

emergence; P.K.] on one hand and specification or controllability on the other: if you increase the control over your system you will suppress self-organization capabilities.” Such a new trade-off principle would indeed constitute a significant break with the traditional design paradigm.27 Given the growing complexity of the systems that engineers have to deal with, it is to be expected that systems engineering will become an ever more important branch of engineering. This growing complexity will pose new challenges to engineering design practice. Whatever the precise nature of this complexity it will, without any doubt, stretch the applicability of traditional methods for designing and controlling technical systems to their limits or even beyond their limits. This means that for these systems the traditional design paradigm with its idea of ‘total design control’ may have to be left behind and alternative design paradigms may have to be developed instead. BIBLIOGRAPHY [ASM, 1997] ASM Handbook: Materials Selection and Design Vol. 20. ASM International, 1997. [Arendt, 1958] H. Arendt. The Human Condition. The University of Chicago Press, 1958. [Barnes et al., 1996] B. Barnes, D. Bloor, and J. Henry. Scientific Knowledge: A Sociological Analysis. Athlone, 1996. [Bertuglia and Vaio, 2005] C. S. Bertuglia and F. Vaio. Nonlinearity, Chaos, and Complexity: The Dynamics of Natural and Social Systems. Oxford University Press, 2005. [Bijker et al., 1987] W. E. Bijker, T. P. Hughes, and T. Pinch, eds. d, The Social Construction of Technological Systems: New Directions in the Sociology and History of Ttechnology. The MIT Press, 1987. [Bratman, 1987] M. Bratman. Intention, Plans, and Practical Reasoning. Harvard University Press, 1987. [Buchli and Santini, 2005] J. Buchli and C. C. Santini. Complexity Engineering: Harnessing Emergent Phenomena as Opportunities for Engineering. Santa Fe Institute, 2005. [Chandrasekaran, 2005] B. Chandrasekaran. Representing function: Relating functional representation and functional modelling streams. Artificial Intelligence for Engineering Design, Analysis, and Manufacturing, 19, 65-74, 2005. [Chandrasekaran and Josephson, 2000] B. Chandrasekaran and J. R. Josephson. Function in device representation. Engineering with Computers, 16, 162-177, 2000. [Chittaro and Kumar, 1998] L. Chittaro and A. N. Kumar. Reasoning about function and its applications to engineering. Artificial Intelligence in Engineering, 12, 331-336, 1998. [Cross, 1989/1994] N. Cross. Engineering Design Methods: Strategies for Product Design. John Wiley & Sons, 1994/1989. [Deguet et al., 2005] J. Deguet, Y. Demazeau, and L. Magnin. Elements about the emergence issue: a survey of emergence definitions. Proceedings of the ECCS 2005, Paris, 2005. [Dennett, 1987] D. C. Dennett. The Intentional Stance. The MIT Press, 1987. [Dym, 1994] C. L. Dym. Engineering Design: A Synthesis of Views. Cambridge University Press, 1994. [Dym and Little, 2000] C. L. Dym and P. Little. Engineering Design: A Project-based Introduction. John Wiley & Sons, 2000. [Feh´ er, 1993] M. Feh´ er. The natural and the artificial. Periodica Polytechnica: Humanities and Social Sciences, 1, 67-76, 1993. [Franssen, 2005] M. Franssen. Arrow’s theorem, multi-criteria decision problems and multiattribute design problems in engineering design. Research in Engineering Design, 16, 42-56, 2005. 27 For a more detailed analysis of the notion of emergence and its relation to the control issue in engineering, see [Kroes, 2009].

540

Peter Kroes

[Hacking, 1983] I. Hacking. Representing and Intervening: Introductory Topics in the Philosophy of Natural Science. Cambridge University Press, 1983. [Houkes and Vermaas, 2004] W. N. Houkes and P. E. Vermaas. Actions versus functions: A plea for an alternative metaphysics of artifacts. Monist, 87, 52-71, 2004. [Houkes et al., 2002] W. N. Houkes, P. E. Vermaas, K. Dorst, and M. J. De Vries. Design and use as plans: an action-theoretical account. Design Studies, 23, 303-320, 2002. [Hubka and Eder, 1996] V. Hubka and W. E. Eder. Design Science: Introduction to the Needs, Scope, and Organization of Engineering Design Knowledge. Springer, 1996. [Hughes et al., 2007] J. Hughes, P. Kroes, and S. Zwart. A semantics for means-end relations. Synthese, 158, 207-231, 2007. [Johnson, date unknown] C. W. Johnson. What are emergent properties and how do they affect the engineering of complex systems? http://www.dcs.gla.ac.uk/ johnson/papers/emergence.pdf. [Kasser and Palmer, 2005] J. E. Kasser and K. D. Palmer. Reducing and managing complexity by changing the boundaries of the system. Proceedings CSER 2005, Hoboken, NJ, USA, 2005. [Kroes, 1996] P. Kroes. Technical and contextual constraints in design: an essay on determinants of technological change. In The Role of Design in the Shaping of Technology. J. Perrin and D. Vinck, eds., 43-76, 1996. [Kroes, 2006] P. Kroes. Coherence of structural and functional descriptions of technical artefacts. Studies in History and Philosophy of Science, 37, 137-151, 2006. [Kroes, 2009] P. Kroes. Technical artifacts, engineering practice, and emergence. In Functions in Biological and Artificial Worlds: Comparative Philosophical Perspectives. U. Krohs and P. Kroes, eds., 277-292, MIT Press, 2009. [Kroes et al., 2006] P. Kroes, M. Franssen, I. van de Poel, and M. Ottens. Treating sociotechnical systems as engineering systems: some conceptual problems. Systems Research and Behavioral Science, 23, 803-814, 2006. [Kroes, 2003] P. Kroes. Physics, experiments and the concept of nature. In The Philosophy of Scientific Experimentation. H. Radder, ed., 68-86, University of Pittsburgh Press, 2003. [McLaughlin, 2001] P. McLaughlin. What Functions Explain: Functional Explanation and SelfReproducing Systems. Cambridge University Press, 2001. [Mitcham, 1994] C. Mitcham. Thinking Through Technology: The Path Between Engineering and Philosophy. The University of Chicago Press, 1994. [Ottens et al., 2006] M. Ottens, M. Franssen, P. Kroes, and I van de Poel. Modelling infrastructures as socio-technical systems. Int. J. Critical Infrastructures, 2, 133-145, 2006. [Otto and Wood, 2001] K. N. Otto and K. L. Wood. Product Design: Techniques in Reverse Engineering and New Product Development. Prentice Hall, 2001. [Pahl and Beitz, 1996] G. Pahl and W. Beitz. Engineering Design: A Systematic Approach. Springer Verlag, 1996. [Paley et al., 2006] W. Paley, M. Eddy and D. M. Knight. Natural Theology: Or, Evidence of the Existence and Attributes of the Deity, Collected from the Adappearances of Nature. Oxford University Press, 2006. [Perlman, 2004] M. Perlman. The modern philosophical resurrection of teleology. The Monist, 97, 3-51, 2004. [Petroski, 1992] H. Petroski. The Evolution of Useful Things. Alfred A. Knopf, 1992. [Pollock, 2002] J. L. Pollock. The logical foundations of means-end reasoning. In Common Sense, Reasoning, and Rationality. R. Elio, ed., Oxford University Press, 2002. [Ratzsch, 2005] D. Ratzsch. Teleological arguments for God’s existence. In The Stanford Encyclopedia of Philosophy (Fall 2005 edition). E. N. Zalta, ed., http://plato.stanford.edu/ archives/fall2005/entries/teleological-arguments/, 2005. [Reymen, 2001] I. Reymen. Improving Design Processes through Structured Reflection: A Domain-independent Approach. Eindhoven, University of Technology Eindhoven, PhD Thesis, 2001. [Rittel and Webber, 1984] H. W. J. Rittel and M. M. Webber. Planning problems are wicked problems. In Developments in Design Methodology. N. Cross, ed., 135-144, John Wiley & Sons, 1984. [Roozenburg and Eekels, 1995] N. F. M. Roozenburg and J. Eekels. Product Design: Fundamentals and Methods. John Wiley & Sons, 1995. [Rosenman and Gero, 1994] M. A. Rosenman and J. S. Gero. The what, the how, and the why in design. Applied Artificial Intelligence, 8, 199–218, 1994.

Foundational Issues of Engineering Design

541

[Russell, 2005] P. Russell. Hume on Religion. In The Stanford Encyclopedia of Philosophy (Fall 2005 edition). E. N. Zalta, ed., http://plato.stanford.edu/arvhives/win2005/entries/humereligion/, 2005. [Ryle, 1984] G. Ryle. The Concept of Mind. University of Chicago Press, 1984. [Sage, 1992] A. P. Sage. Systems Engineering. John Wiley & Sons, 1992. [Segerberg, 1992] K. Segerberg. Getting started: beginnings in the logic of action. Studia Logica, 51, 347-378, 1992. [Simon, 1984] H. A. Simon. The structure of ill-structured problems. In Developments in Design Methodology. N. Cross, ed., 145-166, John Wiley & Sons, 1984. [Simon, 1996] H. A. Simon. The Sciences of the Artificial. MIT Press, 1996 (1969). [Staudenmaier, 1985] J. M. Staudenmaier. Technology’s Storytellers: Reweaving the Human Fabric. The MIT Press, 1985. [Thomasson, 2007] A. L. Thomasson. Artifacts and human concepts. In Creations of the Mind: Essays on Artifacts and their Representations. S. Laurence and E. Margolis, eds., 52-73, Oxford University Press, 2007. [Vermaas and Houkes, 2006] P. E. Vermaas and W. Houkes. Technical functions: a drawbridge between the intentional and structural natures of technical artefacts. Studies in History and Philosophy of Science, 37, 5-18, 2006. [Vincenti, 1990] W. G. Vincenti. What Engineers Know and How They Know It. Johns Hopkins U.P, 1990. [von Wright, 1963] G. H. von Wright. Practical inference. The Philosophical Review, 72, 159179, 1963. [von Wright, 1972] G. H. von Wright. On the so-called practical inference. Acta Sociologica, 15, 39-53, 1972.

This page intentionally left blank

COMPUTATIONAL REPRESENTATIONS OF FUNCTION IN ENGINEERING DESIGN William H. Wood Function is what we want from any technical system; it is what the system does for us, the reason why it was designed, realized, and introduced into our world. Understanding function requires understanding not only the physical environment in which a system operates but also its human socioeconomic context. Design and function are intricately linked — armed with knowledge and experience in both, the human designer bridges the physical/technical and social/economic worlds. Toward mitigating the increasing complexity of both of these worlds, designers increasingly rely on computational tools. While tools that ‘understand’ the physical world have become invaluable in the practice of engineering design (complete cars are designed and their technical performance modeled computationally), computational support to bridge the human side of design is still in its infancy. In this chapter, we explore the current state of computer representations of function. We start by attempting to demarcate the human notion of function from the technical behavior of the artifacts that accomplish that function. Toward refining the transition between the two, we examine the empirical ‘science’ of reverse engineering: the methods used to perform it, the information that is extracted using it, and the application of this information to future designs. In this process, the notion of functionality is expanded to account for the modern engineering practice of life cycle engineering — looking at function from the socioeconomic perspectives of all people ‘touched’ by a design throughout its life. Abstraction is used to traverse the spectrum between the human and technical aspects of function. 1

FUNCTION VERSUS BEHAVIOR

Throughout this chapter, we discuss function in relation to an artifact — a physical object that interacts with the rest of the physical world (including humans) through the laws of physics. The effect of these interactions has been dubbed the behavior of an artifact. Any physical object has an infinity of behaviors as it interacts with an uncountable set of other objects in the world; viewed in this way, behaviors tend to be very generic and neutral. Eventually, however, artifacts find their way into contexts in which they add value to the human experience. Chairs take up space, change local heat transfer, exert gravitational forces, and exhibit many other behaviors that can be said of any other object; but when placed in the right position on a solid surface, they transfer the weight of a human to the Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

544

William H. Wood

ground while at the same time supporting the human body at a level at which it can eat/work/play/relax/etc. It is these latter behaviors that correspond to the human notion of function. Function is the subset of behavior that adds value to the human context. Function is the subset of behaviors that we design into an artifact; the rest of the behaviors are simply ‘along for the ride’. It is useful to discuss function from three main points of view. The most generic treatment of function is closest to the socioeconomic frame — affordance [Gibson, 1979; Maier and Fadel, 2002]. The affordance model of function places an actor within an environment. By avoiding preconceptions about the word function and abstracting the human context to actors and environment, the notion of affordances removes the intent of the designer and with it any preconceptions about what an artifact is supposed to do. In this framework, a chair affords sitting, affords concentration for work, affords comfort, etc. It also affords reaching high shelves, affords ventilation by propping a door open, and might afford freedom of movement by folding. The concept of affordances focuses on the actor and the environment and not on the intent of the designer. Rather than the subset of behaviors in the subset of contexts that drove its specification, it examines all useful behaviors of an artifact in all contexts. As such, affordances can be difficult to apply in a generative sense but can prove a valuable construct in identifying and evaluating useful secondary behaviors that might, at some point, become functions. A slightly more design-centric view of function is that espoused by Gero et al. [Gero, 1990; Rosenman and Gero, 1998; Gero and Kannengiesser, 2002] in their function-behavior-structure (FBS) model of design. Under FBS, the process begins with the intent of the designer — an exploration of ways in which value can be added to a specified human context. Where experience predicts that value can be added by the behavior of an artifact, the design of that artifact is undertaken. This ‘design’ is a mapping from intended behavior (i.e., function) to the structure of an artifact taking place, again, largely through the experience of a designer. Once completed, the structure of the artifact is then used to predict behavior — not only that intended by the designer but also other ‘side effects’ of the chosen structure. Undesirable side effects are removed or mitigated through changes to the structure. Where small structural changes cannot resolve undesirable behavior, the structure is rejected and alternate structures considered. The FBS model draws on the dichotomy between the performance of a design in a human context and its interactions with the physical world. Function is intentional on the part of the designer, it is defined within a human context, it adds value; behavior is neutral, a physical consequence of a design’s structure. The debate about this duality of artifacts continues [Kroes, 2002], nevertheless the FBS model at least provides a basic framework for further development of computational representations of function and the processes that operate over them. The third view is function as a compilation of behavior. Divorced from the rich physical context in which mechanical and architectural behavior takes place, electronic behavior and structure are almost completely aligned. Advances in this field

Computational Representations of Function in Engineering Design

545

have been startling — from the development of application specific ICs (ASICs) to field programmable gate arrays (FPGAs) and systems on a chip (SoCs), a single designer has the power to construct, with the help of extensive libraries of structural units (including microprocessors), a custom computer. On the surface this appears to be a huge advance in design [Whitney, 1996], writing at the dawn of this technology, unfavorably compares advances in mechanical design automation to those in electronics). On deeper examination, the computer so constructed still has little functionality without the program that it executes which, itself, has little true function outside of a human context. So, while the details remain different and the design ‘power’ (in terms of the sheer number of components a single designer can manage) far exceeds that available to mechanical/architectural designers, the result is still much the same — function is an intentional, human concept, developed by the experience of the designer to add value in a human context. The major shift here from Gero et al.’s perspective is that these custom electronics exhibit little behavior outside of that which is intended (an oversimplification to anyone who has attempted to debug these systems, especially as elevated clocking speeds produce unintended, unmodeled behavior). Perhaps the main lesson to draw from the electronics world is the power that functional abstraction produces. An electronics engineer can string together precompiled functional units ranging from flip flops (a few gates) to microprocessors (100,000+ gates) rather than create a VLSI design from the gate level up. Abstraction is central to representing and computing function; design is simplified when function and behavior are represented at the same levels of abstraction. This discussion amplifies the dual nature of technical artifacts - they operate in both physical and human socioeconomic contexts. It also reinforces that design is a human-technical process in which the experience of the designer both as a human and as a student of physical behavior is required. Design might be thought of as a multistage inverse problem: presented with a problematic context, a designer first identifies a behavior or set of behaviors that might add value. This requires inverting a mapping from context to value-adding behavior that a designer has acquired through experience. The second step, according to Gero et al., is then to invert a physics-based mapping between structure and behavior to identify candidate structures to generate the candidate desired behaviors. In both of these mappings, a designer’s experience is vital in drawing generalities from specific experience and in identifying links from one context to another. Abstraction plays a key role in a designer’s ability to encapsulate a context — to hide details of little consequence while focusing on and expanding details that join one context to another, one design solution to another. 2 DESIGNING FUNCTION It is difficult to fix a specific level of abstraction at which one can differentiate between function and behavior. With broad strokes we define two separate map-

546

William H. Wood

pings that go on during the design process: The first mapping is that needed to move from the socioeconomic context of a design to sets of functions (i.e., intended behaviors) likely to provide value in this context. The rise of ‘Voice of the Customer’ [Pugh, 1991] techniques in engineering design emphasizes the importance of capturing the human context of design. Methods like QFD [Hauser and Clausing, 1988] operate over direct customer utterances, attempting to map them onto objective measures of design performance. It is significant that, during this process, utterances relating to assumed solutions are abstracted to more general notions of performance (e.g., instead of talking about how much power a drill has, engineers might establish time/value metrics for the creation of holes or the setting/removal of fasteners). The human context often includes artifacts with which a design must interact and human processes within which this interaction takes place. While QFD strives to establish the value of artifact behavior in the design’s context, the measures it proposes on artifacts and processes also begin to identify key functions (i.e., intended behaviors) of the design. Material Energy Information

Function

Material Energy Information

Electricity Fruit

Fruit

Peel Fruit, peeled

B

A Electricity

Peel Fruit

Convert to Mech

Move Cutter

Hold Fruit

Rotate Fruit

Remove Peel

Peel Fruit, peeled

C

Figure 1. Function Structures: A. Template. B. Fruit Peeler Black Box. C. Fruit Peeler Design The second design mapping relates function to structure. Pahl and Beitz [Pahl and Beitz, 1988] propose a systematic design method based on the functionstructure representation. A function-structure is a block diagram representation of the system in which energy/force, material, and information flow into each functional unit, are transformed by it, and flow out and into other functions. Function-structures pass through several abstraction stages: a top-level ‘black box’ that represents the system as a single functional unit is gradually decomposed into networks where flows are separated and operated upon by successively less abstract

Computational Representations of Function in Engineering Design

547

functions. Figure 1 shows a general template for a function structure as well as two abstraction levels for an electric fruit peeler. Decomposition terminates when all functional transformations can be carried out by a set of well-defined ‘solution principles’. In terms of Gero et al.’s FBS model, the resulting network of solution principles defines the structure of the design, whose behavior is defined by the ways in which they interact through their topological and spatial layout. This systematic method spreads the mapping from intended behavior to structure over a wide range of abstractions, in the process blurring the distinction between functions and structure. The solution principles each embody structural, physical entities, but the topology of the function-structure also represents aspects of structure. So instead of a crisp mapping from behavior to structure, function-structures provide a continuous evolution of structure from intended behavior. 3

COMPUTING FUNCTION

We have introduced function as the physical behavior of an artifact intended to produce value in a socioeconomic context. For designers, this implies two separate mappings: from value to intended behavior and from intended behavior to physical structure. We now look at computational representations intended to address each of these mappings.

3.1

Mapping customer needs to function

Computing intended behavior from value is in its infancy. The most mature of this work is that of McAdams et al. [McAdams, et al., 1999; McAdams and Wood, 2002], as part of a broad project to improve the computation of function. Just as Gero et al. call on the experience of the designer for mapping from value to behavior and behavior to structure, computational research calls on experience collected into a design database. The emphasis here is on the dual representations of customer needs and design function. A set of existing products is reverse engineered to extract information about both product value — what were the important customer needs that shaped the design (from a post-hoc QFD-style analysis of the design), and product function/structure — what functions are carried out, how do they relate to each other. Need and function are then related to each other through a weighting scheme that is essentially part of the reverse engineering process. Otto and Wood [1998] prescribe a ‘remove and operate’ mode of reverse engineering in which the main system flows are identified, components along that flow are successively removed and the artifact ‘operated’ conceptually. The differences in operating behavior with and without the component are then ascribed to the function of that component. At the same time, the differences in the value of the artifact with and without a function are estimated and a value ascribed to each function as well. In the example of a cordless drill, the voltage regulator might be removed, replaced by a simple on-off switch. The ability of an on/off drill to meet customer needs is then assessed and the product specifications

548

William H. Wood

degraded by the change (e.g., accuracy of fastener depth, ease of starting a hole) are associated with that function. This process is relatively straightforward for a single artifact whose functions and flows are captured linguistically, but computing over representations of multiple artifacts reverse engineered by multiple people is more problematic. Toward resolving this difficulty, Stone and Wood [2000] offer a hierarchical basis language for standardizing the encoding of both function and flow. Here function is generally described with action verbs (e.g., branch, channel, connect, control magnitude, convert, provision, signal, support) and flow with nouns (e.g., material: human, gas, liquid, solid; signal: status, control; energy: human, acoustic, biological, chemical, electrical, electromagnetic, hydraulic, magnetic, mechanical, pneumatic, radioactive, thermal). Reverse engineering produces a set of verb-object phrases linked together by the object. Using this encoding along with the ‘remove and operate’ mode of reverse engineering, Kurfman et al. [2003] find up to 70% agreement among reverse engineers for the same artifacts and up to 50% agreement among designers identifying functions for the same design context.

3.2 Mapping function to structure: abstract Verma and Wood [2006] apply this same reverse engineering methodology, expanding the notion of artifact function to include manufacturing, another important ‘customer’ from the life cycle of an artifact. The design of physical artifacts is always a compromise between its functionality in the eyes of the end user and the efficiency of its production by the manufacturer. In an effort to identify the level of abstraction at which manufacturing ‘functionality’ resides, products were reverse engineered at the subsystem (based on functional decomposition), part (a unit either formed by the manufacturer or assembled as a single unit into the final artifact), and geometric feature (a basic CAD entity like ‘boss’) levels of detail. The function of each item was then determined and assigned to one of four categories: primary user (functions directly adding value in the user context), secondary user (functions that support the main functionality but are not essential), assembly (functions that ease assembly but have no impact on user function), and forming (functions that ease part forming processes but do not directly affect user function). Function was recorded in both linguistic and functional basis languages; multiple function assignments for each subsystem/part/feature were allowed. Findings from this work include: 1. Subsystem decomposition is not reliable across reverse engineers. Matching the physical and functional abstraction levels is the main issue — some modules perform a single main function, others simply compare a subassembly useful for manufacturing that performs multiple functions. 2. Feature decomposition is not reliable across reverse engineers. Again abstraction is the main difficulty, in this case the abstraction of a set of individual geometric entities into a recognizable feature: is a domed top on a pin a

Computational Representations of Function in Engineering Design

549

separate feature from the pin or is it an assembly feature to aid in locating a mating part? 3. Part decomposition is generally reliable across reverse engineers. Given a few simple instructions, all reverse engineers are able to identify the parts in a product. Most also agree on the label for the flows that a part connects to; naming how it actually transforms those flows is less reliable. In fact, a large measure of the reliability reported by Kurfman et al. can be attributed to easily recognizable system input and output flows. Each is associated with an ‘import’ or ‘export’ function; these collectively account for about 25% of all functions for the relatively simple products typical of the database. With a 25% head start, little more than half of the remaining functions are assigned reliably. 4. Manufacturing drives part geometry. Over half of all part features have no relationship to user functionality (primary or secondary). Most recognizable geometric detail is, in fact, difficult to relate to function at all — much of it is associated with the mechanical ‘ground’ components (housings, frames, base plates, etc.) that provide spatial reference and react forces generated by the main functional units. Because they often form the backbone of the product but have no direct functional impact, these components are ready targets for design for manufacture techniques that seek to minimize part count in a product. 5. Manufacturing functions are generic. Tapered pins aid in part location; fillets ease flow of material into part molds, etc. The trade between manufacturability and user function happens mainly in the secondary functions, primary user functions are rarely affected by manufacturing concerns. Rather than capture a set of generic rules from product cases, reverse engineering can likely be more effective by capturing the geometry/manufacturing/assembly information related to a specific function. Stahovich and Kara [2001] also study the relationship between geometric feature and function. Their system operates not over physical artifacts but their computer solid geometry representations. They apply the same ‘remove and operate’ paradigm for inducing function, but at lower levels of abstraction by performing physics-based behavioral simulations on parts before and after the removal of identified geometric entities. Several issues arise out of this work: a geometric entity is identified as part of the constructive solid geometry tree for a part. Because only features that are added or subtracted from a base part can be removed for simulation, function ascribed to the basic part or specific dimensions of it may not be induced. Additionally, with much of the geometric detail related to assembly and forming functionality that is not simulated, much of the geometry will remain a mystery to the system. Finally, because removing any one geometric feature might result in null behavior, the mapping from geometry to behavior can prove difficult. As a result, these feature-based efforts reinforce the practice of human-performed

550

William H. Wood

reverse engineering at the part level of abstraction. Given a practical approach to inducing function from existing artifacts, the two remaining steps involve the storage of this information and its retrieval and use in a new design context. Rotational Mechanical Energy Solid object

Solid object

Separate solid (a)

Solid object

Rotational Energy Fruit/vegetable

Fruit/vegetable

Peel fruit/ vegetable

Peel

(b)

Figure 2. (a) function-structure chosen from the functional model of a fruit/vegetable peeler using the functional basis. It is difficult to interpret whether a solid is being cut to pieces of two different sizes or the peel is being removed; (b) the same function structure picked up from the functional model of a fruit/vegetable peeler using unrestricted text. Storing function captured from reverse engineering is a straightforward proposition — store each function structure block by function and associate its input and output flows with it. At issue is the language of expression for function and flow. Stone et al.’s functional basis provides a hierarchical set of terms used to describe both function and flow. On the function side, there are eight main function classes, each with associated subclasses and synonyms. Encoding takes place at the lowest possible level of abstraction (the synonym), with paths to higher levels of functional abstraction directly encoded in the language. An electric motor ‘converts electrical energy to mechanical rotation’, an encoder ‘senses mechanical rotation’. Capturing flow information is similar: the three main flow types (material, energy, and information) are further decomposed into subclasses (energy: electrical, mechanical, thermal, etc.) which are further subdivided (mechanical energy: translation, rotation, pneumatic/hydraulic, etc.). The functional basis is primarily descriptive; no simulation models are associated with functional units, no input or output flow types are prescribed for a given function. In isolation, the representation creates functional ‘sentences’ using a prescribed grammar and vocabulary. Connected together by their input and output flows, functional chains produce ‘paragraphs’ that define the overall subsystem and system function. While the grammar remains fixed under the function-structure model, we can vary the language of expression both within and beyond the functional basis. Within the functional basis, the level of abstraction for each functional sentence can be varied: limit abstracts to regulate which further abstracts to control magnitude. Abstraction can be varied on function, on flow, or on both. Further, we might choose not to use the functional basis at all, instead using free text for the function structure vocabulary. Verma and Wood [2003] explore these possibilities using metrics from information retrieval to evaluate their performance. Within a case base of about 50 reverse engineered artifacts (approximately 1000 total functions), functional units drawn at random were presented to the database for retrieval of ‘like’ functions. Relevance of the retrieval set was determined by

551

Computational Representations of Function in Engineering Design

Retrieval results for different search strategies 1.0 0.9

Average Precision

0.8

Legend

0.7

Freetext

0.6

Functional Basis

F.B. Family

Flow Function Both

0.5 0.4 0.3 0.2 0.1 0.0 0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

Average Recall

Figure 3. Averaged retrieval results over ten queries. Pareto-optimality of precision and recall should be considered in the comparison (optimal point at upper right).

a panel of design experts using the general guidance: Would knowledge of the retrieved functional unit aid in the design of the target functional unit? The two common measures of retrieval performance are precision (i.e., the fraction of retrieved functions relevant to the target function) and recall (i.e., the fraction of all relevant functions that are in the retrieval set). Three levels of abstraction were used for the retrieval experiment: free text, low-level functional basis, (both illustrated in Figure 2 for a fruit/vegetable peeler) and abstract functional basis ‘family’ (i.e., including the hierarchical ‘parents and grandparents’ of both function and flow). The expectation is that the most generic expression of function — the most abstract level of functional basis expression — would have the highest recall but the lowest precision and that the most specific expression — free text — would have the highest precision but the lowest recall. Tests were run retrieving by function, flow, and function-flow ‘sentences’, varying the abstraction of expression. Figure 3 shows an interesting pattern of results for which ideal performance is in the upper right corner: 1. As expected, free text produces good precision but at low recall. 2. Increasing abstraction from free text to functional basis and then to functional basis family further improves recall, but does not produce the expected decrease in precision for flow or function+flow. 3. The best performance is reached for highly abstract functions and flows, driven largely by the high recall performance from the ‘flow’ component.

552

William H. Wood

Electricity Electricity θx

Sense

(a)

Light Convert

Interrupt

Light

Sense

0/1

Count

dx

dx θx

Sense

0/1

(b)

Figure 4. Function structures for generating displacement (dx) of a computer mouse given the angular displacement (θx) of the ‘ball’: (a) an abstract function; (b) the actual functional unit in the ball mouse, reverse engineered at the ‘parts’ level. These findings apply to the retrieval of function stored and queried at the ‘parts’ level of design abstraction. In a parallel study of retrieval across design abstraction levels, Gietka et al. [2002] query the parts-level database using more abstract representations of function: the ‘black box’ level where inputs, function, and outputs are listed but not connected and ‘design’ where preliminary function/flow paths within the system are specified. They find extremely poor retrieval performance (close to zero recall and precision) for most representations but good recall on flow-only queries (albeit with low precision). Figure 4 helps explain this result by showing the black-box level ‘sense’ function within a mechanical computer mouse at the ‘design’ and ‘parts’ levels of abstraction. The reverse engineering at the ‘parts’ level is in line with Gero’s notion of behavior derived from structure; examining the actual object provides a mechanistic frame from which to construct function. Without knowledge of the underlying mechanics, the ‘design’ function structure encodes intended behavior. The main links between the two representations are the identification of the primary flows into and out of the subsystem; the structure, connectivity, and even presence of additional flows within the parts level representation all differ between the two levels (and across different designs). Verma and Wood further examine this issue by developing and testing methods for extracting abstract structures from ‘parts’ level function-structures encoded in the functional basis vocabulary. Within a function structure, the ‘structure’ comes largely from flow information: how flows enter the system, new flows created within the system through transformations, where flows join or split. All of this establishes the overall topology of the function-structure without much concern for the underlying functional description. Several approaches to aggregating low-level function into more abstract structures were studied: 1. Product Modularity — Stone et al. [1999] propose heuristic rules for extracting potential physical modules from function-structures. The idea is to identify substructures whose functions are readily isolated from the rest of the design. 2. Product Specification — Similar to the approach of McAdams and Wood, aggregation focuses on functional chains closely related to key measures of

553

Computational Representations of Function in Engineering Design

design performance. The idea here is that functions that have a large influence on customer satisfaction are the most critical parts of a design. 3. Input/Output — Functional units are collapsed from the outside-in along the flows that enter and exit the system as a whole. These capture the overall structure of the design by identifying interaction points for the main system flows. 4. Internal Flows — Flows that are present only in the interior of the system are aggregated from the inside-out. Internal flows are not part of the overall design requirement, they are introduced by the designer and, as such, represent design decision-making. 5. Distance — Functions are aggregated around each functional unit at successively larger distances. This brute force aggregation method is a superset of all of the other methods.

RETRIEVAL COMPARISON AT BLACK BOX LEVEL 1

0.8

PRECISION

PRECISION

RETRIEVAL COMPARISON AT DESIGN LEVEL 1

0.6 0.4 0.2

IFB

0.8

MOD 0.6

PSFD

0.4

IO NoAgg

0.2

0

0 0

0.2

0.4 0.6 RECALL

0.8

1

0

0.2

0.4 0.6 RECALL

0.8

1

Figure 5. Comparison of retrieval between the different aggregation techniques at both the design and the black box level. The same legend is valid for both (IF = Ingernal-Flows Based method, MOD = Modules based, PSFD = Product Specs Functionality Decomposition, IO = Input-Output flows based methods, NoAgg = without aggregation. In the functional aggregation process, the main casualty is ‘function’: as functional chains and networks are encapsulated, the terms describing the individual functions lose meaning. Based on parts-level retrieval studies, however, this may not be as big a problem as it would seem. The two sets of studies shown in Figure 5 were performed on the database of aggregated function-structures (now with about 30,000 total ‘functional’ units) toward retrieval using more abstract functions as queries. Building up from the lowest level, the first used functions from less abstract design-level function structures (similar to the Gietka et al. study above); for these, the IF aggregation proves superior to all other methods. The second used very abstract, black-box level functions for which retrieval performance was

554

William H. Wood

good for both input/output (IO) aggregation and internal flows aggregation (IF, a superset of IO). Together, these studies show that retrieval by flow pattern over IF aggregated functions derived from parts-level reverse engineering is a viable means of computerizing the engineering ‘function’ experience that Gero’s FBS model of design requires. Figure 4 provides a simple demonstration of IF functional aggregation. Two design subsystems with the same functionality are reverse engineered to the ‘parts’ level of abstraction. Function-structure (a) shows an off-the-shelf encoder; functionstructure (b) is derived from a design in which a set of customized parts is used to measure the rotation of a shaft. Each accomplishes the same function. By collapsing the flows internal to function-structure (b), we get the same pattern of input/output flows as in function-structure (a). By directing the designer to both function-structures, we are supporting the make/buy decision making at the core of detail design. Without aggregation, these two would not generally respond to the same query. Also apparent is the importance of the input/output flow pattern: rotation comes in, information leaves. Clearly the functional term ‘measure’ (which does not appear in function-structure (b) and could not reasonably be induced in its aggregation) is implied by the flow structure of the aggregated function. When attempting to capture and encode function, as Stone et al. set out to do, it seems obvious that one should concentrate on function. The above results indicate that function is more difficult for reverse engineers to agree upon than flows are and that, as function-structures are aggregated from their low-level, reverse engineering encoding, function is difficult to combine. Focusing more directly on the flow component of function-structures, a final study seeks to improve upon a glaring weakness in flow representation within the functional basis: ‘solid objects’. Solids represent about 70% of the material flows in the reverse engineering database, but the functional basis cannot easily differentiate one solid object from another. Within artifacts, solids can pass through (in which case they are handled by the artifact) or they can be modified by the artifact (e.g., the potato separated into a peeled potato and peel in Figure 2). It makes sense that solid objects of similar size and shape would be handled by similar means. It also makes sense that these objects would respond physically to this handling in similar ways. Verma et al. add three parameters to each solid: size (calculated as overall volume); shape (based on group technology shape complexity classifications of revolute, prismatic, or thin-wall artifacts [Swift and Booker, 1997]); and mechanical model (based on the mechanical modeling types of shells, plates, beams, solids, etc). Figure 6 shows retrieval results in the IF-aggregated function database for various functions based on different encodings of their input-output material flows. While it shows no clear best additional parameter for augmenting solid objects, it does show much better precision and recall compared to unaugmented solid flows. Summarizing the computation of function within the Pahl and Beitz functionstructure framework:

Computational Representations of Function in Engineering Design

555

Comparison of different solid classifications

Average Precision

1

0.8

functional basis

0.6

shape size 0.4

mechanics-shape

0.2

0 0

0.2

0.4

0.6

0.8

1

Average Recall

Figure 6. Results of the queries done for solid-material flows from the case-base. The averages of precision and recall over the different classifications are presented. 1. Function-structures provide a flexible formalism that spans abstraction levels from just below customer needs to just above artifact structure. 2. Computing over function-structures is difficult; their flexibility allows for the expression of any function but does not readily support physical modeling of the system. A controlled vocabulary for expressing function-structures aids in computation. 3. Reverse engineering artifacts to create function-structures is most consistent when performed at the ‘parts’ level of abstraction. 4. Parts-level function-structures, regardless of the vocabulary used to encode them, are difficult to retrieve at more abstract design stages. 5. Retrieval studies varying controlled vocabulary, functional abstraction, function aggregation, and flow representation all reinforce the importance of ‘flow’ as opposed to ‘function’ in computing function.

3.3

Mapping function to structure: concrete

Function-structures are used to move from abstract, socioeconomic notions of intended behavior to representations tied closely to physical behavior. Reverse engineering provides a base of decomposed ‘behavior-in-context’ - a set of building

556

William H. Wood

blocks that leads directly to parts whose physical behavior can be predicted. But the case-based design paradigm suggested by Gero et al. and assumed in the above discussion of case retrieval offers only the weakest form of models — abstractions of parts that have been made and combined in the past. As functional computing moves more closely to the actual physical behavior of parts implied by the design cases, it must adopt representations that afford stronger forms of reasoning — physical first principles. Because physical first principles are most useful for simulating the behavior of a fully specified design, representations must bridge between the flows of functionstructures and the components with their behavioral models. This bridge is constructed as part of the reverse engineering process. IF aggregation first ‘collapses’ long chains over which the sub-type of a flow does not change (e.g., rotational energy changing magnitude, direction, and/or position in a design). The remaining ‘core’ structure of the device then contains transformations across flow type (e.g., energy to information) or flow sub-type (e.g., electrical energy to mechanical energy). Such transformations are typically accomplished by physical phenomena (In this context, a phenomenon is an elemental physical effect that changes energy domains, for example the generation of voltage in response to light in a photovoltaic cell). The abstraction process continues by collapsing the remaining transformations until only black box level input and output flows (and their structure) remain. At each point in the process the abstract design is entered into the database as a structure of input/output flow transformations. In a similar way, Sycara et al. [1991] ‘chunk’ portions of design solutions generated in CADET as qualitative relationships over a canonical set of physical flow parameters, reducing design effort by compiling common subsystems as encapsulated behavior. Queries over this database return partial flow structures that can be expanded into more complete designs. The first mode of expansion identifies the physical phenomena used to bridge flow types specified as system inputs/outputs. By introducing flow types not present in the design specification, the choice of phenomena represents significant design decisions not implied directly by design specifications. The reverse engineering case base can be an effective tool for managing the generation of designs: focusing on flow transformations useful in the past produces good starting points for further design work. Without an experiential base to use for guidance, we could interpose an infinite number of physical phenomena between input and output flow. The lack of modeling capability in function-structures generates candidate design topologies without regard to the magnitudes of the flows entering or leaving the phenomena. To complete these designs, single-domain components can be introduced into flows to change the magnitude, direction, flow subclass, or position of the required flows to match those generated or consumed by the designated phenomena. Typical mechanical domain components include gear pairs, v-belts, lead screws, linkages, etc. Kota and Chiou [1992; 1999] propose a method for mechanical mechanism design in which each component is represented by a matrix mapping Cartesian translational and rotational input/output flow transformations for a li-

Computational Representations of Function in Engineering Design

557

brary of components. Secondary information like reversibility, linearity, and gain ranges accompany the basic transformation matrix for basic behavioral modeling of the resulting system. Within a similar Cartesian representation, Chakrabarti and Bligh [1994; 1996a; 1996b] add spatial information like axis offsets to the mechanical flow behavior modeling. CADET’s qualitative physics models also operate over Cartesian variables. Table 1 shows representations for a catalog of mechanical components that transform rotation into translation, changing the sub-type of mechanical energy (this is a subset of about 40 components identified in the mechanical energy domain). In this case, a phenomenon transforming electrical energy into mechanical energy (i.e., an electric motor) has been identified as a possible design solution for a force feedback computer mouse (additional linear motor phenomenon is generated as well, but the case base ‘prefers’ rotational motors). The rotational energy must be transformed into translational energy along the two axes of motion for the mouse. The three representations shown each help compose components and simulate system behavior mechanically, spatially, and/or qualitatively. Stringing together multiple components (rotational-rotational before these or translational-translational after these) can repair inadequate system force or displacement behavior. Table 1. Configuration Design Building Blocks Functional Unit

Key

Kota et al. In: [Tx

T y Tz

Rx

Ry

Rz

Out: [Tx

T y Tz

Rx R y

Rz

C: Crank-Slider

lin rev

0 0 1 0 0]

1 0 0 0 0]

]

§ ª0 0 0º · ¨ « »¸ ¨1 0 1 «0 0 0 » ¸ ¨ «¬1 1 0»¼ ¸¹ ©

In: [0

0 0 1 0 0]

Out: [1

0 0 0 0 0]

§ ª0 0 0º · ¨ « »¸ ¨1 1 0 «0 0 0» ¸ ¨ «¬0 0 1 ¼» ¸¹ ©

In: [0

0 0 1 0 0]

Out: [0

1 0 0 0 0]

C:

]

[I / O ])

Out: [0

C:

Rack&Pinion

(cont

In: [0

C:

Lead Screw

Charkrabarti Bligh

§ ª0 0 0º · ¨ « »¸ ¨1 1 1 «0 0 0» ¸ ¨ ¸ ¬«0 0 1¼» ¹ ©

Axes z x y

Rotation Translation Inline Offset

&

CADET A = B+: A is an increasing function of B

Ty = Rx+

Tx = Rx+

Ty = Rx+

558

William H. Wood

Bond graphs offer an alternative to these representations for modeling mechanical energy. Using an extension of electrical circuits, bond graphs model generalized energy flow using mechanical/fluid analogs to resistance, capacitance, and inductance as well as motors, and transformers. They connect these basic elements with arcs representing ‘current’ and/or ‘voltage’, producing a cross-domain network whose behavior is readily simulated. Hoover and Rinderle [Hoover and Rinderle, 1989] use a grammar to string together bond graph structures, testing the results against desired behavior. Limiting representation to energy flows allows bond graphs to model system transient behavior, a capability unmatched by function-structures. The price paid for this power is a difficulty matching the ‘intent’ part of functionality. To address this, Ulrich and Seering [Ulrich and Seering, 1989] merge bond graphs with function structures to develop novel sensors and actuators. Although bond graphs fall firmly on the side of calculating behavior rather than representing function, they do provide useful insight into energy flows: dividing energetic variables into effort (voltage, force, pressure) and flow (current, velocity, fluid flow) categories is adopted by the functional basis representation. In general, function-structures provide better models for systems with ‘flow’ types of energy because the direction of this flow can be predicted; ‘effort’ often acts in parallel across several subfunctions making it more difficult to model as a function-structure flow. The distinction between representations useful for computing function as intended behavior and those that compute behavior from structure is an important one in design. Using as building blocks single-domain components, such as those in Table 1, produces topologies of parameterized components whose models can be composed to produce overall system-behavior models. Final design behavior requires additional parametric design to select the correct set of actual components (a process of optimization) responding to the intended behavior (expressed as an objective function). Composing components and evaluating their ideal behavior is only a step on the path to a design; actual behavior depends strongly on how these functional units are connected to each other and/or to a common reference.

3.4 Mapping concrete function to design details In addition to intended behavior from the function-structure standpoint, additional notions of function are found throughout the product life cycle. The artifact must be manufactured; customer value in manufacturing generally means prescribing shapes and materials that are both easy to produce and easy to assemble. The artifact must be distributed; customer value here means that it be easy to ship, warehouse, and display. The artifact must be maintained and repaired; customer value entails fewer, simpler maintenance procedures as well as easy diagnosis and repair of problems. Finally the artifact must be retired; customer value means long service life (perhaps with planned upgrades) and low environmental impact (reducing toxic materials, increasing recycling content, making materials easy to separate, etc.). Most of these aspects of functionality are not captured at

Computational Representations of Function in Engineering Design

559

the function-structure stage. Most are not modeled using first principles. Most require structure in the form of final part geometry, material, and assembly sequence in order to predict behavior from structure. For many of the life cycle customers ‘parts are parts’ — end user function plays little role in performance. We will focus on manufacturing as an example of a life-cycle ‘customer’. Functionality in manufacturing is generic: parts must be formed from raw materials and they must be assembled into a final artifact. The basic processes for shaping parts are deformation, solidification, and material removal; these processes each imply additional functionality; for example in solidification processes material is molten, introduced into a mold, solidifies, is removed from the mold, and cleaned and processed into a final part. Part geometry must be designed so that each of these steps is accomplished effectively. Once parts are formed, they are assembled into the final artifact; generic assembly processes identify a part, grasp it, orient it, insert it, and fasten it to the artifact. Again, part geometry plays a significant role in the ease with which each step is accomplished. Design details drive manufacturing functionality; representation of these details is typically in the computational geometry underlying CAD software. Just as with text at the highest level of abstraction, computational geometry can represent virtually any shape that can be imagined; not all such shapes can be manufactured. Thus, the generation step must be biased toward the creation of manufacturable shape by operating over a representation that embodies real-world experience. Again, reverse engineering can provide this experience in the form of part geometry, processing, and assembly information. Design details serve (in our current discussion) two masters: system function and manufacturing function. The input toward generating design details is a set of components (including partial geometry) and their topology. From the system function point of view, the input to detail design is a topology of components (including partial geometry) as well as constraints on motion relative to each other and to a (typically) undefined mechanical ‘ground’. From the manufacturing standpoint, each of these components must be formed and then assembled into the artifact. Roth [1987] has catalogued connections among mechanical components by modeling ‘contact’ between them. Contact matrices similar to the transformation matrices proposed by Kota et al. define the behavior of a given geometry: three rows correspond to the x, yand z axes, four columns denote overall freedom of motion (by a binary entry: 1 for fixed, 0 for free) in positive and negative translation and positive and negative rotation for each axis. At a finer level of detail, contact type matrices replace the generic free/fixed entries with more specific codes for the type of freedom/fixture: form, gravity, friction, same part, stiffnEss, etc. Just as at higher levels of abstraction, representations for generating designs (contact) are more abstract then representations for modeling their behavior (contact type). Roth’s representations help generate the most detailed level of function by providing a mapping from motion to shape. Again, there is no general theory for performing this mapping, so reverse engineered parts are cataloged by contact

560

William H. Wood

matrix, contact types defined for each interface they support or motion they allow (e.g., flexure joints within a single part). These interfaces and functions then play an important role in realizing the functionality of mechanisms specified at higher levels of abstraction by providing kinematic joints both within and between components. Because manufacturing function is generic, behavior is captured in terms of the forming process (e.g., casting/molding, forging, machining), primary and secondary process axes (e.g., mold separation axis, fixturing plane for machining,), number of toleranced features [Fagade and Kazmer, 1999], and primary and secondary assembly axes. All spatial information is defined relative to the functional axes for the contact matrix to assure portable part orientation. Table 2 shows a typical catalog representation for a single degree of freedom rotational joint such as that required in the crank slider mechanism from Table 1. Table 2. Detail Design Building Blocks g

g

Single DOF Joint Connections ª1 1 0 0º

Contact Matrix = ««1 1 1 1»» «¬1 1 1 1»¼

Contact-Type ªs «s « «¬ s

s

E

s

s

s

s

ªf «f « ¬« f

f

r

f

f

f

f

ªf «f « «¬ f

f

r

f

f

f

f

Form

§ ªx + ¨ « ¨ Legend : « y + ¨ «¬ z + ©

# Parts

Eº s »» s »¼

1

rº f »» f ¼»

2

rº f »» f »¼

5

x−

xx +

y− z−

yy + zz +

xx − º · ¸ yy −»» ¸ ¸ zz − »¼ ¹

Process Axis Connect Ass’y Axis Part / Ground

Manufacturing and assembly behavior (i.e., part forming complexity and ease of assembly) can then be ‘simulated’ without knowledge of final part geometry through a simple process of counting the number of parts required, the number of features for each part, and the number of primary and secondary processing and assembly axes. Detail design entries are cataloged by contact matrix but each has its own contact type, part count, part forming axis, and assembly axis. The contact matrix interfaces with component function, the contact type contains

Computational Representations of Function in Engineering Design

561

both detailed physical and manufacturing behavior of each reverse engineered joint within the system. Detail design feeds physical behavior back up to inform the ideal components models with real behavior. For example, the top entry in Table 2 uses a single part for the joint in the crank-slider. This flexure brings with it a spring constant — a force feedback mouse with a springy crank slider would naturally self-center, an undesirable physical behavior that is a side effect of good manufacturing behavior. All other options produce varying degrees of friction or play in the joint, causing a drag on the motion of the mouse or backlash in its motion. 4

SUMMARY

We have traced the representation of function from its origins in customer needs, to the black box of input/output flows that these needs imply, to various levels of abstraction in function-structures, to the identification of physical phenomena for cross-domain transformations, to first-principle generative models for intra-domain flow transformations, to part forming and assembly models for detail design. Representations within each stage exist at multiple abstractions: an abstract representation responds to intended behavior by generating possible structure, generally through a mapping based on experience; each possible structure then generates abstractions of design behavior so that mismatches between intent and behavior can be identified and potentially rectified. Representations at different stages offer different tradeoffs between the ability to generate designs and the ability to model the designs generated. At the most abstract levels, modeling can only verify proper inputs and outputs — actual system behavior is hidden in the semantic ‘meanings’ of the functions. Because modeling relates only to flows at this stage — what is the use of capturing ‘function’ ? Furthermore, the main methods for capturing function do so at such a detailed level that overall system function (the forest) is hidden by the detailed functions required to accomplish it (the trees). It seems rather odd that, even at this highest level of abstraction it is difficult to encode ‘function’ in the language of the designer and difficult to apply ‘function’ as inferred by the reverse engineer in a rigorous way. At lower levels of abstraction, ‘function’ is almost completely lost as representations apply more formal representations modeling the transformation of input flow to output flow. The question then begs — is function as intended behavior useful at any level or does intended behavior just pose goals to be met through search and verified through structural simulation? 5

CONCLUSIONS

Gero defines function as intended behavior and proposes a design process in which experience is used to: identify behavior that can provide value within a specified

562

William H. Wood

socioeconomic context, compose systems that promise to deliver such behavior, and, from the structure of these systems, predict which aspects of system behavior are contrary to providing value. The underlying notion is that structure is the basic level of abstraction and that behavior can be predicted from structure. Some have noted inconsistency in the description of Gero’s function-behaviorstructure design model [Dorst and Vermaas, 2005], finding a blurred distinction between behavioral and structural representations. One of the motivations for producing computational models of the design process is that the rigid representations and formal manipulations required for computation often elucidate the vagaries of a very human process. In the case of design, the human process entails a natural progression in which intended abstract behavior leads to abstract structure. This structure yields additional insights into behavior that help identify candidates for further design refinement at lower levels of abstraction — a continued evolution of function, structure, and behavior at each level of abstraction. Representations within engineering design use abstraction to hide a progression of need-function-structure-behavior chains. The specific representations required of a design are highly problem/domain/solution dependent — there can be no generic design representation because representational bias is required [Mitchell, 1997] to distill experience and reverse its mapping back onto structure. Search in computational design is defined by two measures: completeness — the ability to generate all possible design solutions; and soundness — the ability to avoid generating designs that are not solutions to the given situation. Only in pure mathematical worlds can one have both — in design we must trade between completeness and soundness. The recursive generate-test design process is a fitting one for computational search: favor completeness by initially searching in an abstract design space so that as many designs as possible are generated and mapped onto less abstract solutions, ensure a measure of soundness by ‘simulating’ these solutions and discarding infeasible or ineffective designs. This is a case of good news/bad news for computational design. The good news is that we can identify a general process for design computation and take steps toward realizing it in circumscribed domains at defined levels of abstraction. In so doing, we can hope to improve design performance by affording search over far larger sets of solutions than can be explored by a human designer. The bad news is that the representations required for this process, at least in the intermediate levels, are domain specific. In addition, the knowledge-level ‘fuel’ of design comes not from well-defined theories but from experience gained by evaluating ‘real-world’ simulations of design solutions.

BIBLIOGRAPHY [Chakrabarti and Bligh, 1994] A. Chakrabarti and T. P. Bligh. An approach to functional synthesis of solutions in mechanical conceptual design. Part i: Introduction and knowledge representation. Research in Engineering Design, 6, 127-141, 1994. [Chakrabarti and Bligh, 1996a] A. Chakrabarti and T. P. Bligh. Approach to functional syn-

Computational Representations of Function in Engineering Design

563

thesis of solutions in mechanical conceptual design. Part ii : Kind synthesis. Research in Engineering Design, 8, 52 - 62, 1996. [Chakrabarti and Bligh, 1996b] A. Chakrabarti and T. P. Bligh. Approach to functional synthesis of solutions in mechanical conceptual design. Part iii : Spatial configuration. Research in Engineering Design, 8, 116 - 124, 1996. [Chiou and Kota, 1999] S.-J. Chiou and S. Kota. Automated conceptual design of mechanisms. Mechanism and Machine Theory, 34, 467-495, 1999. [Dorst and Vermaas, 2005] K. Dorst and P. Vermaas. John gero’s function-behavior-structure model of designing: A critical analysis. Research in Engienering Design, 16, 17-26, 2005. [Fagade and Kazmer, 1999] A. Fagade and D. Kazmer. Optimal component consolidation in molded product design: 1999 ASME Design for Manufacture Conference, DETC1999/DFM8921, ASME, 1999 [Gero, 1990] J. Gero. Design prototypes: A knowledge representation schema for design. AI Magazine, 11, 26-36, 1990. [Geor and Kannengiesser, 2002] J. Gero and U. Kannengiesser. The situated function-behaviorstructure framework. In Artificial intelligence in design ’02., Gero, J., 89-104, Kluwer, 2002. [Gibson, 1979] J. Gibson. The theory of affordances. In The ecological approach to visual perception., Houghton Mifflin., 1979. [Gietka et al., 2002] P. Gietka, M. Verma, and W. H. Wood. Functional modeling, reverse engineering, and design reuse: 14th International Conference on Design Theory and Methodology, DETC2002/DTM-34019ASME, 2002 [Hauser and Clausing, 1988] J. R. Hauser and D. Clausing. The house of quality. Harvard Business Review, May-June, 63-73, 1988. [Hoover and Rinderle, 1989] S. P. Hoover and J. R. Rinderle. A synthesis strategy for mechanical devices. Research in Engineering Design, 1, 87-103, 1989. [Kota and Chiou, 1992] S. Kota and S.-J. Chiou. Conceptual design of mechanisms based on computational synthesis and simulation of kinematic building blocks. Research in Engineering Design, 4, 75-87, 1992. [Kroes, 2002] P. Kroes. Design methodology and the nature of technical artefacts. Design Studies, 23, 287-303, 2002. [Kurfman et al., 2003] M. A. Kurfman, M. E. Stock, R. B. Stone, J. Rajan, and K. L. Wood. Experimental studies assessing the repeatability of a functional modeling derivation method. Journal of Mechanical Design, 125, 682-693, 2003. [Maier and Fadel, 2002] J. Maier and G. Fadel. Comparing function and affordance as bases for design: 14 Annual International Conference on Design Theory and Methodology, DETC2002/DTM-34029.ASME, 2002 [McAdams et al., 1999] D. McAdams, R. Stone, and K. Wood. Functional interdependence and product similarity based on customer needs. Research in Engienering Design, 11, 1999. [McAdams and Wood, 2002] D. McAdams and K. Wood. A quantitative similarity metric for design-by-analogy. Journal of Mechanical Design, 124, 173-182, 2002. [Mitchell, 1997] T. Mitchell. Machine learning. McGraw-Hill, 1997. [Otto and Wood, 1998] K. N. Otto and K. L. Wood. Product evolution : A reverse engineering and redesign methodology. Research in Engineering Design, 10, 226-243, 1998. [Pahl and Beitz, 1988] G. Pahl and W. Beitz. Engineering design- a systematic approach. Springer-Verlag, 1988. [Pugh, 1991] S. Pugh. Total design: Integrated methods for successful product engineering. Addison-Wesley, 1991. [Rosenman and Gero, 1998] M. Rosenman and J. Gero. Purpose and function in design: From the socio-cultural to the techno-physical. Design Studies, 19, 161-186, 1998. [Roth, 1987] K. Roth. Design models and design catalogs. International Conference on Engineering Design (ICED’87), 60-66, 1987. [Stahovich and Kara, 2001] T. F. Stahovich and L.B. Kara. A representation for comparing simulations and computing the purpose of geometric features,. AI EDAM, 15, 189-201, 2001. [Stone and Wood, 2000] R. B. Stone and K. L. Wood. Development of functional basis of design. Journal of Mechanical Design, 122, 359-370, 2000. [Stone et al., 1999] R. B. Stone, K. L. Wood, and R. H. Crawford. A heuristic method for identifying modules for product architectures. Design Studies, 21, 5-31, 1999. [Swift and Booker, 1997] K. G. Swift and J. D. Booker. Process selection from design to manufacture. Arnold, 1997.

564

William H. Wood

[Sycara et al., 1991] K. Sycara, D. Navinchandra, et al. Cadet: A case-based synthesis tool for engineering design. International Journal of Expert Systems, 4, 157-188, 1991. [Ulrich and Seering, 1989] K. T. Ulrich and W. P. Seering. Synthesis of schematic descriptions in mechanical design. Research in Engineering Design, 1, 3-18, 1989. [Verma and Wood, 2003] M. Verma and W. Wood. Functional modeling: Toward a common language for design and reverse engineering: 2003 ASME International Conference on Design Theory and Methodology, DETC2003/DTM-48660, ASME, 2003 [Verma and Wood, to appear] M. Verma and W. Wood. Toward case-based functional design: Matching reverse engineering practice with the design process. to appear Design Studies, 2006. [Whitney, 1996] D. E. Whitney. Why mechanical design cannot be like vlsi design, 1996.

RATIONALITY IN DESIGN Peter Kroes, Maarten Franssen and Louis Bucciarelli

1

INTRODUCTION

Engineering design has many faces, ranging from the almost exclusively functionally oriented design of new materials and technical devices in fields like mechanical and electrical engineering, to strongly aesthetically oriented design projects that may be encountered in industrial and architectural design.1 Likewise, rationality has many faces, some of which are oriented to specific fields or disciplines, such as economic, scientific and technological rationality, whereas others refer to general conceptions of rationality, such as theoretical and practical rationality. On the one hand, this state of affairs makes a discussion of rationality in engineering design not an easy matter; on the other hand it may explain why there are so many diverging opinions on the role of rationality in engineering design. Some authors take rationality to be of paramount importance for engineering design; it is a basic assumption of one of the major paradigms in design methodology that engineering-design practice is or ought to be a form of rational problem solving [Simon, 1969/1996; Dorst, 1997]. But there is also widespread recognition of the importance for innovative designs of artistic and/or creative aspects that are assumed to be not susceptible to rational problem-solving methods. Our aim is to present a systematic overview of the different kinds of rationality issues that may come up with regard to engineering-design practice. The main question we explore is in what sense and to what extent a design process can be considered a rational process. It is a premise of much work done in the field of design methodology and engineering design itself that rationality plays a significant role in design processes, not only at the level of the organization of design processes, but also at the level of the design of products. The underlying idea is that many of the decisions that are made regarding design — regardless of whether they concern the set-up and execution of the design process or the object of design itself — can be justified on the basis of reasons (arguments). These reasons are supposed to show that, given the aim of the design process and given the various possible courses of action, a specific course of action is to be preferred above all possible other ones because it leads more directly or effectively to the aim pursued. 1 In this paper we will concentrate on the design of material technical artefacts; we see, however, no reason why the results of our analysis would not be applicable to the design of technical processes and services involving technical artefacts.

Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

566

Peter Kroes, Maarten Franssen and Louis Bucciarelli

This course of action is, then, the most ‘rational’ way to proceed,2 and a ‘rational’ engineer is supposed to choose that action. However, as we will show, engineering practice is more complicated than this simple picture suggests.

2

ENGINEERING DESIGN

Engineering design, as defined by Accreditation Board for Engineering and Technology (ABET), “is the process of devising a system, component, or process to meet desired needs. It is a decision-making process (often iterative), in which the basic science and mathematics and engineering sciences are applied to convert resources optimally to meet a stated objective.” 3 The ‘desired needs’ referred to in this quotation are usually called customer needs. In a process of translation, specification and reduction, these needs are transformed into functional requirements and these again into design specifications.4 The list of design specifications – usually a list of physical parameters — may be taken to be the stated objective of an engineering-design project. If the designed object meets all the specifications, it is supposed to be able to perform the desired function. Whether that is indeed the case depends on whether the list of design specifications adequately captures the functional requirements. If, moreover, there is a good match between the customer needs and the list of functional requirements, the designed object will contribute to satisfying the customer needs. The above definition characterizes the things that are designed by engineers very broadly as “systems, components, or processes”. Indeed such things range from mass-produced computers to unique oil platforms, from telephones to high rise buildings, from components to complex systems, from micro-organisms to software, and so forth. Correspondingly there is also much variety in engineeringdesign practices. In some practices the design phase includes the actual making and testing of prototypes of the designed object, in others, the making of a specimen of the designed object is not considered part of the design phase. In some, aesthetic criteria are of paramount importance, in others not. Some design projects may be performed by a single designer, others require large, multidisciplinary teams of design engineers. There is also much variety in the kind of design problems to be solved. Vincenti [1990, p. 60], for example, distinguishes between normal and radical design and between design tasks high and low in the design hierarchy. In spite of this variety, a number of different kinds of activities may be distinguished in engineering-design processes, namely 2 The measure of rationality of an action may come in degrees; a person or action may be rational to a certain extent. Whenever in the following we will refer to persons/actions as being rational or not, this has to be interpreted as including all cases where they are considered rational to a certain degree. 3 http://www.me.unlv.edu/Undergraduate/coursenotes/meg497/ABETdefinition.htm; accessed September 28, 2006. 4 See the contribution by De Vries in this Volume.

Rationality in Design

567

• Translation of customer needs into functional requirements and these again into design specifications; • Generation of a range of viable options; • Selection of one of the options; and • Justification of the final design proposal. These activities are not necessarily linearly ordered in time; design processes may contain many iterations and feedback loops between these different kinds of activities. A more careful look at the outcome of design processes shows that it is not simply a (description of a) technical artifact, but that it also comprises the manual, that is, instructions that tell the user how to handle the technical artifact such that it will perform its technical function correctly. Both the manual and the technical artifact can be seen as elements of a use plan. A use plan is a considered series of actions to achieve a certain end with the help of a technical artifact necessary for executing the use plan (for more details about use plans, see [Houkes et al., 2002]). The designing of technical artifacts by engineers can therefore be seen as linked to the specification, or even design, of use plans. If attention is restricted to the object to be designed, the engineering-design process starts with a phase in which a functional description of this object is ‘translated’ into a structural description.5 A purely functional description of an object ‘black-boxes’ its internal structure; it is oriented towards the environment of the object and describes it in terms of desired input-output relations. In contrast, a structural description concentrates on the physical characteristics of the object. This may take two forms. The outcome of the design process is usually a description of the object, the blueprint of the artifact, which specifies everything that must be known in order to manufacture it. This description is often limited to the specification of materials to be used and to geometrical properties of the various components. Implicit in this description is a structural description of the artifact in another sense, which contains all of its physical and chemical properties that allow the designing engineers to predict how the artifact will behave under various input conditions. How much of the content of the black box, i.e. which part of the many physical characteristics of the object under design, is in actual design practice already fixed at the beginning of the design process depends strongly on the nature (radical or normal) of the design task. Finally, as the ABET definition describes it, engineering design is a decisionmaking process. This involves decision making on different levels, at different stages and about different kinds of issues. Decisions have to be made on setting the objective of the design process, on redefining the objective when the original objective turns out not to be feasible, on how to spend the available resources, on how many options are to be pursued, on which options to drop because they are 5 See

the contribution by Kroes in this Volume; see also [Kroes, 2002].

568

Peter Kroes, Maarten Franssen and Louis Bucciarelli

too problematic, on which options to develop further because they are promising, on what criteria of evaluation to use for making the final choice, on which option to choose, and so forth. These different kinds of often ill-defined and interconnected decision situations, together with unpredictable changes in the context of the design process that may affect resources and time schedules, make engineeringdesign practices often messy and unruly. What room for rational decision making is there in such practices?

3 APPLYING STANDARDS OF RATIONALITY TO ENGINEERING DESIGN Engineers and design methodologists are interested in studying rationality in design processes primarily because they hope and expect that by explicating and articulating the various roles of rationality and by applying rationality in systematic procedures for design, they will be able to improve design practice.6 The intuitive notion of rationality underlying this line of thought is related to the idea that there are better and worse ways — relatively systematic and relatively chaotic ways — of solving engineering-design problems and of making engineering-design decisions. According to such an interpretation of rationality, rationalizing design practice will enhance the likelihood that the design process terminates in a successful design. With this rough idea of rationality in mind, let us start by asking how the rationality of the way actual design processes are performed or set up may be analyzed. On this issue it may be helpful to borrow the notion of rational reconstruction from the philosophy of science. In discussions about rationality in science it has become more or less standard practice to distinguish between rational reconstructions of science and actual scientific practices. This distinction goes back to Carnap. He describes a rational reconstruction of the formation of a concept as “a schematized description of an imaginary procedure, consisting of rationally prescribed steps, which would lead to essentially the same results as the actual psychological process” [Carnap, 1963, p. 16]. A rational reconstruction of the formation of a theory in science, for instance, is a rationally prescribed series of steps, which may involve performing experiments and choices among sets of alternative theories, that leads ‘logically’ to the same end result that the historical process ended in. These rationally prescribed steps are often not clearly identifiable in the actual actions and decisions of the scientists who have contributed to the discovery of the theory. This does not make their actions and actual reasoning processes by definition irrational. On this point we have to take account of the difference between rational reasoning and rational behavior (for more details, see Subsection 4.3). Rational reconstruction concerns the latter: a rational reconstruction describes the most rational series of steps leading to the given end result (as the goal of the process). 6 For the alleged advantages of a systematic approach to engineering design, see for instance [Pahl and Beitz, 1996, pp. 11 and 499-500].

Rationality in Design

569

Whether the actions and actual reasoning of the scientists involved are rational or not is another matter, to be decided on the basis of the norms of theoretical and practical reasoning. The rationality of the actual scientific process, as a series of steps leading up to the final theory, may be assessed by comparing these steps with the rationally reconstructed steps. In a similar vein rational reconstructions may be used for analyzing rationality in design processes. The design steps and decisions leading up to the final design in an actual design process may be very different from the steps and decisions prescribed by its rational reconstruction. Again, this does not imply that the designers involved in the real design process were acting or reasoning irrationally. In fact it may occur that they have to evaluate some of the intermediate steps and decisions they take without having a very clear idea of the end solution of the design problem. So for them it may have been rational to introduce steps and make decisions that at the end of the process, in the light of the solution they came up with, turned out to be wrong or superfluous and will therefore not figure in the rational reconstruction of the process. Starting from the end result of a design process, a rational reconstruction of that design process filters out as much as possible its ‘irrelevant’ elements and attempts to reconstruct how that design process would have proceeded if all steps and decisions had been taken from the point of view of how they lead up to the final outcome. An assumption underlying rational reconstructions of design processes is that various methods for solving design problems may be compared in terms of their success; some methods are better than others and there may even be a ‘best’ way for solving design problems. This best way would be a or the rationally reconstructed way, satisfying the prescripts or norms of rational problem solving. A basic question concerning rationality in engineering design is whether these rational reconstructions of engineering-design practice make sense at all. If so, then immediately the question pops up of how they relate to actual design processes. Advocates of rational reconstructions take a normative stance with regard to actual design practice and accordingly tend to interpret theories of rationality in a normative way. They come up with prescriptive schemes of how design problems ought to be solved, following the canons of theories of rational problem solving. Those who reject rational reconstructions usually take a descriptive stance toward design practice, and interpret theories of rationality in a descriptive way: starting from a descriptive stance towards theories of rational reasoning or behavior they analyze and describe what kind of behavior is considered to be rational by people involved in design practice. They may come up with detailed analyses of how designers actually solve design problems, of how they establish whether a proposed solution is a good solution or not and even of what designers consider to be rational problem solving methods, but they deny that there is any fixed Archimedean point, that is, a theory of rational problem solving, from which this practice may be critically assessed and consequently improved. So the issue of whether rational reconstructions of engineering-design processes make sense is tied in with the issue of whether theories of rationality are normative or descriptive.

570

Peter Kroes, Maarten Franssen and Louis Bucciarelli

A problem that is key to all discussions about rationality in design concerns the criteria for judging the overall success of the outcome of an engineering-design process. What are these criteria? And to what extent is it possible to operationalize these criteria in objective ways? The intuitive notion of rationality used so far is strongly instrumental in nature: it is about the effective and efficient realization of goals (see Section 5 for a more extensive discussion). A precondition for the instrumentally rational assessment of actions is the availability of clearly defined goals, so that the outcomes of actions may be evaluated with respect to these goals. Otherwise it will not be possible to establish unambiguously whether an action has contributed effectively to the realization of these goals.7 In this context, a distinction is often made between well-defined and ill-defined, or well-structured and ill-structured problems. For ill-defined problems it is unclear what has to be achieved, and what determines whether a particular action constitutes a solution to the problem. In practice, ill-structured problems abound; it is one thing to demand, for instance, that a design solution should be a market success, innovative or beautiful, it is another thing to come up with objective (inter-subjective) scores for market success, innovativeness and beauty.8 In analyzing the goal of design processes, various levels may have to be distinguished. Within commercial firms, design processes are embedded within broader processes that aim at profit and continuity of the firm. At the firm level, therefore, the success of design processes will be primarily judged by the commercial success of the designed product. Although design engineers might all agree on this very general goal, it is not the kind of goal they have to deal with in their daily design activities. At the level of the design project, their goal usually is to come up with a technical artifact that meets a list of functional requirements, that is, a list of criteria that a design should meet. In design projects with participants from different disciplines, with various interests in the design at hand, the explicit and detailed statement of a single common goal, beyond meeting the functional requirements, may be difficult because of conflicting interests but seems also not necessary since the list of requirements may act as a common cornerstone for evaluating the success of a design.9 The list of functional requirements, again, has to be transformed into a list of design specifications, up to the specifications for detailed components. The latter act as (sub)goals for detailed design activities. At this level, termed the level of object worlds by Bucciarelli [1996], (sub)goals may conflict, in fact, usually do. In those situations, trade-offs between conflicting 7 Cf. Nozick, in discussing the application of expected-utility theory: “[A] goal is an evaluative criterion such that, given any outcome, there is a clear, determinate, and feasibly reached answer to the question of whether that outcome satisfies that criterion. Yet in life it may be unclear whether and to what extent an outcome satisfies a goal” [Nozick, 1993, p. 166]. 8 As long as designing is done by one single individual who is also going to judge the outcome of his/her design, there seems to be no need for insisting on objective (intersubjective) scoring methods. However, as soon as more people are involved in the design process and in judging its outcome, such methods become indispensable for reaching consensus. 9 Of course, in the process of establishing the list of specifications, these different underlying goals, often conflicting, may show up and may make the reaching of an agreement on this list a difficult matter.

Rationality in Design

571

specifications will have to be made. This makes judging the success of a design a particularly complicated matter. Contrary to what may have been suggested above, even a full list of functional requirements does not always furnish a clear-cut criterion for judging the success of a design. This is so for two reasons: first, there are important differences among the various engineering-design traditions in the kind of requirements they have to deal with, and second, it is implicitly assumed that the list of requirements is given and fixed at the beginning of design processes. Regarding the first point, some engineering-design traditions, for instance the design of electrical components, are dominated by lists of requirements for the objects to be designed that focus on form, fit and function, on codes and standards that have to be complied with, on costs etc. For many, if not most of these requirements, standardized tests are available, in terms of the operationalizations of the specifications that serve to measure quantitatively the performance of proposed solutions. In such engineering-design contexts, reliable and generally accepted (inter-subjectively valid) measures of success are in principle at hand. This may pave the way for rational choices among proposed design solutions. But even in these cases, as we shall see later on in Subsection 6.3, problems may arise when trade-offs have to be made between a device’s performances on different criteria. In other design practices, situations may be much more complicated. Take architectural design, for example, the design of a house. What is the function of a house and how is this function laid down in the list of functional requirements? In contrast to the function of an electrical component, which may be characterized as a technical function, the function of a house has important psychological and social characteristics, which are difficult to describe unambiguously and/or objectively. Of course, the list of requirements for the house will also contain ‘hard’ elements, such as the total floor space in square meters, the cost per square meter and compliance with relevant codes and standards. To make matters even more complicated, aesthetic criteria may play a prominent role in evaluating the various options proposed by the architect. In such situations the prospects for arriving at a rational choice among the alternatives, in particular if a decision has to be reached by several people jointly, seem dim. A significant difference with the design of an electrical component described above is that in this context the question who gets a say in making the final choice becomes of paramount importance. It illustrates that subjective elements play a much more important role in these situations. In Section 5 it will be argued, however, that subjective elements do not stand in the way of a rational choice as long as this choice is made by a single individual; his/her preferences, subjective as they are, are then the only ones that count. But if more individuals with different preferences are involved, a rational choice between the alternatives becomes problematic because different preference orderings have to be aggregated or because some individuals have to review and change their preferences. The second problem with using the list of functional requirements as a standard against which the worth of a design process is to be measured, concerns the fact

572

Peter Kroes, Maarten Franssen and Louis Bucciarelli

that this list often is readjusted during the design process. In those cases, the goal pursued, conceived of as meeting the list of requirements, changes ‘on the fly’. These readjustments may be due to changes in knowledge or to changes in the context of the design process, but they may also be necessary because the design processes started off with an ill-structured problem [Simon, 1984; Nozick, 1993]. In the latter case, one of the tasks to be accomplished during the design process is to turn the original design problem into a well-structured problem. To do this in a rational way is possible only if the theory of rationality extends to the fixation of ends. It has been denied by many, however, that it extends that far.10 Nevertheless, this is what engineers are doing in daily practice all the time and it seems far-fetched to claim that this kind of activity is not amenable to rational evaluation at all.11 Before discussing rationality issues in engineering design in more detail, first an overview of views concerning the nature of rationality will be presented, which will make clear that there is no overarching account of rationality that can be taken for granted.

4 ASPECTS OF RATIONALITY

4.1 Broad and narrow notions of rationality Rationality can be viewed broadly and narrowly. The broad interpretation is the one expressed by the famous dictum of Aristotle that man is a rational animal. By this is meant that humans, in contrast to most animals, do not merely show behavior, that is, externally observable movements of their body, which can be explained completely by reference to the physical make-up of their bodies and the operation of causal laws. Instead, human beings act, that is, their behavior typically is the manifestation of actions that are voluntarily and deliberatively chosen for certain purposes, to realize their ends or satisfy their desires. They do so, moreover, on the basis of a (coherent) set of beliefs about the world, and these beliefs are in their turn formed so as to match their perception and other beliefs already entertained. In this broad conception, rationality is very close to intentionality: to say that people are rational is to say that they can be described intentionally, that is, that beliefs, desires, and (intentions for) actions can be ascribed to them that are largely coherent both in the sense of being coherent 10 It may be the case that turning an ill-structured design problem into a well-structured one does not affect the goal (the list of functional requirements). Even in that case it is debatable whether the decisions made in arriving at a well-structured problem are subject to the canons of instrumental rationality. 11 The fixation of ends may be subject to constraints of various different types and decisions about ends may be ‘rationally’ evaluated in so far as these constraints are taken into account. For instance, the decision to design a perpetuum mobile can hardly be called rational, since it violates generally accepted beliefs about physical constraints. Similarly, the decision to design a car for a price that is, given prevailing social and economic constraints, totally unrealistic is not rational (see also Subsection 4.4 below).

Rationality in Design

573

in themselves (a coherent set of beliefs or of aims) and in the sense in which actions are justifiable by particular beliefs and desires. Vice versa, to be able to ascribe beliefs and desires or aims to people, we must assume that they are broadly rational. This is what Donald Davidson [1984] has summarized in his ‘principle of charity’. In our ordinary dealings with other human beings, rationality in a broad sense is therefore presupposed. In this broad sense, the notion of rationality has both a descriptive and a normative dimension. In form, the broad supposition of rationality of other people is of a descriptive kind, and the expectations that it creates are similarly descriptive, but at the same time, we judge people by the extent to which they deviate from it. We expect and we may expect that someone typically forms a belief about the external world (including the people that inhabit it) not merely on a whim or because it suits him or her, but on the basis of some form of evidence. We expect and we may expect that if someone interferes with the world this typically happens not merely on the basis of an impulse but for at least some reason. The normative dimension gains in importance to the extent that more depends on someone’s rationality, if not for the person in question then at least for the people who are potentially affected by this person’s behavior. Science and technology are domains where much depends on the rationality of the people involved in it. In a more narrow sense, the assessment of two basic aspects of the broad view of rationality is sharpened as much a possible: the formation of beliefs and the choice of actions. Narrow rationality is, therefore, much more clearly than broad rationality, a normative concept. Theories of rationality in the narrow sense aim, foremost, to give an answer to the questions what to believe and what to do, given particular circumstances. When restricted to the question what to believe, the kind of rationality involved is called theoretical rationality, and when restricted to the question what to do, practical rationality. It is an important issue, whether rationality indeed covers these and only these two aspects, and whether even these two different aspects deserve to be distinguished.12 From the picture of rationality as sketched above, it might be thought there is also a form of rationality restricted to the question what to desire or what to aim for. However, this is not the case. There is an important tradition in philosophy that holds that rationality sets at most formal restrictions to what we can desire or aim for, but not substantial ones. We should not desire the impossible, or should desire in such a way that it can indeed be said that we desire something, and what it is that we desire, but apart from this, we need never justify why we desire or aim for the things that we happen to desire or aim for. We will return to this issue in Subsection 4.4. Occasionally the term ‘rational’ is used to characterize something that is not either an action or a belief or a desire. Engineers sometimes speak of a rational 12 Concerning rational action, a distinction between the rationality of intending to act in a certain way and the rationality of performing the corresponding action plays a role in some puzzles (Newcomb’s problem, Kavka’s toxin puzzle; see for instance [Nozick, 1993]) but this distinction does not play a role in engineering design.

574

Peter Kroes, Maarten Franssen and Louis Bucciarelli

artifact or a rational design in the sense of design concept or design solution. Indeed one could choose to define a rational artifact in a derived sense as an artifact resulting from a rationally performed design process. Doing so leads to problems, however. If a particular action — for example, turning left, or buying a ticket — is termed rational, it can only be so in the particular circumstances in which the action is performed. This usually poses no problems since we are rarely presented with actions outside of their context. In engineering, however, being presented with a product out of the context in which it was developed or manufactured is the rule rather than the exception. A particular technical artifact may seem quite fit for a job that a user has in mind for it, but that in itself does not make it a rational product, since for it to be that it should also have been designed for this job. Even if this is the case, however, this is still not enough, because it may be that another artifact would have been even better for this function, and if nothing had prevented the designer from choosing that design option, this makes the present product not a rational one. On the other hand, the existence of uncertainty may make it the case that an artifact that seems to be perfect for its job was just the lucky outcome of a flunked design job, or that an artifact that fails to perform its intended function was the unlucky result of a design process that included nothing but rational decisions. To speak of a rational design, therefore, may more often confuse rather than help our understanding, and it should therefore be discouraged. If, however, ‘design’ refers not to a product but to a plan for realizing a technical artifact with a particular functionality, than a design may be called rational or irrational, since a plan may be taken to be a considered series of actions.

4.2 Theoretical and practical rationality The notion of rationality touches upon almost all aspects of human intentional and intelligent behavior, and it is therefore necessary to introduce some finer distinctions that focus on particular aspects.13 In the previous subsection it was already mentioned that a generally recognized classification distinguishes between theoretical and practical rationality. This distinction reflects a distinction in what rationality is considered to be about. Nozick [1993, p. xii], for example, distinguishes between rationality as a tool for acquiring knowledge and rationality as a tool for improving our actions. Theoretical rationality, then, concerns what beliefs about the world it is rational to entertain, whereas practical rationality concerns what actions are rational to perform. Both forms of rationality are of paramount importance to engineering practice. Since this practice is primarily aimed at action in the sense of changing our physical environment, practical rationality is of direct significance, but theoretical rationality is equally relevant, because engineering without knowledge of the actual state of what is to be changed or of means-end relations is hardly conceivable. Within an engineering context reliable knowledge is a necessary (but not a sufficient) precondition for effective and efficient action. 13 For

a recent handbook on the notion of rationality, see [Mele and Rawling, 2004].

Rationality in Design

575

However, in this chapter, we will restrict ourselves mainly to issues concerning practical rationality.14 The soundness of the distinction between theoretical and practical rationality has been questioned. If we see the outcome of deliberation about what to believe as the adoption of a particular belief, should this not be seen as an action? And is then not all rationality practical rationality? This would require the introduction of desires or goals that direct our adoption of beliefs. This is problematic, however, because we can have many goals, and some goals may well have us adopt false beliefs. The subsumption of theoretical rationality under practical rationality would, presumably, require the adoption of the universal goal of believing all and only those things that are true, and would therefore introduce a special rational aim or desire alongside all other ones. This approach is taken by some [Levi, 1980; Nozick, 1993] but is not standard. This notwithstanding, theoretical rationality and practical rationality are not independent. Part of practical rationality must be the formation of beliefs about the situation one is in, about the options that are open to one, and about the consequences of these options. In contrast, deliberation about what to believe can continue without there being an occasion or a necessity to do something. This distinction reflects the distinction between science and technology. In science, research is usually directed purely to the formation of knowledge, in the form of true or justified beliefs, without a context of action that these beliefs must support. Research is also of paramount importance in technology, but here it is typically directed at precisely the knowledge that is required to support particular actions, i.e., design decisions. In contrast, it has also been argued that practical rationality must be subsumed under theoretical rationality. Some hold that if we deliberate about what to do, deliberation ends in the belief that a particular course of action is the right one to choose. A real distinction between theoretical and practical rationality would require that deliberation about what to do ends in an action rather than in a belief about an action. It is unclear, however, how we can have ‘mixed’ reasoning schemes, where the premises consist of or include theoretical items like beliefs and the conclusion is an item of another kind: either an act or an intention to perform an act.15 The issue of what practical deliberation results in, exactly, cannot be considered solved. In philosophy the basic division of rationality in theoretical and practical is, nevertheless, almost universally adopted.

4.3

Theories of rational reasoning versus rational behavior

Apart from the distinction between theoretical and practical rationality, which address different aspects of human experience, a distinction between two kinds 14 Theoretical rationality has been studied extensively within epistemology and the philosophy of sciences. Within the field of practical rationality, issues concerning moral action are studied in ethics. Here we are not specifically interested in moral action but in action in general. Only recently the analysis of practical reasoning has become a topic of more intensive research; Von Wright [1963] has been very influential; for a survey of various approaches, see [Segerberg, 1992]. 15 This ignores the role of desires and the like as a possible third kind of item.

576

Peter Kroes, Maarten Franssen and Louis Bucciarelli

of theories of rationality has to be made (cf. [Hampton, 1994, p. 197; Nozick, 1993, p. 65]). One kind is about the mental processes leading up to or preceding a belief or an action. These theories of rationality are about human reasoning, be it theoretical or practical; they analyze how human beings reason or ought to reason. They will be referred to as theories of rational reasoning. The other kind assesses human actions in the light of given ends or desires. These theories describe which action among all actions that are open to a person is rational in the sense of being the action that best realizes the person’s ends or best satisfies his or her desires, irrespective of the reasoning processes that led to or will lead to the action. Theories that focus on identifying best actions will be called here theories of rational behavior. An example of a theory of rational behavior is rational-choice theory, since it determines which action among a set of options maximizes expected utility. The distinction between theories of rational reasoning and theories of rational behavior is important for, as Nozick remarks [1993, p. 65], “an action might reach goals or maximize expected utility without having been arrived at rationally. It could be stumbled upon by accident or done inadvertently or result from a series of miscalculations that cancel each other out. That action then, would have been the best thing to do (given that person’s goals), but it would not have been done rationally.” Indeed, that action would not be rational from the point of view of theories of rational reasoning, but it would be rational from the point of view of standard rational-choice theory in the sense of doing the best thing.16 But the reverse may also be the case. Due to incorrect information, for example, a person may reason correctly that she ought to perform a certain action, and she may act accordingly, whereas that action is not rational in the sense of maximizing expected utility.17 In discussing rationality in engineering design, these two senses of rationality have to be clearly distinguished. For instance, a design engineer may opt for a particular solution for a design problem on the basis of her knowledge about available solutions. Her reasoning resulting in this choice may be rational given the options known to her, but it may not lead to the solution that maximizes expected utility, because on the basis of some false beliefs she rejects a solution that would have maximized her expected utility.18 The rational-reasoning approach to rationality, therefore, can be said to take an internal perspective with regard to the acting agent, whereas the rational-behavior approach takes an external point of view [Bratman, 1987, p. 46].

16 Whether or not theories of rational behaviour are really theories of rationality is a controversial matter. According to Nozick [1993, p. 65], “[D]ecision theory by itself is a theory of best action, not of rational action”; on this point he seems to agree with [Hampton, 1994]. 17 Whether or not people reason according to the principle of maximizing expected utility of rational choice theory, and thus come up with choices which are rational in both of our senses, is an empirical matter; (cf. [Nozick, 1993, p. 65; Hampton, 1994, p. 197]). 18 It is assumed here that the engineer was rationally justified in holding the false beliefs. Note that here ‘being rationally justified’ refers to the rational-reasoning sense of rationality; in the rational-behavior sense, it will only seldom be rational, i.e. beneficial, to hold false beliefs.

Rationality in Design

4.4

577

Fixation of means versus fixation of ends

A long-standing issue in discussions about rationality concerns whether the norms of rationality only apply to the fixation of means, once ends are given, or whether the fixation of ends may also be subject to rational evaluation. The most dominant conception of rationality restricts the domain of rationality to the effective and efficient realization of ends or desires, where the efficiency or effectiveness of a particular realization is judged against the background of one’s beliefs. This conception is called instrumental rationality, since a person’s rationality serves as an instrument in the service of this person’s ends or desires. The person’s beliefs can be seen as part of the instrument and therefore themselves subject to rational criticism and candidates for being replaced by better ones. In line with Hume’s dictum “Reason is, and ought only to be, the slave of the passions. . . ” [Hume, 1969 (1739-40), p. 462], the instrumental conception of rationality assumes that the fixation of goals or ends falls outside the province of rationality (reason). In this conception, it is not contrary to reason, as Hume [ibidem, p. 463] famously remarked, “to prefer the destruction of the whole world to the scratching of my finger”. Instrumental rationality deals only with the most effective and efficient means to reach given ends. As Simon, a modern defender of the instrumental conception of rationality puts it [1983, pp. 7-8]: “[R]eason is wholly instrumental. It cannot tell us where to go; at best it can tell us how to get there. It is a gun for hire that can be employed in the service of whatever goals we have, good or bad.” Even adherents of rationality as instrumental rationality must, however, grant the possibility of some rational criticism of ends. It would, for example, generally be considered irrational to hold ends that are impossible on the basis of one’s beliefs, for example to have as an end to be the first man on the moon (after 1969, that is), or to be the first (wo)man to set foot on the sun, or to construct a perpetuum mobile. Accordingly, there must always be room for general consistency conditions on ends that reflect part of what we mean by rationality. In rational-choice theory, which can be considered as the most sophisticated formal articulation of instrumental rationality, the role of ends or desires is played by the preference relation or utility function defined on the domain of possible outcomes of the options for action from which a choice has to be made. This theory requires these preferences to be, minimally, complete and transitive, and to have complete and transitive preferences is therefore, on this account, a precondition of rational action. For beliefs there are similar criteria, stating that the totality of one’s beliefs must be consistent and that one must also believe the logical consequences of what one (already) believes. Such conditions on one’s beliefs are sometimes referred to as representing epistemic rationality, and analogously any conditions posed on one’s ends or preferences could be said to represent orectic rationality.19 19 Whereas the notion of epistemic rationality is current in philosophy, the notion of orectic rationality is not at all. The word ‘orectic’ nonetheless occurs in the Oxford English Dictionary and is derived from the Greek orexis, the more neutral word for desire used by Aristotle, next to epithumia, which means desire in general but also, more restrictedly, lust or hankering. We owe the term to Jonathan Dancy.

578

Peter Kroes, Maarten Franssen and Louis Bucciarelli

Perhaps Edvardsson & Hansson’s conditions on rational goals [2005] or Searle’s [2001] claim that desires, unlike beliefs, need not be consistent, and a discussion of the question whether this extends to goals or ends, are good examples of the latter. It should be emphasized that there is a clear distinction between epistemic and orectic rationality on the one hand and theoretical and practical rationality on the other, as defined here. The former two concern the coherence or consistency of one’s full set of beliefs and desires, regardless of a particular context of action.20 The latter two concern the adoption of specific beliefs or courses of actions, given specific sets of current beliefs and current ends in specific circumstances. In particular they concern the updating of beliefs given new evidence, or the adjustment of plans, given the realization of intermediate states. The notions of theoretical and practical rationality, therefore, fully cover the content or substance of rationality according to the instrumental view.

4.5 Procedural and substantive rationality The issue of rationality and the fixation of means and ends is closely related to the distinction between procedural and substantive rationality, as made by Hooker and Streumer [2004, p. 60]. According to proceduralists, the fixation of ends (or desires) either pertains to the fixation of non-instrumental, fundamental ends, which falls out of the scope of (practical) rationality, or to the fixation of instrumental ends (instrumental for realizing the fundamental ends), which are subject to rational criticism. Substantivists, on the contrary, assume that an agent may be open to rational criticism also in cases in which ends cannot be reached from prior (fundamental) ends. According to substantivists, all setting of ends is subject to rational scrutiny. This distinction runs closely parallel to the distinction between internalism and externalism about reasons for acting introduced by Williams (see [Wallace, 2003]). In relation to rational-choice theory, the notion of procedural rationality is used in a completely different sense. Based on the work of Simon, the notion of procedural rationality is associated with the notions of bounded rationality and satisficing [Hargreaves et al., 1992]. Taking into account that we live in a world with uncertainty and cognitive complexity and that humans have limited information processing capacities, Simon [1978, p. 9] introduced the distinction between substantive and procedural rationality in the following way: “In such a world, we must give an account not only of substantive rationality — the extent to which appropriate courses of action are chosen — but also procedural rationality — the effectiveness, in the light of human cognitive powers and limitations, of the procedures used to choose actions.” Substantive rationality, according to Simon, is the rationality of an omniscient agent that perceives the world in all its relevant de20 Note that the requirement that one should not aim for what one considers impossible is not a consistency condition on the set of ends alone, since what must be judged to be impossible is determined by one’s beliefs.

Rationality in Design

579

tails and is thereby capable of making the truly optimal choice. But human beings are not omniscient agents that have knowledge of all relevant details of the state of the world. In actual situations, preparatory decisions have to be made about whether to continue gathering additional information about the current situation or whether sufficient information is available for making an optimal decision, or one that is sufficiently good, or about whether more options have to be added to the choice set or whether the current set is likely to contain the best possible option, or a sufficiently good option. In procedural rationality the emphasis lies on the rational scrutiny to which the procedures that are used to make these preparatory decisions can be subjected. Needless to say that, given the uncertainty and cognitive complexity of many engineering-design projects, this form of rationality may be of great importance for engineering-design practice. 5 THE INSTRUMENTAL CONCEPTION OF RATIONALITY

5.1

Rational-choice theory

It was already mentioned that engineering-design problems are overwhelmingly end-focused, since the basic goal is to arrive at the blueprint or prototype of an artifact that meets certain requirements. This matches the instrumental conception of rationality. The articulation of theories of instrumental rationality has taken at least two forms. One is the theory of means-ends reasoning, the other, perhaps the better known one, is the theory of rational choice, which will be discussed first. A discussion of means-ends reasoning and a comparison of the two approaches will follow this discussion. The theory of rational choice has mainly been developed as a core theory for the science of economics, but finds wide application also in areas like operations research, risk analysis, organizational theory, and so forth. Apart from rationalchoice theory, the name expected-utility theory is also much in use. This theory models the problem of deciding upon a course of action as a problem of choosing the best among a set of given possible courses of action or options for choice. Where these options come from is not something that the theory of rational choice has anything to say about. To be applicable, the theory simply requires a set of options to be defined. In order to be able to choose the best option, the consequences or outcomes of each option have to be listed and valued. The preference order or preference measure on the set of possible outcomes exhausts the articulation of the decision maker’s ends or desires; more preferred outcomes can be taken to be ‘closer’ to his or her ends. Once a list of all possible outcomes is available, what determines what is the best option to choose, and therefore the rational choice, is how these possible outcomes are evaluated relative to one another. What is minimally required is a preference order of all possible outcomes. In order for a preference order of outcomes to exist, the binary relation of comparison between two outcomes — if I had to choose between two particular outcomes, which one would I prefer — must

580

Peter Kroes, Maarten Franssen and Louis Bucciarelli

be complete and transitive. Transitivity is the property that when outcome a is judged to be superior to outcome b and outcome b superior to outcome c, then in a direct comparison of outcomes a and c, outcome a is considered better than or preferable to outcome c. Completeness means that for any two outcomes it must be clear which of the two is better or whether they are perhaps equally good. The latter possibility is called indifference, which has to be sharply distinguished from the case where two outcomes are declared incomparable. With indifference the outcomes are considered equally good or bad and accordingly it is considered irrelevant which of the two options is chosen, and how this choice is made, as long as some option is chosen. In the case of incomparability, a person feels it does matter which of the two options is chosen but is, perhaps only for the time being, unable to fix his or her preference. A preference order of outcomes suffices to solve a problem of decision making under certainty, i.e. a problem where a choice for an option leads with certainty to one particular outcome, the situation where the option has been realized. Since options relate one-one to outcomes, the rational choice is to choose the option that results in the best or most preferred outcome. This will not do for decision making under uncertainty or risk, however. In general, decision making under uncertainty or risk is a situation where choosing an option can lead to several mutually exclusive outcomes and the decision maker cannot know beforehand which of these possible outcomes will in fact be the result of his or her choice. The distinction between the two forms is that in a case of decision making under risk the probabilities of the realization of the various possible outcomes can be assumed to be known or given, whereas with decision making under uncertainty this is not so. The approach to decision making under risk is considered to be the major contribution of expected-utility theory. Prima facie engineering design seems not to be the place for its application. The design options that engineers must choose between do not lead to a particular design only in a percentage of all cases, when in fact chosen. However, a particular design, once manufactured, may well perform as intended only in a percentage of all cases. This is the more likely while still in the early phases of a design task, when only prototypes are available. Thus, there will be opportunity for the application of models for decision making under risk in the early phases of design, in order to estimate the relative worth of further developing suggestions for possible design solutions, given that it is uncertain whether they will indeed lead to feasible solutions within a given time and with limited resources to spend. Additionally, there will be opportunity for its application in assessing various design solutions with respect to possible future failures, which can never be ruled out completely. A major problem in engineering design is making trade-offs, that is, weighing the advantages and disadvantages of one design solution against another. Whether or not this problem involves uncertainties as well, it is not one for expected-utility theory to solve. This theory is still relevant here, however, since the weighing of the relative values of design options makes sense only if minimally rough quantitative values are available, and rational-choice theory indirectly contains a method

Rationality in Design

581

for arriving at a quantitative measurement scale for the relative value of design options, by asking the designer to make pair-wise comparisons of lotteries involving design solutions rather than the straightforward pair-wise comparisons of solutions. This procedure, however, has only recently been introduced in engineering design; see e.g. [Thurston, 1991]. The problem of making trade-offs is taken up again in Subsection 6.3.

5.2

Means-ends reasoning

Rational-choice theory contrasts sharply with means-ends reasoning as an approach to instrumental rationality.21 In the language of means and ends, rationalchoice theory starts with the means — the courses of action open to the person facing the problem situation — as given, and then proceeds toward the possible ends — the possible outcomes of the options. These possible outcomes correspond to the ends, because their evaluation by the person in question determines which means — that is, which option — is the rational one to choose. In the case of choice under risk this is actually determined by two measures, one measuring the extent of uncertainty, the other measuring strength of preference, but only the latter measure is evaluative. In means-ends reasoning, one starts with the end — a state of the world that one wishes to realize — and then proceeds toward the means — the possible courses of action that lead to the end. The object of meansends analysis is to generate a list of potential means and to establish that the means would indeed result in the end, or to investigate the circumstances under which they would result in the end, or the form in which they result in the ends, since the state of the world resulting from adopting a particular course of action will usually include more than just the end aimed for.

5.3

Means-ends reasoning as complementary to rational choice

In rationally proceeding in a design process, or in rationally reconstructing one, means-ends reasoning and rational-choice theory are not so much competing approaches but rather complementary ones. Means-ends reasoning will be instrumental in generating a set of potential design solutions, or courses of action toward a solution, whereas rational-choice theory can be used to justify the choice for a particular design solution as the best one from that set, given an evaluation of the results of adopting any of the feasible solutions. In contrast to rational-choice theory, no commonly accepted or standard theory of means-ends reasoning exists. Means-ends reasoning places central what in rational-choice theory is only a sideline: our knowledge of the causal connections between means and ends, or between courses of action and their results. Rationalchoice theory assumes these causal connections to be known or given, minimally in the form of probabilities, and sees its job as specifying what rationally follows from these connections and the evaluation of the outcomes by the decision-maker. 21 For

a discussion of means-ends reasoning, see the contribution by Hughes in this Volume.

582

Peter Kroes, Maarten Franssen and Louis Bucciarelli

Although, in accord with the starting point of instrumental rationality, the fixation of ends lies outside of the theory, formal constraints on the way a rational person may entertain his or her ends play an important part. In means-ends reasoning these constraints are at best sidelines. The starting point is usually one particular end that is to be realized, no further questions asked. It is only in evaluating a means that questions of what further results from applying the means apart from the desired end may come up for evaluation. Means-ends reasoning can therefore be seen as the most explicit articulation of instrumental rationality, in the sense that it addresses the how-question with regard to realising an end directly.

5.4 ‘Reasoning’ about ends On both approaches we can hold on to the basic tenet of instrumental rationality — that instrumental rationality fixes means but not ends, or that preferences are exogenous to the theory of rational choice — while at the same time extending the scope of rational deliberation to include deliberation about ends. On means-ends reasoning, this can be done by introducing a hierarchy of ends, where ends at a lower level, for which at that level means are sought, are treated as means to further ends at a higher level. This is referred to as ends-means reversal: what are initially treated as ends are after some more reflection seen to be means. This puts considerable pressure on an adequate conceptualization of decision problems, since at all intermediate levels one must conceive of entities such that they indeed can be means as well as ends. In rational-choice theory, there are various possibilities. Since the role of ends is played by the relative values attached to the possible outcomes of actions, the way to elaborate the say of the outcomes in the determination of the choice is to introduce increasingly refined descriptions of states of the world resulting from chosen courses of action. These refined descriptions may then lead to a revision of the relative values attributed to the outcomes. Another way is to introduce dynamic representations of decision situations and to push these descriptions forward in time, thus gradually enlarging the scope of considerations deemed relevant.

5.5 Criticism of and alternatives to instrumental rationality Standard rational-choice theory has come under attack from different directions. One kind of criticism, put forward by Simon and already mentioned in Subsection 4.5, points out that the theory puts severe demands on the “choosing organism” with regard to access to information and its computational abilities. It must be able to specify exactly the nature of the outcomes, the pay-offs or amounts of satisfaction afforded by each outcome, these have to be completely ordered, and it must be able to predict with certainty or with definite probabilities the outcomes. Simon doubts whether in actual situations of human choice of any complexity this information is at hand and whether the required calculations “can be, or are in fact, performed” [1955, p. 104]. It is, of course, possible to try to fill in the gaps in

Rationality in Design

583

information, but information gathering has its own costs, and there will be a point at which the costs of gathering more information will be higher than the possible gain in utility due to a better solution of the original choice problem. So, rationalchoice theory has to be simplified, to be useful in practice. Simplifications can be achieved by using procedures such as rules of thumb as a basis for decisions and by replacing the notion of unrestricted rationality based on unlimited information and the principle of optimization by the notion of ‘bounded rationality’ that takes account of the limitations in knowledge and computation capacities. Satisficing is a key notion for bounded rationality; aspiration levels are set and the first option that satisfies the aspiration level is chosen [Simon, 1982]. Thus, procedures have to be set for how decisions have to be made, which means that substantive rationality has to be replaced by procedural rationality. Simon’s criticism of rational-choice theory remains squarely within the instrumental camp: it does not question the instrumental nature of rationality. The instrumental conception of rationality itself, however, has also been criticized for being too limited in scope. It is questioned, for example, whether it can deal with the situations of strategic interaction between people as modeled in game theory. Other approaches to rationality, such as Habermas’ communicative rationality [1984; 1987] and Dreyfus & Dreyfus’s deliberative rationality [1986; 1992], reject the idea that instrumental rationality exhausts the whole notion of rationality. Another attempt at an approach to make the deliberation about ends fall under a broad concept of rationality was undertaken by Richardson [1994]. Repeating some of Richardson’s critiques, Searle [2001] has also attacked the exclusion of rational deliberation about ends from theories of instrumental rationality, in particular rational-choice theory. This theory assumes that a well-ordered preference schedule is given in advance, whereas in real life “a well-ordered set of preferences is typically the result of successful deliberation, and is not its precondition” (pp. 30-31). To make this assumption is to deny rational deliberation what in real human decision making is its prime subject matter (pp. 125-126): “Most of the difficulty of rational deliberation is to decide what you really want, and what you really want to do. [. . . ] The really hard part of practical reason is to figure out what the ends are in the first place.” Searle, however, does not offer us a theory or a scheme for how to figure this out, and it remains unclear what conceptions are possible of what could be termed telical rationality alongside theoretical rationality and practical rationality, or what could be termed final rationality alongside instrumental rationality.22

22 The Oxford English Dictionary has ‘telic’, from the Greek telikos, which, though failing in Aristotle, is attested a few times in ancient Greek in the sense of ‘pertaining to the ends, or the ultimate end, of human life or action’. It derives from telos, ‘end, goal’. ‘Final’ similarly derives from finis, the Latin equivalent of telos.

584

Peter Kroes, Maarten Franssen and Louis Bucciarelli

6 RATIONALITY IN ENGINEERING-DESIGN PRACTICE

6.1 Creativity and rationality in engineering design Engineering design is about the creation of new technical things, the bringing into being of what did not exist before. It is a creative activity and requires creative thinking and doing at some level or in some respects. How does engineering design as a creative activity relate to engineering design as a rational activity (supposing that it is, to some extent, a rational activity)? Is there a tension between creativity and rationality in design, or can they, or even need they, go together? For a start, the vulgar image of creativity in engineering design hardly does justice to actual practice. The drama of an individual steeped in a quest, who suddenly, in a flash, conjures up a brilliant idea, which bears the promise of solving the problem he has been wrestling with, barely scratches the surface. Ideas are important, but in engineering practice ideas have to be put to work; workable ideas are essential. Locating all creativity in the brilliant flash misses the point that the implementation of ideas, their translation into hardware, requires the continual exercise of creative and constructive thinking all along the way. Only in the final, detailing stages design problems may be solvable on the basis of existing routine, handbooks or rulebooks. The foregoing also applies to more ‘artistic’ forms of engineering design like industrial or architectural design. In these domains, design may come close to the creation of a work of art, but here also ideas have to be transformed into real objects, a process which may require a lot of creativity of its own. But if creativity is indispensable for engineering design, so is rationality. At the early stages various ideas, creative or not, have to be judged against available resources, customer needs and functional requirements, in house state of the art of technology, production facilities and so on. Later on all kinds of decisions have to be made about e.g. how to divide the overall function of the artifact in sub-functions and which components to choose for realizing these sub-functions. Coming up with alternative proposals may require creative thinking, but once the proposals are on the table, choices have to be made and these are not (to be) made at whim, but partly based on their rational appraisal. Engineering design, as a problem-solving activity, requires creativity and rationality and no tension seems to be involved. The above remarks suggest a view on the role of creativity and rationality in engineering design very similar to Popper’s view on their role in science. According to Popper [1959, pp. 31-32] there is no ‘logic’, in the sense of rational method, of scientific discovery.23 The rationality of science resides completely in subjecting new ideas, conjectures, to severe criticism. The way scientific ideas are generated is not amenable to rational analysis, only the justification of these ideas is, by evaluating their logical consequences in the light of the available evidence. In Popper’s 23 As has often been observed, the English translation, The logic of scientific discovery, of the German title Logik der Forschung (1934) is very misleading.

Rationality in Design

585

vision of science, creativity and rationality have become decoupled. Whatever tension there might be between creativity and rationality, there is nothing problematic about taking science to be both a creative and a rational process, since creativity and rationality have their role to play in different contexts: creativity rules in the context of discovery, whereas rationality holds sway in the context of justification.24 Application of these views to engineering design leads to the view that there is no ‘logic of technological invention’; rationality plays no role in the creation of new ideas for solving design problems. It only plays a role in choosing among these ideas. Again, as in the case of science, there would be nothing problematic in claiming that engineering design is a creative and a rational activity, since creativity and rationality play a role in different contexts. There are, however, serious problems with such interpretations of the role of rationality and creativity in science and technology. The distinction between context of discovery and context of justification is taken for granted. The validity of this distinction, particularly the idea that the context of justification is independent of the context of discovery, has been put into question [Hoyningen-Huene, 1987]. If indeed it is not possible, in principle, to separate the context of justification from the context of discovery, then the above view on the role of creativity and rationality in design processes has to be reconsidered. An additional problem with this view is that it appears to acknowledge only one form of rationality, namely instrumental rationality, more in particular its articulation as rational-choice theory. As discussed in Section 5, in this theory it is presupposed that all options are given; it is a procedure for establishing which of the alternatives maximizes expected utility. By its own choice, then, this form of rationality has no role to play in creating new options or alternatives. But as we observed above, critics of instrumental rationality precisely maintain that in this domain forms of rationality may be operative as well, be it other forms. This opens up the possibility that creativity and non-instrumental forms of rationality are compatible, not simply because they operate in different contexts, but because rationality contributes in a constructive way to the generation of new ideas or new options. According to these critics, creativity and rationality may not just tolerate each other but may even go hand in hand in engineering design, although they have little to say about the precise way in which they interact. The problems with extending theories of instrumental rationality to cover the creation of new options or the creations of new means to existing options may be one of the reasons why this is often considered the domain of creative thinking, not constrained by rational considerations. But this latter position leads to a rather unsatisfactory picture of the role of rationality in engineering design. Not only the processes by which new options are added to the engineers’ agenda (artefacts with new functionalities, or with better specifications than their predecessors), but also the processes leading up to new ways of realizing existing artefacts (on the basis of 24 The distinction between the context of discovery and the context of justification was introduced by Reichenbach [1938].

586

Peter Kroes, Maarten Franssen and Louis Bucciarelli

new operational principles, with less components, greater ease of production and maintenance etc.), all these processes would fall outside the realm of rational evaluation. But it is precisely in those situations that decisions have to be made about what, given the assessments of all the prevailing constraints and uncertainties, may be possible to achieve and worthwhile to pursue. The constraints involved make this a creative process: without constraints on possible solutions, no creativity. Some of these decisions, or assessments, may be based on more or less rational expectations, others on blind hope. It is in the search for new technological possibilities that creativity and rationality appear to go hand in hand. Paraphrasing Kant, we might say that creativity without rationality is blind, whereas rationality without creativity is barren.

6.2 The engineering picture of the design process Let us now examine more closely some suggestions put forward by engineers and design methodologists for rationally structuring engineering-design processes. Consider the phase diagram for solving design problems of Figure 1. Phase diagrams like this one abound in the literature. Most of them claim implicitly or explicitly universal validity within the field of engineering design, since only one technology-independent phase diagram is usually presented.25 They neatly distinguish various subtasks and the order in which they are to be performed (with the possibility of iterations). In the manuals that go with such phase diagrams, subtasks are usually specified in detail and the kind of design tools to be used in the various subtasks are described. And although the authors of such diagrams are usually careful to stress that following these phase diagrams does not guarantee that the resulting design will be successful, the spirit behind these proposals is nevertheless that implementing these phase diagram will improve design performance. Whether or not these phase diagrams are applied in practice, they do establish a vocabulary and language for addressing any particular design task. Engineers know that specifications need to be specified at the outset of a project; they know that the process is near completion when the design is ‘frozen’ and definitive layouts and documentation need be produced. But the block diagram suggests a more mechanical, linear progression than is generally the case. Although iteration is allowed (see the vertical block of text on the extreme left) back-stepping appears to be constrained to retracing one’s steps, one step at a time. They are misleading too in that they provide hardly any indication of the relative difficulty or intensity of effort required to complete each phase. Is the ‘search 25 Suh [2001, p. 11] presents a simple phase diagram consisting of four domains, the costumer, functional, physical and process domain. He explicitly claims universal validity of this phase diagram [ibidem, p. 12]: “All designs fit into these four domains. Therefore, all design activities, be they product design or software design, can be generalized in terms of the same principles. Because of the logical structure of the design world, the generalized design principles can be applied to all design applications and we can consider all design issues that arise in the four domains systematically and, if necessary, concurrently.”

Rationality in Design

587

Define Task Customer needs Clarify specifications

Phase I Determine Functional Requirements

Generate Conceptual Designs Compare and select options Phase II

Select Best Option Develop Prototype

Implement Design Detail design for production

Phase III

Figure 1. Phase diagram of engineering design

588

Peter Kroes, Maarten Franssen and Louis Bucciarelli

for solution principles and their combinations’ as straight forward a task as ‘complete overall layout’ ? Is the job of dividing the design into ‘realizable modules’ as free of ambiguity as the next step, i.e., ‘develop layout’ of these modules? And what resources are required to accomplish each step? In particular, is but one person responsible for all phases; or is it a team of four or five working within the firm; or a globally dispersed assembly of one, two hundred participants; or does the diagram represent the workings of a computer algorithm? As instructional devices, they are highly idealized representations of what transpires in designing and, as such, downplay, if not ignore completely, the negotiation and exchange among participants required to bring a design to completion. An intriguing feature of these phase-diagrams is that they address only the process level of designing and completely abstract from the substance of design problems. Compared to rational reconstructions of design processes, these schemes exhibit a much greater level of generality. Rational reconstructions are always rational reconstructions of a particular design. This is not the case for these phase diagrams; as observed above, they are even technology independent. They exhibit what could be called process rationality: the most rational way to structure the design process qua process for successful design.26 As such, process rationality is a form of instrumental rationality: the phase-diagrams represent the best procedure or sequence of various steps to bring about a goal or a desired state of affairs. But, what is the connection, if any, between a successful implementation of these phase-diagrams and a successful design as outcome of that process? In design methodology it is usually simply assumed that following these phase diagrams will contribute to better designs.27 Nevertheless, this point needs clarification. A successful implementation appears neither necessary nor sufficient for a successful design. The history of engineering abounds with examples supporting this claim. The problem with these phase-diagrams is that they are based on a very abstract notion of a design process, which results in an extremely abstract form of process rationality. But what is the relation, if any, between rational decisions that affect the structure of the design process and rational decisions that affect the object of design? The latter determine the properties of the object of design; in other words, the latter ultimately determine the selection of the optimal/best option from all possible options. As long as this relation remains obscure, it is not clear at all what the value of such prescriptive phase-diagrams for engineering-design practice might be.

26 Note that process rationality as described here is different from procedural rationality as defined in [Hargreaves et al., 1992, p. 17 ff]; see also the discussion in Subsection 4.5. 27 See, for instance, [Roozenburg and Eekels, 1995, p. 93]: “This is the way design processes are structured. That establishment leads almost imperatively to the statement: effective design processes should be structured in this manner. The cycle, which is found descriptively, changes into a norm for effective designing. We can therefore also consider the basic design cycle as a prescriptive model for designing.” Note the naturalistic fallacy in this passage.

Rationality in Design

6.3 6.3.1

589

Rationality and design decisions Design as a rational decision process of selecting the best choice

When it comes to specifying a view of the process of rational design, in particular a normative or prescriptive view that can be applied in design practice, it is common to view the design process as one of rational decision making, where the designer’s task is to pick one particular design as the best one to realize. Given that rationalchoice theory assumes these options to be given (see Subsection 5.1), this may seem to reinstall the intuition that the creative aspect of design escapes from considerations of instrumental rationality. However, there are ways of bringing the process of generating options within the scope of instrumental rationality. This process can be modeled as a decision-making process on a higher level, where the choice is among types or families of solutions, or among search routines, or whatever is used. Several types or families of solutions or several search routines will then generate a set of best first candidates to be compared directly for the final decision. Since the design process must result in a particular design, that is, a blueprint or a prototype of the product to be manufactured and marketed, the options to be decided among are of the form “make the product as follows:. . . ” If such an action can be taken to result with certainty in the corresponding product, we have a case of decision making under certainty and the decision-making problem is trivial: decide to make the best product. The design problem then amounts to establishing a ranking of all possible design concepts, ranging from the best to the worst. To accomplish this is hardly a trivial matter, of course, and can even be considered one of the major aspects of the design task. Indeed, as will be shown below, it is a task that can itself be conceived as a problem in rational decision making. 6.3.2

Rational decision making of design teams

Already in the case of a single designer, therefore, the availability of a qualitative or quantitative preference scale for design concepts is not a straightforward matter. Many design tasks, however, if not most, are undertaken by a team of designers, or even by teams of designers. When several designers have to decide jointly on which design action to perform what is required in order to solve their problem using rational-choice theory is a joint preference scale. The problem of constructing a joint preference scale out of several individual preference scales can itself be seen as a problem in rational decision making. The problem is known from the theory of social choice, where the issue is to choose one option out of a set of proposed options jointly, as a collective, as happens, for example, in elections. From the perspective of rational decision making, the central issue is what the best translation is of a collection of preference orders of options into a single preference order of the same options, such that the resulting order can be taken as to represent the ‘social’ preferences or the preferences of the ‘collective’. Note

590

Peter Kroes, Maarten Franssen and Louis Bucciarelli

that from the conception of a preference order and from the way one arrives at qualitative and quantitative preference scales, it follows that one can never add individual preference values to arrive at a collective value. Individual preference scales measure separate quantities for each individual, and to add the values of two separate scales is like adding meters to kilograms. Indeed, the endeavor to construct a procedure that aggregates multiple individual rankings into a collective one has its starting point in the recognition of this fact. In 1950 it was proved by the economist Arrow that no general solution exists to this problem of aggregating preferences into an overall order. More precisely, Arrow proved that no function exists, taking ordered sets of preferences as input and delivering exactly one preference order as output, that satisfies a number of requirements that are all deemed necessary for the procedure to be called rational (see [Arrow, 1963]). In other words, we do not generally know how to give meaning to the notion of a collective preference scale when only individual scales are given, and hence the theory of rational choice is not generally applicable when multiple designers are at work. It may be thought that instead of constructing the overall preference order through a rational procedure, one had better ask ‘the team’ directly for its preferences. This will give a straightforward answer when the individual designers are unanimous in their preferences, but what would the team say when its members disagree? To answer as a team in the first place, i.e. with one voice, the team members must start discussing the answer that they will give. In Section 7 more will be said on the social dimension of engineering design introduced by the necessity to negotiate design decisions whenever designing is a team effort (and it almost universally is nowadays). Here it will only be noted that going about such a discussion and negotiation process can also be done in more and less rational ways. In particular, methods from game theory can be applied in order to show the members of a team how they can rationally — in the instrumental utility-maximizing sense – ‘solve’ a case where they must jointly decide on a design solution while their evaluations of the various possible design solutions differ [Franssen and Bucciarelli, 2004]. Depending on the method used, there is nothing against accepting the resulting preferences as rational team preferences for the particular case at hand. Nevertheless, if the consecutive results of many such negotiations are interpreted as jointly instantiating a function that translates all sorts of combinations of individual preferences into team preferences, this function must violate at least one of the requirements that Arrow thought such a function should satisfy in order to be acceptable as a universally applicable procedure.28 28 If a discussion or negotiation round results in consensus on team preferences, it could be argued that the ensuing unanimity reflects a unanimity among the team members that is caused by the fact that some team members have adjusted their original preferences, and accordingly any relation between the final consensus and the team members’ original preferences is lost. This would not be a correct view, however. Presumably after the discussion individuals would still insist, in private, that they, personally, have a different opinion on what the best design is. What the team members agree upon is rather what preference order best represents the distribution of individual opinions and can therefore be promoted to as the opinion of the team.

Rationality in Design

6.3.3

591

Multi-criteria rational decision making

Arrow’s famous theorem also plays a role when the process of arriving at a single preference order for a given set of design solutions is put to rational scrutiny. Usually, the evaluation of design concepts is subject to a great number of criteria, which originate in the functional requirements at the start of the design process. Such functional requirements need not always be translated into design specifications that have the character of constraints that are either met or not met, for example, ‘being able to withstand a pressure of 2000 Pa’. Often the functional requirements lead to criteria on which various design concepts must be scored and compared, for example, ‘being as light as possible’, ‘being as robust as possible’, and so forth. When there are several of these criteria, a comparison of a number of possible design solutions gives rise to as many preference orders of all possible solutions as there are criteria. A design solution that is ranked relatively high on the criterion of weight may be ranked relatively low on the criterion of robustness, and vice versa. In order to determine which is the overall best design solution, an individual designer now faces the problem of aggregating these various preference orders into a single one that ‘best’ represents the set of individual rankings. It is the extent to which the aggregate order ‘faithfully’ reflects the set of individual rankings that determines whether the design solution that is topmost in the aggregate ranking is indeed the best design, since the individual rankings precisely reflect what the designer considers important and relevant concerning the design problem. Not only is this problem isomorphic to the problem of aggregating the rankings of individual people into an overall collective or social ranking, as Figures 2 and 3 show, but the sort of requirements that an aggregative procedure must minimally satisfy in order that the procedure can count as rational are identical. And it was precisely proven by Arrow that no aggregating procedure can satisfy all of these requirements. Therefore, as a consequence of Arrow’s theorem, no algorithmic solution for this problem is available. Rank order of individual x1 a1 a2 a3 ... . .

Rank order of individual x2 aj ak al ... . .

Rank order of individual x3 ar as at ... . .

......... . . . . . . . .

. . . . . . . .

Collective rank order

?

Figure 2. Schematic presentation of a decision problem with multiple decision makers

592

Peter Kroes, Maarten Franssen and Louis Bucciarelli

Rank order by criterion X1 a1 a2 a3 ... . .

Rank order by criterion X2 aj ak al ... . .

Rank order by criterion X3 ar as at ... . .

.........

. . . . . . . .

. . . . . . . .

Overall rank order

?

Figure 3. Schematic presentation of a problem of multi-criteria decision making It might be thought that the reason why no algorithmic solution for this problem exists is that the information to be aggregated is too sparse, so to speak, consisting only of rank orderings of the options according to the various criteria. If quantitative measures of the relative worth of the options were available, one per criterion, an algorithmic solution might well be possible. This, however, is not the case: several impossibility theorems for the aggregation of interval-scale measures of value into a single rank order or measure, analogous to Arrow’s original one, have been proved; for more details see [Franssen, 2005]. For aggregation to be in general possible, either measurement on the same scale or ratio-scale measurability of all values is required, and most design problems do not satisfy the conditions for this to be the case. Obviously, the recourse to negotiation that was mentioned for team decision making is not available in the multi-criteria problem. Any negotiations among the criteria — who takes a step back, whose opinion matters more — must be done by proxy, so to speak, by the designer him- or herself. It is the designer who must make trade-offs between the criteria. The framework for this sort of problem developed by Arrow makes clear that there is no general rational procedure for making tradeoffs in engineering design. Of course a designing engineer may adopt all kinds of routines and rules-of-thumb for arriving at trade-offs, but these methods can hardly serve to justify, from the point of view of a theory of rationality, the tradeoffs actually made. On this problem remarkably little theoretical work seems to have been done; for a more detailed argument, see again [Franssen, 2005]. The most difficult case, finally, is where a design team has to solve a multicriteria design problem. The complexity of this problem, from a rational point of view, will now be appreciated. In the next and final section we will address some of the consequences of this complexity.

Rationality in Design

593

7 DESIGN AS A SOCIAL PROCESS

7.1

Social versus societal aspects of designing

Solving engineering-design problems is usually not an activity that takes place in the splendid isolation of a single individual facing that problem. A designer is typically a member of a design team that operates within a firm, which itself operates within a specific social context. This embedding of engineering design within broader social contexts may have far-reaching consequences for the way engineering-design problems are solved. In analyzing the role of these social factors in engineering design, in particular in relation to rationality issues, several ways in which the word ‘social’ is used have to be distinguished. One way is exemplified by Vincenti when he states [1992, p. 32] that [A]t the levels of project definition and conceptual design — that is, at the top of the hierarchy — social factors obviously have wide scope for influence. Vincenti uses the notion of social primarily in the sense of non-technical; social factors include all factors that are not of a technical kind. He remarks that the object of design is often embedded within a broader socio-technical system and contributes to the functioning of this system. The social aspect of socio-technical systems covers all other factors (such as laws, social institutions etc.) that are necessary for technical systems to perform their function. Closer in to design practice, one can speak of prevailing societal norms and beliefs as they affect corporate thinking about sustainable designs, environmental effects, or financial returns to their stockholders. Still, this is not the ‘social’ that concerns us here. The scope of our social is the immediate milieu of participants in design as they go about their work within the firm. We are interested in how engineers and other participants interact and relate as they negotiate their analyses and proposals; upon what bases do they decide; how do they justify their claims; what norms govern their exchange, their discourse? (See also [Vincenti, 1992, p. 33].) Of particular interest is how participants proceed when it is clear that instrumental rationality will not suffice. Take a situation in which it has become clear to all members of a design team that the original list of design specifications cannot be satisfied because some of these specifications are in conflict. So trade-offs have to be made, which means that the list of design specifications has to be adjusted. What then? What other form of reasoning, or of rational action, can be relied upon to reach agreement about changes in design specifications or more generally to bring a design to its realization? ‘Social’ is to be understood, then, as a label for another form of rational action; what this ‘other’ form of rationality might be is our concern. But first, we sketch our vision of design as a social process.

594

Peter Kroes, Maarten Franssen and Louis Bucciarelli

7.2 Object worlds The design of products and systems generally requires the coordination of a group of individuals from different specialties — e.g., structural analysis, electronics, informatics, thermal engineering, etc. — who work on different but interrelated features of the system. Each participant in the design inhabits his or her own world of professional practice — a world of standard models of the way things work from the disciplinary perspective of the design participant’s domain; unwritten rules and particular metaphors enlighten and enliven the efforts of inhabitants. There are specialized computational algorithms, singular ways of picturing states and processes. Each participant works with a particular system of units and with variables of particular dimensions — certain ranges of values perhaps. Dynamic processes, if that is their concern, unfold with respect to a particular time scale — for someone’s world it may be milliseconds, in another’s, hours or days. One’s object world is situated with respect to a particular infrastructure with its own unique instruments, reference texts, prototypical bits of hardware, special tools, suppliers’ catalogues, codes, and regulations. Within each of these worlds one ‘speaks’ a different dialect all neat and tidy, precise. We say that different participants work within different object worlds [Bucciarelli, 1996]. A structural engineer inhabits a different world from the electronics engineer working on the same design. Within these object worlds, instrumental rationality reigns supreme. Problem solving is what engineers do within these worlds. They make abstractions of artifactual ‘behavior’, reducing appearances — better put as “seeing through” appearances — to reveal the operating principles that explain how things work, e.g., “this behaves as if it were...”. The structural engineer looks at the aircraft and focuses on the wing structure. The engine people look at the whole craft, its weight, the desired range and optimize their design of the power-plant accordingly. The aerodynamicist sees the flow field around the craft and strives to minimize drag relative to lift by giving the craft a proper geometry. Controls people live in yet another world (see Figure 4). Within these worlds, quantitative, theoretical/empirical based ‘models’ are relied upon to produce useful information, in quantitative form, describing the behavior of the object from one, particular perspective. Within each object world, these instrumental models and methods can even be employed to optimize the particular performance parameters relevant to that world. In each of these worlds, problems are defined such that they may be solved (unambiguously) with the help of the tools available in these worlds; in other words, object world problems are well-defined or well-structured problems. The situation becomes different as soon as decisions have to be made that cut across these different object worlds. The problem is that the optimum aerodynamic shape will not harmonize with the optimum structural design, or the optimum turbine design, or the optimum controls design in that each will prescribe different values for at least some parameters that are of common concern. Here, it appears, we reach the limits of applicability of instrumental modeling and calculative ra-

Rationality in Design

595

Figure 4. Different object world views on an airplane; drawings by C. W. Miller (adapted from [Nicolai, 1975])

596

Peter Kroes, Maarten Franssen and Louis Bucciarelli

tionality. Here lies the real complexity of today’s engineering task; it derives from the fact that each participant, with different competencies, responsibilities, and interests sees the object of design differently: one object, different object worlds. Their proposals and analyses are based upon different paradigmatic engineering traditions and the results of their instrumental analyses and proposals are couched in different categorical terms. And there is no over-arching, instrumental way to reconcile these results in strictly object world terms. This makes designing a social process not wholly amenable to instrumental resolution — a deepening of the problems of collective choice as already exposed in Subsection 5.3. Engineers have to articulate their object-world results so that others, those who inhabit other worlds, can establish meaning with respect to their own perspective. So what engineers do includes much more than rational problem solving and constructing efficient means to reach desired ends; the engineering life world provides a much richer experience once one acknowledges the social action that is part and parcel of designing. Within object worlds, instrumental rationality carries the day; but within the more open world of exchange and negotiation, validity claims concerning this or that feature of design are not restricted to calculative, objective statements that describe the functioning of the object from one or another perspective.

7.3 The limits of rationality Dreyfus and Dreyfus [1986] make a distinction between ‘calculative rationality’ and ‘deliberative rationality’. The former is roughly equivalent to instrumental rationality; the latter, less rule bound, is rooted in the intuitive abilities of the ‘expert’. But this does not really help us in our search for a way of framing, as rational, what goes on when participants in design gather together and struggle to meld their claims and proposals. For such design deliberations are the deliberations of a collection of experts, each proficient in a different realm relevant to the design task. They may very well rely upon their intuition as well as instrumental calculation when working within their respective object worlds, but generally no one individual stands above all the rest, plays the role of proficient dictator, when negotiations across worlds are required. A more relevant notion is that developed by Habermas [1984; 1987] whose theory of communicative action explicitly recognizes the possibility for social grounding of rationality, shifting our attention away from focusing solely on the conceptual, cognitive performance of the individual to the social exercise of reason. His move is doubly relevant in that he takes normative statements making claims about ‘rightness’ (regulative speech acts), and subjective statements expressing truthfulness or sincerity (avowal speech acts,) as well as objective, propositional claims (constative speech acts), as legitimate modes of rational discourse. All of this [Habermas, 1984, p. 17]: ...is oriented to achieving, sustaining, and renewing consensus — and indeed a consensus that rests on the intersubjective recognition of crit-

Rationality in Design

597

icizable validity claims. The rationality inherent in this practice is seen in the fact that a communicatively achieved agreement must be based in the end on reasons. And the rationality of those who participate in this communicative practice is determined by whether, if necessary, they could, under suitable circumstances, provide reasons for their expressions. Habermas’ concern is with social rationality in the large; our concern is local, micro. Yet his picture of rational exchange fits very well with what one sees within the firm, with engineers designing. An engineer makes the (constative) claim that part x will fail if the temperature goes above y degrees; another argues (regulative) that more should be spent upon prototype testing in order ensure the safety of the product in the hands of potential users. A third participant in design expresses delight (avowal) at the simplicity of the design of the latching mechanism. Validity claims in the form of propositional truths, normative rightness and/or subjective, sincere expression all enter into the mix of decision making across object worlds, in achieving consensus about both form and function. It is clear that trust is essential for communicative rationality to ring true. In a design task, any one participant trusts another to work within his or her object world with the professional competence appropriate to that world and to be able to articulate one’s results and proposals in a way accessible to the others. Trust is one dimension of ‘pre-understanding’ [Habermas, 1984, p. 100]: Every process of reaching understanding takes place against the background of a culturally ingrained preunderstanding. This background knowledge remains unproblematic as a whole; only that part of the stock of knowledge that participants make use of and thematize at a given time is put to the test. To the extent that definitions of situations are negotiated by participants themselves, this thematic segment of the lifeworld is at their disposal with the negotiation of each new definition of the situation.29 In one respect, Habermas’ communicative rationality is like Dreyfus and Dreyfus’ deliberative rationality in that if you were to ask, after the fact, how the participants in design came to the consensus that they achieved, you would find that different participants would give different reasons (heuristics, rules) for the particular decision made — just as the ‘expert’, when queried by a ‘knowledge engineer’ intending to capture the expert’s heuristics and store them away in his computer, gives multiple reasons for why he or she did as they did — which leads Dreyfus and Dreyfus to give up on calculative rationality and promote intuition as primary. There is a looseness to our social form of rationality evident in this possibility. That different participants’ reasons for why a decision was made are grounded 29 It is tempting to claim that Habermas here has in mind the image of design as an iterative process.

598

Peter Kroes, Maarten Franssen and Louis Bucciarelli

in their work within different, independent, and in some sense, incommensurable object worlds implies that there may be no shared reasons among the participants for making a choice for a particular design. This social form of rational choice is more or less forced upon them by Arrow’s impossibility theorem. Participants recognize the orthogonality of their different claims and proposals, but they also understand, theoretically rational as they are, that there is no globally rational way to compromise their claims and proposals. That still leaves them with locally rational ways of coming to a decision that each can live with. To conclude from the orthogonality of their claims and proposals that they must refrain from any control over the design decision would be the truly irrational choice. From the perspective of engineering design as a social process, the notion of an optimum or best design becomes highly problematic. Although local optimization might be obtained within individual object worlds, there is no overarching, instrumentally rational method for optimization of the whole. At the same time, we recognize that participants in design often reach the state where all are quite pleased with the outcome. How might the process whereby participants achieve consensus on a particular final design, chosen from among some few or many possible configurations, be construed as rational? What method do they, should they use? One way to approach this problem is by assuming that participants settle for a ‘satisficing’ design; lack of information, inaccessibility of and/or limitations on resources — especially when one admits the incommensurability of object worlds — all conspire to deny doing better. But satisficing is unsatisfying as a concept, even when wrapped together with the notion of ‘bounded rationality’ (see Subsection 4.5). The theory of satisficing offers no resources for determining in a rational way the aspiration level for acceptable solutions or for adjusting this level either upwards or downwards in a rational way. All forms of rationality are bounded to a degree, some more than others. The question is, how to move further, how to proceed given the bounds as they are? What criteria might be called upon to rationally direct one’s progress? And how do you know when to stop designing? Another way to deal with this issue is to focus on Pareto optimality. This notion can be construed as a way to both guide and set criteria for closure. To wit: A design can be said to be optimal in the Pareto sense, if any further design change would move at least one participant to object. Any proposed design move otherwise, bringing benefit to at least one participant (no doubt the one proposing that move) and generating no negative reaction on the part of others, would be a step in the right direction — a step toward a ‘better’ design. Proposed moves put forward by coalitions would also be accepted if they contribute positively (that is, meet no resistance from other participants) to the design. This is not as ‘soft’ or fuzzy a proposal as it first appears since, following Habermas, an individual’s proposed design move, as a validity claim, would be subject to critique and “...based in the end on reasons”. Finally, just as there are limits to the applicability of instrumental rationality, there are clearly also limits to the applicability of our social form of rationality.

Rationality in Design

599

Although all participants in a design team may have an interest in coming up with a definite proposal for a design,30 it may in practice turn out to be impossible to reach consensus on which alternative to choose. The fact that many of these participants live in different object worlds and the corresponding lack of shared reasons among the participants may make this situation more the rule than the exception. So, what to do if no consensus emerges? From the point of view of engineering-design practice sec we have then reached a dead end. A rather unattractive way out would be to simply throw a dice. Engineering design, however, takes place is wider contexts, and these contexts may provide reasons for preferring one particular design option above the others. So decisions may be made on the basis of considerations and reasons that are not directly related to the object of design. The broader context in which engineering design is embedded may bring on to the scene considerations or reasons of a personal nature, of hierarchy within a firm, of power, of trust among participants etc. From a broader perspective there may be good reasons for selecting one design option as against other options, but these good reasons do not make the selected option a better option from a strictly engineering-design point of view. Within these wider contexts, rationality issues may again show up (when is trust in a person justified or rational etc.); these issues, however, fall outside the scope of rationality within engineering design proper. BIBLIOGRAPHY [Arrow, 1963] K. A. Arrow. Social Choice and Individual Values. 2nd ed. John Wiley, 1963 (1951). [Bratman, 1987] M. Bratman. Intention, Plans, and Practical Reasoning. Harvard University Press, 1987. [Bucciarelli, 1996] L. L. Bucciarelli. Designing Engineers. MIT Press, 1996. [Bucciarelli, 2003] L. L. Bucciarelli. Engineering Philosophy. Delft University Press, 2003. [Carnap, 1963] R. Carnap. Autobiography. In The Philosophy of Rudolf Carnap (The library of living philosophers, vol. 11), P.A. Schilpp, ed., pp. 1-84. Open Court, 1963. [Davidson, 1984] D. Davidson. Inquiries into Truth and Interpretation. Clarendon Press, 1984. [Dorst, 1997] C. H. Dorst. Describing Design: A Comparison of Paradigms. Ph.D. thesis, Delft University of Technology, 1997. [Dreyfus and Dreyfus, 1986] H. L. Dreyfus and S. E. Dreyfus. Mind over Machine: The Power of Human Intuition and Expertise in the Era of the Computer. The Free Press, 1986. [Dreyfus and Dreyfus, 1992] H. L. Dreyfus and S. E. Dreyfus. What Computers Still Can’t Do: A Critique of Artificial Reason. MIT Press, 1992. [Edvardsson and Hansoon, 2005] K. Edvardsson and S. O. Hansson. When is a goal rational? Social Choice and Welfare, 24, 343-361, 2005. [Franssen, 2005] M. Franssen. Arrow’s theorem, multi-criteria decision problems and multiattribute design problems in engineering design. Research in Engineering Design, 16, 42-56, 2005. [Franssen and Bucciarelli, 2004] M. Franssen and L. L. Bucciarelli. On rationality in engineering design. Journal of Mechanical Design, 126, 945-949, 2004. [Habermas, 1984] J. Habermas. The Theory of Communicative Action, Vol 1: Reason and the Rationalization of Society. Boston, Beacon Press, 1984. 30 Even this does not always have to be the case; some participant may have ‘strategic’ reasons for letting a design project fail.

600

Peter Kroes, Maarten Franssen and Louis Bucciarelli

[Habermas, 1987] J. Habermas. The Theory of Communicative Action, Vol. 2: Lifeworld and System — A Critique of Functionalist Reason. Beacon Press, 1987. [Hampton, 1984] J. Hampton. The failure of expected-utility theory as a theory of reason. Economics and Philosophy, 10, 195-242, 1984. [Hargreaves et al., 1992] H. S. Hargreaves, M. Hollis, B. Lyons, R. Sugden, and A. Weale. The Theory of Choice: A Critical Guide. Blackwell, 1992. [Hooker and Streumer, 2004] B. Hooker and B. Streumer. Procedural and substantive practical rationality. In The Oxford Handbook of Rationality. A. R. Mele and P. Rawling, eds., pp. 57–74. Oxford University Press, 2004. [Houkes et al., 2002] W. N. Houkes, P. E. Vermaas, C. H. Dorst, and M. J. de Vries. Design and use as plans: an action-theoretical account. Design Studies, 23, 303-320, 2002. [Hoyningen-Huene, 1987] P. Hoyningen-Huene. Context of discovery and context of justification. Studies in History and Philosophy of Science, 18, 501-515, 1987. [Hume, 1969] D. Hume. A Treatise of Human Nature. Penguin, 1969 (1739-40). [Kroes, 2002] P. Kroes. Design methodology and the nature of technical artefacts. Design Studies, 23, 287-302, 2002. [Levi, 1980] I. Levi. The Enterprise of Knowledge. MIT Press, 1980. [Mele and Rawling, 2004] A. R. Mele and P. Rawling, eds. The Oxford Handbook of Rationality. Oxford University Press, 2004. [Nicolai, 1975] L. M. Nicolai. Fundamentals of Aircraft Design. University of Dayton School of Engineering, 1975. [Nozick, 1993] R. Nozick. The Nature of Rationality. Princeton University Press, 1993. [Pahl and Beitz, 1996] G. Pahl and W. Beitz. Engineering Design: A Systematic Approach. Springer Verlag, 1996. [Popper, 1959] K. R. Popper. The Logic of Scientific Discovery. Hutchinson, 1959. [Reichenbach, 1938] H. Reichenbach. Experience and Prediction: An Analysis of the Foundations and the Structure of Knowledge. University of Chicago Press, 1938. [Richardson, 1994] H. Richardson. Practical Reasoning about Final Ends. Cambridge University Press, 1994. [Roozenburg and Eekels, 1995] N. F. M. Roozenburg and J. Eekels. Product Design: Fundamentals and Methods. Wiley, 1995. [Searle, 2001] J. R. Searle. Rationality in Action. MIT Press, 2001. [Segerberg, 1992] K. Segerberg. Getting started: beginnings in the logic of action. Studia Logica, 51, 347-378, 1992. [Simon, 1969] H. A. Simon. A behavioral model of rational choice. Quarterly Journal of Economics, 69, 99-118, 1969. [Simon, 1968] H. A. Simon. Rationality as process and as product of thought. American Economic Review, 68, 1-16, 1968. [Simon, 1982] H. A. Simon. Models of Bounded Rationality. MIT Press, 1982. [Simon, 1983] H. A. Simon. Reason in Human Affairs. Stanford University Press, 1983. [Simon, 1984] H. A. Simon. The structure of ill-structured problems. In Developments in Design Methodology, N. Cross, ed., pp. 145-166. Wiley, 1984. [Simon, 1996] H. A. Simon. The Sciences of the Artificial. 3rd ed. MIT Press, 1996 (1969). [Suh, 2001] N. P. Suh. Axiomatic Design: Advances and Applications. Oxford University Press, 2001. [Thurston, 1991] D. L. Thurston. A formal method for subjective design evaluation with multiple attributes. Research in Engineering Design, 3, 105-122, 1991. [Vicenti, 1990] W. G. Vincenti. What Engineers Know and How They Know It. Johns Hopkins University Press, 1990. [Vincenti, 1992] W. G. Vincenti. Engineering knowledge, type of design, and level of hierarchy: further thoughts about what engineers know... In Technological Development and Science in the Industrial Age, P. Kroes and M. Bakker, eds., pp. 17-34. Kluwer Academic Publishers, 1992. [Wallace, 2003] R. J. Wallace. Practical reason. The Stanford Encyclopedia of Philosophy (Winter 2003 Edition), E.N. Zalta, ed., http://plato.stanford.edu/archives/win2003/entries/practicalreason/. [von Wright, 1963] G. H. von Wright. Practical inference. The Philosophical Review, 72, 159179, 1963.

DESIGNING SOCIO-TECHNICAL SYSTEMS Johannes M. Bauer and Paulien M. Herder

1

INTRODUCTION

A diverse body of research in engineering and the social sciences documents the working of systems that require technical artifacts and social arrangements to function. Single plants, firms, or entire industrial sectors constitute socio-technical systems if technological components and social arrangements are so intertwined that their design requires the joint optimization of technological and social variables. The concept of a socio-technical system originated in studies of coal mining in post-World War II Britain [Trist and Bamforth, 1951; Emery, 1959; Trist, 1981]. In contrast to previous studies that had often considered technology as an independent force to which labor had to adapt, the organizational and labor studies influenced by the socio-technical approach emphasized the close interdependence of the social and technical subsystems. Detailed empirical studies formed the starting point for the development of design principles for socio-technical systems, such as compatibility between design process and its objectives; minimal critical specification of tasks, roles, and objectives; and the control of variances as close to the point of origin [Cherns, 1976]. Part expression of the art of design and part normative statements of values, these principles formed an initial set of guidelines for the design of socio-technical systems. Although an analytically precise definition is difficult to formulate, for the purposes of this chapter socio-technical systems will be operationalized as arrangements of multiple purposive actors and material artifacts interacting in ways that require analyzing the total system and not just the constituent subsystems (see [Ropohl, 1999] for a more detailed discussion). Depending on the level of analysis and the research questions asked, each subsystem can be further disintegrated to dissect its logic and internal dynamics. Each subsystem aims to meet its own objectives, by using its own means, but is also in an interdependent relation with other subsystems. For example, technology was designed and built by purposive agents, acting within specific institutional settings, who continue to directly and indirectly shape its future. Likewise, social arrangements, for example, the setting up of decentralized energy trading markets, are in part contingent upon technological advances that support and enable them. As a result of this interdependence, technology and social arrangements co-evolve, each enabling and constraining, but not fully determining, the other sub-system [Murmann, 2003]. This interdependent relation unfolds in real, irreversible time, often resulting in a unique path Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

602

Johannes M. Bauer and Paulien M. Herder

of socio-technical development with inter-temporal dependencies (“path dependence”) [Arthur, 1994; David, 2000]. Infrastructure sectors can be considered as a particular class of socio-technical system [Kroes et al., 2006; Ottens et al., 2006]: technology is central for their operations and both organizational as well as sectoral forms of social control are established to ascertain a range of public values associated with their operation, such as ubiquitous and affordable supply. Engineering and social design issues arise at multiple levels of these socio-technical systems. Modern societies are heavily dependent on the services of multiple infrastructures (a term originally used by the military to refer to supporting transportation and logistical functions). Without reliable and sufficient supply of energy and water social and economic life would collapse quickly. Transportation and communication systems are required to coordinate the ever more differentiated tasks and the related flows of goods, services, and people that go hand in hand with increased specialization. The services of other infrastructure systems, such as sewerage, or waste removal, are similarly indispensable for a high quality of life and the overall well-being of society. The technical and social organization of infrastructure industries is strongly influenced by public values [Bozeman, 2007]. Many of these public values remained remarkably stable over time but the ways in which they are pursued have changed substantially. Worldwide, the traditional system built around strong government intervention and monopolistic industry organization has been superseded by market-based approaches in which government assumes regulatory rather than owner-operator functions. Infrastructure liberalization (the opening of market entry for new service providers) and sector unbundling (the separation of the stages of the value chain, for instance, electricity generation, transmission and distribution) have increased the number of participants and created multi-stakeholder environments. The reasons for these changes are manifold but illustrate the interdependence of the technical and social subsystems. Without significant changes in technology, such as the deep diffusion of information and communication technologies that facilitate decentralized control and management of complicated infrastructure systems, policy changes would not have been feasible. On the other hand, without changing sector organization some of the latent innovation potential might not have been realized. These transformations have increased the social and engineering complexity of infrastructure systems and hence the criticality of their design. Despite the new challenges for the design of infrastructure systems, appropriate comprehensive design processes and methods are still lacking. Important design decisions at the technical and social level are, consequently, often made without a clear view of the overall implications of these decisions for the development of the socio-technical system as a whole. In recent history, this is vividly illustrated by the severe problems and disruptions during the early phases of electricity reform in California during 2000-2001 [de Bruijne, forthcoming]. There is not even a consensus as to the prospects and limits of all-inclusive design in socio-technical systems. The majority of the disciplines that presently influence infrastructure de-

Designing Socio-Technical Systems

603

sign — in particular engineering, economics, management science, law, and public administration — assume, explicitly or more often tacitly, that effective solutions to infrastructure design issues can be found and implemented. At the other end of the spectrum is the view that large infrastructure systems cannot be designed in a rational manner at all. Due the overwhelming complexity and challenges of the optimization problem, it is argued, comprehensive design and control in the classical sense are beyond the reach of social planners and policy makers. In between these opposite positions are authors with a more nuanced view of the overall controllability of socio-technical systems. The desire to devise comprehensive solutions is looked at skeptically as “constructivist fallacy”: neither the information to design such systems nor the capacity to systematically explore all interrelations and contingencies is available. Nonetheless, a piecemeal, more localized and incremental, approach is deemed possible with ample room for deliberate design choices both in the social and technical subsystems. This chapter examines the state of research and knowledge on these issues. In addition to providing a broad framework, it reviews principles for the design of such systems that have been developed by a variety of disciplines during the past decades. The next section discusses the scope of economic, legal, and social design considerations that a prescriptive theory of infrastructure design will have to address. Section three reviews different theoretical frameworks that have or could be used to conceptualize infrastructure design issues. The implications of these considerations for the design of socio-technical systems are taken up in the fourth section. Conclusions and a brief outlook are presented in the final section. Two Intermezzi, one on Syngas and one on the Internet, illustrate the conceputal arguments with particular cases. 2

SCOPE OF DESIGN ISSUES IN SOCIO-TECHNICAL SYSTEMS

Before theoretical frameworks will be addressed in more detail in the next section, it is necessary to clarify key structural features of socio-technical systems and the scope of design issues that arise in such systems. Social and technical subsystems are intertwined and each has multiple layers that are designed and evolve on different time scales. Multi-level systems have been more widely studied by social scientists than by engineers, for example, in institutional approaches, some dating back to the late eighteenth century. Williamson [2000] offered a useful model that allows treating different types of social and institutional arrangements in an integrated fashion. The four layers in this framework are analytically defined. In multi-layer systems, top-down and bottom-up causation interact: upper levels enable and impose constraints on lower levels and vice versa. The approach can be expanded to model the technical and social sub-systems simultaneously, as depicted in Table 1. Changes in the institutional and technical arrangements at these layers, whether the outcome of purposive design choices or emergent phenomena, follow different time patterns. At the lowest layer of the system, continuous decisions regarding

604

Johannes M. Bauer and Paulien M. Herder

resource allocation and operation are made. One layer up, the governance structure of a society (the “play of the game”), most importantly various contractual arrangements, is specified. Specific methods of regulation (e.g., cost-of-service versus price cap regulation), ownership decisions (e.g., state, private, or hybrid), and market design will be defined at this layer, ideally aligning governance structures with the nature of transactions. Markets, hierarchies, and networks are important forms of the broad range of available governance structures. Change at that level unfolds over periods of one to ten years. The next higher layer defines the institutional environment, the “formal rules of the game”. Important design decisions at this layer encompass, among others, the organization of a polity, the organization of sector-specific regulation, and the general definition of property rights. Change on this layer is even slower than at the governance layer with some processes lasting up to a century. Finally, the layer of social embeddedness reflects informal institutions, customs, traditions, norms and religion. Change can take very long time periods, even hundreds or thousands of years. These arrangements are often not designed but emerge from interactions at lower levels of the system. Table 1. Layers and time scales in socio-technical systems Time scale Embeddedness Changes 102 to 103 years often non-calculative Institutional environment Changes 10 to 102 years, institutional setting Governance Changes 1 to 10 years design of efficient government regime Operation and Management Continuous adjustments

Social subsystem Informal institutions, customs, traditions norms, religion

Technical subsystem Informal conventions embedded in the technical artifacts

Formal rules of the game (property, polity, judiciary, . . . ) Play of the game (cotracts, governance of transactions)

Technical standards, design conventions technological paradigms Protocols and routines governing operational decisions and (best available) technology

Prices, quantities incentives

Operational choices

Note: inspired by [Williamson, 2000]. A correspondence can be established between the layers of the social system and the structure of technical artifacts (see Table 1). At the lowest layer of the technical subsystem, continuous operational decisions are made in response to its state. The nature of these decisions is dependent on the specific technical system. Managers of electricity grids need to balance load and supply; controllers of transport systems need to organize traffic flows; and control algorithms in communication

Designing Socio-Technical Systems

605

systems need to route and prioritize traffic according to the quality-of-service requirements of various applications. These decisions may be taken by human agents and hence be directly linked to the social system, or they may be automated based on pre-specified routines and technical protocols. In this latter case, they are indirectly linked to the social system. At the next higher layer, decisions as to how these technical artifacts are designed are made. These include both the architecture of the physical systems as well as the control processes for these systems. At the third layer, corresponding to the institutional environment in the social subsystem, decisions relating to the broad parameters of the technical solutions are taken. These may include arrangements such as patent laws, national and international standard-setting mechanisms, and the adoption of conventions for the design of technologies. The highest layer reflects tacit technical conventions and prior design decisions, as described by Hughes [1983] as characteristic for later stages of the development of a technology. In this socio-technical multi-layer system, bottom-up and top down enabling and constraining relations co-exist with horizontal ones between the respective social and technical layers. Moreover, “diagonal” forms of influence connect higher social and technical layers with lower layers in the respective other system and vice versa. Design decisions are made at all layers but the scope for such choices is generally broader at the lower layers. Consequently, in higher layers of the socio-technical system, deliberate design decisions become less prevalent and emergent characteristics become more important. Continuous and specific design choices are made at the operational and management layer. These design decisions are constrained by design choices at the governance layer. Design decisions are also made at that layer although the decision-makers typically are different. Rather than individuals and managers in organizations, governance decisions are made by agents in government agencies, standards bodies, non-government organizations, business associations, and other stakeholders that legitimately make collective, multilaterally or bilaterally binding, decisions. In turn, they are enabled and constrained by design decisions at the next-higher institutional layer. For example, the constitution of a nation or statutes may privilege certain forms of ownership of infrastructure networks or stipulate the mandate of regulatory agencies. Constitutions are typically designed so that they can only be changed with qualified majorities, adding additional inertia to change at this layer. At the highest layer, most characteristics are emergent. Emergence refers here to not explicitly intended or unexpected characteristics or behavior of the system. Although the notion of emergence is subject to much debate, see for example [Kroes, 2009] and [Mayntz, 2008a], it is helpful in contrasting the deliberate design decisions at lower levels with the non-deliberate outcomes and the resulting unpredictable behavior at higher levels. However, the source of what is labeled “emergence” may just be lack of thorough system knowledge; more complete theories and models may allow explaining these phenomena. Table 2 documents, in an exemplary fashion, elements of the matrix of design decisions that arise at the various layers. The fact that such design choices exist

606

Johannes M. Bauer and Paulien M. Herder

Table 2. Design decisions and emergence in socio-technical systems Time scale Embeddedness Mostly emergent Institutional environment Emergent/deliberate

Governance Deliberate/emergent

Operation and Management Deliberate

Social subsystem Tacit conventions and prior decisions Division of powers; assignment of jurisdiction; legal framework; general definition of property rights Ownership; form, organisation, and methods of regulation; market design (entry, number of licencees, etc. Regulation of prices and conditions, antitrust enforcement, social regulation

Technical subsystem Tacit conventions and prior decisions Selection of standards, technology selection architecture

Design of specific technical artifacts, protoocls and routines to govern operational decisions Execution of operational decisions

does not mean that they are actually made in a deliberate way, as they may also be done in routine or spontaneous fashion. Nor does it imply that they are in any form optimal. With each piecemeal choice, the conditions for subsequent decisions are being altered. These alterations may be reversible, reversible at a cost, or fully irreversible. Unless a decision is fully reversible, past choices will constrain the options for future decisions. The space of theoretically possible design options ranges from one, for instance if a chemical process works only in one particular way, to many alternatives, for example, with regard to the topology of networks or the organization of the governance of infrastructure services. Due to the constraints imposed by the components of the socio-technical system on each other, only a subset of this theoretically possible space is within the realm of feasible choice options. From that space, specific choices are made that, taken together, constitute a specific configuration of socio-technical design choices. To make these choices, an understanding of the working of the system and normative criteria guiding the design, such as efficiency or robustness, are necessary. However, as all purposive decisions are made in social settings, the process of decision-making and the participating stakeholders will also influence the outcomes. A rich political science literature explores these effects (for an overview see [Sabatier, 2003] and for an integrative analytical treatment [Tsebelis, 2002]). In the social domain, the combination of choices forms a particular institutional arrangement (or institutional

Designing Socio-Technical Systems

607

design). In the technical domain one could refer to a specific technical arrangement (or artifact design). Taken together, social and technical design realizations form a highly differentiated and complicated socio-technical arrangement (socio-technical design) with corresponding unique performance characteristics. INTERMEZZO 1 The Port of Rotterdam, The Netherlands, has a large petrochemical cluster that processes incoming crude oil into numerous end products. In the coming decades the cluster may find itself increasingly at risk of not being supplied with enough coal and crude oil, on which it so heavily relies. In order to safeguard the competitiveness of the cluster as a whole, it is important to reduce the dependency on fossil fuels by increasing feedstock flexibility [Herder et al., 2008]. As a solution to the feedstock inflexibility problem an industrial cluster feeding on synthesis gas has been proposed and designed [Stikkelman et al., 2006]. Synthesis gas (or syngas in short) is a mixture of carbon monoxide and hydrogen and is widely used for methanol and ammonia synthesis. Syngas is produced by gasifying carbon-containing feedstock such as coal, biomass, organic waste, crude oil, and natural gas. In addition to serving as a generic feedstock to power plants, hydrogen and carbon monoxide are important building blocks and intermediates in the petrochemical industry. Moreover, syngas is the main feedstock to produce FischerTropsch liquid transport fuels: in stead of refining crude oil to create petrol and diesel, these fuels are chemically synthesized from the building blocks in syngas. The designer fuels contain less to no sulphur and hence are more environmentally friendly than conventional diesel and petrol. Carbon monoxide and hydrogen also find other applications, for example in the direct reduction of iron, in which iron ore is reduced to metallic iron without using energy intensive blast furnaces. It is obvious that for the design of this energy infrastructure, the physical as well as a social subsystem has to be designed, and that both subsystems can be considered to be complex (with emergent behaviour, deep uncertainty, strong interaction between physical and social subsystem, many actors). Referring to Table 2, only the “Operation and Management” level and the “Governance” level are addressed in this design. In the proposed physical system’s design, which is approached from a technological determinism’s framework more than from a social shaping theory, network topologies such as ring, central bus and star networks can be considered. For the governance of this system, three archetypical structure types can be recognized: hierarchy, market or network structures. After confrontation of the physical with the social subsystem choices, a central bus system with a network governances structure was chosen as the basis for further design activities [Apotheker et al., 2007]. The final proposed design consists of a double bus network, with two different qualities of syngas, due to technical “Operation and Management” (lowest level, Table 1) considerations.

608

Johannes M. Bauer and Paulien M. Herder

The economic subsystem for this energy infrastructure (i.e. local syngas market design) comprises transaction systems through bilateral contracts or syngas trading on a syngas spot market. These design options are restricted by the technical design choices made, such as the double bus topology and the qualities of the syngas. Finally, applying Hughes’ theory on the development of large-scale systems and moving to larger time-scales in Table 1, it is obvious that the initial stages of the development of this energy infrastructure are shaped by engineers and entrepreneurs. Network topology, syngas qualities, decisions to design and construct large-scale gasifiers to produce syngas are among the most important decisions. Then, when the production and use of syngas takes off, the energy infrastructure may slowly expand and evolve mainly by its own momentum. 3

FRAMEWORKS FOR THE DESIGN OF SOCIO-TECHNICAL SYSTEMS

This section reviews several approaches that have been or could be used to design functional aspects of socio-technical systems (we will not discuss aesthetic design aspects). We briefly discuss, inter alia, constrained optimization approaches, systems approaches, and complexity theory. These approaches are not mutually exclusive but often complement each other. They differ with respect to their disciplinary foundations, their paradigmatic structure (the methods used and questions asked), their basic stance with regard to the possibility of deliberate socio-technical system design, and the specific forms in which such designs can be realized. Sociotechnical design issues often pose “wicked” [Rittel and Webber, 1973; Conklin, 2006], poorly defined and evolving problems. One way to address them is to narrow the problem space until design issues can be formulated as simpler problems (“puzzles”). As this will not always be possible, reliance on dynamic adaptive approaches may be the only workable approach.

3.1 Constrained optimization approaches A wide spectrum of methods to solve engineering and social design issues can be considered constrained optimization approaches. These methods have in common that a complicated and unwieldy problem is reduced to a manageable scale by focusing on variables that can be controlled or influenced. Other relevant factors are treated as independent, exogenous variables. Constrained optimization then maximizes or minimizes an objective function subject to possible values of the independent variables. In socio-technical systems nearly all decisions are constrained rather than unconstrained optimization problems. Constraints arise, among others, from physical features of the artifact; information constraints of the decision makers (incomplete information, asymmetrically distributed information, various forms of uncertainty); limitations of the decision-making process; constraints imposed by the multiple layers of socio-technical systems on each other; and constraints emanating from past choices that are not fully reversible. It may

Designing Socio-Technical Systems

609

be possible to use the simplifying (methodological) assumption that some of these exogenous factors do not change to explore relations between limited numbers of them (ceteris paribus clause). Various mathematical and other tools, ranging from linear and non-linear programming to computational modeling and scenario analysis can be utilized in developing solutions to constrained optimization problems. In this framework, a design (an engineering design, an institutional design, etc.) is effective if it is necessary and sufficient to cause a desired or prevent an undesired outcome. Sufficiency implies that, whenever a design is present, a certain outcome is also observable. Necessity means that this particular design can be observed whenever an outcome is present; however, it may also have other effects. This approach was elegantly formalized by Tinbergen [1952] and Theil [1964] for the field of economic policy. However, it can be restated to represent the essence of the constrained optimization approach to the design of socio-technical systems. Adopting the notation of Eggertsson [1998] the approach may be represented in the following way. A generic socio-technical design decision has four aspects: an objective function, a model of the system to be influenced, design variables, and factors external to the system. The objective function W = W (x) expresses societal preferences and/or engineering goals. The most general interpretation is that W captures the overall valuation of different states by society, in other words, a social welfare function. A model of the system x = f (a, z) specifies theoretical and empirical relations between instrument (design) variables a, outcomes x, and variables z that can be considered external to the system. Such instruments could be policy measures under the discretion of a policy-maker. For example, a regulatory agency may set the price for use of the electricity transmission grid or for access to local telecommunication networks. These instrument variables are part of a larger set of available choice options A(a ∈ A) that typically also include other instruments not relevant for a specific case. The external variables and parameters z are those aspects of the system that cannot, at a specific point in time, be controlled by the decision maker and are hence treated as exogenous to the design decision. For many short-term decisions, in particular at the operational level, z will include the characteristics of the installed technology base and the existing institutional setting. In the medium and long-run, technology and institutional arrangements will be at least partially endogenous, shaped by design choices. Depending on the structure of the problem, different methods, including analytical or computational methods, will be best suited to determine the values of instruments that maximize the objective function W (x∗ ). x∗ are the desired, optimal outcomes that maximize the respective objective function. The goal of socio-technical design is to find optimal instruments a∗ , which are dependent on desired outcomes x∗ and given external conditions z. More formally, a∗ = g(x∗ , z), that is, the choice of a∗ generates outcomes x∗ that maximize the objective function W ∗ = W (x∗ ) given the external conditions z. The constrained optimization view often tacitly assumes a division of labor between policy-makers, who determine W (x) and experts, who reveal the relevant

610

Johannes M. Bauer and Paulien M. Herder

theoretical and empirical relations f (a, z) and assist in the choice of the optimal policy instrument(s). In practice, this separation of roles is rarely maintained as experts are involved in setting goals and even the choice of instruments is not valueneutral (as claimed by many proponents of the means-end paradigm, which is one particular expression of the constrained optimization approach). This framework is often expressed in a mechanic and deterministic way [Mor¸c¨ol, 2002] but it can also be formulated in a probabilistic fashion to reflect incomplete information and uncertainty [Morgan and Henrion, 1990]. The default assumption is that it is possible to control and steer a socio-technical system. However, in principle the approach also allows for situations in which no sufficient instrument is known or where not all the necessary conditions for successful control may be met. In this case, the design problem has no known workable solution. More recent contributions have modified the basic model to take complications, in particular in the social subsystem, such as incomplete information, uncertainty and opportunistic behavior of agents, into account. These approaches abandon the view of policy makers and social designers as omniscient, omnipotent, and benevolent actors [Dixit, 1996]. Rather, all stakeholders are seen as motivated, at least in part, by their own self-interest. Under conditions of imperfect information, principles (e.g., policy-makers) typically have different information available than agents (e.g., managers of a regulated firm). A key challenge for design is to devise governance structures and processes that are incentive compatible (that is, truthfully reveal information only known to them). In this newer literature, in particular the research on mechanism design, the design of instruments and institutional arrangements becomes a more complicated, but not an impossible problem (e.g., [Hurwicz and Reiter, 2006; Laffont and Tirole, 1993]). One of the potential shortcomings of the approach is the assumption that the regularities underlying the working of the socio-technical system are immutable. This may be correct with regard to fundamental physical and possibly some social laws but is at least questionable with regard to other aspects of design, as deliberate choices, in particular at the upper layers of the system, often are made with the intent to change the working of the system. Institutional theories in the social sciences have long recognized the problem that individual decisions or markets are embedded and enabled by complex systems of tacit and formal rules (see, for example, [North, 1990; Ostrom, 2005; Greif, 2006; Zak, 2008]). Another aspect of this debate is the notion of performativity in economic sociology, pointing out that the world represented in theories and models is itself shaped by measures based on such theories [Callon, 1998; Aspers, 2007]. Seen from this perspective, the constrained optimization view does not pay sufficient attention to the fundamental endogeneity of the workings of social systems. However, in spite of these weaknesses, the model may be a workable approximation to find improvements over the status quo ante in situations that can be dealt with in a piecemeal way.

Designing Socio-Technical Systems

3.2

611

System approaches

Since the 1940s, social scientists and engineers have looked at the effects of largescale technology. Beginning with critical studies like those by Mumford [1963; 1967], the initial focus was on the uncontrollability and potentially devastating impact of large technology. Later the emphasis shifted on questions of whether social or technological forces were prime movers and the controllability of sociotechnical systems. Technological determinism and social shaping theory constitute nearly opposite positions, claiming a dominant effect of either the technical or the social subsystem on the trajectory of the entire socio-technical system. Proponents of technological determinism assert that the evolution of technology, which is largely seen as a discovery of existing laws and processes, determines social structures [Chandler, 1995]. Social structures and processes can only adapt to successive generations of technology. In strict versions even the design choices in the technical subsystem are limited, as they follow from the technological principles. In less radical formulations, technology allows design choices but these technical choices, in turn, determine the evolution of the social subsystem. Social shaping theory, on the other hand, emphasizes the decisive role of social factors in the evolution and in particular the application of technologies [MacKenzie and Wajcman, 1985; Williams and Edge, 1996]. It is argued that technologies are always socially embedded and that critical choices emanate from the social subsystem. Much of social shaping theory focuses on the role of the state and government. However, the influence of social factors is also, for example, seen in the organization of R&D, standardization, and the development of applications and services. Whereas technology is not irrelevant, it is malleable and strongly shaped by social forces. This approach was further developed in the now highly popular science and technology studies (STS) school, which considers social and technological factors as a seamless web of interrelationships [Bijker, Hughes, and Pinch, 1987]. A middle ground in these discussions is occupied by theories originating from the study of large technical systems and the factors influencing their course [Hughes, 1983; Mayntz and Hughes, 1988; Hughes, 2004; Mayntz, 2008c]. In Hughes’ model, design choices by engineers and individual entrepreneurs are decisive during the early stages of the development of a large technical system. As the system expands to ever wider geographic reach it develops its own inner logic (“momentum”) and design choices are less influential. The approach offers a useful metaphor and organizing framework to examine the evolution of network infrastructure industries (see the discussion in [Joerges, 1988] and [Sawhney, 2001]). Subsequent studies found that the specific historical trajectories of large technical systems do not seem to follow just one pattern but that different paths exist for different infrastructures and different contexts [Joerges, 1999]. Earlier approaches to the theory of large technical systems did acknowledge but not fully integrate the interaction between the technical and the social subsystems. For example, Hughes [1983] explores the interactions between technical artifacts and the social system. Perrow [1994] is

612

Johannes M. Bauer and Paulien M. Herder

even more explicit in his attention to the social aspects and in particular the role of interests and power. More recent approaches have explicitly integrated the role of agency and contexts in influencing outcomes (see, for example, [Sawhney, 2003; Ottens, et al., 2004; Geels, 2005]). Comprehensive system thinking dates back to writers in the eighteenth century, who compared societies to organisms (e.g., [La Mettrie, 1748/1912]). Cybernetics [Wiener, 1948] inspired Parsons’ [1951] structural functionalism. Further attempts at a systematic theory were made with general systems theory [Bertalanffy, 1968] and mathematical systems theory [Mesarovic and Takahara, 1975]. In Germany, Luhmann [1995] and his collaborators developed a unique version of systems theory with a strong emphasis on communication processes within and between subsystems. All these approaches have in common that the reproduction of the system imposes certain functional requirements. Effective design is only possible in as far as it is compatible with system logic and functional requirements [Schneider and Bauer, 2007]. Systems models attempt to understand the dynamic processes generated by the interaction of component subsystems, which in turn may consist of interacting subsystems. In that sense they are a good match to the problem structure of multi-layer systems found in socio-technical systems. Whereas system theory is not predominantly a theory of design, its insights can inform the actions shaping socio-technical systems, at least at a conceptual level. For example, it points out that differently structured systems may yield similar overall performance characteristics (“functional equivalence”, see [Ropohl, 1999]). This would suggest that no overall superior design of a socio-technical infrastructure system, for example, a fully deregulated market organization, may exist. Rather, alternative approaches will have different implications for system performance. The notion of System-of-Systems (SoS) is another response to the need to better capture the social aspects of technical systems and to better account for actor behavior in socio-technical systems [Sage, 2001; DeLaurentis, 2004; Boardman, 2006]. The SoS concept is not just a “box-in-a-box” model. DeLaurentis [2004] argues that SoS have the following three traits that distinguish them from regular systems: (1) they are geographically distributed; (2) their overall functionality is primarily dependent on linkages between distributed systems; and (3) the systems are heterogeneous, especially because of the inclusion of sentient systems, such as thinking and evolving individuals or organizations. The SoS paradigm requires designers to consider the system that is studied or designed from a higher system level, i.e. the upper layers in Table 2, since these are the layers where impacts of changes at the lower layers of the system are most prominently observed. An important consequence of the system’s heterogeneity is that higher system levels often display unpredictable behavior. Decision making and designing in the SoS paradigm requires an approach that cuts across various domains, combining, for instance, economic decisions with engineering design and policy making without losing the strengths of either modeling domain [De Bruijn and Herder, 2009]. Currently, an important bottleneck for proper SoS design is the lack of a common framework or lexicon [DeLaurentis, 2004]. Using a common lexicon will allow

Designing Socio-Technical Systems

613

designers to switch perspectives in a timely fashion instead of trying to force the paradigm of one domain into the straight jacket of another one. Agent-based modeling and “serious games” are emerging SoS modeling and design tools [De Bruijn and Herder, 2009]. As relatively new methods, the first tends to oversimplify actors’ behavior in the SoS whereas serious gaming is likely to unduly downgrade the engineering complicatedness of the SoS.

3.3

Complexity theory

Complexity theory originated in the physical and biological sciences and was successively applied to social systems in an attempt to understand dynamic processes which were difficult to explain with prevailing equilibrium models [Rosser, 1999; Beinhocker, 2006]. It has only recently been applied to problems related to the governance and design of socio-technical technical systems (see [Longstaff, 2003; Mitleton-Kelly, 2003; Cherry, 2007; Schneider and Bauer, 2007; Bauer and Schneider, 2008; Duit and Galaz, 2008]). Scholars in this tradition recognize that such systems can operate in different states. For example, Kauffman [1993; 1995] distinguishes order, edge-of-chaos, and a chaotic state. Orderly regimes can be stable or oscillate between two or more positions. Whereas orderly regimes are predictable, the state of edge-of-chaos and chaotic regimes cannot be forecasted with accuracy. Nonetheless, the general position of the system may be known [Mor¸c¨ol, 2002, p. 156]. Complex systems may undergo phase transitions. Orderly systems may become chaotic; conversely, chaotic systems can become orderly. Complex systems often exhibit non-linear dynamic behavior. They show a high degree of diversity and agents in the system are connected via multiple flows over networks of nodes and connectors [Holland, 1995; Colander, 2000; Axelrod, 1997]. This may lead to emergent behavior, i.e. overall complicated system behavior that transpires out of simple lower system level behaviors and rules. In socio-technical systems, complexity is introduced predominantly in the social subsystem but it also may be found in the engineering aspects. Due to the multiplicity of links in complex adaptive systems, the limited ability of actors to influence the overall conditions of the system, the adaptation of actors to changing system conditions, and the unpredictability of the system, effective socio-technical designs are difficult if not impossible to determine. As designs and interventions are rarely based on a full understanding of all the relevant interactions and dynamic effects, specific choices often also have unanticipated effects. Only in rare circumstances (“leverage points”) will it be possible to design and implement effective comprehensive designs although even in these cases the full implications of choices may only be realized in hindsight. One of these leverage points is the overhaul of the legal and regulatory framework of a sector (“constitutive moments”, see [Starr, 2004]). In most other conditions, specific designs will at best “nudge” the overall system in the desired direction [Brock and Colander, 2000], with the overall effect modified by positive and negative feedbacks.

614

Johannes M. Bauer and Paulien M. Herder

The complexity lens does not necessarily provide radically new and different answers to the problem of socio-economic design but it contributes additional insights. It has not yet developed a fully articulated prescriptive framework for the design of systems. Although the notions of unexpected outcomes and non-linear phenomena are common in traditional engineering disciplines (see [Kroes, 2009]), complexity theory broadens this perspective considerably. Like system theory, it highlights the importance of the overall rules, within which a sector evolves, on its performance, without claiming that there is one preferred set of rules (e.g., an “unregulated” market). Complexity is a matter of degree. If an industry is in static equilibrium or in a steady state expansion path, insights gained from complexity theory would converge with the results of constrained optimization models. However, if these conditions do not hold — and recently deregulated infrastructure industries are most likely not in such an equilibrium state — it points to aspects that are often overlooked by other approaches. The emphasis on unpredictability challenges traditional notions of design. In extreme cases, purposive design will not be possible. The theory of complex adaptive systems does, however, yield insights that can be used for the design of systems even in these situations. First, it contributes to the design process, where it encourages designers to systematically model all feedback effects and tenaciously look for possibly overlooked interrelations that might cause unintended consequences. Such systematic explorations are greatly facilitated by computer-based modeling techniques (e.g., [Koza, 2000; Sherman, 2000; Sawyer, 2005; Epstein, 2006]). Second, if alternative designs are available, it encourages such choices that create more resilient systems that can rebound from “normal accidents”, in particular in tightly coupled systems [Perrow, 1994] or by designing more modular organizations, processes, and products [Perrow, 2008]. One well-known example of the success of such design is the global Internet (see Intermezzo 2, p. 615). Third, the theory of complex adaptive systems identifies several processes to improce performance (“fitness” in the terminology of [Kauffman, 1983; 1995]). An “adaptive walk” strategy varies single features of the design and observes its effects on system performance. Only changes in design that improve performance are retained. Such strategies will gradually approximate a local optimum but may be insufficient to reach an alternative, possibly superior optimum if it would require incurring temporary efficiency losses. For example, realizing a more efficient overall energy supply system may require short-term inefficiencies during the reorganization of the system. In such cases, “patching”, the assignment of tasks to distributed units combined with some overarching coordination mechanism, might be a feasible strategy. For example, federalism can be considered a form of a patching mechanism: individual states may serve as laboratories for new policies from which a federal government can then pick successful approaches that are applied to the whole system [Cherry, 2008]. Such an approach may have desirable properties and enable the system to reach higher than just local optima.

Designing Socio-Technical Systems

615

INTERMEZZO 2: DESIGN AND EMERGENCE IN THE INTERNET The Internet is a multi-layered global network of networks. Its physical base is a heterogeneous and diverse set of specialized and general purpose communications networks. These comprise, for example, global, regional, and national backbone networks as well as local access networks. Whereas the backbone networks are fast digital electronic and/or optical networks, a larger variety of technologies is used in the access networks. Such access platforms can range from traditional twisted pair telephone lines (limited to fairly low access speeds) to various forms of wired and wireless broadband technologies allowing much higher data rates. Important wireline access technologies include digital subscriber line (DSL), cable modems, and fiber optical networks. Wireless access technologies comprise mobile, nomadic, and stationary platforms. This multitude of technical means of communications is integrated into a seamless, end-to-end, web by logical protocols — most importantly the TCP/IP suite of protocols — that reside on these technical artifacts. During its initial stages, although funded from government sources, the conventions at the heart of the logical Internet infrastructure emerged from voluntary forms of coordination among the pioneers of computing and data communications. Design choices, such as the end-to-end principle (resulting in a network that is essentially a dumb information transport infrastructure allowing the “intelligence” and applications to reside on the fringe of that network) or the numbering conventions of nodes on the network, were pragmatic responses to specific problems. As the network grew beyond a limited number of nodes, these early design principles were retained and shaped the rapidly expanding network. When the initial government operated network in the U.S. was privatized and increasingly operated by commercial enterprises in the 1990s, the informal governance mechanisms of the Internet were augmented by a more formal structure [Mueller, 2003]. Initially, this was achieved with the creation of the non-profit, U.S.-based Internet Corporation for Assigned Names and Numbers (ICANN). ICANN is assisted in its tasks, which include technical and operational aspects of the Internet as well as numbering and domain name conventions, by two supporting organizations, the Address Supporting Organization (ASO) and the Domain Name Supporting Organization (DNSO). Domain name administration is accomplished at the operational level by many private sector registrars. These are coordinated by five Regional Internet Registries (RIRs) such as RIPE for the European region or AfriNIC for Africa, which, in turn, cooperate in the Internet Assigned Numbers Authority (IANA). In the two U.N. sponsored World Summits on the Information Society (WSIS) in 2003 and 2005, a new global governance structure was added, assigning policy development to the Internet Governance Forum (IGF), which is organized as a multi-stakeholder policy dialogue. The Internet is also affected by design choices at the level of the supporting access networks. Operational choices are made by a large number of commercial firms, non-profit organizations, and government operators. These are in varying degrees regulated by national regulatory agencies, regional bodies, such as

616

Johannes M. Bauer and Paulien M. Herder

the European Commission, and international agencies such as the International Telecommunication Union (ITU) or the World Intellectual Property Organization (WIPO). These organizations cooperate with national organizations directly in working groups that define standards and operational principles as well as in policy-setting regional and global conferences, whose results are adopted with modifications into national laws and regulations. Content traveling on the Internet is furthermore heavily influenced by national political systems and laws governing the freedom of speech. Moreover, it is shaped by increasing concerns about information security [Zittrain, 2008]. The resulting socio-technical system was and is thus shaped by many decentralized decisions, coordinated and integrated at different layers of the system. Decisions at higher levels initially resulted from bottom-up forms of coordination. As the Internet grew in complexity, increasingly higher levels of governance were added adding a top-down direction of governance. In this process, past choices created many forms of path dependency, influencing subsequent choice options. The overall system emerges from these sequences of decisions but no single actor or group of actors controls the overall evolutionary path.

3.4 A comparative assessment These theories have widely differing consequences for the design of socio-technical systems. The dominant constrained optimization approach tacitly assumes that socio-technical systems can be controlled and that sufficient solutions to a design problem can be found and implemented. Depending on the diagnosis of the primary problem different solutions or mixes of solutions will be devised. Systems and complexity approaches are more cautious as to whether socio-technical systems can be fully controlled. Design of such systems is seen as an adaptive, incremental process, plagued by unanticipated events. Nevertheless, even these approaches see considerable room for deliberate design of social and technical aspects and the improvement of designs in physical and virtual (simulated) trial and error processes. With few exceptions it is typically recognized that design decisions are made under limited information and will have to be adapted as effects become visible and/or external conditions change. To realize overarching public values, SoS and in particular complexity approaches tend to see a larger and more effective role in designing the meta-conditions, the “order” of a sector rather than specific interventions at the operational level of socio-technical systems as manifested in the institutional and governance layers of the system as described in Tables 1 and 2. All approaches see ample room for artifact design. The constrained optimization approach may have been a reasonable simplification while infrastructure systems were organized as (state) monopolies. This setup gave social planners and designers broad control over the course of the industry. Even if planning and design mistakes were made, it was usually possible to come up with consistent approaches (if at the price of lower efficiency and higher cost). The reforms that started in the 1960s in the U.S. and in the 1980s in other parts of

Designing Socio-Technical Systems

617

the world have replaced the historical monopoly approach with a more open and competitive market environment. These measures have also complicated the coordination requirements and, for various reasons, reduced the effective control span of any of the players, including policy-makers. Socio-economic design decisions in the new environment will be made in a sequence of more partial and limited decisions. Only if the overall design problem can be segmented in a way such that every incremental local improvement will also contribute to improvements in the global performance of the system will the constrained optimization approach yield reliable outcomes. In general, the new reality of socio-technical infrastructure design is better reflected in multi-stakeholder system models and complexity theory. In practice it is also reflected in a shift from outcome-oriented forms of design to process-oriented forms of designing both institutional and technical system aspects.

4 NORMATIVE FOUNDATIONS, DESIGN GOALS, AND IMPLEMENTATION The previous section has identified overarching frameworks and explored whether and to what extent socio-technical systems can be designed and controlled. None of the reviewed approaches rejects the notion that deliberate choices can be made to design aspects of socio-technical systems although they diverge with respect to the ability of agents to influence the overall system and its dynamics. As purposive acts, design decisions are necessarily based on visions of the goals that should be realized [Bromley, 2006], even if that vision may not be articulated fully, and how it should be realized. This section reviews selected design goals and associated design/decision variables for socio-technical systems as formulated by engineers, economists, lawyers, and social planners. We also briefly discuss the relations between these goals and how possible tensions may be reconciled, if at all.

4.1

Overarching objectives

Design goals area formulated in multiple ways and amalgamated into more or less coherent systems of objectives. In infrastructure industries, important overarching and specific goals are settled in a political and social discourse, typically by players with different information and power to influence the outcomes. Such “public values” reflect a “normative consensus about (a) the rights, benefits, and prerogatives to which citizens should and should not be entitled; (b) the obligations of citizens to society, the state, and one another; and (c) the principles on which governments and policies should be based” [Bozeman, 2007, p. 17]. In that sense, public values rather than the more ambitious and vague notion of the “public interest” reflect the guiding visions of a social entity, such as local communities, regions, nations or a super-national regimes. Public values are not stable but change over time in response to general societal values, technological change, and stakeholder interests.

618

Johannes M. Bauer and Paulien M. Herder

Public values have a dual nature: they provide orientation but they may also be invoked opportunistically to justify actions that are motivated by private and special rather than public goals. Infrastructure industries, like other social and economic activities, abound with such opportunistic behavior of all private and public stakeholders (see [ten Heuvelhof et al., forthcoming] for a more extensive discussion). Because of opportunistic behavior and the limitations and challenges of socio-technical design, the practical implementation of public values and their specific operationalization may deviate from the intended effects. In that sense, socio-technical designers may fail to achieve stated and consented public values. This should predominantly be judged based on the outcomes of specific design choices rather than expressed motivations. Table 3 summarizes important specific design goals. Some of these goals are derived from public values and the associated public discourse. Part of this discourse draws on the findings of disciplines relevant for socio-technical design, such as engineering, economics, and law. Contributions from these disciplines are particularly important when broader goals (such as equitable supply) are operationalized as more specific objectives (for example, a specific universal service funding model). Economics enjoys a unique position among these disciplines. Many social and engineering decisions can be framed in terms of the benefits and costs associated with a specific course of action. Therefore, the economic approach offers a generic framework capable of dealing with engineering and social design issues in a unified framework. At least in principle, as long as a problem can be expressed in benefit-cost terms, economic analysis can deal with quantitative and qualitative aspects of socio-technical decisions in a commensurable way. Each engineering optimization problem has a dual economic optimization problem. Likewise, each solution to a social design problem has economic consequences and can also be expressed as an economic optimization problem. With the normative concepts of efficiency and welfare optimization, economics also has broad yardsticks to access design outcomes. Consequently, economics has played a major role in the infrastructure reform debates of the past decades. However, despite its theoretical elegance, in practice economic reasoning has serious limitations due to the ubiquitous prevalence of uncertainty, incomplete information, and the intangible nature of some public values that is often too elusive to determine costs and benefits. If costs and benefits cannot be expressed in monetary terms other forms of multi-factor optimization can be employed.

4.2

Specific design goals

From an engineering perspective, multiple specific design goals have been formulated, many of them related to the fundamental importance of the services of socio-technical infrastructure systems for society. These include technical efficiency, robustness, flexibility, safety, stability/security, resilience, modularity and controllability. Technical efficiency refers to the rate of the artifact to transfer

Designing Socio-Technical Systems

619

Table 3. Typical design goals and variables for socio-technical systems

Typical design goals

Typical design variables (examples)

Technical <————————————–> Social Engineering Economic Legal/Politicial • Technical • Efficiency • Constitutionality efficiency (technical, • Legality • Robustness productive, • Accountability • Flexibility allocative, • Transparency • Safety dynamic) • Justice • Stability/ • Adaptability/ • Equity security resilience • Universality • Resilience • Stability/ • Control of • Modularity security political power • Controllability • Universality • Sustainability • Sustainability • Control of market power • Technology • Market design • Laws • Network • Product/service • Regulatory topology • Production framework • Capacity/ method • Rights and throughput • Price regulation obligations • Feedstock • Organization of • Basic rights • Dimensions regulation system • Material • Competition law • Universal service • Standards obligations/ • Operating fund conditions • Divestiture of assets

620

Johannes M. Bauer and Paulien M. Herder

inputs into outputs, for example the processing of gas into electricity, or the transformation of voice into a digital signal. Through technology selection and the choice of the right operating conditions, this efficiency is typically maximized, whereby other design goals are often treated as constraints. The productively efficient (lowest cost) solution can be found among the technically efficient solutions, by assessing inputs and outputs at their economic value. Robustness and flexibility are concerned with the system’s capability to respond to changes in its environment. Robust systems, realized for example by overdimensioning an artifact, are able to continue functioning in the new environment without changing their inner layout, technology or workings. Flexible systems on the other hand respond and adapt to the changed environment, for example by changing operating conditions. Over-dimensioning may result in problems of high sunk costs (costs that cannot be recovered should a project be terminated). One way to adapt the system without being trapped by large sunk cost is by introducing and selecting the right standards, as standardization allows for modularization. Systems that are built from smaller modules can be changed and replaced and upgraded relatively easy without having to upset the entire system. In modular design, the most efficient way of interfacing is through the use of standards. Most of these goals relate to the operations and the design of the technical artifacts but some may require complementary social arrangements to be implemented effectively. For example, standardization might best be pursued if a standard, once developed, is mandated rather than adopted on a voluntary basis. The design of socio-technical systems is also strongly influenced by goals originating in political science and jurisprudence. The most important of these goals include constitutionality, legality, accountability, transparency, justice, equity, and universality of access. These goals have process-oriented aspects as well as substantive aspects. For example, constitutionality of an arrangement may imply that it corresponds to the substantive provisions of the respective constitution (e.g., with respect to the taking of private property in pursuit of public interest goals) or it may have procedural requirements (e.g., that a measure is formulated following constitutionally prescribed processes). Legal objectives are of particular importance for the design of the organizations entrusted with developing specific policies for socio-technical systems (e.g., regulatory agencies, ministerial departments, competition authorities). Designing the overall legal and regulatory framework (the “order” or “constitution”) of a market and an entire economic sector is one of the most important design tasks confronted presently in socio-technical systems. The school of constitutional economics has devoted particular emphasis to these issues (see [Vanberg, 2005] for a succinct discussion of the position) although it tends to underestimate the role of the other components of the socio-technical arrangement. Much of the regulatory reform debate, which is predominantly directed to redesigning the social framework of infrastructure industries, is based on economic concepts. Important goals stated by economists for the design of the institutional and sectoral arrangements are technical efficiency, productive efficiency, al-

Designing Socio-Technical Systems

621

locative efficiency, dynamic efficiency (innovation), adaptability/resilience, stability/security, universality of access to services, and sustainability. Some of these goals overlap with engineering objectives. This is not surprising, as many economic goals will either have to be implemented using a specific engineering solution, or can be achieved by choosing from engineering and economic solutions. Unless a technically feasible and cost-effective engineering solution is available, economic policy proposals are futile. Productive efficiency refers to producing any given output with the least-cost combination of inputs. Allocative efficiency requires, intuitively stated, that the mix of goods and services produced matches their valuation by consumers. Dynamic efficiency refers to the innovation rate in a system and the associated inter-temporal resource allocation decisions. Economists have traditionally focused on these efficiency goals and subjugated all design decisions to meeting these criteria. It is a fundamental theorem of welfare economics that under certain ideal conditions decentralized decisions by individual actors in competitive markets will optimize welfare, at least in the sense of a Pareto-optimum, a state in which nobody can be made better off without making somebody else worse off (e.g., [Friedman, 2002; Just et al., 2004]). Under other conditions, for example, the prevalence of externalities, public good characteristics, the existence of natural monopoly characteristics, this result does not hold and social intervention may move the system closer to the optimum. Moreover, even if a decentralized system works in principle, forms of market deficiency may require interventions to assure certain public values, such as universality of access to infrastructure services. Many of these policy choices will violate the relatively stringent assumptions of the Pareto criterion and have distributional impacts, hence create winners and losers compared to the status quo ante. Other welfare criteria, such as the KaldorHicks compensation test have been developed (see [Just et al., 2004]). The latter asks whether the beneficiaries of a decision were better off even after they were to compensate the losers of a decision (without requiring that such transfers actually take place). This is essentially the criterion underpinning cost-benefit analysis. However, as policy takes place within imperfect institutions and is implemented by imperfect actors, policy design itself may be flawed. Government or governance failure may jeopardize well-intended policy designs. Thus, under real world conditions, socio-technical design has to find an appropriate balance between imperfect markets and imperfect government. To a large part, sector design decisions, including whether to allow competition, how to support competition in market segments with strong monopolistic tendencies, and how to define the rights and obligations of the different actors, are made based on economic rationales. These design choices should also draw on legal and other bases, including political science thinking, when devising solutions to the assignment of duties to different organizations, the organization of the processes that support decision-making on an ongoing basis, and the methods of conflict resolution to adopt. Which basic rights system should be adopted (private property, commons, or open access) and how liability rules should be defined, if any, are

622

Johannes M. Bauer and Paulien M. Herder

also fudamental decisions. In a seminal paper, Coase [1960] pointed out that the assignment and specification of rights was irrelevant in the absence of transaction costs, as negotiations would allow finding an optimal solution. However, it is now widely recognized that under real-world circumstances, where transaction costs play a role, institutional choices do matter and have direct implications for overall sector performance and evolution.

4.3 Implementation After World War II, the predominant view was that government could control technology and social processes. However, with the failure of important programs during the 1970s, such as the fights against poverty, unemployment, and business cycles, this view was superseded by a more humble perspective, recognizing the limits to controlling socio-technical systems via government intervention. Social scientists also became more aware of the fact that, probably partially in response to government deficits, government control was increasingly complemented and in some cases replaced by other forms of social coordination, including self-regulation by stakeholders, co-regulation in which the public and the private sector collaborate, and interest group representation (e.g., in business associations and public interest groups) [Streeck and Schmitter, 1985; Latzer, et al., 2002]. The focus shifted from government to governance, an umbrella term referring to these multiple forms of purposive social coordination. In these emerging arrangements distinguishing between the object and the subject of governance is more difficult [Mayntz, 2008c]. In addition to the changes in the forms of local and national governance, sociotechnical design issues affecting infrastructure industries are increasingly addressed at super-national regional and global levels, although not all infrastructure industries are internationalized to the same extent. A wide range of global agreements and governance arrangements exist in information and communications infrastructures and transportation but in other areas, such as energy, more limited multilateral and regional arrangements continue to prevail. Regional organizations like the European Union have become strong players shaping infrastructure networks as have global organizations such as the International Telecommunication Union (ITU) or the World Trade Organization (WTO). At that level, it is even more complicated to identify the object and the subject of design decisions. For example, in intergovernmental organizations, the subjects and objects of decisions are identical (national governments). Many issues are discussed within the civic sector, non-governmental organizations that have no clear jurisdiction or power of enforcement [Mayntz, 2008b]. Consequently, the larger the number of players and the weaker their ability to design and control, the more likely it is that the set of policies that simultaneously meets all these requirements is small or even empty. Despite this blurring of traditional forms of government control, there is substantial evidence that large nation states remain key players [Drezner, 2007]. Against these dual backgrounds of the changing forms of social control and growing criticality of infrastructures for society, the issue of their controllability

Designing Socio-Technical Systems

623

deserves revisiting. One reason for the reduced controllability is the outcome of previous deliberate design choices: in most sectors some market segments have been partially or even fully deregulated. Whereas technical and social aspects could, in principle, be collectively designed, a choice was made to curtail such planning and shift decisions to decentralized firms and users. Measures directed toward the remaining regulated parts may thus be undermined by actions in the deregulated part of the industry. This is reinforced by adaptations in technology and the market organization of infrastructure industries, as visible, among others, in the phenomenon of convergence (the provision of multiple infrastructure services by one organization and the increasing (inter)dependence between infrastructure industries). The overall trajectory of the system is hence not controlled by engineers and social planners but emerges from the interaction of multiple stakeholders. Infrastructure industries have, therefore, evolved from a controllable monopoly system to a less controllable adaptive dynamic system [Bauer, 2004]. The explicit or implicit assumption of policy-makers was that the anticipated higher efficiencies of the more dynamic system outweigh its possible costs. Such costs include the heightened coordination requirements, the possible dynamic frictions and inconsistencies, greater difficulties of safeguarding public values, as well as the reduced ability to influence the overall evolution of the infrastructure system. The challenge for socio-technical design is to seek technical and social mechanisms that might influence the balance between benefits and costs in an advantageous way. Given the magnitude of the current reorganization, such an overall assessment is, at the time of writing, in many areas still outstanding. Another reason for potentially decreased control is that in the new environment sustainable socio-technical designs may be more difficult to find. Sustainable policies are the subset of measures that are capable of achieving the desired goals, are politically feasible, and economically viable. In other words they must be compatible with the constraints and interests of all stakeholders [Cherry and Bauer, 2004]. In a multi-stakeholder environment it is more difficult to identify policies that simultaneously satisfy all relevant constraints. This does not mean that policies and other design solutions will not be adopted at all but it increases the likelihood that challenges to policies will happen continuously, forcing decision-makers to frequently modify and adapt measures to changing circumstances and interest constellations. This difficulty may also affect finding a sustainable position with regard to the trade-off between short and long term effects of infrastructure reform. In sectors with highly durable, capital-intensive technology, the fluidity introduced by the multi-stakeholder environment may have undesirable consequences and may affect the incentives for investment and risk taking negatively. Unless appropriate adjustments are made, some of the stated goals of supporting dynamic efficiency and innovation may be inadvertently undermined due to these feedback effects. Worse, the relevant trade-offs between different goals may not be known with sufficient accuracy. In such situations, design choices will become real-world experiments with outcomes known only ex post. A design challenge is to better

624

Johannes M. Bauer and Paulien M. Herder

understand these dynamic trade-offs in advance, for instance, by using computer modeling techniques that shift the experiment into virtual design space, and to devise feasible and sustainable solutions. Several issues arise from these transformations: First, it is important to understand whether the multiplicity of stated goals can be realized at all. Goal conflicts and incompatibilities between different aspects of the socio-technical design will have to be examined with renewed vigor to find sustainable and overall consistent solutions. Second, socio-technical designers have to find out whether issues can be addressed at one level or whether changes in one area require cascading adjustments in related areas and thus can only be effectively addressed at multiple levels (as distinguished in Table 2) simultaneously. This has immediate implications for social and engineering design choices and the ability to implement them. Goals that can be realized at an individual layer call for different design approaches than those that require action at multiple layers. In some cases, for example, information security, meeting the goal at the individual firm level also implies that the goal is met at the sector level. Design can therefore focus on the individual firm level. This does not exclude that additional benefits might be realized by coordinating approaches also at a higher layer. In some cases, the most effective level to implement a design decision may be the sector level. For example standardization is best achieved at a level higher than an individual firm. In most cases, however, goals will need to be met at the level of the individual firm or actor and at the level of the entire industry simultaneously. Take, for example, the case of technical efficiency. The operation of an individual firm is technically efficient if it uses the least amount of resources to achieve a certain output level (or achieve the highest possible output with a given amount of resources). From individual firm technical efficiency does not necessarily follow that the entire sector is technically efficient. In fact, all individual firms could be sub-optimally small, leaving economies of scale and hence improvements of technical efficiency unutilized. Similar arguments may hold in the case of positive or negative externalities, where optimal individual level decisions nonetheless aggregate to socially sub-optimal outcomes. Third, once the structure of the problem and the principal design responses are known, the optimal socio-technical mix of measures to implement a goal or a set of goals as well as the layer on which it is best pursued need to be determined. Design methods and processes for socio-technical systems would, ideally, facilitate such a comprehensive perspective (see also the similar arguments in [Andrew and Petkov, 2007]). In our time scale model (see Table 1), engineering objectives will often be pursued at the operational and governance level. During the time of monopoly organization, many public values, such as non-discriminatory pricing across geographic regions, used to be pursued at the operational level. However, the more decentralized market organization that emerged during the past two decades does curtail many forms of effective control at that layer. As a consequence and also in response to new political visions as to the appropriate role of government in the economy (a change that has sometimes been

Designing Socio-Technical Systems

625

identified as a move from government to the regulation state) economic design decisions are now more often located at the second, governance layer. Legal goals are most often implemented at the second and third layers, the governance and institutional layers. There is increasing evidence that infrastructure systems (like many other socio-technical systems and social processes) can only be governed with a multiplexity of forms and instruments. Traditional government hierarchical control coexists with other forms of governance. Due to the multiplexity of governance, the entire system is in continuous motion, ever in need of adapting to changing circumstances and outcomes. 5 CONCLUSIONS This chapter has reviewed issues related to the design of socio-technical systems in general and of one particular class of such systems, infrastructure networks, in particular. The socio-technical systems approach was initially developed in the context of organizational studies. One of the key insights of the approach was that social and technical aspects of such systems needed to be optimized jointly. In principle, this method can be applied to the issues raised by infrastructure networks. At present, no overarching approach to designing such systems is available and a socio-technical systems approach might assist in closing that gap. However, the issues reach beyond devising a method of joint planning and design and might be rooted in inherent limits of designing large complex systems. The chapter first reviewed the multiplicity of design decisions that need to be made in such systems. Our review of alternative conceptual frameworks that might inform such choices revealed a bifurcation between theories that do not principally question the ability of planners and designers to shape socio-technical systems and theories that question full controllability. In the first group are models of constrained maximization, which currently dominate public policy formation. However, a very influential group of scholars argues that such deliberate design is not possible, that is it a constructivist fallacy. This stance does not deny that deliberate design choices can be made and need to be made but it questions the possibility of a comprehensive, outcome-oriented planning process. Instead, it emphasizes that technical and social design is much more limited in its overall impacts. It may be most effective in creating a metaframework, consisting of basic rules for standards, technologies, law, property, contract, and so forth, that allow decentralized, self-organizing forces to unfold. In this view, the overall outcomes and trajectory of the system are emergent, resulting from behavior at lower system levels, and cannot be fully planned or designed. Practical experience has generated ample evidence that design choices and planning can make a significant difference. Even if they may not be able to fully determine the overall trajectory of a socio-technical system, they matter and very often with serious consequences. Given the complexity of socio-technical arrangements, it is increasingly daunting to understand the correspondence between design choices and system responses. It would be desirable, to deepen conceptual

626

Johannes M. Bauer and Paulien M. Herder

and empirical knowledge of these correspondences. Not only should this assist in finding superior designs, it will also helps distinguishing situations in which deliberate design is possible and effective from those situations in which it may not be. Evolutionary models that allow for learning and adaptation are probably a promising step in this direction. The dominant paradigm informing design choices continues to use static maximization methods. However, socio-technical systems are dynamic, evolving systems and static maximization may be dynamically suboptimal or have ambiguous short and long term consequences. Rational designers of aspects of socio-technical systems would address these dynamic trade-offs explicitly. However, due to a lack of dynamic analyses this is rarely the case. Socio-technical designers can rely on a variety of available and emerging dynamic simulation and optimization models. Presently, their use is more widespread in transportation and energy but they are more broadly applicable. Given the complex nature of socio-technical infrastructure systems, these models will often not produce one right answer, but may support them in better grasping the correspondences between socio-technical design choices and a range of possible outcomes. Such an approach would allow to help avoid inconsistent socio-technical designs that have plagued infrastructure reform. Moreover, it should allow identifying the range of consistent socio-technical arrangements. If these are in the set of feasible options, the approach will facilitate selection of the most appropriate designs, including robust, resilient or “no regret” courses of action. BIBLIOGRAPHY [Andrew and Petkov, 2003] T. N. Andrew and D. Petkov. The need for a systems thinking approach to the planning of rural telecommunications infrastructure. Telecommunications Policy, 27, 75-93, 2003. [Apotheker et al., 2007] D. Apotheker, D. J. van der Elst, M. Gaillard, M. van den Heuvel, and W. van Lelyveld SPM4910 -Sepam Design Project: The Design of a Syngas Infrastructure in the Port of Rotterdam, Faculteit TBM, TU Delft, 2007. [Arthur, 1994] W. B. Arthur. Increasing Returns and Path Dependence in the Economy. University of Michigan Press, 1994. [Aspers, 2007] P. Aspers. Theory, reality, and performativity in markets. American Journal of Economics and Sociology, 66, 379-398, 2007. [Axelrod, 1997] R. M. Axelrod. The Complexity of Cooperation: Agent-Based Models of Competition and Collaboration. Princeton University Press, 1997. [Bauer, 2002] J. M. Bauer. Public utilities in the perspective of the “Gemeinwirtschaftslehre”. In An Institutionalist Approach to Public Utilities Regulation, W. Samuels, W.G. Shepherd and E. Miller, eds., pp 82-94, Michigan State University Press, 2002. [Bauer, 2004] J. M. Bauer. Harnessing the swarm: prospects and limits of communications policy in an era of ubiquitous networks and disruptive technologies,” Communications & Strategies, 54, 19-43, 2004. [Bauer and Schneider, 2008] J. M. Bauer and V. Schneider. Lessons from complexity theory for the Governance of large technical systems. Paper presented at the conference Complexity and Large Technical Systems, Meersburg, Germany, May 30-31, 2008. [Beinhocker, 2006] E. D. Beinhocker. The Origin of Wealth: Evolution, Complexity and the Radical Remaking of Economics. Harvard Business School Press, 2006. [Bertalanffy, 1969] L. von Bertalanffy. General System Theory; Foundations, Development, Applications. G. Braziller, 1969.

Designing Socio-Technical Systems

627

[Bijker et al., 1987] W. Bijker, T. P. Hughes and T. J. Pinch, eds. The Social Construction of Technological Systems: New Directions in the Sociology and History of Technology. MIT Press, 1987. [Boardman and Sauser, 2006] B. Boardman and B. Sauser. System of systems – the meaning of of, Proceedings of the IEEE/SMC International Conference, 2006. [Bozeman, 2007] B. Bozeman. Public Values and Public Interest: Counterbalancing Economic Individualism. Georgetown University Press, 2007. [Brock and Colander, 2000] W. A. Brock and D. Colander. Complexity and policy. In The Complexity Vision and the Teaching of Economics, D. Colander, ed., pp. 73-96, Edward Elgar, 2000. [Bromley, 2006] D. W. Bromley. Sufficient Reason: Volitional Pragmatism and the Meaning of Economic Institutions. Princeton University Press, 2006. [Callon, 1998] M. Callon, ed. The Laws of the Market. Blackwell Publishers, 1998. [Chandler, 1995] D. Chandler. Technological or Media Determinism. Aberystwyth, 1995. [Cherns, 1976] A. Cherns. The principles of sociotechnical design. Human Relations, 29, 783792, 1976. [Cherry, 2007] B. A. Cherry. The telecommunications economy and regulation as coevolving complex systems: implications for federalism. Federal Communications Law Journal, 59, 369402, 2007. [Coase, 1960] R. Coase. The problem of social cost. Journal of Law and Economics, 3, 1-40, 1960. [Colander, 2000] D. Colander, ed. The Complexity Vision and the Teaching of Economics. Edward Elgar, 2000. [Conklin, 2005] J. Conklin. Dialogue Mapping: Building Shared Understanding of Wicked Problems. New York, Wiley, 2005. [Davis, 2000] P. A. Davis. Path dependence, its critics and the quest for “historical economics”. In Evolution and Path Dependence in Economic Ideas: Past and Present, P. Garrouste and S. Ioannides, eds., Edward Elgar, 2000. [De Bruijn and Herder, 2009] J. A. De Bruijn and P. M. Herder. Systems and Actor Perspectives on Socio-Techucal Systems. IEEE Transations on Systems Man and Cybernetics, Part A — Systems and Humans, 2009. [De Bruijne, forthcoming] M. De Bruijne. Enron vs. FERC. In Strategic Behaviour in Network industries: A Multidisciplinary Approach, ten Heuvelhof, E., de Jong, M., Kars, M. and Stout, H. eds., Edward Elgar, forthcoming. [DeLaurentis and Callaway, 2004] D. DeLaurentis and R. K. Callaway. A system-of-systems perspective for public policy decisions. Review of Policy Research, 21, 829-837, 2004. [Dixit, 1996] A. K. Dixit. The Making of Economic Policy: A Transaction Cost Politics Perspective. MIT Press, 1996. [Drezner, 2007] D. W. Drezner. All Politics is Global: Explaining International Regulatory Regimes. Princeton University Press, 2007. [Duit and Galaz, 2008] A. Duit and V. Galaz. Governance and complexity-emerging issues for governance theory. Governance: An International Journal of Policy, Administration, and Institutions, 21, 311-335, 2008. [Eggertsson, 1998] T. Eggertsson. Limits to institutional reform. Scandinavian Journal of Economics, 100, 335-357, 1998. [Emery, 1959] F. E. Emery. Characteristics of Socio-Technical Systems. Tavistock Institute, 1959. [Epstein, 2006] J. M. Epstein. Generative Social Science: Studies in Agent-Based Computational Modeling. Princeton University Press, 2006. [Friedman, 2002] L. S. Friedman. The Microeconomics of Public Policy Analysis. Princeton University Press, 2002. [Geels, 2005] F. W. Geels. The dynamics of transitions in socio-technical systems: a multi-level analysis of the transition pathway from horse-drawn carriages to automobiles (1860-1930). Technology Analysis & Strategic Management, 17, 445-476, 2005. [Greif, 2006] A. Greif. Institutions and the Path to the Modern Economy: Lessons from Medieval Trade. Cambridge University Press, 2006. [Herder et al., 2008] P. M. Herder, I. Bouwmans, R. M. Stikkelman, G. P. J. Dijkema, and M. P. C. Weijnen. Designing infrastructures using a complex systems perspective. Journal of Design Research, 7, 17-34, 2008.

628

Johannes M. Bauer and Paulien M. Herder

[Holland, 1995] J. Holland. Hidden Order: How Adaptation Builds Complexity. Addison-Wesley, 1995. [Hughes, 1983] T. P. Hughes. Networks of Power: Electrification in Western Society, 18801930. Johns Hopkins University Press, 1983. [Hughes, 2004] T. P. Hughes. (2004) Human-built World: How to Think About Technology and Culture. University of Chicago Press, 2004. [Hurwicz and Reiter, 2006] L. Hurwicz and S. Reiter. Designing Economic Mechanisms, Cambridge University Press, 2006. [Joerges, 1988] B. Joerges. Large technical systems: concepts and issues. In The Development of Large Technical Systems, R. Mayntz and T.R. Hughes, eds., pp. 9-36, Campus, 1988. [Just et al., 2004] R. E. Just, D. L. Hueth, and A. Schmitz. The Welfare Economics of Public Policy: A Practical Approach to Project and Policy Evaluation. Edward Elgar, 2004. [Kauffman, 1983] S. A. Kauffman. The Origins of Order: Self Organization and Selection in Evolution. Oxford University Press, 1983. [Kauffman, 1995] S. A. Kauffman. At Home in the Universe: The Search for Laws of SelfOrganization and Complexity. Oxford University Press, 1995. [Koza, 2000] J. R. Koza. Genetic Programming: On the Programming of Computers by Means of Natural Selection. MIT Press, 2000. [Kroes, 2009] P. A. Kroes. Technical artifacts, engineering practice and emergence. In: Functions in Biological and Artificial World: Comparative Philosophical Perspectives, U. Krohs and P. Kroes, eds., pp. 277-292, MIT Press, 2009. [Kroes et al., 2006] P. Kroes, M. P. M. Franssen, I. R. Van de Poel, and M. M. Ottens. Treating socio-technical systems as engineering systems: some conceptual problems. Systems Research and Behavioral Science, 23, 803-814, 2006. [de La Mettrie, 1748/1912] J. O. de La Mettrie. Man a Machine. Open Court, 1748/1912. [Laffont and Tirole, 1993] J.-J. Laffont and J. Tirole. A Theory of Incentives in Procurement and Regulation, MIT Press, 1993. [Latzer et al., 2002] M. Latzer, N. Just, F. Saurwein, and P. Slominski. Selbst- und KoRegulierung im Mediamatiksektor. Alternative Regulierungsformen zwischen Staat und Markt. Westdeutscher Verlag, 2002. [Longstaff, 2003] P. H. Longstaff. The puzzle of competition in the communications sector: Can complex systems be regulated or managed? Program on Information Resources Policy Working Paper, Harvard University, 2003. [Luhmann, 1995] N. Luhmann. Social Systems. Stanford University Press, 1995. [MacKenzie and Wajcman, 1985] D. MacKenzie and J. Wajcman. The Social Shaping of Technology: How the Refrigerator Got its Hum. Open University Press, 1985. [Mayntz, 2008a] R. Mayntz. Emergence in philosophy and social theory. Unpublished manuscript. 2008. [Mayntz, 2008b] R. Mayntz. Governance in an internationalizing world. In G. F. Schuppert and M. Z¨ urn, eds., Governance in einer sich wandelnden Welt. Politische Vierteljahresschrift, 41, 2008. [Mayntz, 2008c] R. Mayntz. The changing governance of large technical infrastructure systems. Paper presented at the conference Complexity and Large Technical Systems, Meersburg, Germany, May 30-31, 2008. [Mayntz and Hughes, 1988] R. Mayntz and T. P. Hughes. The Development of Large Technical Systems. Campus, 1988. [Mesarovic and Takahara, 1975] M. Mesarovic and Y. Takahara. General Systems Theory: Mathematical Foundations. New York, Academic Press, 1975. [Mitleton-Kelly, 2003] E. Mitleton-Kelly. Ten principles of complexity and enabling infrastructures. In: E. Mitleton-Kelly, ed., Complex Systems and Evolutionary Perspectives on Organizations: The Application of Complexity Theory to Organizations, pp. 23-50, Pergamon, 2003. [Mor¸c¨ ol, 2002] G. Mor¸c¨ ol. A New Mind for Policy Analysis: Toward a Post-Newtonian and Postpositivist Epistemology and Methodology. Praeger, 2002. [Morgan and Henrion, 1990] M. G. Morgan and M. Henrion. Uncertainty: A Guide to Dealing with Uncertainty in Quantitative Risk and Policy Analysis. Cambridge University Press, 1990. [Mueller, 2003] M. Mueller. Ruling the Root. MIT Press, 2003. [Mumford, 1963] L. Mumford. Technics and Civilization. Harcourt, Brace & World, 1963. [Mumford, 1967] L. Mumford. The Myth of the Machine. Harcourt, Brace & World, 1967.

Designing Socio-Technical Systems

629

[Murmann, 2003] J. P. Murmann. Knowledge and Competitive Advantage: The Coevolution of Firms, Technology, and National Institutions. Cambridge University Press, 2003. [North, 1990] D. C. North. Institutions, Institutional Change, and Economic Performance. Cambridge University Press, 1990. [Ostrom, 2005] E. Ostrom. Understanding Institutional Diversity. Princeton University Press, 2005. [Ottens et al., 2006] M. M. Ottens, P. A. Kroes, M. P. M. Franssen, and I. R. Van de Poel. Modeling engineering systems as socio-technical systems. International Journal of Critical Infrastructures, 2, 133-145, 2006. [Parsons, 1951] T. Parsons. The Social System. The Free Press, 1951. [Perrow, 1994] C. Perrow. Normal Accidents: Living with High-risk Technologies. Princeton University press, 1994. [Perrow, 2008] C. Perrow. Modeling firms in the global economy: new forms, new concentration. Paper presented at the conference Complexity and Large Technical Systems, Meersburg, Germany, May 30-31, 2008. [Ropohl, 1999] G. Ropohl. Philosophy of socio-technical systems. Philosophy & Technology, 4, 59-71, 1999. [Ritchey, 2002] T. Ritchey. Modeling complex socio-technical systems using morphological analysis. Adapted from an address to the Swedish Parliamentary IT Commission, Stockholm, 2002. [Rittel and Webber, 1973] H. Rittel and M. Webber. Dilemmas in a general theory of planning. Policy Sciences, 4, 155-169, 1973. [Rosser, 1999] J. B. Rosser, Jr. On the complexities of complex economic dynamics. Journal of Economic Perspectives, 13, 169-192, 1999. [Sabatier, 2007] P. A. Sabatier, ed., Theories of the Political Process. 2nd ed., Westview Press, 2007. [Sage and Cuppan, 2001] A. Sage and C. D. Cuppan. On the systems engineering and management of systems of systems and federations of systems. Information, Knowledge, Systems Management, 2, 325-345, 2001. [Sawyer, 2005] R. K. Sawyer. Social Emergence: Societies as Complex Systems. Cambridge University Press, 2005. [Sawhney, 2001] H. Sawhney. Dynamics of infrastructure development: the role of metaphors, political will and sunk investment. Media, Culture and Society, 23, 33-51, 2001. [Sawhney, 2003] H. Sawhney. Wi-Fi networks and the rerun of the cycle. Info, 5, 25-33, 2003. [Schneider and Bauer, 2007] V. Schneider and J. M. Bauer. Governance: Prospects of Complexity Theory in Revisiting System Theory. Paper presented at the 65th Annual Conference of the Midwest Political Science Association, Chicago, 2007. [Starr, 2004] P. Starr. The Creation of the Media: Political Origins of Modern Communications. Basic Books, 2004. [Stikkelman et al., 2006] R. M. Stikkelman, G. P. J. Dijkema, P. M. Herder, and A. F. Correlj´ e. Van bloedgroep olie naar bloedgroep syngas. Petrovision, September, 62-67, 2006. [Streeck and Schmitter, 1985] W. Streeck and P. C. Schmitter. Private Interest Government: Between Market and State. Sage, 1985. [ten Heuvelhof et al., forthcoming] E. ten Heuvelhof, M. de Jong, M. Kars, and H. Stout, eds. Strategic Behaviour in Network industries: A Multidisciplinary Approach. Edward Elgar, forthcoming. [Theil, 1964] H. Theil. Optimal Decisions Rules for Government and Industry. North Holland, 1964. [Tinbergen, 1952] J. Tinbergen. On the Theory of Economic Policy. North-Holland, 1952. [Trist, 1981] E. L. Trist. The evolution of socio-technical systems. In Perspectives on Organization Design and Behavior, A.H. van de Ven and W.F. Joyce, eds., John Wiley, 1981. [Trist and Bamforth, 1951] E. L. Trist and K. W. Bamforth. Some social and psychologial consequences of the Longwall method of coal-getting. Human Relations, 4, 3-38, 1951. [Tsebelis, 2002] G. Tsebelis. Veto Players: How Political Institutions Work. Princeton University Press, 2002. [Vanberg, 2005] V. J. Vanberg. Market and state: the perspective of constitutional political economy. Journal of Institutional Economics, 1, 23-49, 2005. [Wiener, 1948] N. Wiener. Cybernetics or Control and Communication in the Animal and the Machine. Technology Press, 1948.

630

Johannes M. Bauer and Paulien M. Herder

[Williams and Edge, 1996] R. Williams and D. Edge. The social shaping of technology. In Information and Communication Technologies: Visions and Realities, W.H. Dutton, ed., pp. 53-67, Oxford University Press, 1996. [Williamson, 2000] O. E. Williamson. The New Institutional Economics: taking stock, looking ahead. Journal of Economic Literature, 38, 595-613, 2000. [Zak, 2008] P. J. Zak, ed. Moral Markets: The Critical Role of Values in the Economy. Princeton University Press, 2008. [Zittrain, 2008] J. Zittrain. The Future of the Internet and How to Stop It. Yale University Press, 2008.

Part IV

Modelling in Engineering Sciences

This page intentionally left blank

INTRODUCTION TO PART IV Sjoerd D. Zwart, associate editor A crucial characteristic, and one that sets technology and the engineering sciences apart from the natural sciences, is the fact that engineers are not only concerned with the production side of knowledge for the sake of artifacts but also with goal-oriented action based on that same knowledge. The engineering orientation towards action has important consequences for the philosophy of technology and for the engineering sciences. These consequences are especially apparent in Parts III, IV and V of this Handbook. Part III focuses on an important difference between science and technology in the form of design. Design has to do with the construction of artifacts or technical processes. Part V is also very connected to action because it deals with norms and values in technology and engineering. These norms and values steer action. The present part of the Handbook, Part IV, focuses on modeling in the engineering sciences. How is this modeling aspect related to action, the reader might ask, since models and modeling have been substantially researched in the field of the philosophy of the natural sciences? And how does the focus on action affect the character of engineering models and the ways in which these models are produced and used? Traditionally, philosophers of science discerned at least three important features of models: the representational facet, the abstraction from details side, and finally the goal or purpose of the model or the modeling activity (e.g., Stachowiak [1973, p.131–132]). It is this last characteristic, the goal of a model, which might shed light on the special character of engineering models, as in engineering the ultimate purpose of modeling is to realize reliable artifacts or technical processes. This contrasts substantially with the natural sciences where, conceptually at least, the aim underlying the modeling activities is to gain knowledge for knowledge’s sake. The practical aim of models in the engineering sciences will become apparent from almost all the contributions included in Part IV. It especially comes to the fore in the chapters on models as epistemic tools, model-based reasoning and scale models. Apart from the possible differences between modeling goals in the engineering and the natural sciences, we have at least two other reasons for wanting to focus on modeling in the present part: the first relates to the disappearance of the ‘technology as applied science’ paradigm, and the second has to do with renewed interest in the use and function of models in science. Traditionally, philosophers of science took technology to be nothing more than applied science [Bunge, 1966] and were therefore not very interested in the engineering sciences. Detailed research into the history of technology carried out by, amongst others, Edwin Layton, John Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

634

Sjoerd D. Zwart, associate editor

Staudenmaier and Walter Vincenti did, however, lead to the claim that technological knowlegde is relatively detached from scientific knowledge (see for more details the chapter by Houkes in Part II of this Handbook). This independence and autonomy challenged the notion that technology is an applied science and urged philosophers of technology to open the black box and see for themselves how engineering knowledge is generated. The second reason is the renewed interest in the use and function of models involved in knowledge production. It all started with [Morgan and Morrison, 1999] and it went on to produce many interesting approaches to and unconventional views on the functions and uses of models and modeling. Still, much of this renewed interest in models fails to consider the distinction between specific engineering uses of models, such as design and artifact behavior, and the use of models in the natural sciences. (See for a clear overview [Frigg and Hartmann, 2006].) The focus of the present part on modeling in the engineering sciences should therefore be contemplated against the background of the two developments just sketched. The eight chapters in Part IV revolve around three focal points concerning models: conceptual issues, empirical issues and methodological issues. In the first two chapters, we address the conceptual issues surrounding the notion of a model. The two chapters can be seen as complimentary. The first chapter is dedicated to an historical overview of the notion of a model. There Roland M¨ uller surveys the range of entities known as models ranging from the tools that are used to support manufacturing and design to various educational and mathematical models. In the second chapter, Wilfrid Hodges provides a systematic analysis of functional modeling and mathematical models. He considers the role of semantic models serving as intermediate entities between theories and systems in the case of various kinds of scientific and engineering practices. One aspect of the renewed interest in models has directly to do with the status of the semantic view of theories and the claim that the semantic view disregards the autonomous character of models. The second major focus of Part IV is on the use made of models by engineers when they design or develop further understanding of technical phenomena. Mieke Boon and Tarja Knuuttila conceive of models in the engineering sciences as epistemic tools and so present their approach as an alternative to the semantic view of models. From their pragmatic perspective the authors focus primarily on the modeling activity rather than on the models themselves. They illustrate their approach by referring to Carnot’s heat engine, which according to the authors is an important example of an engineering model with practical implications. In the fourth chapter on ‘Model-Based Reasoning in Interdisciplinary Engineering’, Nancy Nersessian and Christopher Patton study what the researchers themselves call the ‘model-systems’ of two biomedical engineering research laboratories (a tissue engineering and a neural engineering laboratory). The claim is made that model-based reasoning differs substantially from formal deductive reasoning, because it also depends on content and is thus not purely formal. The fifth chapter discusses engineering scaling methodologies and reveals that one of the founding fathers, William Froude, did not base his well-known scaling methodology on di-

Introduction to Part IV

635

mension analysis, as is commonly assumed. The important questions addressed are: What is the basis of Froude’s scaling method if the author himself did not found it on dimensionless numbers? What is the status of model laws within scale modeling? Do these laws have empirical content or are they merely analytical? The third and final focal point of Part IV is the methodological issues regarding modeling. In the sixth chapter, Susan Sterrett considers dimensional analysis in similarity-based reasoning. Although similarity and dimensional analysis remain important in philosophy, science and technology, they have not received the attention they deserve in contemporary philosophy of science studies. The chapter starts by explaining ratios, physical similarity and quantities, and arrives—via dimensions and coherent systems of units—at dimensionless homogeneity; it ends with Buckingham’s Π-theorem, and with partial similarity. In the process the author also deals with interesting questions such as, for instance, ‘what has logical priority: scales, units of measurement or the quantities measured?’ She bases the answer on Lodge’s concept of ‘quantity equations.’ The seventh chapter is closely connected to the previous one as it concentrates on measurement theory in engineering. Its author, Patrick Suppes, who is of course one of founding fathers of the representation view of measurement theory, characterizes a measurement procedure using two fundamental problems. The first is the representation problem which dictates that the structure of the phenomena must be the same as the structure of a set of numbers. Since the answer to the representation question does not uniquely determine the theory’s structure, an invariance theorem must be proved for the representation and that answers the second fundamental question relating to the determination of the measurement procedure’s scale type. After that the author gives the representation and invariance theorems for four scale types. The second part of this chapter deals with measurement errors, a compulsory topic for any theory of measurement applied to engineering practice. Since many technological explanations make use of models and simulations, these explanations form the core subject-matter of the last chapter of Part IV. Unlike scientific explanations, technological explanations have, in the past, attracted very little attention from philosophers of science. Therefore, in many respects this chapter is as much an attempt to explore the territory as to supply an adequate theory of technological explanation. At first sight, one might expect technological explanations to be similar to scientific explanations since both concern the physical processes that explain natural phenomena or the way a technical artefact works. Joseph Pitt argues, however, that technical explanations involve much more. For instance, if we want to explain how an artifact has come to be what it is we cannot confine ourselves to efficient causes, (to use the Aristotelian terminology), in the way that has become customary in scientific explanation. In addition, technological explanations need to allude to other Aristotelian causes: the material, formal, and final causes. The reason for this lies in the involvement of a multiplicity of factors, including social ones. Clearly, Part IV does not cover all the issues relevant to engineering models. At least three further areas come to mind. Firstly, there are many interesting

636

Sjoerd D. Zwart, associate editor

philosophical questions related to the numerical and approximating methods and techniques that pervade engineering sciences and practices. An important example is the finite element method (FEM), which has been developed to cope with the intractability of the analytical solution to the equations that describe the physics of airplane foils. Secondly, engineering models are often used as blueprints for the design, manufacturing and maintenance of technological artifacts and processes. In that respect the unified modeling language (UML) forms an interesting subject of philosophical analysis as UML functions as a general language for these blueprints. Thirdly, important topics for philosophical analysis are to be found in the domain of engineering simulation techniques of the kind frequently deployed for explanatory purposes. For the broader range of subjects and further philosophical reflection the reader is referred to the editorial introduction. BIBLIOGRAPHY [Bunge, 1966] M. Bunge. Technology as Applied Science. Technology and Culture, 7:329–347, 1966. [Frigg and Hartmann, 2006] R. Frigg and S. Hartmann. ‘Models in Science’. The Stanford Encyclopedia of Philosophy (2006 Edition), Edward N. Zalta (ed.), http://plato.stanford. edu/entries/models-science/. [Morgan and Morrison, 1999] M. S. Morgan and M. Morrison. Models as Mediators. Cambridge University Press, Cambridge, 1999. [Stachowiak, 1973] H. Stachowiak. Allgemeine Modelltheorie. Springer, Wien, 1973.

THE NOTION OF A MODEL: A HISTORICAL OVERVIEW Roland M¨ uller “The meaning of a word is its use in the language.” “What we do is to bring words back from their metaphysical to their ordinary use.” “A main source of our failure to understand is that we do not command a clear view of the use of our words.” Philosophical Investigations [Wittgenstein, 1953, §§43, pp. 116, 122].

1

INTRODUCTION

Even if we limit ourselves to science and technology, the range of uses of the word ‘model’ is vast — a lexicographer’s paradise. To bring this range under control, theoretically one should analyse how many meanings the word ‘model’ has, and how they are related to each other. But in practice this may be impossible to do, for at least the following three reasons. First, we can classify models either according to what kinds of object they are, or according to what they are intended to do. These two classifications often diverge. Two scientists can agree that something is a model, but use it for quite different purposes — for example to explain and to predict. Two engineers can use very different devices — for example a mechanical prototype and a computer simulation, both of them called ‘models’ — for one and the same purpose. Second, the boundaries between one kind of model and another are very often blurred. This is partly because the construction and use of models can depend quite deeply on habits and cultures. Also it is because models lie close to the creative edge of thought; any advance in science or technology is liable to throw up new applications of the word ‘model’. In short, the word ‘model’ is transdisciplinary and has an open range of application. Third, the literature on ‘models’ during the last half century contains several classifications that depend on philosophical doctrines which are at the very least contentious. For example some writers have suggested that a theory about some aspect of the world must be intended either as a true description or as an ‘analogy’. Other writers have suggested that we have a choice between taking models to be Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

638

Roland M¨ uller

‘descriptions’ and making them ‘non-linguistic’. (Frigg and Hartmann state the dichotomy [2008, sect. 1.1]; some authors take it to great lengths.) In this chapter we will survey the range of things called ‘model’, beginning with aids to manufacture and design, and moving on through scientific explanation to educational and mathematical models. Within these broad groupings we will base our arrangement on the historical development, rather than on more recent philosophical analyses. The chapter by Hodges is in some sense a complement to this chapter: it ignores the history and addresses some of the questions that arise if one tries to make a general theory of models [?]. We concentrate on the word ‘model’ itself, though we should add two reservations. First, many of the things called ‘model’ also have other names such as ‘structure’, ‘representation’, ‘hypothesis’, ‘plan’, ‘simulation’ etc. etc. We will note some of these as we proceed. Second, there are various things that are very similar to things called models and could probably have been called models, though in fact they aren’t or weren’t. In fact many things that later came to be called ‘models’ were widely used before they got this name. This is particularly true in the early history of the notion; today a new kind of model is likely to be called a ‘model’ from the outset. We have given a number of examples of these not-yet-baptised models, since they help to show why the need for the word ‘model’ arose. Many more examples can be found in the author’s archive [M¨ uller, 2008]. 2

THE ORIGIN OF THE WORD ‘MODEL’

The line of descent of the word ‘model’ has not been undisputed. According to mainstream etymological theory the word ‘model’ descends via the diminutives ‘modellus’ and ‘modulus’ from the Latin word ‘modus’, which means ‘measure’. ‘Modus’ derives from the Indo-European root ‘med-’ [Pokorny, 1949; Shipley, 1984] that also means ‘measure’, ‘to take appropriate mesures’. According to others, however, with whom the present author agrees, model does not derive from ‘modellus’, but from ‘modulus’. After all, long before 1300 we find in German ‘Modul’ and ‘Model’ (e g. in the courtly love songs), and in French ‘modle’ and ‘molle’, descending immediately from ‘modulus’ [von Wartburg, 1966]. From around the middle of the first century BC, the diminutive ‘modulus’ and its verbal form ‘modulor’ begin to appear in Latin literature. Their basic meaning at this date is still ‘measure’ (as in Horace and Varro), but there is also a particular association with music, with reference to measures of time and pitch [Lewis, Short, 1879; Glare, 1982]. The well-known architect Vitruvius uses ‘modulus’ in his Ten Books on Architecture (around 23 BC) usually as an architectural standard, namely the radius of a column [Vitruvius, 2004]; this seems to be the first recorded engineering use of the word. In his Two Books on the Water Supply of the City of Rome (100 AD) the politician and writer Sextus Julius Frontinus [2004] uses the word ‘modulus’ around thirty times as “a pipe of specified diameter used to control the rate of flow of

The Notion of a Model

639

water, an adjutage” [Glare, 1982]. It appears that ‘moduli’ are standardized pipes available in twenty-five sizes. Around AD 200, Tertullian uses ‘modulus’ with various meanings, in particular the manipulable figures employed by sculptors, for example as a basis for a marble sculpture (“inde circino et plumbeis modulis praeparatio simulacri, in marmor, in lutum uel aes uel argentum, uel quodcumque placuit deum fieri, transmigratura”; NAT. I, 12, 9 [2004]). From the later Roman Empire to the late Middle Ages, the word ‘modulus’ and its derivatives tend to fade from view in the literary record, and our evidence for their progress into modern European languages is patchy. French (von Wartburg, [1966]) reduced ‘modulus’ to ‘modle’, ‘mole’ and ‘moule’, which came into English as ‘mould’ (as in cheese-mould). English later took the word ‘modulus’ directly from Latin, producing ‘module’ and the scientific term ‘modulus’. But also the double diminutive form ‘modell-’ began to appear, as witnessed by the Italian ‘modello’. Thus in 14th century Italian we find ‘modelo’ (baking tin or mould) and ‘modello’ (drawing), the latter with respect to the construction of the Dome of Florence (see [Bernzen, 1986; M¨ uller, 1983]). From the early 15th century we also find ‘modello’ used for 3D-architectural models, especially for Brunelleschi’s various models for the Cupola of the Dome of Florence and for the elevators and cranes necessary to build it. We will see below how Alberti used ‘modello’ in the mid 15th century in art and architecture. The Italian ‘modello’ came into French by 1542 as ‘modelle’ and ‘mod`ele’, and into English by 1570 as ‘model’ or ‘modell’, and the variety of meanings expanded rapidly into the present-day abundance. The range of meanings of ‘model’ in the early seventeenth century is already striking. Thus Shakespeare Much ado about nothing (1613) I.3: Borachio I can give you intelligence of an intended marriage. Don John Will it serve for any Modell to build mischiefe on? Note both the literal sense (model of a building) and the metaphor (modelling one’s behaviour on something). In 1627 the editor of Francis Bacon’s New Atlantis opens his foreword with the words: This Fable my Lord devised, to the end that He might exhibite therein, a Modell or Description of a Colledge, instituted for the Interpreting of Nature, and the Producing of Great and Marveilous Workes for the Benefit of Men [. . . ] Certainely, the Modell is more Vast, and High, then can possibly be imitated in all things; Notwithstanding most Things therein are within Mens Power to effect [Bacon, 1627; spelling of the 1628 edition]. Bacon himself (writing in Latin in 1620 [Bacon, 2004, i.110 and 124]) uses ‘modulus’ both for printing type, and for attempts to ‘copy the world in the human mind’. Here we are already close to modern notions of scientific modelling;

640

Roland M¨ uller

but Bacon is not the earliest example. In 1576 the astronomer Thomas Digges had referred to Copernicus’ use of a ‘new Theorick or model of the world’. The word ‘model’ and its cognates were common in the main European languages up to the present day, in all the meanings that we have noted. But since the mid 1940s there has been an explosive growth of their use in science and technology. We can see this from the statistics of the archives of the Deutsche Bibliothek in Frankfurt, concerning titles of books — mostly in German — that use the words ‘Modell’, ‘Modellierung’, ‘Modellversuche’ and the like. We find the following number of titles per year: 1950: 30 1960: 50 1970: 350 1980: 700 1990: 1100 2000: 1950 2004: 2300 In the ten years 1990–1999 there appeared more than 17,000 titles which include these words. The online-catalogue of the largest British libraries (COPAC) has for the same period of time more than 16,000 English titles (admittedly with a lot of doublets). It would be interesting to know the reasons for this explosion. Words become trendy for reasons that are often hard to discover. Perhaps the main reason has been the growth of interest in methodology (for education, for theoretical analysis, for impressing grant-giving bodies, . . . ), and ‘model’ is one of the most methodological words we know. While the word ‘model’ blossomed, the word ‘module’ went its own way. Since World War II it has come to be used regularly for a semi-independent or standardized system, generally one that occurs as a component of a larger system. We find it used for furniture, and for electronic parts that are ‘individually fabricated subassemblies’. We have had modules as sections of a spacecraft since the Apollo missions around 1960. Modules as units of a training programme have been with us since 1966. But this weakened meaning of ‘module’ is probably not new; it’s hard to think of anything more precise that the mathematician Richard Dedekind could have had in mind when in [Dirichlet, 1871] he coined the name ‘Modul’ for certain substructures of fields. Dedekind’s term was generalised in the 1920s when the branch of algebra known as module theory was founded. A system is described as ‘modular’ when it is built of separate parts, each of which has its own clearly defined function. This usage seems to have appeared in the early 1970s in electronics and computer science, where the importance of modular design is that a part of a system or program can be replaced by another unit that performs the same function, without disrupting the rest of the system. For example the patent [Cheney and Kuczura, 1976] emphasises that its ‘modularity’ allows ‘facile growth and additions of feature-oriented service packages’. The notion of modularity became important in cognitive science, where for example

The Notion of a Model

641

one asked whether the brain has modules that are specific for language. Fodor [1983] served as a focus for these discussions. We note briefly that the original sense ‘measure’ can still be found. Examples are Young’s modulus (a measure of elasticity), and Le Corbusier’s [1951] architectural yardstick ‘Modulor’. 3 MOULDS Hollow forms have been used for more than 10,000 years to form bricks, pastries, metals, etc. The use of convex forms to create patterns, however, is apparently more recent. But at least we can trace it to the use of sticks to impress letters into clay tablets, as in Babylonian cuneiform script [Glassner, 2000], and the Greek tablets written in Linear A and Linear B. In Japan a kind of printed textiles were known in the first century AD. Real prints on bright, uncoloured linen are known since the 4th century from Egypt and since the 7th century from Europe and the Coptic areas of North Africa; likewise Chinese colour prints on silk. In 175 AD Chinese scholars began to cut the main works of classical Chinese literature in stone plates. From these plates, thousands of poor copies were made: dampened paper was pressed onto the plates and then smoothed out with a brush and ink, so that the carved signs stood out in white among the black background. Printing with mobile type characters is said to have been practised in China already around 1040. In Europe Johannes Gutenberg introduced it around 1440. After printing the whole bible he lost goods and chattels in a court case and died as a broken man [Scholderer, 1963]. ‘Modellus’ is found in Britain as early as the fourteenth century; it is conjectured to refer to a ‘vessel, mould’ cheese-mould in [Latham, 1965].1 By the beginning of the 16th century the use of ‘mould’ or ‘model’ for both hollow moulds and convex printing forms was well established. Thus in Randle Cotgrave’s Dictionary French-English [1611] we read: Modeler: To modell, forme, fashion, plot, cast in a mould. Modelle (f.): A modell, patterne, mould, plot, forme, frame. Moule (m.): A mould (wherein a thing is cast, formed, or forged;) Chandelles de moule. Candles made in moulds; (great) Christmas candles. Moul´e: m.´ee. (f.) Moulded; cast, or framed in a mould. Mouler. To mould, or cast in a mould; to frame, or forge by mould; also, to appoint a mould for, prescribe a size unto. Moulle. as Moule. In the famous French Encyclop´edie [Diderot and D’Alembert, 1765, Vol. 10], we read under ‘mod`ele’: 1 This conjecture, however, has disappeared from the ‘modellus’ entry in [Latham and Howlett, 2001].

642

Roland M¨ uller

dans les ouvrages de fonte, le mod`ele est en quelque fa¸con l’ouvrage mˆeme, dont le m´etal prend la forme; la mati`ere seule en fait la diff´erence; une couche de ciment & de terre, de la forme de la cloche qu’on veut fondre, & de la mˆeme ´epaisseur que la cloche doit avoir. Moulds are still known as moulds today, but sometimes also as models. For example in Germany, baking tins, printing forms, stamps and moulds are partly named ‘Model’, partly ‘Modell’ (in Switzerland also ‘Foermli’, i.e. little forms). The embossed or formed things however are seldom called ‘Modell’ (a rare example is artistic castings) or ‘Model’ (e.g. in Switzerland ‘model bacon’ and ‘es Moedeli Anke’, a piece of butter in the form of an ingot). We call attention to one feature of moulds, which applies also to some but not all of the things known as models. A mould can be used indefinitely many times, to create many objects. In fact this is often their main purpose, as with printing type. All the objects that come from the same mould will be similar; in some contexts we say they are ‘the same model’. This usage also applies more generally to things that are made according to the same specification, even when there is no physical mould. It goes back to about 1840. It was used for industrial products manufactured in great quantities [Landes, 1968]. One of the first manufacturers to introduce serial production of objects was Samuel Colt. He sold more than 330,000 of his model ‘pocket’ — also called ‘Wells & Fargo’ — between 1849 and 1875. English has spoken of ‘Parisian models’ (in fashion) since 1859, and ‘models’ of automobiles since 1900. Of Henry Ford’s legendary ‘Model T’ 15 million exemplars were produced between 1908 and 1927 [Clymer, 1955]. The word ‘mould’ still occurs in connection with modelling; for example Boumans [1999] described a component of modelling that he called ‘mathematical moulding’. 4

MODELS AS A STEP IN DESIGN

From very early times, artists have made preliminary sketches or mockups before producing their final work. Before modern times, one of the most fully reported examples is the modelling for the Dome of Florence, mentioned in section 2 above. We know for example that models were made for various parts of the Dome — even for the elevators and cranes — and that some parts had alternative modules, so that a choice could be made between them [Bernzen, 1986; M¨ uller, 1997; Saalman, 1980; Ferguson, 1992, p. 66]. This usage of the word ‘model’ leaves some aspects of the meaning unclear. Is it crucial that the artist copies the model, or does the model simply help him to clarify a design that lies elsewhere, either in the real world or in his mind? Does the model have to be something that he constructs? Does it have to be small, like the architect’s model, but unlike a preliminary sketch for a painting? Later writers resolved these ambiguities, but not always in the same way. Suppose for

The Notion of a Model

643

example that I make a sculpture of your wife, and as a first step I make a plaster cast. Which is my model, your wife or the plaster cast? ‘The plaster cast’, says Leon Battista Alberti in the mid 15th century [Alberti, 1847, Book i p. 180]. For Alberti the artist can work either from ‘vivo’ or from ‘modello’. ‘Your wife’, says the French Encyclop´edie, in [Diderot and D’Alembert, 1765, vol. 10] under the word ‘mod`ele’: tout ce qu’on regarde comme original, & dont on se propose d’ex´ecuter la copie. Ce mot se prend au simple & au figur´e, au physique & au moral. Example: une femme, mod`ele pr´ecieux pour un peintre . . . The models who advertise clothing are presumably descendents of the Encyclop´edie’s ‘femme, mod`ele pr´ecieux’. Besides tightening up the definitions, writers on artistic and design models gradually built up a theory of these models. We illustrate this from four authors: Alberti around 1450/60, Martini a few years later, Sturtevant in 1612 and Leibniz in 1669. We have already cited Alberti’s work on sculpture. At about the same time, in fact roughly when the Dome of Florence was being finished, Alberti also wrote a book on architecture. The most important advantages of the use of models, he says, are vividness, manipulability and improvability. And there you may easily and freely add, retrench, alter, renew, and in short change every Thing from one End to tother, till all and every one of the Parts are just as you would have them, and without Fault. [Alberti, 1755, Book II, chap. I, p. 22] He also recommends producing copies, so that the original model is preserved, even when the architect plays with changes in the copies. Be sure to have a compleat Model of the Whole, by which examine every minute Part of your future Structure eight, nine, ten Times over, and again, after different Intermissions of Times. [ibid, Book IX, chap. VIII, p. 203] Francesco di Giorgio Martini was a paintor, a sculptor, an architect and an engineer. Writing around 1480/90, he made the cognitive aspects of model-building explicit: Whereas it is difficult to demonstrate everything through drawings, nor is it at all possible to express many things in words, . . . so it is necessary to make a model of nearly every object. [Martini, 1967, I, p. 142] In 1612 the engineer Simon Sturtevant published a remarkable book on Metallical Inventions, also known as Heuretica [Sturtevant, 1612]. In this book he defines Heuretica as ‘the Art of inuentions, teaching how to find new, and to iudge of the old’. This doctrine of invention consists of a real and a technical part. The

644

Roland M¨ uller

first consists of ‘the instruments and reall things which belong to the inuentions’, the latter concerns ‘the dexterous habit and faculty’ of the craftsmen. Inventions themselves can be differentiated by ‘magnitude . . . greatnesse or quantity’. Three kinds of models result: moddle, protoplast and mechanick. The ‘moddle’ is a ‘Mechanick’, which represents and shows the basic parts and contours of an invention without really functioning. For example we cannot expect a model of a windmill to grind corn. Such a model can be smaller, or also — when details need to be shown — larger than the object being represented. It can be drawn or painted (and then is ‘superficiall’), or ‘reall’ like a model ship. The ‘Protoplast’ — today’s prototype — fulfills all the functions of the final device and works productively, but is open for further refinements and adaptations to special conditions. The first Protoplast of a group of machines or apparatuses is the ‘Archetype of the Protoplast’, e.g. the first windmill which was able to grind corn. In an additional chapter, Sturtevant gives ‘Cannons or Rules seruing to iudge of the goodnesse’ of an invention or an improvement, wherefore he develops a differentiated theory of equivalence using the criteria Equi-sufficiencie, Equi-cheapness, Equi-excellency. In 1669 the mathematician and philosopher Leibniz praised the construction of small scale ‘modulis’ in order to design fortresses in his sketch of the ‘Ars inveniendi’. Then he mentions the collections of models, which were very popular in his time: ‘de Theatro Naturae et Artis seu de Modulis rerum ipsarum conservatoriis’ [Leibniz, 1903, p. 163]. Shortly after he proposed in his Atlas universalis a department of objects, which ‘can be presented to the eye’: ‘mechanical devices, including machines and models of every kind’ [ibid, p. 223]. Leibniz was no doubt influenced by the French Academy of Sciences, which had started to collect models from inventors hoping to gain official recognition for their products. A catalogue of these models was published in seven volumes: Machines et inventions approuv´ees ´ par l’Acad´emie royale des sciences, depuis son Etablissement jusqu’au pr´esent, Paris, 1735–77; the first volume contains the objects submitted before 1700. More recent collections of patent models in various countries are developments of this same idea. 5

MODERN TOOLS FOR THE DESIGNER

In the wake of the industrial revolution, the nineteenth and twentieth centuries brought a step change in the sophistication of tools to help the designer. We mention three important examples, namely scale modelling, computer simulation and stepwise refinement. (a) Scale modelling. The designer of an ocean liner or an aeroplane needs to check that the vehicle will behave as intended. An obvious kind of test is to build a scale model and try it out in a wind tunnel or a water tank. But as Galileo already explained in connection with animals of different sizes [1638, Second Day] some physical properties depend on length, some on area and some on volume; so a copy that is shrunken in strict proportion will not behave in the same way as the full-scale vehicle.

The Notion of a Model

645

An early pioneer of scale modelling of ships was William Froude (1810–1879), who built very accurate scale models and tested them in tanks of his own design, using his own purpose-built measuring instruments. Today he is remembered above all for his mathematical analysis of the dimensions of some of the key quantities, and a method for predicting the friction of a full scale vessel using the measurements of the friction on small scale. This is how the Froude number got its name — Froude himself never used this number in his published writings. Another dimensionless number in fluid dynamics is the Reynolds number, introduced by Osborne Reynolds in [1883]. Dimensionless numbers are crucially important for scale modelling, because they allow the modeller to use different scalings in different dimensions and still be sure that the behaviour of the final object can be accurately read off from the behaviour of the model. The mathematics involved is known today as dimensional analysis; it was famously explored by Lord Rayleigh in [1915]. Textbooks include [Bridgman, 1922], [Weber, 1930] and [Langhaar, 1967]. See also Zwart’s chapter in this Volume. (b) Computer simulation and computer-aided design In some sense, computer simulation goes in the opposite direction to scale modelling. They share the feature that they rely on knowing the laws obeyed by the object being modelled. But whereas scale modelling finds a different object obeying those laws, computer simulation works directly with the laws themselves and not with an object that obeys them. A computer is generally cheaper to buy and run than a wind tunnel, and you can simulate on a computer things that it would be unethical to do to a living organism. The earliest example of simulation on a digital computer consisted in using the computer to solve equations that it was impracticable to solve in any other way. In 1946 Stanislaw Ulam had the idea that one could approximate solutions of some combinatorial or differential equations by reading the equations as descriptions of a process where some random operation is applied very many times — like dropping a pin onto a lined sheet of paper, to quote an example often given. Computers are very good at performing a random operation very many times. He discussed the idea with John von Neumann, who was at that date deeply involved in developing the first generation of digital computers. They named the approach ‘Monte Carlo method’. The result was a series of computer simulations of thermonuclear processes [Eckhardt, 1987]. In fact many real-life processes really are the accumulated results of performing a limited range of random operations very many times. Weather forecasting is an obvious example, and it was a computer simulation of a weather system that set Edward N. Lorenz on the road to chaos theory. [Lorenz, 1963; this paper cites ‘computations . . . performed on a Royal McBee LGP-30 electronic computing machine’ at a speed of one second per iteration]. The idea of using computers for simulation of complex processes spread very fast. In [1960] D. G. Malcolm submitted a Bibliography on the use of simulation in management analysis. In his opening paragraph he said: ‘In system simulation the computer is typically utilized in problem-solving associated with the specific tasks

646

Roland M¨ uller

of designing better systems, understanding the workings of operative systems, and studying man-machine interactions’. So computer-aided design was already in full swing within some ten years of digital computers becoming publicly available. An important aspect of computers is that they not only solve equations; they also display the results on the screen. Not only does the screen create a visual model; the user can interact with the computer to test the results of adjustments. By the mid 1960s these ideas were already well in hand for studying the structure of complicated molecules [Levinthal, 1966]. Virtual reality is a further development of similar techniques. When computer modelling involves visualisation, it tends to rely on a vast range of expertise, as Donna J. Cox has emphasised with her notion of a ‘Renaissance Team’ that brings together artists, scientists, and technologists to produce images and movies of scientific data [Cox, 1988]. Many traditional and beloved techniques of drafting and designing ‘models’ for buildings, ships or automobiles, electrical circuits or injection moulds have been replaced since 1960 by Computer Aided Design (CAD). This kind of computer use developed from US military research on space travel. Later it was made available to the public. In 1964 IBM developed the first CAD Computer, the ‘System 2250’, [Bissell, 1990]. A first general introduction to the field is given by Charles Russell Mischke [1968]. (c) Stepwise refinement. Stepwise refinement is a tool of a very different sort from (a) and (b). It is a methodology for managing the process of matching up the designed object and its specification. The name comes from software engineering [Wirth, 1971], though similar methodologies apply in other areas under names such as ‘functional decomposition’. As the name indicates, the idea is to start with a specification of the system to be constructed, separate out the different requirements that need to be met, and work incrementally towards more concrete descriptions of units that will meet the requirements.

6

MODELS OF THE WORLD

We saw that the use of the word ‘model’ for scientific descriptions of the world goes back to the sixteenth century. As scientific research expanded and took on new forms, new kinds of ‘model’ appeared. Historical turning-points are always a little arbitrary, but it may be safe to point to three key moments. The first was in the middle of the nineteenth century, when many leading physicists came to regard ‘models’ as an essential part of scientific methodology. The second was roughly a hundred years later, after the Second World War, when philosophers of science realised that some coherent account of the role of models in scientific progress was needed. The third was the cognitive turn, say around 1980, when psychologists and others used the notion of a model to explain how each of us — even the least scientific of us — comes to make sense of the world. These three moments conveniently divide the history of scientific modelling into four segments.

The Notion of a Model

6.1

647

Early scientific modelling

In [1576] Thomas Digges wrote about Copernicus: But in this our age one rare witte hath by long studie, painfull practise, and rare invention delivered a new Theorick or model of the world, shewing that the earth resteth not in the Center of the whole world, but only in the Center of thys our mortal world or Globe of Elements. What was it in Copernicus’ work that counted as a model? In Copernicus’ own words, the astronomer cannot by any line of reasoning reach the true causes of [the celestial] movements, [so it is his job] to think up or construct whatever causes or hypotheses he pleases such that, by the assumption of these causes, those same movements can be calculated . . . [Hawking, 2002, p. 7] We will see that words like these were used in later centuries to distinguish between a theory (which aims to state the facts about the world) and a model (which gives us some kind of information or insight without claiming to be factual). But in Copernicus’ case a more likely reason for putting these words up front in his introduction was to reassure the Catholic Church that he wasn’t going to startle the horses. We know what happened to Galileo when he showed less humility. In fact for the remaining several hundred pages of his work Copernicus writes exactly as if he was aiming to establish the facts of the case, and their causes, and Digges seems to have read him this way too. As we have already mentioned in section 2, by 1620 Francis Bacon is also using the word ‘model’ for a copy that we make of something already out there in the world. For I lay foundations in the human intellect for a true pattern of the world (verum exemplar Mundi) as we actually find it and not as someone’s own private reason hands it down to him. And this cannot be achieved unless we undertake a most painstaking dissection and anatomy of the world. But I proclaim that the botched and (if you like) apish patterns of worlds (‘Modulos vero ineptos Mundorum’) which men’s fancies have thrown together into philosophical systems should be utterly destroyed. [Bacon, 1620, p. 124] Bacon’s use of ‘modulus’ here is interesting. Like the architect’s model, Bacon’s model is similar to something ‘out there’ and is closer to hand than the thing it resembles. Also like the architect’s model it is a human construction. Bacon’s reasons for choosing the word ‘model’ probably involve some or all of these properties. But unlike the architect’s model, the world (the thing ‘out there’) comes first, and the model is created as a copy of it. Also Bacon’s models and exemplars are ‘in the human mind’; they are not physical objects like the architect’s model. But already Copernicus had used

648

Roland M¨ uller

physical objects of a kind, namely around a hundred printed diagrams. The second half of the seventeenth century was a boom time for scientific diagrams. We can cite Leeuwenhoek’s biological pictures [Schierbeek, 1959] or Hooke’s illustrations of crystal structure [1665]. These authors didn’t describe their diagrams as models; but Ford [1993, p. 137] comments ‘Hooke shows a series of what we might now term ‘molecular models’, relating their orientation to the facets that must result.’ One way to make sense of a phenomenon is to describe how a machine could be constructed to produce or mimic the phenomenon. Today a large part of the programme of Artificial Intelligence rests on just such an idea. But the idea goes back at least to 1616, when William Harvey in his handwritten notes on the circulation of the blood claims to have demonstrated ‘that the blood flows in continuous stream of the lungs into the aorta as by two valves of a water pump used to lift water’, [Van Leeuwen et al., 1946, p. 75] see also [Vonessen, 1989]. In the 18th century the state and the economy often were regarded as machines. The Cameralist J. H. G. Justi wrote in [1764, p. 86-87]: ‘A perfect established state has to be exactly like a machine in which all wheels and transmissions fit into each other with the utmost precision.’ And the Cameralist August Ludwig von Schl¨ ozer proposed in [1793, p. 3]: ‘The most instructive way of teaching politics is to regard the state as an artificial, extremely complex machine serving a certain purpose.’ This kind of language foreshadows the nineteenth century use of machines in scientific methodology, which we turn to next.

6.2 Models within scientific methodology The 19th century saw many mechanical ‘explanations’ of natural phenomena. They include Sigmund Freud’s use, in [1895], of a hydraulic model in his sketch of a psychology (published 1950) for his first demonstration of the drive dynamics of the ‘psychic apparatus’. Famously, James Clerk Maxwell suggested various mechanical devices to explain the apparent behaviour of electromagnetic radiation. Exactly what Maxwell constructed, and what was purely in his head, is not always easy to establish; but some of his early work with gelatine models may have paid off in his contributions to the foundations of photography. However, the 19th century also brought a new ingredient: the beginnings of a theory of explanation by mechanical analogues. Maxwell himself, at the age of 24, introduced the term ‘physical analogy’ in his lectures On Faraday’s Lines of Forces 1855–6: By a physical analogy I mean that partial similarity between the laws of one science and those of another which makes each of them illustrate the other. Already in 1845 the young Scottish physicist William Thomson had proposed ‘the principle of ‘images’ as a means to solve some problems of the distribution of electricity’ [1890, §208, p. 143] and emphasised in a paper on magnetism the ‘analogy’ between a result on induction and a theorem of optics [1890, §157 f.,

The Notion of a Model

649

p. 104]. Nearly at the same time, 1843, Richard Owen specified for biology the important distinction between analogy (i.e. same function) and homology (i.e. same origin) [Bljacher, 1982]. Theoretical discussions of explanation by physical models continued through the 19th century, and in 1902 the physicist Ludwig Boltzmann contributed an article Model to the Encyclopedia Britannica which summed up the achievements of this discussion [Boltzman, 1902]. Boltzmann concentrates on scientific modelling, though he is well aware of the use of models in design and manufacture. He catalogues the different kinds of model according to their structure (for example some are stationary and some move), and describes how some of them give a direct intuition of the phenomena they represent. He remarks in several places that models have ‘resemblance’ or ‘similarities’ or ‘analogies’ to what they are models of, but he makes no attempt to analyse these notions. Already in 1893 the French Physicist Pierre Duhem launched a strong negative backlash against explanation by models. In 1996 he published his view that scientific advance consists of the construction of fully meaningful and exactly stated theories whose truth is confirmed by experiment. To the extent that a model is anything less than such a theory — for example if its connection with the world is only one of analogy — it fails to explain the world. In the nationalistic spirit of the times, Duhem added that mechanical models were attractive to the English because, by appealing to the intuition, these models made detailed logical deductions unnecessary. For example Lord Kelvin used models because they bypass detailed calculation and ‘by appealing only to the imagination, leave it to the imagination to judge whether they resemble what they are supposed to represent’ [Duhem, 1996, p. 115]. Duhem’s views slotted well into the views of those who, like the Vienna Circle in the 1920s and 1930s, regarded the business of science as the construction of formal theories in the language of logic, together with derivation of facts from these theories. During this period there seems to have been no pressure to develop a theory of technological modelling.

6.3

The philosophy of scientific models

In the middle of the twentieth century the notion of models moved into centre stage in the philosophy of science. This seems to have been at least partly a reaction to the Vienna Circle. During the interwar years, logicians had developed some sophisticated theories of how to encode information in logical systems of various kinds. The Vienna Circle had become a focus for applications of these ideas in philosophy. Now logical systems are not just a way of representing information. They also play a role in activities designed to increase information. In other words, they have an epistemological role. Some representatives of the Vienna Circle were quite naive about this aspect of logical systems. They suggested that we gain knowledge by ‘verifying’ [Ayer, 1936] or ‘falsifying’ [Popper, 1935] propositions. To dig deeper,

650

Roland M¨ uller

a number of philosophers wrote books and papers on the question how various logical systems can be used to increase our understanding. Some of the relevant logical systems were called ‘models’, and several authors concentrated on these. Two typical quotations from this period are: It is in fact a great virtue of a good model that it does suggest further questions, taking us beyond the phenomena from which we began, and tempts us to formulate hypotheses which turn out to be experimentally fertile [Toulmin, 1953, p. 38]. Something must now be said concerning [the] use [of theoretical models]. Theoretical models . . . can be used for purposes of explanation, prediction, calculation, systemisation, derivation of laws and so forth [Achinstein, 1965, p. 106]. Two characteristics of this literature seem dated now. First, the increase in knowledge or understanding was nearly always taken to be theoretical or scientific, not technological. Second, the question of what ‘suggests further questions’ or ‘tempts us to formulate hypotheses’ or ‘can be used for derivation of laws’ was taken to be a philosophical question that philosophers can answer by thinking. Hardly anybody at that date thought of looking for empirical evidence to show what kinds of model do in fact help creativity, and how they do it. In short, these writings were contributions to epistemology and not to cognitive science. A common feature of this literature was the classification of different kinds of model. Of course many models are far from being logical systems. But one could take on board a wide range of these models by studying the ‘logic’ of their use. A striking example is Mary Hesse’s defence of models against Duhem, in her booklet Models and Analogies in Science [1963]. Duhem had attacked ‘mechanical models’, which are certainly not logical systems. But for Hesse, mechanical and other models are used as ‘analogues’; the last section of her book is an attempt to build a kind of formal logic of analogy. Some of the terminology introduced in this period is still in use today. For example an influential chapter by Max Black described ‘mathematical models’ and ‘theoretical models’ [Black, 1962, Chapter 13, ‘Models and Archetypes’]. In both cases Black first describes what the relevant models are, and then gives conditions for their use. A typical example of a ‘mathematical model’ is a set of equations designed to represent some phenomenon in the social sciences; Black twice cites the ‘logistic function’. The paradigm example of a ‘theoretical model’ is Maxwell’s representation of an electrical field in terms of the properties of an imaginary incompressible fluid. ‘The theoretical model need not be built; it is enough that it be described.’ [ibid, p. 229]. Black also discusses ‘scale models’ and ‘analogue models’. Another common theme of the period was the relationship between models and theories. In practice the two words were sometimes used interchangeably. For example in the journal Econometrica Donald Davidson and Patrick Suppes [1956]

The Notion of a Model

651

refer on their first page to a ‘formal theory’ proposed by Frank Ramsey, and a few lines later they distinguish their own ‘model’ from Ramsey’s; if they intend any difference between theory and model, they don’t say so. This was not a new phenomenon; we saw earlier that in 1576 Thomas Digges described the ideas of Copernicus as ‘a new Theorick or model of the world’. One of the more detailed discussions of the relation between models and theories is Achinstein’s Models, Analogies, and Theories [1964]. Achinstein provides a number of clues on why, for example, the Bohr model of the hydrogen atom is known as a model and not a theory, whereas statistical mechanics is a theory and not a model. But he hardly takes on board the notion of ‘model of a theory’ that the logicians had recently created (see section 7 below). In fact for this notion he refers to a passage of Carnap [1942, pp. 203ff] which today we can see is a rather clumsy mixture of the notions of a theory and a model of the theory. The questions raised during this period about the philosophical analysis of models are still alive today. Frigg and Hartmann give a good summary of the literature [2008]. However, readers of this volume will be aware that a number of recent writers have taken up the question of the epistemological role of modelling in technology.

6.4

Mental models

In 1983 two influential books appeared, both in cognitive science and both with the title Mental Models (Gentner and Stevens [1983] and Johnson-Laird [1983]). The term ‘mental models’ was rare before these two books, but thanks to them it became widespread. Even when it did appear before 1980, it generally meant nothing of any particular interest to a cognitive scientist. For example in [Gruchy, 1944, p. 220] we read about ‘mental or logical models’ of an economy, which are to be ‘modified into concrete existence’; here the word ‘mental’ serves mainly to distinguish these models from concrete ones. One finds similar language through the 1950s and 1960s. It’s a very old idea that human beings make sense of the world they live in by forming internal representations of features of the world. During the heyday of behaviourism, psychologists relegated this notion to the shadows. But as behaviourism broke down and it became acceptable for psychologists to talk about mental representations, various scientific questions about them came into view. Tolman’s work [1932] on the behaviour of rats in mazes was an early example; he provided evidence that rats form internal representations of the geometry of mazes, and that they form plans for negotiating the mazes. At around the same time, Piaget began his series of works On the child’s conception of . . . , in which he investigated the concepts available to children at different ages (starting with [1929] on living beings and natural objects, and [1930] on causality). In 1943 Kenneth Craik published a book in which he suggested that human beings make sense of the world by forming ‘models’ of it [Craik, 1943]. He meant something more specific than the old truism that we form internal representations.

652

Roland M¨ uller

More precisely he suggested that the human brain is a general-purpose simulation machine. In his view, a key role of thinking is to predict the future; the brain does this by simulating the relevant part of the world and running the simulation. He speculated about what kind of machine has the power to carry out such simulations, and proposed that the brain operates by manipulating symbols (see [Craik, 1943, Ch. V: ‘Hypothesis on the nature of thought’]). If he had been more of a cognitive scientist and less of a philosopher, he might have anticipated the ‘symbol systems’ of Newell and Simon [1976]. By 1980 a number of empirical questions about internal representations of the world had emerged. The relevant literature is still fairly chaotic about what these questions are, and terminology tends to be limited to particular schools. But here follow some examples. First, the word ‘model’ is relevant in two ways. We can discuss the internal ‘models’ that people form in order to understand the world; or we can ‘model’ the ways in which people think about the world. This division of uses comes out clearly in the book of Gentner and Stevens mentioned above. In their opening paragraph Gentner and Stevens say ‘we can model the way in which people imagine liquids moving through time . . . ’ [1983, p. 1]. Here the model is not in the mind, it is of the mind. But Norman starts the first paper in the volume [1983, p. 7] by noting that ‘people form internal, mental models of themselves and of the things with which they are interacting’; these are models in the mind (though also partly of it). The same double take can be traced throughout the book. For example the authors of one chapter cite a paper with the title ‘Models of competence in solving physics problems’, and another paper with the title ‘Mental models of physical mechanisms and their acquisition’. (These two papers are [Larkin et al., 1980] and [Kleer and Brown, 1981]; they are cited at [Gentner and Stevens, 1983, p. 97f].) Making models of how people think is sometimes known as ‘cognitive modeling’ [Polk and Seifert, 2002]. But this kind of model is much older than the name: witness for example the papers A mechanical model of the conditioned reflexes [Baernstein and Hull, 1931] and A model of the synthesis of conditioned reflexes [Bennett and Ward, 1933]. Second, if we limit ourselves to internal models, there is a difference between those internal representations that we are fully conscious of and can report without any guesswork, and those representations that can be uncovered only by patient and carefully motivated experiment. As an example of the first, Nersessian [2002] describes Faraday’s model of the lines of force surrounding a bar magnet. This model is so public that the editors of the volume where Nersessian’s paper appears have put it on their front cover. At the other end of the scale are Johnson-Laird and Byrne’s mental models for logical reasoning; these models are so elusive that Johnson-Laird and his co-workers have to use indirect techniques to elicit them. The techniques include measuring the length of time taken to carry out an inference, and the number of errors made [Johnson-Laird and Byrne, 1991]. More recent work brings in evidence from brain scans [Knauff, 2007]. The difference

The Notion of a Model

653

between public and hidden may be related to an interesting distinction that Nersessian draws between mental models that are held in long-term memory and ones that are devices created in working memory during comprehension and reasoning [Nersessian, 2002, p. 140]. Cognitive scientists who work in Johnson-Laird’s programme often call themselves ‘model theorists’ because of their use of mental models. Ruth Byrne’s ‘Mental Models Website’ [Byrne, 2008] gives further information. An earlier example of mental modelling that was uncovered by experiments using the time taken to solve problems was the famous work of Shepard and Metzler [1971], on rotating three-dimensional shapes to make them coincide. Third, we can classify internal models by the kind of representation that they use. Nersessian [1984] contrasts ‘propositional’ models with ‘iconic’ models; Stenning [2002] argues for a related distinction between ‘indirectly interpreted’ representations and ‘directly interpreted’ ones. Then there are questions about the availability of different kinds of internal model, depending on how sophisticated our understanding is. At a low level these questions link into educational theory: how can we best teach students to form helpful representations? At a higher level they touch on the cognitive science of research: can we improve creativity by using models in a certain way? The publication blurb for Young and Veen [2008] promises that the book will help you ‘quickly get up and running; you’ll be creating mental models right away’. 7 FROM EDUCATIONAL MODELS TO MATHEMATICAL MODELS Some toys are known as models, for example model aeroplanes. But many toys that are not models in this literal sense can still teach a child things about how the world works. Children like playing with toys and looking at pictures. If the toys have shapes that move, or interesting textures, children can learn something about the world by manipulating them. Educational reformers have often emphasised this aspect of toys. For example Campanella placed a reform program for education in his City of the Sun (1602 respectively 1623), which is based decisively on the use of models. Bacon’s New Atlantis [1627] brings something similar. And of course we are all children when we need to learn something. At any level of research we use copies and diagrams. The use of sign systems and their manipulations on paper in the laboratory sciences of chemists in the early 19th century is described by Ursula Klein [1999, p. 153–164; 2003, p 2–3; ibid p. 118]. She sees them not as representations but as productive tools to open up new ways of research. Also the historian of chemistry Christoph Meinel [2004, pp. 243, 25, 270] denies that the molecular models of that time were illustrating theoretical concepts. They were meant to communicate and teach. We can hardly put an exact date on the use of the word ‘model’ for educational tools, because there is no hard and fast division between them and the kinds of model discussed in the previous two sections. An architect’s model for a building

654

Roland M¨ uller

is a kind of toy version that the architect can play with, as Alberti noted [1755, Book II, ch. I, p. 22]. A working model of circulation of the blood can also be used as a teaching aid. But at least by the late 19th century one could find in university Mathematics departments things that were called ‘models’ and were clearly educational devices. It seems that these mathematical models have a double ancestry. The first is from theoretical geometry. In the mid 4th century BC, Plato described the five regular polyhedra (Timaeus 54f, p. 75ff in [Plato, 1971]) and thus earned for them the name ‘Platonic solids’. The thirteenth book of Euclid’s Elements gives instructions for constructing these polyhedra, [Euclid, 1956]; for example his Proposition xiii.14 describes a geometrical procedure for inscribing an octahedron in a given sphere. Euclid’s instructions are part of a theoretical study of regular polyhedra which is believed to go back to Theaetetus in the late 5th century BC (see Heath’s Historical Note on page 438f of [Euclid, 1956]); there seems to be no evidence that the Greeks connected it with any practical engineering problem. The second source is architectural. The Alhambra in Granada (14th c.) contains magnificent examples of the geometrical forms that mathematicians describe as ‘wallpaper patterns’; the builders who created these forms must have had quite sophisticated practical devices for making them. In Early Renaissance Europe it became a custom to incorporate interesting geometrical shapes into buildings, and naturally a theory developed to show builders how to construct these shapes. These two strands had come together by the time that Luca Pacioli published his Summa in [1994] and his De Divina Proportione ([1956], written in 1497 but published in 1509). Pacioli gives practical instructions for constructing the regular polyhedra, and mentions how they figure in Plato’s philosophical speculations. Leonardo da Vinci contributed illustrations to De Divina Proportione; these included numerous pictures of three-dimensional geometrical figures, and the first recorded picture of an icosahedron. Olschki comments on Pacioli’s work that ‘The treatment of the material, as stone and marble, for structural as well as sculptural purposes according to the mathematical principles of the time required precise determination of volume and reliable information for the transformation of a body into another. The times of treating materials by eye were definitely gone,’ [1918, I, p. 218]. Around 1860 a number of mathematicians, among them Julius Pl¨ ucker and Ernst Eduard Kummer, started to make three-dimensional models of complicated mathematical and geometrical curves and surfaces, [Fink, 1890]. Herbert Mehrtens mentions that Pl¨ ucker had the idea of working with models from Faraday, and that the physicist Gustav Magnus had already prepared a model of a wave surface as early as 1840, [Mehrtens, 2004, p. 291-292]. In three dimensions, we can model a solid shape by making an object that has this shape. We can model a two-dimensional surface by making an object whose surface is this surface. The geometry of surfaces made great advances in the late nineteenth century, and models were useful for reporting them. The surfaces were defined by algebraic equations, but one could check some properties directly from

The Notion of a Model

655

the model — for example whether the surface is ruled (i.e. can be generated by moving a straight line through space). But the nineteenth century geometers also discovered some monsters. There were two-dimensional surfaces that could be described algebraically but not built in three-dimensional space; the first example to be discovered was the non-euclidean hyperbolic plane. Later in the century, Poincar´e began the study of the topological classification of surfaces; in this classification length and parallels are ignored, but one is not allowed to tear or glue surfaces. Two monsters here were the real projective plane and the Klein bottle. Neither of these can be modelled as the surface of an object in three-dimensional space. These monsters were a real challenge to the model builders. Lewis Carroll taught children how to make a real projective plane by sewing together three square handkerchiefs, [1996, Ch. VII, p. 521ff]. In his story the task can’t quite be completed: “I’ll sew it up after tea”, says one character. The geometers solved their modelling problem in a magnificent way; the consequences for modern notions of model were fundamental. The solution was a pair of devices that could be used together or separately; they came to be known as ‘pseudomodels’ and ‘abstract models’. In pseudomodels we introduce a systematic and reversible distortion of some feature of what is being modelled. Two early and famous examples are Poincar´e’s models of the hyperbolic plane, known today as the circle model and the half-plane model. In both models, the distance between two points x and y in the hyperbolic plane is not the normal euclidean distance between x and y in the circle or half-plane, but a distance that depends on x and y by a subtle formula (which reappeared soon afterwards in special relativity). For example in the circle model the effect of this distortion is that the shortest path between two points is in fact a segment of a circle. Poincar´e’s circle model can be drawn on the page; as the geometer Coxeter showed, Escher’s picture Circle Limit III [Escher, 1992, p. 97]is a perfect representation of the circle model. In abstract models we give up the attempt to make a physical model. Instead we take an abstract mathematical structure, for example four-dimensional euclidean space, and we give a mathematical definition of the set of points of this structure which form our model. Both the real projective plane and the Klein bottle are easily described as abstract objects in four-dimensional euclidean space. The vocabulary used to describe these advances took some time to settle down. Abstract models were at first known as ‘interpretations’; this terminology may be in debt to the use of the word by Peacock [1834] or his French sources. Beltrami’s contribution to non-euclidean geometry bore the title Saggio di interpretazione della geometria non-euclidea, [1868]. Sometimes ‘interpretation’ was translated into German as ‘Bild’. During the 1920s distorted models of geometrical surfaces and spaces came to be known as ‘pseudospaces’. Later in the 1920s mathematicians in the school of Hilbert describe distorted abstract models as ‘pseudomodels’ or simply ‘models’ (Fraenkel, Von Neumann, Weyl). In 1936 Alfred Tarski took a further step in generalisation, and used ‘model’ to mean any systematically

656

Roland M¨ uller

distorted interpretation, for example where we read ‘left’ and ‘right’ as meaning respectively ‘right’ and ‘left’. (His context was set-theoretical interpretations of formal languages.) Gottlob Frege in 1884, in an early discussion of geometrical pseudomodels, remarked: Conceptual thought alone can after a fashion shake off the yoke [of spatial intuition], when it assumes, say, a space of four dimensions or positive curvature. To study such conceptions . . . is to leave the ground of intuition entirely behind. [1959, §14] Frege was wrong about the positive curvature — the surface of a homely apple is a space of positive curvature. But he was right about the loss of intuition. Then and now, it poses a risk for the human endeavour of studying these spaces. Witness John Nash’s masterpiece on embedding surfaces in spaces of higher dimension [2002, pp. 151–208 and the note on page 209], where a serious error was discovered only half a century after the paper was published. (Fortunately it left the mostquoted results intact.) The effect depends of course on how good a feel we have for the notions in their ‘distorted’ form. During the 1940s it became clear to a number of logicians that a notion was needed which would be a common generalisation of both pseudomodels and classes of algebraic structures (such as groups) that are defined by systems of axioms. Around 1950 two logicians, Alfred Tarski and Abraham Robinson (then a PhD student in London) came to essentially the same generalisation. They seem to have worked independently; Tarski generalised his own ‘models’ (i.e. pseudomodels), while Robinson combined work of Carnap [1942] with Bourbaki’s notion of a ‘structure’. In [1954] Tarski proposed the name ‘theory of models’ for the discipline that dealt with this new notion. Carnap had earlier tentatively used the name ‘theory of systems’ for the theory of models [1942, p. 240]; one can speculate that if Carnap’s terminology had been adopted rather than Tarski’s, the theory of modelling in the second half of the 20th century might have developed very differently. Tarski’s new notion of model was essentially a linguistic device. Language came in at three points. First, Tarski worked with a mathematically defined formal language, using artificial symbols like ‘∧’ for ‘and’. Second, Tarski put into his formal languages some indexical expressions (now known as ‘non-logical constants’) which, when applied to an appropriate structure, would pick out a particular feature of that structure. For example he assumed that the structure would have a sequence of relations, indexed by natural numbers. The corresponding indexical expressions in his formal language would be written, say, as R3 , but the meaning of this symbol was essentially ‘the third relation of the structure’. (Other model theorists have used other conventions for tying the non-logical constants to the structures.) Third was the central feature of model theory: a structure would be defined by formulas of set theory, and these formulas would be used to build up a mathematical definition of the notion ‘Sentence S is true in structure M ’ (i.e.

The Notion of a Model

657

sentence S comes out true when its indexical expressions are interpreted with reference to the structure M ). Robinson, following Carnap, added a fourth linguistic feature: his variant of Tarski’s definition of truth involved adding to the formal language names for individual elements of structures. In short, the statement of Morgan and Morrison [1999, p. 3] that ‘according to Alfred Tarski . . . a model is a non-linguistic entity’ is about as false as it could possibly be. Probably the first appearance in print of ‘model’ in this new sense was the paper of Mostowski [1952], which sets out the gory details of the linguistics. But it is true that moves were made to cut down the reference to language. For informal applications one could easily slip back from ‘∧’ to ‘and’. Also the set-theoretic definitions of structures could at least be done in informal set theory; Patrick Suppes [1960, 2002] recommended this forcefully when he applied Tarski’s models to scientific modelling. But the irreducible linguistic core is the non-logical constants; they serve as labels for certain features of the world that one wants to study. Tarski’s model-theoretic models are often known as semantic models. This name probably comes from the connection with his earlier mathematical truth definition; he had coined the name ‘semantic theory of truth’ for his own justification of it. In [1965] John Addison published some recommendations for terminology and notation in model theory [1965, pp. 438–441]; the main lines of his recommendations are still followed today in mathematical model theory. He remarked that in model theory a model is considered to be a structure related to a given theory, rather than a theory intended to explain a given realm of phenomena. [ibid p. 438] Thus for most model theorists, ‘Structure M is a model of theory T ’ means that the sentences of the theory T , when interpreted as applying to the structure M , are all true. But some mathematicians also use ‘model’ as a synonym for ‘structure’. In 1960 Robinson found himself studying theories where one particular interpretation is ‘intended’ , but other interpretations of the same formal language are used as a mathematical device. So these other interpretations were what had earlier been known as pseudomodels. But he introduced a new name: the intended interpretation was the ‘standard model’ of the relevant theory, and the pseudomodels were ‘nonstandard models’. His application of these notions to the foundations of mathematical analysis is known as ‘nonstandard analysis’. In Tarski’s version of model theory, each structure has a ‘domain’ or ‘universe’, which is a set consisting of all the objects under consideration. A common variation — though it never won Tarski’s approval — is to allow a structure to have several domains. For example in algebra we might regard a vector space as a structure with two domains, one of scalars and one of vectors. The domains are often known as ‘sorts’; a structure with more than one sort is said to be ‘manysorted’, and model theory using several sorts is called ‘many-sorted model theory’.

658

Roland M¨ uller

Some of the most successful applications of many-sorted structures have been in computer science. Pride of place goes to Edgar F. Codd’s ‘relational databases’, as follows. Database theory is concerned with representing large bodies of data in forms that allow insertion of new data, and retrieval of old data. In [1970] Edgar F. Codd introduced one of the most successful styles of database, under the name of ‘relational model of data’; a database which uses the relational model is known today as a ‘relational database’. A relational database is in fact a many-sorted structure. A well-known example is a database of a firm’s customers; this will have a domain of people, a domain of addresses, a domain of products, a domain of order numbers, and so forth. Each order has an order number, a customer who makes it, and a product which is ordered; so it is represented as an ordered triple (order-number, customer, product). The set of all such triples is a relation in the database structure, and it represents the set of orders. (And of course one can add further items, such as dates, costs, addresses . . . .) It may be worth noting that Codd’s paper contains no reference to model theory. (He contrasts the ‘relational model of data’ with graph and network models of data; here ‘model’ means ‘form of representation’.) Apparently he came to his notion by combining his experience of databases with his knowledge of logic. The resulting model-theoretic structures emerged naturally from the engineering requirements. Tarski’s notion of a structure being a ‘model of’ a formal sentence is there in the engineering problem too: a database query is precisely a sentence in a formal language (the database language), and making the query is asking whether the database is a model of the sentence. So languages are to the fore in this application too; in fact Codd’s paper has a section entitled ‘Some linguistic aspects’. The relational database query language SQL is specified by ANSI and ISO standards [ISO/IEC 9075, 2003]. As soon as the new notion of semantic model became available, philosophers of science were quick to see its relevance to their concerns. Tarski had explained the notion of a formal theory being ‘true in’ a structure. There was an obvious analogy with the way in which a formal scientific theory is or is not ‘true of’ some aspect of the real world. The analogy was particularly convincing for philosophers who regarded the ideal scientific theory as a formal theory in some mathematically defined language. To say that the theory is ‘true of’ some aspect of the world just means that if we assign certain features of the world to the non-logical constants of the theory, this makes the world into a model (in Tarski’s sense) of the theory. But there was also a clash of terminology, because scientific theories had often been regarded as a kind of ‘model’ (as in section 6.3 above). One popular solution to this terminological clash has been to think of a theory as standing not for a single model but for a class of models — the class of all its semantic models. One can finesse this idea by limiting to the class of ‘physically relevant’ models. For certain types of theory, particularly those consisting of systems of differential equations, one can pack down a whole family of seman-

The Notion of a Model

659

tic models into a single model together with a choice of real-number parameters. Some of the options are studied in Suppe [1989], Kuipers [2001] and Niiniluoto [1999]; of these, Kuipers has the most to say about the use of theories and models in technology. Suppes [1960] argues that the clash of terminology is only apparent. 8

EPILOGUE

In this chapter we have surveyed the historical development of various notions of ‘model’. Allow me to add that up to now the whole historical dimension of the use and construction of models has not received much attention. A rare exception is Roland M¨ uller [1983; 2000; 2004]. Other authors describe only parts of the history of selected meanings of ‘model’. This is not surprising, since there is no general agreement that the word ‘model’ captures a single unified phenomenon. But I hope I have shown that the uses of the word through history do have a certain coherence. ACKNOWLEDGEMENTS The author and editors would like to express their deepest gratitude to Wilfrid Hodges for his enormous dedication and substantial contribution to the form and content of this chapter. The editorial efforts of Sjoerd Zwart are also greatly acknowledged. BIBLIOGRAPHY [Achinstein, 1964] Peter Achinstein, Models, Analogies, and Theories. Philosophy of Science, 31, 328–350, 1964. [Achinstein, 1965] Peter Achinstein, Theoretical models. British Journal for the Philosophy of Science, 16, 102–120, 1965. [Addison, Henkin, Tarski, 1965] John W. Addison, Leon Henkin, Alfred Tarski (eds.), The Theory of Models. Proceedings oft the 1963 International Symposium at Berkeley, North-Holland, 1965. [Alberti, 1755] Leon Battista Alberti, De Re Aedificatoria Libri X. Alamani, 1485; English translation by James Leoni of the Italian translation by Cosimo Bartoli, Firenze, 1550. (The) Ten Books of Architecture. Reprint of the edition of London 1755; Tiranti, 1955. [Alberti, 1847] Leon Battista Alberti, Della Statua, in Opere Volgari di Leon Battista Alberti vol. 4, Tipografia Galileiana, 1847. [Ayer, 1936] Alfred Jules Ayer, Language, Truth and Logic, Gollancz, 1936. [Bacon, 1620] Francis Bacon, Novum Organum Scientiarum. Billium, 1620. (Edition of 1994). Engl. translation: The Instauratio Magna, Part II: Novum Organum and Associated Texts. Text in Latin with parallel English translation; ed. with introduction, notes, commentaries by Graham Rees with Maria Wakely. Clarendon Press, 2004. [Bacon, 1627] Francis Bacon, New Atlantis (1624). A work unfinished. Edited by W. Rawley. John Havilland for William Lee, 1627. [Baernstein and Hull, 1931] H. D. Baernstein and Clark Leonard Hull, ‘A mechanical model of the conditioned reflexes’, Journal of General Psychology, 5, 99–106, 1931. [Beltrami, 1868] Eugenio Beltrami, ‘Saggio di interpretazione della geometria non-euclidea’, Giornale di Mathematiche, 6, 284–312, 1868.

660

Roland M¨ uller

[Bennett and Ward, 1933] George K. Bennett and Lewis B. Ward, ‘A model of the synthesis of conditioned reflexes’, American Journal of Psychology, 45.2, 339 — 342, 1933, . [Bernzen, 1986] Rolf Bernzen, Die praktische und theoretische Konstruktion des Modellverfahrens. Ein Beitrag zur Fr¨ uhgeschichte der neuzeitlichen Wissenschaft. Peter Lang, 1986. [Bissell, 1990] D. Bissell, The Father of Computer Graphics: Todays graphics systems owe their existence to an innovative graduate school project called Sketchpad. Byte, June, 380-381, 1990. [Black, 1962] Max Black, Models and Metaphors. Studies in Language and Philosophy. Cornell University Press, 1962. [Bljacher, 1982] Leonid J. Bljacher, Geschichte der Biologie. Theorien, Methoden, Institutionen, Kurzbiographien (russ. 1975). Fischer, 1982, 2. ed. 1985. [Boltzman, 1902] Ludwig Boltzmann, ‘Model’, in Encyclopedia Britannica, 1902; reprinted in Ludwig Boltzmann, Theoretical Physics and Philosophical Problems, Brian McGuinness, ed. Reidel, pp. 213–220, 1974. [Boumans, 1999] Marcel Boumans, Built-in justification. In [Morgan and Morrison, 1999], pp. 66–96, 1999. [Bridgman, 1922] Percy Williams Bridgman, Dimensional Analysis. Yale University Press, 1922. [Byrne, 2008] Ruth Byrne, Mental Models Website, http://www.tcd.ie/Psychology/Ruth Byrne/mental models/ [Carnap, 1942] Rudolf Carnap, Introduction to Semantics, Harvard University Press, 1942. [Carroll, 1996] Lewis Carroll, The Complete Illustrated Lewis Carroll, Ware, 1996. [Cheney and Kuczura, 1976] Thomas K. Cheney and Anatol Kuczura, Small Modular Communications Switching System with Distributed Programmable Control, US Patent 3,974,343, 1976. [Clymer, 1955] Floyd Clymer, Henry’s Wonderful Model T. 1908–1927. McGraw-Hill, 1955. [Codd, 1970] E. F. Codd, ‘A relational model of data for large shared data banks’, Communications of the ACM 13 (6), 377 — 387, 1970. [Cotgrave, 1611] Randle Cotgrave, A Dictionarie of the French and English Tongues, Islip, 1611. [Cox, 1988] Donna J. Cox, ‘Renaissance Teams and Scientific Visualization: A Convergence of Art and Science’, Collaboration in Computer Graphics Education, SIGGRAPH 88 Educator’s Workshop Proceedings, pp. 81–104, 1988. [Craik, 1943] Kenneth J .W. Craik, The Nature of Explanation, Cambridge University Press, 1943. [Davidson and Suppes, 1956] Donald Davidson and Patrick Suppes, ‘A finitistic axiomatization of subjective probability and utility’, Econometrica, 24, 264–275, 1956. [Diderot and D’Alembert, 1765] Encyclop´ edie ou Dictionnaire raisonn´ e des Sciences, des Arts et des M´ etiers, Faulche, vol. X, 1765. [Digges, 1576] Leonard Digges, A Prognostication Everlastinge, corrected and augmented by Thomas Digges, 1576. [Dirichlet, 1871] P. G. Lejeune Dirichlet, Vorlesungen u ¨ber Zahlentheorie, mit Zus¨ atzen von R. Dedekind, Vieweg, 1871. Engl. translation: Lectures on Number Theory. A translation by John Stillwell. Vol. 16 of History of Mathematics. American Mathematical Society, Providence, RI, 1999. [Duhem, 1996] Pierre Duhem, The English School and Physical Theories: On a Recent Book by W. Thomson (1893). In Pier Duhem, Essays in History and Philosophy of Science, pp. 50–74, Hackett, 1996. [Eckhardt, 1987] Roger Eckhardt, ‘Stan Ulam, John von Neumann, and the Monte Carlo method’, Los Alamos Science Special Issue, 131 — 136, 1987. [Escher, 1992] M. C. Escher, Escher: The Complete Graphic Work, edited by F. H. Bool, Bruno Ernst, J. R. Kist, J. L. Locher and F. Wierda, Thames and Hudson, 1992. [Euclid, 1956] Euclid, The Thirteen Books of the Elements, Vol. 3, translated and edited by Thomas L. Heath, Dover Publications, 1956. [Ferguson, 1992] Eugene S. Ferguson, Engineering and the Minds Eye. MIT Press, 1992. [Fink, 1890] Karl Fink, Kurzer Abriss einer Geschichte der Elementar-Mathematik. Laupp, 1890; English translation: A Brief History of Mathematics. Open Court Publishing Co., 1910. [Fodor, 1983] Jerry A. Fodor, The Modularity of Mind, MIT Press 1983.

The Notion of a Model

661

[Ford, 1993] Brian J. Ford, Images of Science. A History of Scientific Illustrations. British Library, 1992; Oxford University Press, 1993. [Frege, 1959] Gottlob Frege, Die Grundlagen der Arithmetik, with translation by J. L. Austin, Blackwell, 1959. [Frigg and Hartmann, 2008] Roman Frigg and Stephan Hartmann, ‘Models in Science’, Stanford Encyclopedia of Philosophy, http://plato.stanford.edu. [Freud, 1895] Sigmund Freud, Project for a scientific psychology. In The Standard Edition of the Complete Psychological Works of Sigmund Freud, J. Strachey (ed.) Vol. 1, pp. 283-398. Hogarth Press, 1966. Originally written 1895. [Frontinus, 1899] Sextus Julius Frontinus, De Aquaeductu Urbis Romae. Ed. with introduction and commentary by R. H. Rodgers. Cambridge University Press 2004; Engl. translation: The Two Books on the Water Supply of the City of Rome. A photographic reproduction of the sole original Latin Ms. and its reprint in Latin; also a translation into English and explanatory chapters, by Clemens Herschel. Estes 1899; Longmans 1913; reprint New England Water Works Association, 1973 [Galilei, 1638] Galileo Galilei, Discorsi e Dimostrazioni Matematiche Intorno a ´ due Nuove Scienze, Leida, 1638. [Gentner and Stevens, 1983] Dedre Gentner, Albert L. Stevens (Eds.), Mental Models. Lawrence Erlbaum, 1983. [Glare, 1982] P. G. W. Glare (ed.), Oxford Latin Dictionary, Clarendon Press, 1982. ´ [Glassner, 2000] Jean-Jacques Glassner, Ecrire a ` Sumer: L’Invention du Cuneiforme, Seuil, 2000. [Gruchy, 1944] Allan G. Gruchy, ‘Facts and reality in the social sciences’, Ethics 54, 216–222, 1944. [Hawking, 2002] Stephen Hawking (ed.), On the Shoulders of Giants: The Great Works of Physics and Astronomy, Running Press, 2002. [Hesse, 1963] Mary Brenda Hesse, Models and Analogies in Science. Sheed and Ward, 1963. [Hooke, 1665] Robert Hooke, Micrographia. London, 1665. [ISO/IEC 9075, 2003] ‘Information technology – Database languages – SQL – Part 1: Framework’, International Organization for Standardization ISO/IEC 9075-1:2003. [Johnson-Laird and Byrne, 1991] Philip N. Johnson-Laird and Ruth M. J. Byrne, Deduction, Lawrence Erlbaum, 1991. [Johnson-Laird, 1983] Philip N. Johnson-Laird, Mental Models. Towards a Cognitive Science of Language, Inference, and Consciousness. Cambridge University Press, 1983. [Justi, 1764] Johann Heinrich Gottlob von Justi, Von der wahren Macht der Staaten. In Gesammlete Politische und Finanzschriften u ¨ber wichtige Gegenstnde der Staatskunst, der Kriegswissenschaften und des Cameral- und Finanzwesens. vol. 3, pp. 40–106. Rothen 1764. Facsimile-reprint Scientia Verlag, 1970. [Kleer and Brown, 1981] J. de Kleer and J. S. Brown, ‘Mental models of physical mechanisms and their acquisition’. In Cognitive Skills and their Acquisition, J. Anderson ed., pp. 285–302, Lawrence Erlbaum, 1981. [Klein, 1999] Ursula Klein, Techniques of modelling and paper-tools in classical chemistry. In [Morgan and Morrison, 1999, pp. 146-167], 1999. [Klein, 2003] Ursula Klein, Experiments, Models, Paper Tools. Cultures or Organic Chemistry in the Nineteenth Century. Stanford University Press, 2003. [Knauff, 2007] Markus Knauff, ‘How our brains reason logically’, Topoi 26, 19–36, 2007. [Kuipers, 2001] T. A. F. Kuipers, Structures in Science: Heuristic Patterns Based on Cognitive Structures, Kluwer 2001. [Landes, 1968] David S. Landes, The Unbound Prometheus: Technological Change and Industrial Development in Western Europe from 1750 to the Present, Cambridge University Press, 1968. [Langhaar, 1967] Henry Louis Langhaar, Dimensional Analysis and Theory of Models. Wiley/ Chapman and Hall, 1951, 8th ed. 1967. [Larkin et al., 1980] J. H. Larkin, J. McDermott, D. P. Simon and H. A. Simon, ‘Models of competence in solving physics problems’, Cognitive Science 4, 317–345, 1980. [Latham, 1965] R. E. Latham, Revised Medieval Latin Word-List from British and Irish Sources, Oxford University Press, 1965.

662

Roland M¨ uller

[Latham and Howlett, 2001] R. E. Latham and D. H. Howlett, Dictionary of Medieval Latin from British Sources. Fascicule VI, M, Oxford University Press, 2001. [Le Corbusier, 1951] Le Corbusier, Le Modulor : Essai sur une Mesure Harmonique a ` l’´ echelle Humaine Applicable Universellement a ` l’Architecture et a ` la M´ ecanique, L’Architecture d’Aujourd’hui, 1951. [Leibniz, 1903] Gottfried Wilhelm Leibniz, Opuscules et Fragments In´ edits. Louis Couturat, ed., Olms, 1966. (Reprint of Paris 1903 edition). [Levinthal, 1966] Cyrus Levinthal, ‘Molecular model-building by computer’, Scientific American 214 (6), 42–52, 1966. [Lewis, Short, 1879] Charlton T. Lewis, Charles Short, A Latin Dictionary, Clarendon Press, 1879. [Lorenz, 1963] Edward N. Lorenz, ‘Deterministic nonperiodic flow’, Journal of the Atmospheric Sciences 20, 130–141, 1963. [Malcolm, 1960] Donald G. Malcolm, ‘Bibliography on the use of simulation in management analysis’, Operations Research 8 (2), 169–177, 1960. [Martini, 1967] Francesco di Giorgio Martini, Trattati di Architettura Ingegneria e Arte Militare. A cura di Corrado Maltese. Trascrizione di Livia Maltese Degrassi. 2 vols. Poliofilo, 1967. [Mehrtens, 2004] Herbert Mehrtens, Mathematical Models. In Models: The Third Dimension of Science, Soraya de Chadarevian and Nick Hopwood (eds.), pp. 276–306, Stanford University Press 2004. [Meinel, 2004] Christoph Meinel, Molecules and Croquet Balls. In Models. The Third Dimension of Science, Soraya de Chadarevian and Nick Hopwood, eds., pp. 242-275, Stanford University Press, 2004. [Mischke, 1968] Charles Russell Mischke, An Introduction to Computer-Aided Design. Fundamentals of engineering design. Prentice-Hall, 1968. [Morgan and Morrison, 1999] Mary S. Morgan and Margaret Morrison (eds.), Models as Mediators. Perspectives on Natural and Social Science. Cambridge University Press, 1999. [Mostowski, 1952] Andrzej Mostowski, ‘On models of axiomatic systems’, Fundamenta Mathematicae 39, 133–158, 1952. [M¨ uller, 1983] Roland M¨ uller. Zur Geschichte des Modelldenkens und des Modellbegriffs. In Modelle - Konstruktion der Wirklichkeit, Herbert Stachowiak, ed., pp. 17-86, Fink, 1983. [M¨ uller, 1997] Roland M¨ uller. Innovatives Lernen am Modell. In Innovation gewinnt. Kulturgeschichte und Erfolgsrezepte, Roland M¨ uller, ed., pp. 127–137, Industrielle Organisation/Orell F¨ ussli, 1997. [M¨ uller, 2000] Roland M¨ uller. The Concept of Model and its Triple History. In Scientific Models. Their Historical and Philosophical Relevance, Erwin Neuenschwander, ed., pp. 7-19. Universit¨ at Z¨ urich, Geschichte der Naturwissenschaften, 2000. [M¨ uller, 2004] Roland M¨ uller. Model. The history of the concept and of its use in science. In Yearbook of the Artificial, Vol. 2. Models in Contemporary Sciences, Massimo Negrotti, ed., pp. 239-262, Lang, 2004. [M¨ uller, 2008] Roland M¨ uller: MuellerScience, www.htmmuellerscience.com/ENGLISH/model.htm. [Nash, 2002] John Nash, The Essential John Nash, Harold W. Kuhn and Sylvia Nasar, eds., Princeton University Press 2002. [Nersessian, 1984] Nancy J. Nersessian, Faraday to Einstein. Constructing Meaning in Scientific Theories. Nijhoff/Kluwer, 1984. [Nersessian, 2002] Nancy J. Nersessian, ‘The cognitive basis of model-based reasoning in science’. In The Cognitive Basis of Science, Peter Carruthers, Stephen Stich and Michael Siegal, eds., pp. 133–153, Cambridge University Press 2002. [Newell and Simon, 1976] Allen Newell and Herbert Alexander Simon, ‘Computer science as empirical inquiry: symbols and search’, Communications of the ACM 19, 113–126, 1976. [Niiniluoto, 1999] Ilkka Niiniluoto, Critical Scientific Realism, Oxford University Press 1999. [Norman, 1983] Donald A. Norman, ‘Some observations on mental models’, in [Gentner and Stevens, 1983] pp. 7–14. [Olschki, 1918] Leonardo Olschki, Geschichte der neusprachlichen wissenschaftlichen Literatur. 1. Bd.: Die Literatur der Technik und der angewandten Wissenschaften vom Mittelalter bis zur Renaissance. Winter 1918. [Pacioli, 1994] Luca Pacioli, Summa de Arithmetica, Geometria, Proportioni, et Proportionalita, Balassi Kiad´ o, 1994.

The Notion of a Model

663

[Pacioli, 1956] Luca Pacioli, De Divina Proportione, Fontes Ambrosiani, 1956. [Peacock, 1834] George Peacock, ‘Report on the recent progress and present state of certain branches of analysis’. In Report of the Third Meeting of the British Association for the Advancement of Science, John Murray, pp. 185–352, 1834. [Piaget, 1929] Jean Piaget, The Child’s Conception of the World, Kegan Paul/ Harcourt Brace 1929. [Piaget, 1930] Jean Piaget, The Child’s Conception of Physical Causality, Kegan Paul, 1930. [Plato, 1971] Plato, Timaeus and Critias, translated and edited by Desmond Lee, Penguin, 1971. [Pokorny, 1949]] Julius Pokorny. Indogermanisches Etymologisches W¨ orterbuch, 11th edition, Bern: A. Francke, 1948–1969. [Polk and Seifert, 2002] Thad A. Polk and Colleen M. Seifert, Cognitive Modeling, MIT Press, 2002. [Popper, 1935] Karl Popper, Logik der Forschung, Springer, 1935. [Rayleigh, 1915] Lord Rayleigh, ‘The principle of similitude’, Nature 95, 66–68, 1915. [Reynolds, 1883] Osborne Reynolds. An experimental investigation of the circumstances which determine whether the motion of water shall be direct or sinuous, and of the law of resistance in parallel channels. Philosophical Transactions of the Royal Society, March 1883. Reprinted in: Papers on Mechanical and Physical Subjects. Cambridge University Press, vol. 2, pp. 51ff., 1901. [Saalman, 1980] Howard Saalman, Filippo Brunelleschi. The Cupola of Santa Maria del Fiore. Zwimmer, 1980. [Schierbeek, 1959] A. Schierbeek, Measuring the Invisible World: The Life and Works of Antoni van Leeuwenhoek FRS, Abelard-Shuman, 1959. [Shipley, 1984] J. T. Shipley. The Origins of English Words: A Discursive Dictionary of IndoEuropean Roots. Johns Hopkins University Press, 1984. [von Schl¨ ozer, 1793] August Ludwig von Schl¨ ozer. Allgemeines Stats Recht und Stats Verfassungs Lehre. Vandenhoeck & Ruprecht, 1793. (reprint Keip, 1970). [Scholderer, 1963] Victor Scholderer, Johann Gutenberg, Inventor of Printing, British Museum, 1963. [Shepard and Metzler, 1971] Roger Newland Shepard and Jacqueline Metzler, ‘Mental rotation of three-dimensional objects’, Science 171, 701–703, 1971. [Stenning, 2002] Keith Stenning, Seeing Reason: Image and Language in Learning to Think, Oxford University Press, 2002. [Sturtevant, 1612] Simon Sturtevant, Metallica: or the Treatise of Metallica. Briefly comprehending the doctrine of diverse new metallical inventions. Eld, 1612. [Suppe, 1989] Frederick Roy Suppe, The Semantic Conception of Theories and Scientific Realism, University of Chicago Press, 1989. [Suppes, 1960] Patrick Suppes, ‘A comparison of the meaning and uses of models in mathematics and the empirical sciences’, Synthese 12 287–301, 1960. [Suppes, 2002] Patrick Suppes, Representation and Invariance of Scientific Structures. CSLI Publications, 2002. [Tarski, 1954] Alfred Tarski, Contributions to the theory of models. Indagationes Mathematicae 16, 572–588, 1954; 17, 56–64, 1955. [Tertullianus, 2004] Quintus Septimius Florens Tertullianus, Ad Nationes, Kessinger, 2004. [Thomson, 1890] William Thomson, Gesammelte Abhandlungen zur Lehre von der Elektricit¨ at und dem Magnetismus (Reprint of Papers on Electrostatics and Magnetism). German translation by L. Levy and B. Weinstein. Springer, 1890 (engl. 1872, 2. ed. 1884). [Tolman, 1932] Edward Chace Tolman, Purposive Behaviour in Animals and Man, Century, 1932. [Toulmin, 1953] Stephen Toulmin, The Philosophy of Science, Harper and Row, 1953. [Van Leeuwen et al., 1946] C. H. W. Van Leeuwen, B. Schilder and D. Veltman, Entdecker und Entdeckungen. B¨ uchergilde Gutenberg, 1946 (holl. 1941). [Vitruvius, 2004] Marcus Vitruvius Pollio, De Architectura Libri Decem, Matrix Verlag 2004. [Vonessen, 1989] Franz Vonessen, Herz-Sonne-Symbolik bei Kepler und Harvey. Zur Geistesgeschichte der Naturwissenschaft. Frankenbach, 1989. [von Wartburg, 1966] Walther von Wartburg, Franz¨ osisches Etymologisches W¨ orterbuch 1922– 1967 (keyword: modulus, 1966, 15–19)

664

Roland M¨ uller

[Watkins, 2000] Calvert Watkins, The American Heritage Dictionary of Indo-European Roots, Houghton Mifflin, 2000. ¨ [Weber, 1930] Moritz Weber, Das Allgemeine Ahnlichkeitsprinzip der Physik und sein Zusammenhang mit der Dimensionslehre und der Modellwissenschaft. Jhrb. Schiffbautechn. Ges. 31, 274–354, 1930. [Wittgenstein, 1953] Ludwig Wittgenstein, Philosophical Investigations. Blackwell, 1953. [Wirth, 1971] Niklaus Wirth, Program development by stepwise refinement. Communications of the ACM, 14 (4), 221–227, 1971. [Young and Veen, 2008] Indi Young and Jeff Veen, Mental Models: Aligning Design Strategy with Human Behavior, Rosenfeld, 2008. [Zwart, this volume] Sjoerd Zwart, ‘Scale modelling in engineering sciences’, this volume, part V.

FUNCTIONAL MODELLING AND MATHEMATICAL MODELS: A SEMANTIC ANALYSIS

Wilfrid Hodges We describe to ourselves the world that we live in—both how it is and how we would like to change it. Modelling is how we do this when we want our descriptions to be deliberate, public, precise and adjustable; the resulting descriptions are called models. Humans take an interest in a vast number of aspects of the world, and the criteria for good and bad modelling vary from field to field. But there are some basic ingredients that stay the same throughout. For example in every case there needs to be some correlation between features of the real or planned world and features of the model. This chapter will discuss these constant ingredients. For a taste of the variety of models, see the chapter by Mueller [this volume]. It seems that not much has been written about the semantics of technological modelling, except in the specialist area of software engineering. (And even in that area the philosophical literature is weak. One can still meet philosophers who believe that computer science is about Turing machines.) Being myself a semanticist and not an engineer, I am not best placed to fill the gap. But I thank the Handbook team for their wise advice, and above all Sjoerd Zwart for his guidance on the fields of marine and aerodynamical modelling. These fields contain rich material for future writers on this topic.

1 ABOUT SEMANTICS This chapter is about the two-place relation ‘M is a model of the system S’. (The system is whatever the model models. Some people call it the ‘target’.) Here is a picture of the rest of this chapter: (1)

model M

system S

Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

666

Wilfrid Hodges

Strictly what will concern us is not the two items shown, but what lies in the blank area between them. This blank area contains whatever it is that connects model to system. Common descriptions of this situation are that M ‘represents’ S, and that M ‘gives information about’ S. A semanticist is a person who asks what counts as representing, and how the information about S can be retrieved from M . These are the questions we will study. But first we need to clear some space. There are several other questions that can obstruct our view of the semantical facts.

What the model is made of Some models are made of papier-mach´e or wood. Some models are diagrams. Some models are sets of equations. Some models are computer programs. Some models consist of real-world objects with abstract labels attached. There is no good reason to try to restrict what kinds of object might serve as models. In the 1990s some questions about knots were settled by a process that involved first modelling the knots within elementary embeddings between models of set theory with large cardinals [Dehornoy, 1995]. These models of set theory are way out on the ontological fringes of mathematics; many perfectly sane mathematicians deny that they exist at all. Fortunately Dehornoy was able to write out the results in a form that is convincing for any mainstream mathematician. So the set-theoretic models are not needed for proving the results; but they were needed for discovering the proofs in the first place. A San Francisco newspaper interviewed another set theorist, Julius Barbanel, and asked him ‘How do you discover facts about numbers that are so large they might not exist?’ He answered ‘I do it in the shower’. Quite right too. We can’t let the ontologists restrict our routes to discovery. The reason I mention ontology at all is that in some past centuries the intrusion of ontology has done serious damage to the understanding of semantics. I trust we are beyond that now, and today’s ontologists have neither the wish nor the power to inhibit semantic research. But vigilance is prudent.

The user’s intention ‘An object M is not a model of S unless the creator or user of M intends it to be one. So an account of what it is for M to be a model of S needs to involve people’s intentions.’ Thus speaks an imaginary philosopher. To illustrate her first sentence, this philosopher might cite the heat equation:  2  d ρ d2 ρ d 2 ρ ∂ρ =D (2) + + ∂t dx2 dy 2 dz 2 This equation is used to model any number of things. Even when t is understood to be time, the function ρ(x, y, z, t) could be the density of matter at the point

Functional Modelling and Mathematical Models

667

(x, y, z) at time t resulting from diffusion of a substance introduced at time t = 0. Or it could be the probability density of finding a particle at point (x, y, z) at time t when the particle is moving under Brownian motion. Or it could be the temperature at point (x, y, z) at time t resulting from spread of heat from a hot body. And so on. The heat equation is important precisely because of the large number of kinds of situation that it can be used to model. Now the semanticist asks: What counts as the heat equation being true of (say) the temperature in a system of the third kind above? The answer has to involve some correlation between the symbol ρ in the equation and the temperature in the system. But is it really true that we need to mention some person’s intention in order to say what this correlation is? The common sense view—and I think the correct one—is that it’s the other way round: we need to be able to say what this correlation is in order to say meaningfully that somebody intends it. Where the model is a step in the design of the system S, the opposite question can be asked: is S an accurate exemplification of the model M ? But the same general point applies: we can describe what is the relevant kind of resemblance between S and M without having to refer to anybody intending that kind of resemblance. But our imaginary philosopher was right about one important thing. The question ‘Is M an accurate model of S?’ doesn’t have an answer independent of any correlation between the features of M and those of S. The question has to be short for something more specific, such as: Under the following correlation between features of M and features of S (. . . and here follows a description of the correlation . . . ), does M form an accurate model of S?

(3)

Is there a correlation between features of M and features of S which would make M an accurate model of S?

(4)

or:

These questions do involve a correlation linking M and S. But there is no need for them to mention anybody intending this correlation. There are of course important philosophical questions about the nature of meaning and reference, and how they relate to the intentions of the speaker or writer. But these questions belong to the philosophy of semantics; count them as part of semantics itself and they become intrusive.

Semantic models One of the tools of modern semantics is the notion of a semantic model. This is a technical notion with a rather precise definition. Semantic models don’t belong in the same list as scale models, analogue models, diagram models, mathematical models and so forth. The reason why they are called ‘models’ at all is historical and rather indirect (see Mueller [this volume] Section 7). There is no point in

668

Wilfrid Hodges

discussing them without reference to their definition. We will come to that in due course, but here is a brief summary of what they are. Alfred Tarski laid the foundations of modern semantics in a monograph which he published in Polish in 1933. Today we usually refer to the English translation of the German translation of this work [Tarski, 1983]. This monograph of Tarski was not about models. The word ‘model’ never occurs in it. Tarski set himself the task of giving a purely mathematical definition of the set of true sentences of a language L. The language L had to be already describable in mathematical terms; in fact he required that it was a formal language like those used by logicians, but he also assumed that every sentence of L had a meaning. By ‘purely mathematical’ he meant that his definition of ‘true sentence’ was written in the language of set theory, except that it was allowed to use a description of L, including a correlation between the individual words of L and their meanings. So we see that Tarski’s ideas were all about languages. Together with the object language L, there were the metalanguage in which he gave his definition of ‘true sentence’, and the meta-metalanguage in which he discussed the metalanguage. For mathematical applications Tarski revised his truth definition in the 1950s. He dropped the assumption that the sentences of the object language already mean something. Instead he took an object language L containing a set of meaningless symbols called primitives, and he allowed us to choose any reasonable correlation between primitives and meanings—provided that the meanings are given in a settheoretic form which we will discuss in Section 7 below. Such a correlation is now usually called a structure, though in 1954 Tarski used the name relational system. In Tarski’s terminology, a structure A is a model of the formal sentence T if T becomes a true sentence when A is used to give meanings to the primitives in T . Revised in these terms, Tarski’s definition of truth from the 1930s became a definition of the relation ‘A is a model of T ’. Tarski proposed the name model theory for the study of this relation. (See [Tarski, 1954; Hodges, 2001] for further information.) Structures in the sense of the previous paragraph are also sometimes called models; though careful writers restrict the use of this word to the context ‘model of (a sentence or a set of sentences)’. Other names exist; for example [Van Fraassen, 1990, p. 45] calls them ‘relational structures’. Sometimes they are called interpretations, because they interpret the primitives (as we will see in Section 5). In the context of modelling there is an obvious danger of confusing Tarski’s models with other kinds, and so one refers to Tarski’s models as semantic models. The name is a little ironic, because Tarski’s aim and achievement in his work of the 1930s was to paraphrase away all semantical notions and replace them by set-theoretic ones. But it makes sense if we reckon that Tarski was uncovering the mathematical properties of meanings. In fact Tarski’s student Richard Montague did precisely use a form of Tarski’s model theory to create a theory of meanings in natural languages [Dowty et al., 1981].

Functional Modelling and Mathematical Models

669

Now it should be clear at once that hardly any models in the sense of ‘modelling’ are structures (i.e. correlations between primitives and their meanings). A model railway doesn’t correlate primitives and their meanings; Bohr’s model of the hydrogen atom doesn’t correlate primitives and their meanings; and so on. Hence hardly any models in the sense of ‘modelling’ are semantic models of anything. But semantic models are still very relevant to modelling, because they help to fill the space between M and S in (1). The model M is a kind of generalisation of Tarski’s sentence T , and the system S is a kind of generalisation of Tarski’s structure A. (Note the reversal: in Tarski’s terminology the semantic model belongs with S and not with M ! We have to live with this.) This paper will be mostly about the kinds of generalisation that are needed on the two sides. On the basis of this introduction, I hope the reader will forgive me for making some requests. • Don’t assume that anything sensible can be said about semantic models just on the basis of other uses of the words ‘model’, ‘semantic’ or ‘structure’. • The question whether various other kinds of model ‘are’ semantical models makes sense only as a question about whether the one kind of model can in some sense be encoded in the other. You can’t sensibly discuss this without asking what kinds of encoding would work. • It doesn’t make any sense in the abstract to talk about a semantical model being ‘isomorphic to’ or ‘similar to’ a system that is being modelled. A correlation between primitives and their meanings is not in any straightforward sense even remotely similar to a wind turbine, for example. The same applies to talk of a semantic model being ‘isomorphic to’ a system being modelled. Some theories can be encoded in semantic models, and some semantic models are useful for discussing similarity. But these points can only be established by careful analysis on the basis of the definitions. They are way beyond the reach of general reflections on epistemological outlooks (for example). 2

THREE RUNNING EXAMPLES

It will be helpful to have three running examples of modelling. The first example comes from fundamental science, not from technology; the excuse for including it is that it is very familiar, and also very much simpler than most examples of technological modelling.

The weight on a spring Our first example is a weight suspended on a spring, where the spring is stretched and then released. This example is a regular visitor in discussions of scientific modelling. I rely on Richard Feynman’s account in Sections 21.2 and 24.2 of his

670

Wilfrid Hodges

Lectures on Physics [Feynman et al., 1963]. A weight is suspended on a string which is firmly anchored at the top end. The weight is moved vertically from equilibrium position, and then the system is left to adjust on its own. (Note that the system is not just the spring and the weight, but also the situation of moving and then releasing the weight.) Feynman calculates that if there is no friction, the vertical coordinate x of the centre of mass of the weight will move according to the equation x = A0 cos ω0 t

(5)

where A0 is the amplitude of the oscillation and ω0 is the natural frequency of the system. If we introduce friction through a constant resistive coefficient γ, the equation for x becomes more complicated: x = Ae−γt/2 cos(ωγ t) where ωγ2 = ω02 −

γ2 . 4

(6)

We can read these as equations in the real numbers, unless γ > 2ω0 ; in this case we can either switch to a different formula, or read the equation in the complex numbers and take the displacement of the centre of mass to be the real part of x. (This is the case where the frictional damping prevents any oscillation.)

The development of wind turbines This is an area undergoing intensive development as I write, in response to climate change and anticipated rises in the price of oil [Wizelius, 2007]. A typical wind turbine consists of a high tower at the top of which are (a) a rotor consisting of three blades fixed to a central shaft, so that the shaft turns when the rotor is faced into the wind, (b) an electrical generator driven by the rotating shaft and (c) a yaw mechanism which turns the rotor to face into the wind. Usually the turbine contains one or more gears connecting the axle to the generator, and some control and maintenance devices. The first requirement of a wind turbine is to deliver electricity as efficiently as possible. The efficiency depends crucially on the design of the rotor blades. Scientific experiments on the shapes of blades go back to the mid eighteenth century. Thanks to work of Albert Betz, Hans Glauert and others in the mid twentieth century, there is now a good quantitative theory of rotor design. But the resulting equations don’t lend themselves to easy solution; most researchers in this field use computers to generate approximate solutions. But there are some other major requirements. Above all, the turbine and its tower must not break down. The expected lifetime can be more than two decades. Sudden sharp bursts of wind will put all parts of the system under strain. An electrical fault may cause the load at the generator to drop suddenly, so that the rotor races and breaks. (Mechanical breakdowns were given as a major reason when the leading Californian wind turbine company Kenetech Windpower filed for bankruptcy in 1996.)

Functional Modelling and Mathematical Models

671

Another requirement is easy starting. A rotor that is good for efficient running may be harder to set in motion. Some designers ensure that the rotor blades have segments that respond quickly to light winds, at some loss in overall efficiency. So models of wind turbines tend to be complex in several different senses. (a) There are several requirements to be met; they conflict and a balance has to be struck. (b) The turbine itself is complex, and different kinds of modelling may be needed for different parts (even for different parts of the rotor blades). (c) The environment in which a wind turbine works is very variable. Steady laminar flow of air behaves quite differently from turbulent flow, and sudden violent gusts are different yet again. (Many turbines now have devices that allow them to run at different speeds depending on the weather conditions; so complexity (c) spills over into complexity (b).)

Software design This example is based on C. A. R. Hoare’s construction and verification of a program Find in [Hoare, 1989]. Hoare begins with a description of the purpose of the program Find : . . . to find that element of an array A[1 : N ] whose value is f th in order of magnitude; and to rearrange the array in such a way that this element is placed in A[f ]; and furthermore, all elements with subscripts lower than f have lesser [or equal] values, and all elements with subscripts greater than f have greater [or equal] values.

(7)

(The words in square brackets are missing from Hoare’s text, but his explanation presupposes them.) For example if the array (9, 5, 8, 2, 9, 9, 4, 5, 1, 3, 8, 6, 2, 6)

(8)

is input, and f = 8, the program should output an array (m, m, m, m, m, m, m, 6, n, n, n, n, n, n)

(9)

where the places marked m are filled with 1, 2, 2, 3, 4, 5, 5 in some order and the places marked n are filled with 6, 8, 8, 9, 9, 9 in some order. Hoare builds up a program and shows that it does the required job. Here we won’t need the details, but a broad description of his approach will help. We are given the array A and the number f . For convenience we write f  for the number which comes in the f -th position if A is rearranged into nondecreasing order; though the identity of f  won’t be known until the end of Hoare’s procedure. We begin by choosing a number r between 1 and N , and we put m = 1 and n = N so that A is the array (A[m], . . . , A[n]).

(10)

672

Wilfrid Hodges

We perform an operation on the array which has the following effect. Some numbers in the array A may be moved to different positions, and the number r may be changed. Also either m is increased or n is decreased, so that the new array (A[m], . . . , A[n]) is shorter than what we called (A[m], . . . , A[n]) before the operation. One can show that after this operation, A[1], . . . , A[m − 1] are all  f  and A[n + 1], . . . , A[N ] are all  f  , so that we can leave these ends of the array untouched for the rest of the procedure. We then repeat the operation, but on the new (A[m], . . . , A[n]) and using the new value of r. Again the effect is that we move around some numbers in this shorter array and maybe change r, and we shrink the difference between m and n. We keep doing this until m and n meet. One of Hoare’s main purposes in his paper is to prove rigorously that when his procedure rolls to a halt, the requirements of (7) will be satisfied. Neither Feynman nor Hoare uses the words ‘model’ or ‘modelling’. But in the next section we will see how their examples fall into the general pattern. 3 EXPLANATORY VERSUS FUNCTIONAL MODELLING At a first approximation we can distinguish between (1) explanatory modelling, whose purpose is to explain the workings of something that already exists, and (2) functional modelling, where one makes a model that shows how something can be built to perform a certain function. You might want to add some other kinds, for example predictive modelling. But for our purposes there is not much point in distinguishing between explanatory and predictive models; both of them do their job if they correctly tell us how the world is. The examples in the previous section illustrate these notions. Feynman’s description of the weight on the spring is useless for telling us how to attach a weight to a string, but it does report how weights on strings behave. So it is almost purely explanatory. (I say ‘almost’ because one could dream up a situation where someone needs a thing that behaves in a certain way, and Feynman’s description is used to confirm that the weight on a spring will do what is required.) By contrast Hoare’s description of the program Find is purely functional. It tells us what the program is supposed to achieve, but it gives no information at all about how the world works. The example of wind turbines is more typical of technological modelling. One can see at once that it involves elements of both explanatory and functional modelling. The overall purpose is functional, namely to design a machine that performs the functions of a wind turbine. But a good deal of the modelling along the way will be to explain how the wind works and what effect it has on certain mechanical devices. As with all explanatory modelling, the question arises whether the system really does behave as the model says it does. Showing that the answer is ‘Yes’ is called validating the model. This is not the place to discuss what an explanation is. (See Pitt [this volume].) But it’s clear that the requirements for an explanation of the behaviour of air in

Functional Modelling and Mathematical Models

673

relation to wind turbines are not the same as those for a theoretical physicist. For the engineer, the explanation is needed in order to be sure that the wind turbine will do its job reliably and not cause any damage. An approximate description up to a certain accuracy will be fine for this; the physicist might reject it as simply incorrect, even when the error is known to be small. This is an oversimplification, of course. Physicists do constantly use unreal approximations—think of the smooth pulley. And if engineers use a method known to be sufficiently accurate under certain circumstances, they need to check whether they can still use it when the circumstances change. To quote [Quarton, 1998, p. 7f], . . . The previously accepted simplified method involved the generation of component fatigue load spectra from the mean loads at the rated wind speed of the wind turbine. This approach, although undoubtedly conservative for certain components of a wind turbine, is now recognized as unsafe for others, particularly in the case of large machines. Quarton gives a graph showing the vast difference between the results of the simplified calculation and a rigorous simulation under certain circumstances. We will need to say more about the use of approximations. It will be helpful to have in hand a description of one common ingredient of wind turbine models. This is the beam-element momentum (BEM) model. The model assumes a stream of air flowing through the rotor. At a certain distance in front of the rotor, the air has speed v; at a certain distance behind the rotor the speed has dropped to w. The drop in speed, v − w, indicates the drop in kinetic energy of the air; the turbine picks up this energy and converts it into electrical power and some heat. The BEM model predicts the drop in the kinetic energy of the air. It does so by combining two separate calculations. (1) The first is the momentum calculation. For this, we assume that the drop in air velocity occurs within a disc of infinitesimal thickness—known as the actuator disc—at right angles to the incoming air stream. The momentum calculation gives the components of force on an element of air within the actuator disc, in terms of the velocities v, w and the angular velocity of the rotor. (2) The second is the blade-element calculation; this derives equations for the components of force on an element of air within the disc, in terms of the lift and drag coefficients of the rotor blade (which can be determined experimentally). Equating the expressions for each component of force, we reach equations that relate v and w to characteristics of the rotor. Since the BEM model involves several simplifications, it needs to be validated by experiment. In practice it works well for relatively steady air flow, but turbulence throws it out of line. (See for example Sørensen and Kock [1995] for an analysis and some experimental data.) Even within a single model it may not make sense to distinguish between the functional and the explanatory components. Take for example the following com-

674

Wilfrid Hodges

ment on fishery management: The manager can look at the problem from an ecological point of view (ecologically enlightened manager), who considers the consequences of harvesting on the population size of fish, but does not assume that the fish will evolve to a new adult size. Or the manager can look at the problem from an evolutionary point of view (evolutionarily enlightened manager), who considers both the ecological and evolutionary consequences of harvesting. ([Vincent and Brown, 2005, p. 353])

(11)

The following could happen. Our ecologically enlightened manager builds a model and feeds into it some parameter values that describe a certain species of fish; other parts of the model describe the effects of human intervention, and are used to guide policy. Then her evolutionarily enlightened colleague takes over the model and adds a component that predicts how the characteristics of the species evolve under pressure from the planned environment; in this model the form of the species ceases to be a datum and becomes part of what is planned. Where modelling is used for purposes of social or environmental planning, we should expect that sometimes the parameters of a model could be set to current values, or they could be used to describe the state we want to achieve. Some possible parameter values will just be byproducts of the model—values that we couldn’t achieve or would never want to. Functional modelling often takes place in a commercial setting. The customer wants a system for some purpose. The modeller and the customer agree a description (called a specification) of what the system is intended to do. There are often many stages of refinement before the specification is implemented (i.e. turned into a system). But functional modelling can also be driven purely by research. 4

PICTURE OR TEXT

We come back to our diagram (1) and the question how the model M is connected to the system S. Feynman’s weight on a spring is a convenient place to start. In all three examples below, S is a weight on a spring, either a real one or one that we propose to construct. But the connection takes a different form in each example. (A) The model is a computer mockup in virtual reality. It looks like the real thing, and maybe sounds like the real thing if the programmer was having fun. In this case the connection with the system is resemblance or similarity; we say that the model is pictorial. In functional modelling the modeller will sometimes turn an early stage of the specification into a toy working system, called a prototype. The prototype is a pictorial model of the final system. It shows how the final system will operate, by working more or less like the final system but maybe with some features missing.

Functional Modelling and Mathematical Models

675

(B) The model is a graph, for example [Feynman et al., 1963, p. 24-3]:

(12)

This graph describes the system less directly. Left to right in the graph represents time, up and down represents the vertical distance of the centre of mass of the weight from its resting position. In both dimensions a distance in the graph is proportional to a distance in space or time. A model that can be read in this way, by taking some dimensions in the model as corresponding to some dimensions in the system, is called an analogue model. In hydraulic and aeronautical engineering one often meets scale models. These are analogue models where the dimensions of the final system are accurately scaled up or down (usually down) so that the model is a more convenient size than the final system. But if all the dimensions are scaled down in a ratio r, then the areas are scaled down in ratio r2 and the volumes (and hence the weights) in ratio r3 . So given the laws of physics, how should we scale the time if we want the behaviour of the model to predict the behaviour of the system? Dimensional analysis answers this question (see Zwart’s chapter in this Volume). A model can be both pictorial and analogue. For example the architect’s model is both. But the model in (12) is clearly not a pictorial model; it doesn’t look anything like a weight on a spring. (C) Feynman himself models his system with an equation, (6). His equation is a piece of text which makes a statement about the system. We call it a textual model. Some fields have developed specialist notations for their subject matter. Generally these notations are textual, in the sense that they build up expressions from a finite alphabet, though there may be pictorial reasons why one symbol was chosen rather than another. Often they are meant to be written and read rather than spoken. Musical scores are an obvious example. There is no inherent problem about translating the score of a Beethoven piano sonata symbol by symbol into English (‘The time signature is common time, the first bar begins with a minim at g in the right hand . . . ’) in such a way that the original score could be recovered from the English—though I can’t think why anybody would want to do it. The analogue model (12) doesn’t translate into English in any similar way.

676

Wilfrid Hodges

Another example of a textual notation is Universal Modelling Language (UML), which is often used in early stages of software modelling; it’s less specialist than musical scores but still very limited in what it can express. The diagram notation of Barwise and Etchemendy’s logic software Hyperproof [Barwise and Etchemendy, 1994] looks pictorial but can be read as textual. The upper part of the screen carries pictures of a chessboard with various objects arranged on it, while the lower part carries conventional logical formulas. Builtin rules (Observe and Apply) carry information from pictures to formulas and from formulas to pictures. Although we can say in formulas anything we can say in pictures, and to some extent vice versa, the two kinds of representation encourage different habits of reasoning. Keith Stenning [Stenning, 2002] reports some valuable experiments on how students with different cognitive styles use Hyperproof. His Chapter Two surveys different forms of representation from a cognitive point of view. (See also the chater by Nersessian and Patton in this Volume on model-based reasoning.) Textual models are particularly important for us because in principle Tarski’s semantic analysis applies to them. Tarski himself said ([Tarski, 1984, p. 267): Whoever wishes . . . to pursue the semantics of colloquial language with the help of exact methods will be driven first to undertake the thankless task of a reform of this language. . . . It may, however, be doubted whether the language of everyday life, after being ‘rationalized’ in this way, would still preserve its naturalness and whether it would not rather take on the characteristic features of the formalized languages. Tarski may have intended these remarks to discourage people from extending his semantic theory beyond the case of formalised languages. But today his theory is applied very generally, and the ‘rationalisation’ that he refers to is taken as part of the job of a semanticist. For example the diagrams of Barwise and Etchemendy (above) are studied in this spirit. Very many models are text from the outset, or can be read as text. One important case that we need to consider is computer models. For example models for wind turbines are usually presented as computer programs together with some accompanying theory to justify the programs. For semantic analysis we need to be more precise about exactly what feature of a computer model is the actual model. Let me give my own answer; other analysts may see things differently. The information about the proposed wind turbine is got by running the program. So we should count the model as being the output of the program. The output may include text printed on the screen or saved in a file; in this respect the model is textual. The output may also consist of pictures on the screen, or graphs; in this respect the model is pictorial, and possibly also analogue. Dynamic real-time simulations are certainly analogue; they may include sound as well as graphics. Often the same program can be run with different parameters. (For example in wind turbine modelling one uses programs that simulate wind conditions and

Functional Modelling and Mathematical Models

677

are seeded with a random number. For practical purposes these programs can be thought of as allowing infinitely many parameter values.) Only a few of these runs will ever take place; the rest are virtual outputs. So we have to allow that a textual model can consist of virtual text—or perhaps better, it can consist of a family of different virtual texts. On this reading, the computer program itself is not the model; it is a device for generating the model. Also the background theory supporting the program is not the model—though often it contains explanatory models of other things. Feynman’s example throws up a further object that we need to notice. This is the function f which, for any real numbers A, γ, t, ωγ , takes the value f (A, γ, t, ωγ ) = Ae−γt/2 cos ωγ t.

(13)

This function is an abstract object. For definiteness some people give it a settheoretic form by identifying it with a set of ordered 5-tuples of real numbers. Although the function clearly bears some close relationship to the equation (6), it’s a wholly different kind of object. We can’t put it on a page or a screen, or make it out of wood or plaster of paris. In short it is not ‘accessible’ to us in any direct way. We can only have any cognitive relationship to it through some description of it—for example the equation (6). For this reason I think we should hesitate to call the function a ‘model’ of the spring-weight system. (Later we will see that it’s closer to a semantic model, though it isn’t quite that either.) Nor should we confuse functions in this sense with the ‘function’ of an artefact as in functional modelling (on which see the chapter by Vermaas and Garbacz in this Volume). 5

PRIMITIVES AND INTERPRETATIONS

In order to connect with the system, the text of a model has to contain some expressions that refer to features of the system. For example a description of a spring-weight system may contain a phrase the weight

(14)

No specific weight is mentioned. But when a particular system is under discussion, by default this phrase (14) refers to the weight in the system. Or consider the symbol A in Feynman’s example (6). To attach the equation to the system, we need to take A as measuring the vertical displacement of the weight at time t = 0; but there is nothing in the symbol to tell us this. At most there might be a convention that a symbol in this position refers to a real or complex number. If a piece of text contains a word or symbol that needs interpretation before we can say that the text makes a true or false statement, then (as in Section 1 above) we say that the word or symbol is a primitive. The name comes from the 17th century essay [Pascal, 1963, p. 350], but with a curious twist. According to

678

Wilfrid Hodges

Pascal, we don’t need to give the meanings of our primitives, because everybody already knows what they mean. In today’s terminology, we don’t give permanent meanings for our primitives because they don’t have any; but we do need to explain what they refer to in any application. Another name for primitives in the logical literature is non-logical constant. Linguists sometimes call our primitives indexicals because the user needs to indicate what they refer to (for example by pointing with her index finger ). So one ingredient of modelling is to set up a correlation between certain symbols or expressions (the primitives) and certain features of the system. A correlation that does this is called an interpretation. We will see in Section 7 (and we mentioned already in Section 1) that a structure (in the sense of model theory) is a kind of interpretation. Meanwhile we will examine in this section what needs to be done to the system to extract the relevant features. The system itself hardly ever tells us where and how to connect the primitives; the example of ‘the weight’ in (14) was unusually straightforward. An extremely important aspect of modelling is that one often needs to attach primitives to things that could be regarded as not parts of the system at all. Here follow four kinds of example. Coordinate systems. The symbol A in (6) stands for a real number (or possibly a complex number, as Feynman explains). The particular spring-weight system S doesn’t come with real numbers attached. To relate A to the system we need to put the system into a spatial coordinate system. The case of time is a little different. The symbol t in (6) is not a primitive; it isn’t intended to refer to a particular moment of time. Rather (6) says (in the intended reading of it) that a certain equation holds for all values of t in some range to be supplied. There is no expression in (6) that refers to this range. But to make sense of (6) we need to suppose that there is a tacit reference to the range; if you like, (6) is short for another more explicit piece of text, for example x(t) = Ae−γt/2 cos(ωγ t) 2 where ωγ2 = ω02 − γ4 .

(for all t ∈ J)

(15)

Here the symbol J is a primitive that needs to be interpreted as referring to an interval in the field R of real numbers. To connect J to the system, we need to specify this interval as the set of values in a certain interval of time measured on some given time scale. The symbol x needs to be interpreted as referring to the vertical displacement of the weight at time t, where both the displacement and the time are measured on the scales just mentioned. Incidentally I took the liberty of making the dependence of x on t explicit in (15). On one possible reading, when we use (6) as a model of a spring-weight system, we need not take it as referring implicitly or explicitly to any coordinate system. Instead we can read it as saying ‘There is a coordinate system such that . . . ’. Well yes, this is a possible reading. But I make two remarks on it. First, the reading is

Functional Modelling and Mathematical Models

679

empty unless something is said or implied about what coordinate systems would count. You don’t want to use a logarithmic scale for the time, for example. And second, logicians will note that in order to explain what it means to say ‘There is a coordinate system A such that φ(A)’, we first need an explanation of what it means to say ‘φ(A)’. So this quantified reading of (6) is not a very effective way of sweeping the coordinate systems under the carpet. It’s important to note that the coordinates referred to in a model need not be space-time coordinates. To model vibrations of a complex mechanical system, as for example a wind turbine, it can be convenient to apply a Fourier transform so that one dimension represents frequency and another represents power. (The model is then said to be in the frequency domain.) Observable versus theoretical. Sometimes a model contains explicit primitives referring to things that can’t be read off from the system. Feynman’s example is too simple-minded to illustrate this convincingly. But for example the constant  A is known to be equal to 2T /k where T is the initial energy of the system and k is the spring constant. The values of x can be observed by measuring with a ruler (and remembering that the modelled system includes the initial displacement of the weight). But one could argue that we can only discover the initial energy of the system by making some other measurements and assuming that the formula  A = 2T /k holds—so strictly T is not observed from the system. In a terminology sometimes used in philosophy of science, the function x in (15) is observable, whereas T is theoretical. I don’t know any examples of theoretical items in purely functional modelling; but there is no reason why they shouldn’t exist. They certainly can be found in the explanatory parts of functional modelling. The actuator disc in the BEM model (Section 3 above) is a clear example; infinitesimal objects are theoretical almost by definition. In many cases (ranging from statistical mechanics to the theory of elections) the use of infinity is a convenience and one knows in principle how to eliminate it. There are some indications that the use of the actuator disc in the BEM model is not of this kind, and it contains significant and false physical assumptions. We return to this in the next section. Nancy Cartwright ([1999] passim) attacks the observable/theoretical distinction. Some physical quantities that are supposedly measurable very often aren’t. For example in a Newtonian universe a particle may be acted on simultaneously by pressure and by gravitational and electrostatic forces. We may have theories that tell us the values of these forces. But what we actually see is the movement produced by the resultant of the forces, and there is no way of separating out the components for inspection without making major changes in the physical setup. It seems that Cartwright’s point applies equally to functional modelling of physical mechanisms. For example there is no prospect of deducing the movement of the air from the movement of the wind turbine alone; and introducing other tests of wind speed and direction will interfere with the operation of the turbine.

680

Wilfrid Hodges

External versus internal features. Sometimes functional modellers make a distinction which looks similar to that between observable and theoretical. External functional analysis (or the functional analysis of need), lists the services the product is required to provide irrespective of the means available to it to provide them. . . . Internal functional analysis (or technical functional analysis) enables us to analyse the resources required and the way they are allocated to provide the service required. [Prudhomme et al., 2003]

(16)

But the basis for the external/internal distinction is quite different from the observable/theoretical distinction in philosophy of science. The internal features of the constructed system really are there and must have been observed by the person who constructed it. In more physical kinds of engineering there are plenty of examples of systems with ‘internal’ functional features that everybody can see: for example the piston boxes on a steam engine (unless you think that steam engines were built for the pleasure of train-spotters). Object versus behaviour. In Hoare’s example the system to be constructed is a piece of software in some programming language. But the initial description (7) didn’t say anything about the software; it said what the software is supposed to achieve. Hoare himself presents a series of refinements of this description, that get closer and closer to a piece of software, but at every stage he is still describing the behaviour of the software. Now software itself in a sense describes its own behaviour; it tells you what will happen when it is run. So we can think of the software as a more refined version of the model; both of them are descriptions of the same intended behaviour. This raises the possibility of building the software S from the specification M by textual operations. What kind of refinement is needed to turn M into S? The reason why M won’t work as it stands is that the computer can’t read it. The descriptions are not ones that the computer can read as instructions. In order to turn them into instructions, we need to make the description more concrete, and we need to include descriptions of the data structures that the computer will need for the computation. (These data structures are internal features as defined above.) Computer scientists have worked out the required methodology in great detail; see for example [De Roever and Engelhardt, 1998]. Starting from the initial specification (which might form the basis of a contract with the customer), the software developer rewrites the text several times, each time making it more specific. This process is sometimes known as stepwise refinement. Hoare’s paper helped to found the methodology of this process. For a semanticist the process is interesting because of the direct role that semantics takes in the engineering task itself.

Functional Modelling and Mathematical Models

6

681

TRUTH AND ITS ALTERNATIVES

Suppose we have our model connected to its system via a suitable interpretation. Then reading the primitives according to the interpretation, the model makes a statement about the system. If everything is in order, this statement is true. If the statement isn’t true then either the modeller has incorrectly reported the world, or the implementer failed to copy the model correctly. At this point Tarski’s semantics kicks in to tell us what features of the interpretation are needed in order to define whether the model makes a true statement. The answer is a structure. But before we come to that in the next section below, we notice that in practice we often have to put some reservations on the notion of truth. These reservations are a challenge to the semanticist: do they mean that part of Tarski’s account has to be abandoned or generalised? There are two cases to consider.

Case One: An approximation suffices. The first and most obvious reservation is that physical objects have rough edges. Granted, sometimes we can be extraordinarily precise. Feynman remarks (about an experimental value for Dirac’s constant): If you were to measure the distance from Los Angeles to New York to this accuracy, it would be exact to the thickness of a human hair. That’s how delicately quantum electrodynamics has, in the past fifty years, been checked—both theoretically and experimentally. [Feynman, 1990, p. 7]

(17)

But still there is a grey area, small though it is. With the very best measuring equipment there is no physical action that constitutes putting two point charges at a distance of exactly one centimetre. So there is no precise separation between the truth and the falsity of the statement ‘These two charges are at a distance of one centimetre’. This problem affects both explanatory and functional modelling when the systems contain physical objects. The scenery can change when we move to less physical systems. If our system is software then we can make statements about its behaviour that are either absolutely true or absolutely false, with no middle ground at all. In principle the same applies to any system that allows a digital description, for example a digital recording of a piece of music. By contrast Ramsey in his essay basing probability on subjective beliefs comments I have not worked out the mathematical logic of this in detail, because this would, I think, be rather like working out to seven places of decimals a result only valid to two. [Ramsey, 1978, p. 82]

(18)

Psychological measurers often have to be content with vague boundaries. In practice there is often a reasonable solution to vagueness of boundaries: we choose a degree of inaccuracy that we are prepared to tolerate, and we count a

682

Wilfrid Hodges

description as true if it is true within that degree of inaccuracy. This approach allows us to ignore the difference between two different kinds of error: getting a precise number wrong, and putting a precise number onto a quantity that doesn’t have a precise value. On the other hand it can be a nontrivial matter to tell whether a model does fit the facts within the allowed tolerance. An example is to determine the size of the various relativistic corrections needed for calculating position in a satellite navigation system. They turn out to be so large that a system using Newtonian physics would be completely unacceptable for many practical purposes. (See [Ashby, 2003] for this example, Section 4 of Suppes [this Volume] for errors in measurement, and [Taylor, 1997] for analysis of errors in general.) This case adds a complication to Tarski semantics, but nothing more. Instead of examining the truth of statements like ‘The power generated is 500kW’, we examine the truth of statements like ‘The power generated is within ε of 500kW’ where ε is some fixed small number. This complicates the object language, but we are back in a situation where truth rules: an explanatory model aims to say something true, a functional model invites us to make it true.

Case Two: False but we don’t care. Sometimes we accept a model/ system pair where the model gives an admittedly false description of the system, but we don’t care. This can happen both with explanatory modelling and with functional, when we make an assumption that simplifies the calculations but is known to be incorrect. Incorrect assumptions can be global, but often a model divides the system into several components and makes different simplifying assumptions on the different components. A fairly straightforward example of this is Prandtl’s handling of fluid flow by separating off a thin boundary layer and assuming that friction operates throughout the boundary layer but nowhere else; see [Morrison, 1999, p. 53ff]. The statement that the flow of a fluid splits into two parts, in which different laws apply, is simply false; we allow it because the resulting theory is computationally tractable and leads to numerical values that are close to observed values. A subtler example appears in the BEM model for wind turbines. Here the components are conceptual, not physical. Recall that momentum theory yields equations that relate the force F on elements of the air in the actuator disc to the air velocities V before and behind the rotor, and blade-element theory yields equations that relate F to blade characteristics C. We relate V to C by simultaneously solving these equations. The simplifying assumptions made in the momentum theory are different from those in the blade-element theory; for example the blade-element theory assumes that the lift and drag coefficients determine the forces on the blade elements. Unlike the Prandtl example, there is no space-time separation of the components where the momentum theory applies and those where the blade-element theory applies. But things are not really so different: the two theories describe different objects in F × V × C-space.

Functional Modelling and Mathematical Models

683

van Kuik [2003] provides some evidence that the assumptions on the actuator disc in the momentum theory contain physical falsehoods. Applying physical arguments to singularities is delicate at the best of times, and van Kuik shows that in this case it leads to incompatible results, some of which are falsified by experiment. He concludes with an open question: ‘What is the origin of the inconsistency and what changes are required in the modelling of the momentum balance to remove it?’ Thanks to nonstandard analysis we know that the notion of an infinitesimal disc is not inconsistent in itself, at least if it is handled correctly. But the problem is really to find a physically well-motivated way of using the disc to reach results that agree with experiment under reasonable conditions. Most cases of ‘don’t care’ falsity are also cases where approximations are good enough. We adopt false but simplifying assumptions; the falsehood of the assumptions does affect the results but only to a degree that we can tolerate. The nearest I know to a pure example of ‘don’t care’ falsity is almost any compiler for the programming language Prolog. The requirements on a compiler are precise enough. But the designers of Prolog compilers often ignore one of these requirements, because to implement it precisely would slow down the operation of the compiled software to an unacceptable degree. The requirement is to check that certain variables don’t occur in certain terms—the ‘occur check’. It’s known that software with a defective occur check works correctly most of the time, and Prolog programmers develop a sense of what to avoid if they want to stay clear of trouble with it. (See [Apt and Pellegrini, 1994].) It’s impossible to list in advance the ways in which modellers might find justification for using models that aren’t accurate descriptions of their systems. In 1430 the architect’s model of the Duomo at Florence was destroyed for its ‘dishonesty’ (inhonestas)—what can they have meant? A kind of slippage that I would never have foreseen appears in a recent paper of Klaus Oberauer which compares four psychological theories of how people interpret conditional sentences. Oberauer translates the four theories ‘into formal models’ which contain some features that the authors of the theories would almost certainly disown, and he comments The assumption [that the processes run down the same path . . . ] is not demanded by the theories or by any data, but it is necessary to hold the number of free parameters in the models smaller than the number of data points to be fitted . . . [Oberauer, 2006]

(19)

So the models being compared are ones that demonstrably don’t fit the facts, and arguably don’t agree with the theories that are supposedly being compared. But the results of the paper are striking and suggest a methodology for objective comparisons of psychological models with different primitives.

684

Wilfrid Hodges

7

SEMANTIC MODELS

It’s time to return to the picture (1) and fill in the gap in the middle. I have to confess here that the details seem to me to have much more interest for a semanticist than they do for an engineer. But they are an essential preliminary for understanding some claims of Suppes, Suppe and others that are mentioned regularly in the literature. So we continue. The model on the left in (1) consists of a piece of text, for example our most recent version of Feynman’s example: x(t) = Ae−γt/2 cos(ωγ t) 2 where ωγ2 = ω02 − γ4 .

(for all t ∈ J)

(20)

As it stands, the text is neither true nor false, because the primitives are hanging in the air with no meanings attached to them. What kind of meanings would we need to attach in order to make (20) true or false? This is not just a question about the text itself; there could be some ingenious ways of interpreting (20) that make it true or false but have nothing to do with the intentions behind it. Engineers often laugh at mathematicians for interpreting equations in ways that have no possible physical reality, for example using pipes of negative diameter. But pity the mathematicians: they were given only the equations, not the engineering expertise that motivated the equations. In the last resort it will have to be the engineers’ job to eliminate the garbage. But at least we can pin down what kinds of interpretation would make logical sense. It’s normal practice to assume that each interpretation has a subject matter consisting of some basic objects; logicians call them individuals, and the set of individuals in an interpretation is called the domain. In (20) we perform arithmetical operations using the primitives, so the basic objects had better be numbers of some kind. Feynman himself sometimes takes them to be real numbers and sometimes complex numbers. No matter: either way there is a class of individuals. Inspecting (20), we can see that the primitives A, γ, ω0 and ωγ should be interpreted as standing for individuals. But x has to be read as a function taking individuals to individuals, and J has to be read as a set of individuals. ‘Individual’, ‘function from individuals to individuals’ and ‘set of individuals’ are three examples of types. Other examples of types are ‘ordered pair of individuals’, ‘function taking ordered pairs of individuals to individuals’, and so on. The possible types can be catalogued and then interpreted set-theoretically; see for example [Kamareddine et al., 2004] for a thorough treatment. Sometimes it makes sense to split the domain into different ‘sorts’. For example there might be one sort consisting of numbers (which can be multiplied but not concatenated) and one for strings of symbols (which can be concatenated but not multiplied). There is no need to do this with Feynman’s example, but it might be sensible with Hoare’s. A type structure for a piece of text is an assignment of a type to each primitive of the text, in a way that fits the syntax. (For example in (20) x must have the

Functional Modelling and Mathematical Models

685

type of a function because we apply it to t. Strictly we need to assign types to the variables too, and t will have the type of an individual, so that x has the type of a function on individuals.) A piece of text with a type structure attached is known as a theory. Suppose we have agreed a type structure for (20) so that it becomes a theory. Then we can interpret (20) as follows. First we supply a domain of individuals. For uniformity we can take ‘domain’ to be a primitive, so that supplying this domain is in fact interpreting a primitive. Then we supply, for each other primitive, an object of the right type, based on the given domain. Let’s go for the ‘real number’ reading of (20); then the domain is the set R of real numbers. To each of the primitives A, γ, ω0 and ωγ we assign a real number, to x we assign a function from real numbers to real numbers, and to J we assign a set of real numbers. One way to describe the assignment is to give a lookup table, for example PRIMITIVES domain A γ ω0 ωγ x(t) J

INTERPRETATIONS R 13 5 1.5 1.4 t2 the interval [0, 20] in R

(21)

On the left we list the primitives, and on the right we put their interpretations. Also there are some other symbols that we’ve ignored so far, namely the symbols for arithmetical operations such as addition and exponentiation. We must come back to this point at the end of this section. For the moment, regard these arithmetical primitives as having the usual arithmetical meanings. A lookup table as in (21) is known as a structure (because of its close resemblance to things that mathematicians call structures, for example vector spaces and graphs). Typically for a structure, the objects named on the righthand side of (21) are numbers, sets and functions. Since the structure (21) respects the types of the primitives in (20), we can meaningfully read the theory (20) as a statement about the structure (21). This statement will be either true or false; if it’s true we say that (21) is a semantic model of (20). This is the basic notion of model theory ([Hodges, 1997]), and it is the same notion of semantic model that we briefly introduced in Section 1 above. (For the record, (21) is pretty clearly not a semantic model of (20); the function assigned to the primitive x is quite wrong.) Model theorists tend to assume that their theories are written in artificial languages of logic. One reason for making this assumption is that these artificial languages have their type structures built into them from the start. By contrast if you take a piece of engineering text with any real complexity, it can be a tough job to work out a type structure for it, and you may need to rewrite parts of it to fit the types. For reasons of this kind, many logicians and philosophers of science used to argue that a theory should always be in a formalised logical language.

686

Wilfrid Hodges

But in modelling it’s common to use informal mathematical language to express theories. Patrick Suppes has for many years argued that the formal languages of logic (for example first-order logic) are inappropriate for scientific theories. Almost all systematic scientific theories of any interest or power assume a great deal of mathematics as part of their formal background. There is no simple or elegant way to include this mathematical background in a standard formalization that assumes only the apparatus of elementary logic. This single point has been responsible for the lack of contact between much of the discussion of the structure of scientific theories by philosophers of science and the standard scientific discussions of these theories. ([Suppes, 2002, p. 27])

(22)

Similar arguments have been used in the context of functional modelling too. Part of Suppes’ case is that the theories used in modelling very often have a number system built into them. For example we saw that (20) has some symbols that are meant to be read as standard arithmetical operations. If we regard these symbols as primitives, then we have to include number systems in our structures; this is counterintuitive—the system of real numbers is hardly part of a springweight system. The alternative, which we followed in (21), is to say that these symbols have fixed meanings in the language of the theory. But then we have left the territory of first-order logic, since the languages of first-order logic don’t have fixed symbols for arithmetical operations. Strictly we could use higher-type logics, and in software specifications this is quite common. But as Suppes implies, it’s hard to make these logics ‘simple or elegant’. Most hydrodynamical modellers certainly wouldn’t want to be bothered with such things. 8

ATTACHING THE THEORY TO THE SYSTEM

We still haven’t attached our model/theory (20) to a spring-weight system S. We failed to do it because our lookup table (21) never mentioned S. To make the connection we need a different lookup table that defines the primitives in terms of S. For example: Note at once that for this lookup table to make sense at all, we have to supply S with space and time coordinates. Also S must have other features that are not immediately observable, such as a fundamental frequency, a resistive coefficient and an energy. The table (23) defines a structure. Recall from Section 7 that in a structure the interpretations of the primitives are the objects named on the right, not the actual text on the right. So as in (21), γ is interpreted as a particular real number, though now we have to examine S to see what that number is. Likewise x is interpreted as a function from real numbers to real numbers, though the function is now defined in terms of the behaviour of S, not in purely mathematical terms.

Functional Modelling and Mathematical Models

PRIMITIVES domain A γ ω0 ωγ

x(t) J

687

INTERPRETATIONS R  2T /k where T is the energy of S at time t = 0 and k is the spring constant of S the resistive coefficient of S the fundamental frequency of S (23) the square root of (the square of the fundamental frequency of S minus a quarter of the square of the resistive coefficient of S) the upward vertical displacement of the weight of S at time t the interval [0, 20] in the time scale of S

Note that the structure defined by (23) could very well be exactly the same structure as one of the form (21). This would happen if the numbers, functions etc. named on the righthand side of (21) were exactly the same as those read off from S in (23). The mention of S in (23) affects how we find the numbers, but in principle the same numbers could be named directly. Feynman’s aim was to give a correct description of the behaviour of springweight systems. His theory (20) is a correct description of the behaviour of the particular system S if and only if the structure (23) is a semantic model of (20). We can describe this situation more formally as follows. Let M be a theory with a given set of primitives. The theory can be read as a definition of a class of structures, namely all those structures that make assignments of the correct types to all and only the primitives of M , and are semantic models of M . This class is called the model class of M , in symbols Mod(M ). The model class of any consistent theory is a huge object—a proper class, to use the set theorists’ terminology. If one wanted to use the model class to represent the theory, it would make sense to cut the class down to a manageable set of representative structures. By good fortune there are some kinds of theory which have a single ‘canonical’ semantic model that in some sense contains as much information as the whole model class. Two such cases are worth mentioning. The first case is where the theory simply states what certain numerical functions are, given certain parameters. Feynman’s example is a case in point. The theory (20) states the values of a function x, given certain parameters A, γ, ω0 and J. (The parameter J fixes the domain of the function. The parameter ωγ can be ignored since the theory merely says that it’s an abbreviation of an expression in terms of ω0 and γ.) The trick now is to convert the theory so that the parameters A, γ and ω0 become variables ranging over (say) the positive real numbers. This conversion turns (20) into the theory

688

Wilfrid Hodges

 x(t, A, γ, ω0 , J) = Ae−γt/2 cos( ω02 − γ 2 /4t) (for all t ∈ J; A, γ, ω0  γ/2 > 0; J an interval in R)

(24)

This theory has only one semantic model. The model corresponds to different spring-weight systems as we take different values for A, γ, ω0 and J. Note that the semantic model is almost the same thing as the function in (13); the main differences are the dependence on J (which we hadn’t yet discussed in Section 4), and the attachment of the function to the primitive x. This reduction to a single semantic model is available for many numerical models. In general it applies to the computer models used for modelling wind turbines, if we regard the computed outputs as descriptions of mathematical functions. Different runs of the program with different parameters correspond to different parameter values in the computed functions. A theory in first-order logic, if it has any semantic models with infinitely many individuals, has many such semantic models of different sizes. (This is a consequence of the Upward L¨ owenheim-Skolem Theorem [Hodges, 1997, p. 127].) So first-order theories are almost guaranteed to be unsuitable for this case. The theory (24) is not first-order, because it has expressions with fixed arithmetical meanings. The second case is where the theory is first-order, but of a particular form called universal Horn. A typical universal Horn sentence has the form ‘For all x1 , . . . , xn , if φ1 and . . . and φk then ψ’, where φ1 , . . . , φk , ψ are atomic formulas (for example equations). Universal Horn theories have ‘free’ semantic models whose elements satisfy the minimum possible number of atomic formulas; a free model is determined (up to isomorphism, but there is a canonical choice in the isomorphism class) by the size of its ‘basis’, i.e. its set of generators. If A is a free model of a universal Horn theory M and A has an infinite basis, then we can recover from A the set of all first-order sentences that follow from M . For further details (including the connection with ‘initial models’) see [Hodges, 1993]. At present the main practical applications of universal Horn theories are in computer science. In fact software specifications are often written entirely in universal Horn sentences. (See for example the ‘free specifications’ in CASL [Astesiano et al., 2002].) Frederick Suppe proposes we should apply the name ‘theory’ not to a theory in our sense, but to its model class (more precisely to its class of ‘theory-induced physical systems’). His reason is that

Functional Modelling and Mathematical Models

As actually employed by working scientists, theories admit of a number of alternative linguistic formulations—for example, classical particle mechanics sometimes is given a Lagrangian formulation and other times a Hamiltonian formulation—but it is the same theory regardless which formulation is employed. As such, scientific theories cannot be identified with their linguistic formulations; rather, they are extralinguistic entities which are referred to and described by their various linguistic formulations. ([Suppe, 1977, p. 82ff])

689

(25)

Suppe’s proposal is technically flawed; Lagrangian and Hamiltonian formulations use different primitives, and different primitives implies disjoint model classes. More important, one of the chief purposes of having theories is that we can write them down in journal articles or on computer slides. We can’t do that with model classes, or even with single semantic models when they have infinite domains. In order to think about a model class or a structure at all, we need to describe it in words or pictures, and then we are back with a linguistic theory.

9 ISOMORPHISM AND SIMILARITY Our diagram (1) now has two new kinds of entity in the middle: structure A

defines

interpretation model class

model (or theory) M

system S

(26)

The model M on the left was textual, so we equipped it with a type structure on its primitives and called it a theory. Then we introduced its model class (top left). To relate the theory to the system S on the right, we interpreted each primitive in terms of S. This interpretation gave us a structure A (top right). Now the following three statements all mean the same thing: A is in the model class of M . A is a (semantic) model of M . Read as a description of A, M is true.

690

Wilfrid Hodges

The modelling literature contains two other possible relationships between A and M that are alleged to be helpful. The first is that A is ‘isomorphic to a model of’ M , and the second is that A is ‘similar to a model of’ M . What if anything do these notions add to the picture we already have?

Isomorphism Suppose A and B are two structures with the same primitives and type structure. Suppose also that there is a one-to-one correspondence i between the domain of A and the domain of B. We say that i is an isomorphism from A to B if i makes B a perfect copy of A in the following sense: for each primitive p, if we take the interpretation of p in A and we apply i to it so as to change elements of the domain of A into elements of the domain of B, what we finish up with is exactly the interpretation of p in B. For example suppose the domains of both A and B are the set R of real numbers, and i is the correspondence that makes each real number x correspond to −x. Suppose also that one of the primitives of A is the symbol <, interpreted in A as the relation ‘less than’. Then if i is an isomorphism, the interpretation of < in B must be the relation ‘greater than’, since x is less than y if and only if −x is greater than −y. One of the basic theorems of model theory states that (under conditions that are likely to be satisfied by any example you can expect to meet) if A is isomorphic to B then A and B are models of exactly the same sentences. It follows that A is in the model class of M if and only if A is isomorphic to some structure in the model class of M . So the relation ‘A is isomorphic to a model of M ’ gives us nothing new. There are ways of using the notion of isomorphism to give something new and useful. For example Suppes [this volume] shows the usefulness of representation theorems, which state that in some class M of structures each structure is isomorphic to one in some smaller class B. But I am very sceptical of some other uses that I have seen suggested. For example I have seen it suggested that we should distinguish between ‘numerical’ structures like (21) and ‘real-world’ structures like (23), and say that the numerical structure is a model of the real-world one if they are isomorphic. This seems to be a plain misunderstanding. As we noted in the previous section, structures like (21) are no more or less numerical than the structure (23), and they could in fact be the same structure.

Similarity There is no standard definition of similarity between structures. But when structures contain numerical functions, there are some obvious ways of defining similarity relations. For example, suppose the function symbol f is a primitive and its type structure makes it stand for a function from R to R. Suppose A and B are

Functional Modelling and Mathematical Models

691

two structures that interpret f as the function F and the function G respectively. Then we can choose a small positive real number ε and describe A and B as being ‘similar’ if for all real x, the difference between F (x) and G(x) is less than ε. This device has uses, but some warnings are in order. First, the most obvious use of this device is to do the same work as the approximations that we discussed in Section 6. In that section we dealt with approximations by adjusting the statements made, not the structures they were made about. There are no doubt reasons for preferring one approach to the other, but they are likely to depend on the details of the case in hand. Second, the fact that it’s easy to put a bound ε on the difference between A and B doesn’t guarantee that the bound serves any useful purpose. For example, if functions F and G are similar in this sense, their Fourier transforms need not be, although the functions and their Fourier transforms may contain essentially the same information. Third, not all references to ‘similarity’ in the literature can be read in this way. For example [Giere, 1999, p. 92] says models need only be similar to particular real-world systems in specified respect and to limited degrees of accuracy. But he gives the Earth-Moon system as an example of a ‘real-world system’; since the Earth-Moon system is not a structure, the notion of similarity above doesn’t apply to it. I don’t feel competent to explain what kind of similarity relation Giere has in mind. There is more to be said about the relevance of model classes to modelling, though most discussions in the literature are about scientific rather than technological modelling. See for example [Balzer et al., 1987] (who represent the structuralist tradition), [Niiniluoto, 1999] (who discusses the role of structures in scientific explanation) and [Kuipers, 2001] (a very thorough treatment that takes on board both explanatory and functional modelling). BIBLIOGRAPHY [Apt and Pellegrini, 1994] K. R. Apt and A. Pellegrini. On the occur-check free Prolog programs. ACM Toplas, 16, 687–726, 1994. [Ashby, 2003] N. Ashby. Relativity in the global positioning system. Living Reviews in Relativity, 6, 2003. www.livingreviews.org/. [Astesiano et al., 2002] E. Astesiano, M. Bidoit, H. Kirchner, B. Krieg-Brueckner, P. D. Mosses, D. Sannella, and A. Tarlecki. CASL: the common algebraic specification language. Theoretical Computer Science, 286, 153–196, 2002. [Balzer et al., 1987] W. Balzer, C. U. Moulines, and J. D. Sneed. An Architectonic for Science. Reidel, Dordrecht, 1987. [Barwise and Etchemendy, 1994] J. Barwise and J. Etchemendy. Hyperproof. CSLI, Stanford CA, 1994. [Cartwright, 1999] N. Cartwright. The Dappled World: A Study of the Boundaries of Science. Cambridge University Press, Cambridge, 1999. [Dehornoy, 1995] P. Dehornoy. From large cardinals to braids via distributive algebra. Journal of Knot Theory and Ramifications, 4–1, 33–79, 1995.

692

Wilfrid Hodges

[de Roever and Engelhardt, 1998] W.-P. de Roever and K. Engelhardt. Data Refinement: Model-Oriented Proof Methods and their Comparison. Cambridge University Press, Cambridge, 1998. [Dowty et al., 1981] D. R. Dowty, R. E. Wall, and S. Peters. Introduction to Montague Semantics. Reidel, Dordrecht, 1981. [Feynman, 1990] R. P. Feynman. QED: The Strange Theory of Light and Matter. Penguin Books, London, 1990. [Feynman et al., 1963] R. P. Feynman, R. B. Leighton, and M. Sands. The Feynman Lectures on Physics, Volume I. Addison-Wesley, Reading Mass., 1963. [Giere, 1999] R. N. Giere. Science Without Laws. University of Chicago Press, Chicago, 1999. [Hoare, 1989] C. A. R. Hoare. Proof of a program: Find. In Essays in Computing Science, C.B. Jones ed., pp. 59–74. Prentice Hall, New York, 1989. [Hodges, 1993] W. Hodges. Logical features of Horn clauses. In Handbook of Logic in Artificial Intelligence and Logic Programming, Vol. I: Logical Foundations, D.M. Gabbay, C.J. Hogger and J.A. Robinson eds., pp. 449–503. Clarendon Press, Oxford, 1993. [Hodges, 1997] W. Hodges. A Shorter Model Theory. Cambridge University Press, Cambridge, 1997. [Hodges, 2001] W. Hodges. Articles ‘Model theory’, ‘First-order model theory’ and ‘Tarski’s truth definition’. Stanford Encyclopedia of Philosophy. plato.stanford.edu, 2001. [Kamareddine et al., 2004] F. Kamareddine, T. Laan, and R. Nederpelt. A Modern Perspective on Type Theory. Kluwer, Dordrecht, 2004. [Kuipers, 2001] T. A. F. Kuipers. Structures in Science: Heuristic Patterns Based on Cognitive Structures. Kluwer, Dordrecht, 2001. [Morrison, 1999] M. Morrison, M. Models as autonomous agents In Models as Mediators, M. Morrison and M.S. Morgan, eds., pp. 38–65. Cambridge University Press, Cambridge, 1999. [Niiniluoto, 1999] I. Niiniluoto. Critical Scientific Realism. Oxford University Press, Oxford, 1999. [Oberauer, 2006] K. Oberauer. Reasoning with conditionals: A test of formal models of four theories. Cognitive Psychology, 53, 238–283, 2006. [Pascal, 1963] B. Pascal. De l’esprit g´ eom´ etrique, pp. 348–355. Seuil, Paris, 1963. [Prudhomme et al., 2003] G. Prudhomme, G. P. Zwolinski, and D. Brissaud. Integrating into the design process the needs of those involved in the product life-cycle. Journal of Engineering Design, 14, 333–353, 2003. [Quarton, 1998] D. C. Quarton. The evolution of wind turbine design analysis—A twenty year progress review. Wind Energy, 1, 5–24, 1998. [Ramsey, 1978] F. Ramsey. Foundations: Essays in Philosophy, Logic, Mathematics and Economics, D.H. Mellor ed. Routledge and Kegan Paul, London, 1978. [Sørensen and Kock, 1995] J. N. Sørensen and C. W. Kock. A model for unsteady rotor aerodynamics. Journal of Wind Engineering and Industrial Aerodynamics, 58, 259–275, 1995. [Stenning, 2002] K. Stenning. Seeing Reason: Image and Language in Learning to Think. Oxford University Press, Oxford, 2002. [Suppe, 1977] F. Suppe. The Semantic Conception of Theories and Scientific Realism. University of Illinois Press, Illinois Urbana, 1977. [Suppes, 2002] P. Suppes. Representation and Invariance of Scientific Structures. CSLI, Stanford CA, 2002. [Tarski, 1954] A. Tarski. Contributions to the theory of models I. Indagationes Mathematicae, 16, 572–581, 1954. [Tarski, 1983] A. Tarski. The concept of truth in formalized languages. In Logic, Semantics, Metamathematics, J. Corcoran ed., pp. 152–278. Hackett, Indianapolis, Indiana, 1983. [Taylor, 1997] J. R. Taylor. An Introduction to Error Analysis: The Study of Uncertainties in Physical Measurements. University Science Books, Dulles Virginia, 1997. [van Fraassen, 1990] B. van Fraassen. The Scientific Image. Clarendon Press, Oxford, 1990. [van Kuik, 2003] G. A. M. van Kuik. An inconsistency in the actuator disc momentum theory. Wind Energy, 7, 9–19, 2003. [Vincent and Brown, 2005] T. L. Vincent and J. S. Brown. Evolutionary Game Theory, Natural Selection, and Darwinian Dynamics. Cambridge University Press, Cambridge, 2005. [Wizelius, 2007] T. Wizelius. Developing Wind Power Projects: Theory and Practice. Earthscan, London, 2007.

MODELS AS EPISTEMIC TOOLS IN ENGINEERING SCIENCES

Mieke Boon and Tarja Knuuttila

1

INTRODUCTION

When browsing through scientific journals in the field of engineering sciences, we soon learn that models play a central role in them. Through modelling, engineering sciences strive to understand, predict or optimize the behaviour of devices or the properties of diverse materials, whether actual or possible. The models developed in the engineering sciences should be distinguished from the models produced in engineering. Whereas the latter usually represent the design of a device or its mechanical workings, models in the engineering sciences aim for scientific understanding of the behaviour of different devices or the properties of diverse materials. For instance, chemical engineering is concerned with designing processes for converting materials or chemicals into other materials and chemicals that meet certain functions or purposes. For these processes it uses devices, such as chemical reactors and equipment for separation of substances such as crystallization, precipitation, absorption, filtration and distillation. Scientific research in the field of chemical engineering proposes models of the behaviour of chemical devices. It typically proceeds to study the behaviour of devices by interpreting them in terms of physical phenomena considered to be relevant to their proper or improper functioning, and then modelling these phenomena. Examples of such phenomena are desired and undesirable chemical reactions, the transport of liquids, gasses and solids within the device, the transport of chemical compounds by means of fluid flow or diffusion in the fluid, the transport of heat by convection or conduction, and other physical processes such as absorption, dissolution, ionization, precipitation, vaporization and crystallization. In the scientific literature, authors typically propose a certain type of design of the device — which consists of a configuration (e.g. a schema of its mechanical construction and dimensions) and its chemical and physical conditions — for meeting a certain function, for instance, for producing a compound at a high purity and with a minimum of waste production and energy use. Likewise, electrical engineering is concerned with designing devices — such that convert or transform electrical, electro-magnetic or mechanical input into electrical, electro-magnetic or mechanical output that meets certain functions. As in Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

694

Mieke Boon and Tarja Knuuttila

the case of chemical engineering, scientific research in the field of electrical engineering proposes models of the behaviour of electrical devices, which task differs from the design (e.g. of electrical circuits) of such devices. Also in this scientific field, scientific articles aim to contribute to optimizing the devices with regard to their functioning. Or, to take a third example, materials engineering is concerned with the application of materials with properties (e.g. chemical, electrical or mechanical properties) that meet certain functions. For instance, metals which are resistant to corrosion, ceramics that are superconductive at higher temperatures, and polymers of a particular strength. Materials science is concerned with scientific understanding of materials — either of materials that already exist or of materials that scientists aim to create artificially — which may then indicate ways in how to create or intervene with specific material properties. As the examples above show, the engineering sciences aim at both furthering the development of devices and materials meeting certain functions and optimizing them. Through modelling the engineering scientist seek to gain understanding of the behaviour and properties of various devices and materials. More often than not, this involves conceiving the functioning of the device, often in terms of particular physical phenomena that produce the proper or improper functioning of the device. However, in many cases, the desired properly functioning devices and materials do not exist. In these cases, the scientific models function as tools for producing such devices and materials. To understand engineering sciences and the way they use modelling to optimize and create devices and materials to meet specific functions, we need an account of how scientific models are produced and used in scientific practice. In particular this involves making sense of how models in engineering sciences acquire cognitive value with respect to their very orientation towards the artifactual, in other words, how models enable scientists to reason through constructing and using them. For this task a mere representational approach to models proves too limiting. In the philosophy of science it is generally accepted that scientific models represent some aspects or parts of the world or, more specifically, some real target systems. This idea of models as representations has been given different formulations ranging from semantic to pragmatist accounts of representation. According to the semantic conception of representation models are structures that represent the structural properties of real target systems as they feature in experimental and measurement reports by being either isomorphic or similar to them. From the pragmatist perspective this amounts to approaching research from the finished science point of view — yet it seems more apt to conceive especially engineering scientists as active interveners with the world. Instead of depicting an already existing world, the engineering sciences aim at theories and models that provide understanding of artificially created phenomena. This role of engineering sciences seems to us better accommodated by a pragmatic view on them. Indeed, as we will show below, the pragmatic approach to representation in fact points to a more versatile understanding of models than what a mere representational approach to them grants.

Models as Epistemic Tools in Engineering Sciences

695

In the following we will consider scientific models in engineering as epistemic tools for creating or optimizing concrete devices or materials. From this pragmatist and functional perspective, scientific models appear as things that are used by scientists to do some work, in other words, to fulfil some purposes. Consequently, we approach modelling as a specific scientific practice in which concrete entities, i.e. models, are constructed with the help of specific representational means and used in various ways, for example, for the purposes of scientific reasoning, theory construction and design of other artifacts and instruments. The key to the epistemic value of models does not lie in their being accurate representations of some real target systems but rather in their independent systemic construction that enables scientists to draw inferences and reason through constructing models and manipulating them. Although this way of looking at models makes sense especially in the context of engineering sciences because of their intervening and constructive character, we suggest that it could be applied to other sciences as well.1 In this sense engineering sciences might even serve to highlight some characteristics of scientific modelling and representation in general, especially if engineering sciences are firmly distinguished from engineering (see above). This chapter thus aims also to give an overview of the various accounts of scientific models and representation in the philosophy of science, and to show in which directions these approaches have recently been extended in order to capture the role of scientific models in scientific practices better. We will proceed as follows and start by presenting an overview of the present discussion of models and representation in the philosophy of science, and explicating how the conception of models as epistemic tools fits into this more overall picture (Section Two). In turn this general discussion on models provides us with a background for analysing the Carnot model of the heat-engine, which, as we will argue, can still serve as a paradigmatic case of modelling in engineering sciences (Section Three). The final section draws together the themes of this chapter and points out different topics that an extended understanding of models should take into account (Section Four). 2

SCIENTIFIC MODELS IN PHILOSOPHY OF SCIENCE: FROM REPRESENTATIONS TO EPISTEMIC TOOLS

Judging by their virtual absence from the general philosophical discussion on modelling, models in engineering sciences have not qualified as worthy objects of study. This may be due to the tendency of the philosophers to relegate the engineering sciences to the realm of application. However, there is at present an intense discussion going on in the philosophy of science concerning models and modelling which is largely due to their constantly rising importance in contemporary science. As a result, new accounts of models and their epistemic or cognitive value have been 1 Hacking [1983] argues that science in general should be seen as an intervening rather than representing enterprise. These kinds of claims have also been repeatedly presented in the field of Science and Technology Studies, although not in the form of a philosophical argument.

696

Mieke Boon and Tarja Knuuttila

presented that also seem to suit modelling in the engineering sciences better. In the following we will shortly review this discussion in an attempt to show that the emphasis on representation places excessive limitations on our view of the knowledge-bearing nature of models. As an alternative we suggest that models could be approached as epistemic tools.

2.1 Models as representations The discussion on models in the philosophy of science has heterogeneous beginnings. It testifies to both theoretical and formal, as well as practical aspirations, which can be seen to have different and even conflicting goals [Bailer-Jones, 1999, 32]. Thus in parallel with approaches which focus on the pragmatic and cognitive role of models in scientific enterprise, there have been attempts to establish, within a formal framework, what scientific models are. Of the formal approaches the semantic conception was the most widely held conception of models for several decades, since its emergence in the early sixties. Yet it can be claimed that the very philosophical discussion of models has been importantly motivated by practice-oriented considerations — even the proponents of the semantic theory understood themselves as providing a more realistic picture of theories (see [van Fraassen, 1980, 64]). Although there have thus been differing perspectives on models, philosophers of science have still generally agreed that models are representations and as such they give us knowledge because they represent their supposed external target objects more or less accurately, in relevant respects [Bailer-Jones, 2003; da Costa and French, 2000; French and Ladyman, 1999; Frigg, 2002; Morrison and Morgan, 1999; Su´ arez, 1999; Giere, 2004]. Yet due to their general approach to models, different philosophers have presented widely diverging accounts of representation. The fundamental dividing line goes between those accounts that take representation to be a two-place relation between two things, the model and its target system, and the pragmatist ones according to which also the representation-users and their purposes should be taken into account, thus arguing for at least threeplaced analysis of representation. The conviction that representation can be accounted for by reverting solely to the properties of the model and its target system is part and parcel of the semantic approach to scientific modelling. Recently, the semantic conception has been defended, for instance, by da Costa and French [2000], and French and Ladyman [1999]. According to the semantic conception, models specify structures that are posited as possible representations of either the observable phenomena or, even more ambitiously, the underlying structures of the real target systems. The representational relationship between models and their target systems is analysed in terms of isomorphism: a given structure represents its target system if both are structurally isomorphic to each other [van Fraassen, 1980, pp. 45, 64; Suppe, 1974, pp. 97, 92; French, 2003; French and Ladyman, 1999]. Isomorphism refers to a kind of mapping that can be established between the two that preserves the

Models as Epistemic Tools in Engineering Sciences

697

relations among elements. Consequently, the representational power of a structure derives from its being isomorphic with respect to some real system or a part of it. (Other candidates offered for the analysis of representation by the proponents of the semantic view are similarity [Giere, 1988] and homomorphism [Bartels, 2006].) The above-mentioned theoretical attractiveness of isomorphism vanishes once we realize that the parts of the real world we aim to represent are not “structures” in any obvious way, at least not in the sense required by the semantic account. It is possible to ascribe structures to some parts of the real world, but this involves that these parts of the empirical world are already modelled (or represented) somehow. This has, of course, been noticed by the proponents of the semantic theory. Patrick Suppes [1962] has, for instance, invoked “models of data” in order to account for the fact that isomorphism concerns the relation between structures, not the relation between raw data and theory. Thus the isomorphism required by the semantic account actually concerns the relationship between a theoretical model and an empirical model, the theoretical model being the model that satisfies the equations of the theory [Suppe, 1989, 103–106]. Even if we disregard the problem that the world does not present itself to us in ready-made structures, isomorphism does not give us a satisfactory account of representation, because it does not capture some common features of representation. Firstly, isomorphism has wrong formal properties. For instance, isomorphism denotes a symmetric relation whereas representation does not: we want a model to represent its target system but not vice versa. Secondly, and more fundamentally, isomorphism is a relationship between two structures, whereas scientific representation assumes a relationship between a structure and a real world target system. Structural isomorphism is not sufficient for representation since the same structure can be instantiated by different systems and thus isomorphisms. Isomorphism alone is thus not able to fix the extension of representation. On the other hand, a certain target system need not have an unique structure; depending on the perspective adopted, it can be sliced up differently (see [Frigg, 2006, 56–59]). From the scientific practice point of view, the idea that isomorphism establishes scientific representation seems inadequate, or at least unfruitful. The idea that representation is either an accurate depiction of its object or not a representation at all does not fit our actual representational practices. It is typical of scientific models that they are inaccurate in many ways. Indeed, the important role of idealizations, simplifications, approximations and tractability considerations in modelling seem difficult to account for from the semantic perspective; for further comment on these topics the readers might refer to Hodges’s contribution to this volume. Moreover, it seems unacceptable to consider the cases in which isomorphism between the theoretical structure and intended real target system fails as un-representational.2 The pragmatic approaches, in turn, make representation less a feature of the models and their target systems themselves than an accomplishment of the representation users [Su´ arez, 2004; Giere, 2004; Bailer-Jones, 2003; Frigg, 2006]. These 2 For other properties that we might expect an acceptable concept of representation to satisfy, see Su´ arez [2003] and Frigg [2002; 2006].

698

Mieke Boon and Tarja Knuuttila

studies criticize the assumption that representation could be regarded as a twoplace relationship of correspondence between the representative vehicle and its target. This way of conceiving representation attempts, as Su´ arez [2004] has aptly put it, “to reduce the essentially intentional judgments of representation-users to facts about the source and target objects or systems and their properties” (p. 768). In contrast, the pragmatic approaches point out that no thing is a representation of something else in and of itself; it has to be always used by the scientists to represent some other thing [Teller, 2001; Giere, 2004]. Consequently, what is common to pragmatic approaches is their focus on the intentional activity of scientists as representers and denial that the relationship of representation to what is represented can be based only on the respective properties of the representative vehicle and its target object. Pragmatic approaches to representation solve the problems of the semantic notion of representation mentioned above; the users’ intentions both create the directionality needed to establish a representative relationship and introduce indeterminateness into the representative relationships (since human beings as representers are fallible). But this comes at a price. When representation is grounded primarily on the specific goals and representing activity of humans as opposed to the properties of the representative vehicle and the target object, as a result nothing very substantive can be said about the relationship of representation in general. This has also been explicitly admitted by the proponents of the pragmatic approach (see [Giere, 2004; Su´ arez, 2004]), of whom Su´ arez has gone farthest in arguing for a minimalist account of representation which resists saying anything substantive about the supposed basis on which the representational power of representative vehicles rests, i.e. whether it rests, for instance, on isomorphism, similarity or denotation. According to Su´ arez, such accounts of representation err in trying to “seek for some deeper constituent relation between the source and the target”, which could then explain as a by-product why, firstly, the source is capable of leading a competent user to a consideration of a target, and secondly, why scienarez builds tific representation is able to sustain “surrogate reasoning”. Instead, Su´ his inferential account of representation directly on the very features of surrogate reasoning. The formulation Su´ arez [2004, p. 773] gives to the inferential conception of representation is the following: A represents B only if (i) the representational force of A points towards B, and (ii) A allows competent and informed agents to draw specific inferences regarding B. The “representational force”, according to Su´ arez, is “the capacity of the source to lead a competent and informed user to a consideration of the target”. Thus part (i) of the formulation postulates that the representational uses of the source are a result of intentional activity of competent and informed agents. Part (ii) of the formulation contributes to the objectivity that is required of scientific representation by assuming A to have the constitution that allows agents to correctly

Models as Epistemic Tools in Engineering Sciences

699

draw inferences from B. Yet Su´ arez resists saying anything about what kind of a relation there is supposed to be between the source and the target. Thus it is legitimate to conclude that for him models do not have any uniquely determinate relationship to the real world.3 The pragmatic minimalist approach to representation has rather radical consequences for how we conceive of models. If we accept the minimalist approach to representation, not much is established in claiming that models give us knowledge because they represent their target objects. It is nevertheless important to be clear on what is established by the pragmatist account. In fact, it just points to the impossibility of giving a general substantial analysis of representation that would explain how knowledge, or information, concerning real target systems could be retrieved from the model (cf. Hodges in this volume). The pragmatists do not deny that many scientific representations can be traced back to some target systems, or that they can depict them more or less accurately at least in some respects — the clearest cases of such models being scale models and maps. However, if we adopt a pragmatic approach to models, the focus on representation only starts to seem unnecessarily limiting. Apart from the general philosophical reasons mentioned above, there are also reasons stemming from the scientific practice that make us question the fruitfulness of representational paradigm as regards the epistemic value of models. Not the least of them is the fact that instead of functioning as straightforward representations of some “real” systems, models often depict some tentative mechanisms, processes or solutions that serve as a basis for various inferences, interventions and experimental set-ups. On many occasions, scientific models are used primarily as demonstrations, exemplifications, proofs of existence, etc. The philosophical and empirical points made above are bound to make one ask whether there is any other angle than representation alone from which to approach the knowledge-bearing properties of models. Interestingly, largely apart from the very interest on the topic of models and representation, a new discussion on models has emerged that loosens the epistemic value of models from representing definite target systems and considers them as independent objects. This gesture, we suggest, makes room for the various roles the very same models can play in scientific endeavour and prepares the way for conceiving models as epistemic tools (see also [Portides, 2005]).

2.2

Models as epistemic tools

The idea of models as independent objects or entities has been expressed by several recent authors in various ways. Morrison [1999] and Morrison and Morgan [1999] have considered models as autonomous agents which are through their con3 This is reflected in the way models can be extended from one context of use to a rather different context of use. However, the applicability of models in other contexts is often limited and must be “handled with care”. In our account of models as epistemic tools they incorporate knowledge about where and how they can be used in generating knowledge.

700

Mieke Boon and Tarja Knuuttila

struction partially independent from theory and data. This is because, besides being comprised of both theory and data, models typically also involve “additional ‘outside’ elements” [Morrison and Morgan, 1999, 11]. Boumans [1999] for his part disentangles models from the theory-data framework altogether. In his study on business-cycle he shows how many different “ingredients” the model can be constructed of, such as analogies, metaphors, theoretical notions, mathematical concepts, mathematical techniques, stylised facts, empirical data and finally relevant policy views. From another perspective, Weisberg [2007] and Godfrey-Smith [2006] have also come to the conclusion that models should be treated as independent entities. For them independence means independence from certain real target systems. Thus, instead of conceiving independence in terms of the relationship of models to the theory and world, or data, they release models from representing any definite real target system. According to Weisberg and Godfrey-Smith, modelling can be viewed as a specific theoretical practice of its own that can be characterized through the procedures of indirect representation and analysis that modellers use to study the real-world phenomena. With indirect representation they refer to the way modellers, instead of striving to represent some real target systems directly, rather construct simple, ideal model systems to which only a few properties are attributed. As Godfrey-Smith has aptly put it, modelling can be characterized by the “deliberate detour through merely hypothetical systems” it makes use of [2006, 734]. How, then, are models as independent objects able to give us knowledge? Whereas Godfrey-Smith evokes the “effortless informal facility” with which we can assess similarities between imagined and real-world systems, Weisberg refers to the notion of representation. But reverting to representation would take us back to the problems discussed above. In contrast, what we find the most important point in viewing models as independent things is that it enables us to appreciate their functional characteristics, that is, the different purposes for which they are used in scientific practice. This gives us, we suggest, a clue to how to appreciate the epistemic properties of models from another perspective than that provided by representation. Considering scientific models from the functional perspective requires one to address them as concrete objects that are constructed for certain epistemic purposes and whose cognitive value derives largely from our interaction with them [Knuuttila and Merz, forthcoming]. Consequently, scientific models can be considered as multifunctional epistemic tools [Knuuttila, 2005; Knuuttila and Voutilainen, 2003]. The importance of our interaction with models is recognized by Morrison and Morgan [1999], who stress how we learn from models by constructing and manipulating them. However, it seems to us that they leave this important idea somewhat half-way. Namely, if our aim is to understand how models enable us to learn from the processes of constructing and manipulating them, it is not sufficient that they are considered as autonomous; they also need to be concrete in the sense that they must have a tangible dimension that can be worked on. This concreteness is provided by the material embodiment of a model: the concrete

Models as Epistemic Tools in Engineering Sciences

701

representational means through which a model is achieved gives it the spatial and temporal cohesion that enables its manipulability. This also applies to so-called abstract models: when working with them we typically construct and manipulate external representational means such as diagrams or equations. Herein lies also the rationale for comparing models to experiments: in devising models we construct self-contained artificial systems through which we can make our theoretical conjectures conceivable and workable. This applies as well to mathematical models as to other kinds of models that are more readily seen as having a material dimension. Also the very variation of the different kinds of models used: scale models, pictures, diagrams, different symbolic formulas and mathematical formalisms, suggests that the material dimension of models and the diverse representational means they make use of are crucial for their epistemic functioning. The representational means used have different characteristic limitations and affordances; one can express different kinds of content with symbols than with pictures, for example. From this perspective the diverse external representational means provide external aids for our thinking, which also partly explains what is commonly ascribed as the heuristic value of modelling (see [Giere, 2002]). Cognitive scientists have approached this importance of external representational tools for our cognition through the notion of scaffolding. External representational scaffolding both narrow the space of information search by localizing the most important features of the object in a perceptually salient and manipulable form and enable further inferences by making the previously obscure or scattered information available in a systematic fashion (see e.g. [Larkin and Simon, 1989; Clark, 1997; Zhang, 1997]). Science provides the utmost human activity of creating and using representational tools for cognitive purposes. It is already a remarkable cognitive achievement to be able to express any mechanism, structure or phenomenon of interest in terms of some representational means, including assumptions concerning them that are often translated in a conventional mathematical form. Such articulation enables further theoretical findings as well as new experimental set-ups, but it also imposes its own limitations on what can be done with a certain model. Another aspect of scaffolding provided by models is related to the way they help us to conceive the objects of our interest clearly and to proceed in a more systematic manner. Models are typically constructed in such a way that they constrain the problem at hand — which happens typically by way of idealizations and abstractions — thereby rendering the situation more intelligible and workable. As the real world is just too complex to study as such, models simplify or modify the problems scientists deal with. Thus, modellers typically proceed by turning the constraints (e.g. the specific model assumptions) built into the model into affordances; one devises the model in such a way that one can gain understanding and draw inferences from using or “manipulating” it. Yet the seeming simplicity of models disguises the various elements they incorporate, such as familiar mathematical functions, already established theoretical entities, relevant scientific knowledge, certain generally accepted solution concepts, the intended uses of a

702

Mieke Boon and Tarja Knuuttila

model, the epistemological criteria that are supposed to apply to it and so forth. All these things that are built into a model give it also certain original built-in justification [Boumans, 1999]. These aspects of models also explain, on the one hand, how they allow certain kinds of solutions and inferences, and on the other hand, how they can also lead to unexpected findings, thereby breeding new concepts and problems and opening up novel areas of research (for concept formation in sciences, see [Nersessian, 2008]). We thus suggest that we gain knowledge through models typically by interacting with them, that is by building them, manipulating them, and trying out their different alternative uses — which in turn explains why they are regularly valued for their performance and their results or output. From the functional perspective, rather than trying to represent some selected aspects of a given target system, modellers often proceed in a roundabout way, seeking to build hypothetical model systems in the light of their anticipated results or of certain general features of phenomena they are supposed to bring about. If a model gives us certain expected results or replicates some features of the phenomenon, it provides an interesting starting point for further theoretical and experimental conjectures. This orientation towards the results brought about by models also accounts for why modellers frequently use the same cross-disciplinary computational templates, such as wellknown general equation types, statistical distributions and computational methods (for the notion of computational template, see [Humphreys, 2004]). The overall usability of computational templates is based on the one hand on their generality and the observed similarities between different phenomena and on the other hand on their tractability. The purposes the model is constructed for and the computability considerations often override in modelling the strive for correct representation. Consequently, the very peculiarity of scientific models lies in their being concrete entities that are aimed at accounting for certain phenomena through the detour of constructing artificial entities keeping simultaneously in mind their intended uses and other pragmatic questions such as their computability. This holistic nature of models in fact distinguishes them from more elementary scientific representations such as different visual displays, which often further fragment the object or specimen to reveal its details (see [Lynch, 1990]). Consequently, many scientific models should not be first and foremost considered as accurate representations of some target systems, but rather as epistemic tools. In an engineering context this amounts to finding out how to produce, control, and intervene — or to prevent some properties of materials or behaviour of processes and devices. Scientists in the engineering sciences build models for the purposes of imagining and reasoning about how to improve the performance of the devices, processes or materials of interest. These models involve imaginable properties and processes, and they incorporate measurable physical variables and parameters (e.g. in the case of chemical engineering chemical concentrations, flow rates, temperature, and properties of materials such as diffusion, viscosity, density). Often, these models also incorporate dimensions of typical configurations of certain devices. In the following section we will exemplify the functional ap-

Models as Epistemic Tools in Engineering Sciences

703

proach to models by studying the Carnot model of the heat-engine. We highlight the purpose of the model, the way the original problem motivating the model construction was translated into a phenomenon to be accounted for by way of different constraints and representational means. We also argue that the consequent thermodynamic theory became possible through the construction of this model, and not vice versa, which stresses the epistemic importance of building models and working with them. 3 DEVELOPMENT AND EPISTEMIC USE OF SCIENTIFIC MODELS: THE CASE OF THE CARNOT MODEL OF A HEAT-ENGINE The heat-engine is a classical example of a technological device that was subject to scientific modelling. We will analyse more closely how Carnot and his successors developed the Carnot model of the heat-engine. By this analysis we aim to illustrate that a pragmatic approach presents us with a more adequate picture of models and modelling than a paradigmatic representational view, in particular as regards how, in actual scientific practices, models are justified and why they give us knowledge. Carnot’s Reflexions on the Motive Power of Fire [1824/1986] is particularly interesting as a case of scientific modelling in the engineering sciences because Carnot’s treatise describes how step-by-step he develops a theoretical interpretation of a technological device. His writings expose the explorative reasoning process by which different aspects are built in the scientific model, which illustrates how it was constructed and justified. Scientific articles in our days often hide important parts of the reasoning process by which the model was developed. Nersessian and Patton (this Volume), for instance, meticulously describe many of the aspects scientists take into account in developing their models, many of which will not be part of scientific articles by these scientists. Besides the fact that Carnot exposes how he developed the model, the Carnot model of the heat-engine makes a good case because it is less complex than many of the appealing modern cases in the engineering sciences. Although historical, it still illustrates how engineering sciences approach technological problems. Moreover, it makes a better case than modern examples because many scholars in philosophy of technology are already familiar with the Carnot model (cf. [Kroes, 1995]). Last but not least, we take it that despite the enormous increase in scientific knowledge, mathematical and computational techniques, and scientific instruments, the way in which scientists develop scientific models of devices has not fundamentally changed.

3.1

The Carnot model of the heat-engine

According to the representational view, the Carnot model of the heat-engine is a scientific model that represents the real heat-engine. Our pragmatic view, on the contrary, targets modelling rather than just the entities called models, thus also making place for the role of the scientists and their epistemic purposes in accounting for the epistemic value of models. From this perspective Carnot’s model is a

704

Mieke Boon and Tarja Knuuttila

constructed entity, which gives a theoretical interpretation of the heat-engine in view of particular epistemic purposes. One such important purpose was identifying the theoretical limits of the performance (efficiency, in modern terms) of the heat-engine. The turn to modelling thus actually implies an extended notion of a model: models can be regarded as unfolding entities constructed by scientists with various representational means, in which the epistemic purposes and various other ingredients are incorporated. These aspects of a model will not reveal themselves to non-experts, yet without them the model can not be understood, let alone used. Consequently, a model should not be reduced either to the model description or to the imaginary entity that is set up by this description but rather involves both of them. What creates the two and mediates between them is the human activity of modelling. Such other aspects that are built-into models in the process of modelling are: (1) the idealizations, abstractions and simplifications that make the real target system intelligible and workable, (2) the (theoretical) phenomenon into which the original problem was translated, (3) the particular representational forms with the help of which the imaginary (or hypothetical) target system is represented, (4) the experiential and theoretical knowledge used in its development and justification, (5) the new concepts and principles that may emerge in its development, and (6) the relevant observable or measurable parameters of the real target system which link the scientific model to the real target system. With our analysis of the Carnot model we aim to show that it reduces to neither a diagram nor a theory or an imaginary entity, but consists of diverse aspects that scientists have built into it in the process of modelling. We claim that this intricate content of scientific models, which usually is fully understood only by the scientists working in the field in question, makes models function as epistemic tools. Furthermore, from the perspective on models sketched above, models can be approached as historically evolving entities: what the model consists of, how its content is represented, and how the model can be used in generating knowledge, also developed over the course of time. In a representational view, this change of content of the model is problematic because the Carnot model would not have a clear referent. In a pragmatic view, this change of content is unproblematic. The Carnot model “keeps this content together” and what remains stable in its development is (1) the theoretical interpretation of the heat-engine and (2) the epistemic purpose of finding the theoretical limits of the performance of heat-engines. From the philosophical point of view, a seeming problem of the pragmatic approach is in explaining how models give us knowledge if not by means of a predetermined representational relation with the real target system. An important aim of our analysis of the development of the Carnot model is thus to illustrate how conceiving of models as epistemic tools (as presented in Section 2.2) makes it intelligible how models are justified and how they give us knowledge. The key to this question lies in the activity of modelling. As models are purposefully designed things, they allow scientists to interact with them, which is afforded and limited by the representational means they make use of (which thus determine partly what can be done with the model and what not; examples of representational means

Models as Epistemic Tools in Engineering Sciences

705

are text, pictures, diagrams, graphs, tables, mathematical equations, computer simulations). What is more, we aim to show that scientific models are not only epistemic tools for the purposes of scientific reasoning, theory construction or design of other artifacts and instruments — models also function as epistemic tools of their own making. Scientists develop a model step-by-step, buiding in new aspects by which the content of the model becomes richer and more advanced. As an epistemic tool it ‘affords and limits’ also its own further development, which explains why part of the justification of a model is ‘built-in’: the development and justification of a model typically concur as already accepted pieces of knowledge and the conventional ways of representing them are incorporated into the model. It is important to keep in mind, however, that Carnot himself did not call his theoretical account of the heat-engine a model. The notion of a scientific model in its present sense was not in use those days (cf. [Bailer-Jones, 1999]). Thus it is only with the benefit of hindsight that the scientific community calls it a model of the heat-engine. A case can be made that the theoretical activity had taken the form of modelling towards the end of the twentieth century, modelling itself bearing “a distinctive historical signature” (as Peter Godfrey-Smith, [2006, 726]). Certainly, the theoretical strategy of the engineering sciences consists of modelling. Last but not least, since our aims are philosophical, our analysis is essentially reconstructive in aiming to highlight how the development and use of models can give us knowledge. Although we will present historical facts, we do not strive to present a historical account of how Carnot and his successors actually developed the Carnot model.

3.2

Epistemic purpose of the Carnot model

The French physicist and engineer Sadi Carnot, in his Reflexions on the Motive Power of Fire [1824/1986], gave the first successful theoretical account of heatengines, which we will refer to as ‘the Carnot model of the heat-engine’. Carnot opens his Reflexions with the statement: “It is generally known that heat can be the cause of motion and that it possesses great motive power. The steam engine in widespread use today are visible proof of this” (p. 61) He credits English engineers such as Savery, Newcomen, Smeathon and Watt for the discovery, development and improvement of the heat-engine (p. 63). Figure 1 presents a picture of the mechanical principles of one of the earliest steam engines, the Newcomen steam engine, invented in 1712 by Thomas Newcomen. According to Carnot: “The study of these engines is of the utmost interest [because] their importance is immense, and their use is increasing daily” (ibid. p. 61) He then states the problem and why a theory of its operation is needed: In spite of the many advances that have been made with the heatengine, and the satisfactory state in which it exists today, the theory of its operation is rudimentary, and attempts to improve its performance are still made in an almost haphazard way.

706

Mieke Boon and Tarja Knuuttila

Figure 1. Schematic Newcomen steam engine that presents its mechanical principles. Steam is light-grey and water is dark-grey. Valves between boiler and condenser (the cylinder) move from open to closed. This schema presents three different moments of a cycle. (a) Valve between boiler and cylinder is open. Steam from boiler enters the cylinder and pushes piston upwards. (b) Piston has arrived at its highest position. Valve between cylinder and boiler is closed. Valve between cold sink and cylinder opens and water from a reservoir of cold water sprays in the cylinder causing condensation of steam in the cylinder. (c) Both valves are closed. Piston moves down, which is due to its own weight and the reduced pressure of steam in the cylinder. At the lowest position the valve between boiler and cylinder opens. Water runs back to boiler. Cycle repeats. Figure taken from http://en/wikipedia.org/wiki/Newcomen steam engine. The question whether the motive power of heat [i.e. the useful effect that an engine is capable of producing] is limited or whether it is boundless has been frequently discussed. Can we set a limit to the improvement of the heat-engine, a limit which, by the very nature of the things, cannot in any way be surpassed? Or conversely, is it possible for the process of improvement to go on indefinitely? For a long time there have also been attempts to discover whether there might be working substances preferable to steam for the development of the motive power of fire; and that is a question still debated today. Might air, for example, have great advantages in this respect? In the following pages, we propose to examine these questions carefully. (ibid. p. 63)

Models as Epistemic Tools in Engineering Sciences

707

This introduction by Carnot to his theoretical work illustrates that, unlike many of the ‘basic’ sciences, the engineering sciences usually start from questions related to practical problems and applications, for instance, the problem of how the functioning of a device can be improved. One of the technological problems of steam engines in the early nineteenth century was how to improve their performance, which meant how to reduce the quantity of coal needed for producing an amount of motive power. This so-called duty (now called ‘efficiency’) of steam engines was expressed as the pounds of water that were pumped to one foot height per bushel of coal. Engineers wanted to know whether the performance of steam engines could be improved by the use of steam at a higher pressure and/or by replacing steam by other vapours or gases. Carnot translated the practical problem of how to improve the performance of these engines to a theoretical question about the limits to the performance of the heat-engine determined ‘by the very nature of the things’. Generally, the first step in developing a scientific model of a device such as the heat-engine, involves conceiving of its functioning in terms of particular physical phenomena that produce its proper or improper functioning. Carnot assumed that “In order to grasp in a completely general way the principle governing the production of motion by heat, it is necessary to consider the problem independently of any mechanism or any particular working substance” (ibid. p. 64). Hence, Carnot conceived of the functioning of the heat-engine, not primarily in terms of its mechanical working such as represented in Figure 1, but as a device that produces motion by heat. The phenomenon of interest produced by the heat-engine, according to Carnot, is “the production of motion by heat”. This conception of the phenomenon of interest is already part of the development of the scientific model because this phenomenon is not simply observed but discerned or conceptualized by scientists. As a consequence, the description of the phenomenon cannot be easily understood as a representation that stands in a correspondence or similarity relation with the real target system (e.g. the steam engine). Rather, it presents a particular way of ‘seeing’ or ‘imagining’ the real device. In Carnot’s writings, many other examples can be found of conceptions of phenomena that he discerns and that function as epistemic tools to the development of the scientific model rather than being claims about phenomena that exist or could be observed somehow in the real heat-engine. Examples are: the phenomenon that ‘a difference in the temperature of two bodies A and B brings about motive power’; and the phenomenon that ‘a transfer of caloric from A to B brings about motive power’. A scientist who postulates a phenomenon is not obliged to believe that it exists as a real (ontological) occurrence that could be observed had we better instruments. Instead, a scientist must have reasons to believe that the model can be used as an epistemic tool in reasoning about the real target system, in particular with regard to the epistemic purpose of the model. In brief, producing the preliminary Carnot model requires conceiving of the real heat-engine in view of the epistemic purpose of the model. Imagining it that way

708

Mieke Boon and Tarja Knuuttila

involves relevant empirical and theoretical knowledge that allow making abstractions and conceptualizing particular features of the real target system in view of the epistemic purpose. This first modelling step results in a preliminary Carnot model that consists of the conception of the real heat-engine as an abstract device that produces motion by heat. Clearly, the model is not yet satisfactory since it does not sufficiently explain the theoretical limits of the performance of this device. Nevertheless, the preliminary model of the real heat-engine must be such that it allows (i.e. affords and limits) further development of the Carnot model.

3.3 Modelling a hypothetical device The development of the Carnot model of the heat-engine proceeds by fleshing out the abstract device that produces motion by heat in terms of a preliminary model. In Reflexions, Carnot pictures a hypothetical device that produces the phenomenon of interest, i.e. “motion by heat”. The hypothetical device consists of a cylindrical vessel closed with a movable piston that encloses a constant amount of gas; this gas can be either thermally isolated, or contacted with a body at a constant high temperature that acts as a heat source, or with a body at a constant low temperature that acts as heat sink. This device produces motion by heat because it goes through a specific cycle by which the piston moves up and down. Carnot describes the working of this hypothetical device as follows, making use of the diagram in Figure 2: Let us picture an elastic fluid, air for example, enclosed in the cylindrical vessel abcd in Figure 3 [see Figure 2 below]. In this figure, cd is a movable diaphragm or piston fitted inside the cylinder, and the two bodies A and B are each maintained at a constant temperature, that of A being higher that that of B. Let us now imagine the following sequence of operations: (1) The body A is placed in contact with the air enclosed in the volume abcd, ... As a result of this contact, the air assumes the temperature of the body A. At this point, cd marks the actual position of the piston. (2) The piston rises gradually to the position ef. Contact between the body A and the air is maintained throughout, so that the temperature of the air remains unchanged during the expansion. The body A provides the caloric that is needed in order to keep the temperature constant. (3) A is removed, so that the air is no longer in contact with any body that can act as a source of caloric. But the piston continues to move, rising from the position ef to gh The air expands without absorbing caloric, and its temperature falls. Let us suppose that the temperature continues to fall until it is equal to that of B, whereupon the piston stops at the position gh.

Models as Epistemic Tools in Engineering Sciences

709

Figure 2. Axial cross section of the hypothetical device, which is part of the Carnot model of the heat-engine. In this diagram, abcd is a cylindrical vessel, cd is a movable piston, and A and B are constant–temperature bodies. The vessel may be placed in contact with either body or removed from both (as it is here). Figure taken from [Carnot, 1824, p. 17].

(4) The air is placed in contact with the body B. It is then compressed by returning the piston from its position gh to cd. During this process, the air maintains a constant temperature, since it remains in contact with B and gives up caloric to it. (5) The body B is taken away, and the compression of the air is continued. Since the air is now isolated, its temperature rises. Compression continues until the temperature of the air reaches that of the body A, by which time the piston has moved from the position of cd to ik. (6) The air is placed once again in contact with the body A, and the piston returns from ik to ef ; the temperature remains constant. (7) The third of the stages just described is repeated, followed by stages 4, 5, 6, 3, 4, 5, 6, 3, 4, 5, and so on.” (ibid. p. 74-75, our italics)

710

Mieke Boon and Tarja Knuuttila

Hence, Carnot developed the preliminary model by building in this more extended conception of a device that that produces motion by heat. In modern language, the successive stages in operating the device are called a thermodynamic or Carnot cycle. Once the cycle is ‘running’, operation (6) replaces (1) and (2), therefore, the cycle actually consists of four operations: 3, 4, 5, 6. Accordingly, the hypothetical device produces a thermodynamic phenomenon (as we now call it), which is “the production of motion by heat”.

3.4

Representational versus pragmatic views on models.

Carnot’s conception of (the operation of) a hypothetical device that produces motion by heat, also entails several other aspects. Firstly, it explains how operations with this hypothetical device (such as placing the air in contact with body A or B, or compressing the air in the cylinder) produce observable and measurable phenomena (such as changes in temperature, pressure and volume of the air in the cylinder). This is how the Carnot model gives knowledge of observable and measurable parameters. As a consequence, the Carnot model is connected with the real world target system by means of observable and measurable parameters entailed by the model, whereas a representational view of models would attribute this connection to a representational relation between the model and the real heat-engine as it is in itself — which, with the model in its given state, is hard to imagine. Secondly, the model entails imaginary phenomena that could not possibly be observed or measured (such as transfer of caloric). Such descriptions of imaginary phenomena in the model could not be justified as part of the model if their justification depended on a representational relation with observable or measurable occurrences. From the pragmatic perspective, positing imaginary phenomena is justified if it enables further reasoning, as long as it does not generate contradictions. Indeed, the conception of ‘transfer of caloric’ was later rejected, but not because it was somehow discovered that caloric did not exist, but because reasoning upon it led to contradictions. Also, it is obvious that the Carnot model — which, next to the hypothetical device (operations 3, 4, 5, 6) includes a diagram of this device (Figure 2) — does not represent the mechanical working of the real heat-engine as described and pictured in Figure 1. Moreover, Carnot’s conception intentionally neglects all possible losses of energy in a real heat-engine due to the mechanical working thereof, such as loss by friction of the moving piston, loss of steam past the piston, and loss of heat by conduction between parts of the engine at different temperatures. Carnot assumed that these losses should be neglected in order to arrive at a model that explains a limit to the performance of heat-engines which, by the very nature of the things, cannot be surpassed. Hence, the epistemic purpose of the Carnot model justified the neglect of the mechanical workings and related shortcomings of the real heat-engine. Finally, in modern accounts of the Carnot model, the hypothetical device that produces motion by heat is often called the ideal heat-engine. These accounts

Models as Epistemic Tools in Engineering Sciences

711

usually embrace the idea that the Carnot model is the ideal heat-engine, which leads us to consider models as abstract entities (for models as abstract entities, see [Giere, 1999]). However, this seems to aggravate the philosophical problem of representation: How are we supposed to relate an imaginary entity to a real target system? Especially as the representational view remains silent about the actual means of representation. Conceiving models as epistemic tools pays explicit attention to the use of external representational means, attributing to this dimension of modelling part of its epistemic value. When it is realized that models are used for making inferences and reasoning, the urgency of grounding the epistemic value of models to a supposed representational relation between the model and some external target system (or its rendering in terms of a data model) vanishes. Instead, the results of a model and its behaviour are related to measurements, experimental results and other existing theoretical knowledge in a subtle process of triangulation. From the pragmatic perspective modelling proceeds certainly by representing, i.e. using representational means for conveying and creating meaning, yet this need not establish any determinable, representational relationship between some real system (or its rendering) and the hypothetical system thus introduced. Thus, when modelling involves the construction of an imaginary object, one does not have to assume that for it to afford us knowledge it would need to replicate accurately some aspects of some real target systems. The epistemic value of modelling is accounted for by referring to the tool-like characteristics of models rather than referring to any supposed representational relationships. In sum, even though models are constructed by making use of representational means, they need not be conceived of as representations of any definite real target systems. However, as argued above, the pragmatist analyses of representation show that invoking representation does not per se establish much as regards the epistemic (or cognitive) value of models. In the preceding sections we took the Carnot model of the heat-engine as our practical example whereby we aimed to argue and illustrate that the pragmatic approach leads to a more intelligible account of modelling in the engineering sciences than the representational paradigm. The remainder of Section 3 aims at reconstructing in more detail how the Carnot model of the heat engine was developed, and, in particular, how Carnot and his successors built, step-by-step, various aspects into their model such as experiential and theoretical knowledge, theoretical principles and concepts, and using new representational means, by which process the model was also partly justified. Furthermore, we aim to answer why and how modelling in this case, which apparently proceeds via detour of a hypothetical device far removed from actual heat engines, nevertheless affords us knowledge about them. In Sections 3.5 and 3.6, we will first illustrate that the development of the model of the heat-engine by Carnot and his successors also proceeded in tandem with the development of the representational means and theoretical concepts used.

712

Mieke Boon and Tarja Knuuttila

3.5 Representational means for developing the Carnot model An important aspect of developing scientific models is the representational means scientists have at their disposal. Representational means provide the material embodiment of the model that provides its spatial and temporal cohesion that scientists can work on. From Carnot’s Reflexions, it becomes obvious that Carnot’s representational means were limited. The use of diagrammatic and mathematical representations as we know them today would have made his laborious reasoning much easier for him, and more accessible to the reader. Carnot only used text, a few equations and calculations, and some tables with experimental data and calculations from formula. The only type of diagram he presented is Figure 2. Only Carnot’s successors developed several representational means that allowed re-formulation and further development of the Carnot model into the form as we know it today. Figure 3, for instance, is a modern block-diagram of the phenomenon (i.e. the production of motion by heat). This modern diagram represents the heat-engine as a device that converts heat to mechanical work, where heat, Q, flows from a furnace (e.g. a boiler) at high temperature TH through the fluid of the “working body” (e.g. steam or air) and into the cold sink (e.g. a condenser) at TC , thus forcing the working substance to do mechanical work, W , on the surroundings, via cycles of compressions and expansions of the fluid. It is striking that Carnot neither used symbols (H and Q) in the representation of the phenomenon nor arrows for representing the directions of work and heat between bodies.

TH

QH

QC

TC

W

Figure 3. A modern diagram of the Carnot engine, which presents the production of work by heat. In a modern conception, Carnot’s ideal heat-engine produces work, W , by heat, Q, by means of a thermodynamic cycle of a gas contacted with a hot reservoir at temperature TH , and a cold reservoir at temperature TC . This cycle is now called the Carnot cycle (see also Figure 4). Figure is an adaptation of http://en.wikipedia.org/wiki/Carnot heat engine. Similar to Figure 3, the top diagram in Figure 4 shows a modern representational means for representing the ‘operation’ of the ideal heat-engine, which expands on Carnot’s diagram (Figure 2). The pictures (1), (2), (3), (4) in this block-diagram represents the working of the ideal heat-engine, i.e. the four ‘operations’ 6, 3, 4, 5, respectively, described by Carnot (i.e. the four stages of a — thermodynamic

Models as Epistemic Tools in Engineering Sciences

713

— cycle of the gas in a cylinder under a freely moving piston). Here also Carnot’s presentation is enriched with the use of arrows (while it could have been enriched with the use of symbols Q and W as well): upward and downward arrows inside the cylinder depict heat that flows in and out of the cylinder, respectively, while upward and downward arrows outside the cylinder depict work exerted by, and on the piston, respectively.

!

 

!

!

!





 

 

  









 

 



        



         

     



Figure 4. A modern diagram of the hypothetical device that produces motion from heat. The top diagram presents the ‘operation’ of the device, which is a cylinder filled with a constant amount of gas (dark-grey) and closed with a movable piston. The pictures in this diagram, numbered (1), (2), (3), (4) represent respectively the four ‘operations’ 6, 3, 4 and 5 described by Carnot [1824/1986, 74-75]. The lower diagram presents the P-V diagram of the Carnot cycle. Another important type of diagram that became part of the Carnot model only ´ after Carnot had published his Reflexions, was invented by Benoˆıt Paul Emile Clapeyron, likewise a French engineer and physicist, who, ten years later, pre-

714

Mieke Boon and Tarja Knuuttila

sented the cycle as a closed curve in a graph of the pressure P of the gas in the cylinder against its volume V. At present, this graph, of which a version is presented in the lower diagram of Figure 4, is called a P-V diagram of the Carnot cycle. Additionally, modern conceptions of the ideal heat-engine expand on Carnot’s description of the cycle, i.e. the ‘operations’ described in 6,3,4,5 by introducing new thermodynamic concepts, such as ‘reversible isothermal expansion’, that were only developed by Carnot’s successors. These new concepts allowed for a more efficient and precise description of the Carnot cycle.4 Finally, new mathematical approaches became part of the Carnot model through the work of Rudolf Julius Emanuel Clausius, a German physicist and mathematician, who in 1865 published The Mechanical Theory of Heat — with its Applications to the Steam Engine and to Physical Properties of Bodies. Besides other things, he developed a model of the conversion of heat, Q, to work, W, by using differential calculus as a representational means for representing the Carnot cycle. This new representational means allowed, for instance, to mathematically describe reversible processes. In the first chapter (Mathematical Introduction), Clausius explains the mathematical apparatus. In subsequent chapters he develops the ‘mechanical theory of heat’ by using this apparatus for constructing mathematical equations that represent, for instance, the Carnot cycle. Hence, the cycle of the gas in the heat-engine proposed by Carnot was represented in a completely new way, and the resulting mathematical model affords and confines particular new ways of reasoning and manipulating with the model. Such inventions of representational means did not only improve the understanding but also played an indispensable role in the further development of the Carnot model to ever clearer and richer models of heat-engines by his successors.

4 Today, the Carnot-cycle is for instance described as follows (the numbers 1, 2, 3, 4 in this description refer to the numbers in the two diagrams in Figure 4): 1. (= Carnot’s operation 6) Reversible isothermal expansion of the gas at the “hot” temperature, TH (i.e. isothermal heat addition). The gas expansion is propelled by absorption of heat Q1 from the high temperature reservoir. During this step the expanding gas causes the piston to do work W1 on the surroundings. 2. (= Carnot’s operation 3) Reversible adiabatic (i.e. isentropic) expansion of the gas (i.e. no heat is transferred to or from the gas in the cylinder: Q2 =0). In this step the piston and cylinder are thermally insulated, so that no heat is gained or lost. The gas continues to expand, doing work W2 on the surroundings. The gas expansion causes it to cool to the “cold” temperature, TC . 3. (= Carnot’s operation 4) Reversible isothermal compression of the gas at the “cold” temperature, TC (i.e. isothermal heat rejection). During this step the surroundings do work W3 on the gas, causing heat Q3 to flow out of the gas to the low temperature reservoir. 4. (= Carnot’s operation 5) Reversible adiabatic (i.e. isentropic) compression of the gas (i.e. no heat is transferred to or from the gas in the cylinder: Q4 =0). The piston and cylinder are thermally insulated. During this step the surroundings do work W4 on the gas, compressing it and causing the temperature to rise to TH . At this point the gas is in the same state as at the start of this cycle.

Models as Epistemic Tools in Engineering Sciences

3.6

715

Theoretical knowledge and concepts for developing the Carnot model

In order to understand Carnot’s way of reasoning in developing the model, we also must master some of the theoretical and experimental knowledge he was familiar with, as well as concepts that were unknown to him while familiar to us. An outline is presented only in as far as it helps to illustrate how knowledge at the time (and the lack of it) is part of how the model is developed; by no means does it aim to present a complete outline. Important is the conception of heat. In Carnot’s time, the prevalent theory of heat was the caloric theory which supposed that heat was a sort of weightless, invisible fluid that flowed from hotter to colder bodies. It was also assumed that caloric is a substance, which, like matter, is indestructible. Only by the mid-19th century was the caloric theory replaced by a theory of heat (using the notion ‘quantity of heat’, referred to as Q) by the work of scientists such as Clausius, James Joule, William Thomson (Lord Kelvin), and James Clerk Maxwell. Clausius and Thomson rejected the idea that heat is a substance (i.e. caloric), because it led to contradictions in the Carnot model. Clausius explains: “Heat is not invariable in quantity; but . . . when mechanical work is produced by heat, heat must be consumed, and that, on the contrary by the expenditure of work a corresponding quantity of heat can be produced.” Therefore, in the new mechanical theory of heat, the nature of heat is “not a substance but a motion”. Clausius argues that “According to this theory, the causal relation involved in the process of the production of work by heat is quite different from that which Carnot assumed. Mechanical work ensues from the conversion of existing heat into work, just in the same manners as, by the ordinary laws of mechanics, force is overcome, and work thereby produced” [Clausius, 1865, p. 268]. Hence, whereas Carnot believed that work is produced by the fall of a quantity of heat from a higher to a lower temperature, the mechanical theory of heat argues that work is produced by motion. A brief intermezzo about this change of the theory of heat can illustrate somewhat further why it is important to understand scientific models in terms of how scientists interpret and structure what they observe or experience (cf. Boon, forthcoming). Carnot and his predecessors interpreted heat as a substance. Adopting the idea that heat is a substance means that one ‘imagines’ heat as an indestructible thing. Subsequently, they used this conception of heat in their further reasoning and modelling. Carnot thus imagined heat as a fluid that can be carried from one body to another (where it is carried by other fluids such as steam), without being consumed or produced. Replacing this conception of heat by the idea that heat is motion that acts as a force is a tremendous intellectual achievement. Obviously, Clausius and other successors of Carnot did not observe that heat actually appears not to be a substance but a motion; instead, they found that this conception of heat leads to contradictions, which forced them to find a new conception.

716

Mieke Boon and Tarja Knuuttila

In developing the model also experiential and theoretical knowledge was built in. Carnot was familiar with the gas-laws of Boyle-Marriote (which states that at constant temperature, the absolute pressure and the volume of gas are inversely proportional), the Charles law (which states that at constant pressure, the volume of a given amount of gas and the temperature in Kelvin are proportional), GayLussac’s law (which states that the pressure of a fixed amount of gas at a fixed volume is proportional to its temperature in Kelvin), and Dalton’s law (which states that the total pressure exerted by a gaseous mixture is equal to the sum of the partial pressures of each individual component in a gas mixture). See [Carnot, 1824/1986, p. 78]. The ideal gas law as we know it today, and which includes Avogadro’s principle (which asserts that equal volumes of ideal or perfect gases, at the same temperature and pressure, contain the same number of particles, or molecules), was only stated by Clapeyron in 1834 (i.e. after the publication of Carnot’s work).5

3.7 Why and how the Carnot model of heat-engines yields knowledge In exploring why and how the Carnot model gives knowledge, we will focus on the question how Carnot arrived at the description of the ideal heat-engine (presented in Section 3.3), which he pictures as a fixed amount of air in a cylinder closed with a piston that performs “a sequence of four operations” 3, 4, 5, 6. In our reconstruction, we will ignore many of Carnot’s refined and intelligent arguments;6 we will also ignore arguments that are grounded on his conception of heat as an indestructible substance (i.e. his use of ‘caloric’). Accordingly, our reconstruction of Carnot’s modelling returns to the preliminary Carnot model of the heat-engine, which entails the abstract device that produces 5 At that time (around 1824), the basic laws of thermodynamics had not been formulated either. Nevertheless, Carnot is often called the father of thermodynamics. Around 1850, Clausius and Thomson formulated the first and the second law of thermodynamics (abandoning the caloric theory), which state (1) the conservation of energy, and (2) that heat cannot of itself pass from a colder to a warmer body (formulated by Clausius [1854, p. 116; 1865, p. 270]); the modern version of the second law reads as follows: the entropy of an isolated system which is not in equilibrium will tend to increase over time, approaching a maximum value at equilibrium. This notion of entropy was not known to Carnot but developed and named by Clausius, who, besides other things, wanted to understand in a fundamental way why “heat cannot of itself pass from a colder to a warmer body”. He introduced this concept in order to account for the heat-loss when “heat of one temperature is transformed in the heat of another temperature” (ibid p.217, and p. 357). The entropy, S, of a body is the ratio between heat, Q, and temperature, T, while the change of entropy is the dissipative energy use, or irreversible heat loss, during a change of state: « „ 1 1 − . ΔS = Q T2 T1 6 For instance, arguments for the theorem that the magnitude of the work produced is “independent of the nature of the substances through which the production of work and the transfer of heat are effected.”Carnot’s proof of the necessity of such a relation is based on the axiom that it is impossible to create a moving force out of nothing, or in other words, that perpetual motion is impossible. (cf. [Carnot, 1824/1986, pp. 69-70; Clausius, 1865, p 268]).

Models as Epistemic Tools in Engineering Sciences

717

motion by heat and the epistemic purpose of the modelling (i.e. identifying the theoretical limits of the performance of the heat-engine). Carnot carried on with the modelling by using this abstract device and theoretical knowledge of heat (caloric) for interpreting the working of a steam engine: So what exactly happens in a steam engine of the kind now in use? Caloric produced in the furnace by combustion passes through the walls of the boiler and creates steam, becoming in a sense part of it. The steam bears the caloric along with it, transporting it first into the cylinder, where it fulfils a certain function, and then into the condenser. There, the steam is liquefied by contact with the cold water it encounters. In this way, at the end of the whole process, the cold water in the condenser absorbs the caloric produced by the initial combustion: it is heated by the steam just as if it had been in direct contact with the furnace. The steam serves simply as a means of transporting the caloric, ... we are considering the movement of the steam is put to use. (ibid. p. 64) This interpretation explains how in a steam engine heat (caloric) is transported. In this way, the abstract device that produces motion from heat has become more substantial. Carnot concludes that the steam simply serves as a means of transporting the caloric (heat), and that “the production of motive power in a steam engine is due not to an actual consumption of caloric but to its passage from a hot body to a cold one” (ibid. 65). Clearly, the modelling just described did not primarily aim at a faithful representation of the mechanical working of real heat-engine. Moreover, not much can be deduced from the Carnot model at this point. This is one of the reasons for regarding scientific models as ‘epistemic tools’ rather than representations. Tools afford but also confine what can be done with them without deductively determining the result since this result also depends on aspects built in by the cognitive agent (see above). How this is done, in turn, depends on epistemic purposes and specific background knowledge of cognitive agents. Accordingly, the modelling aims at producing an epistemic tool that affords reasoning about the production of motion by heat in an ideal heat-engine. Carnot proceeded in his modelling endeavour by introducing propositions and principles that relate the transport of heat (caloric) and the production of motive power to other relevant parameters such as temperature, volume, and compression or expansion of the gas in the modelled steam engine. His development of propositions and principles is reconstructed and summarized in the list below (ibid. pp. 64-67, selecting, paraphrasing and numbering of principles by the authors): a) An experiential principle is that equilibrium restores wherever a difference in temperature exists, which means that b) heat (caloric) will always flow from a hot body to a cold body until the two bodies have the same temperature, by which equilibrium is restored.

718

Mieke Boon and Tarja Knuuttila

Therefore, c) in steam engines motive power is produced by the re-establishment of the equilibrium of caloric, not by consumption of caloric, and d) whenever there is a difference in temperature, motive power can be produced. while the converse is also true, that is, e) wherever there is power which can be expended, it is possible to bring about a difference in temperature and to disturb the equilibrium of caloric. f) The heat-engine is any engine that is driven by caloric. g) It is an experimental fact that the temperature of gaseous substances rises when they are compressed, and falls when they are expanded. h) An obvious principle is that heat can only be a source of motion in so far as it causes substances to undergo changes in volume or shape. Next, these principles guide Carnot in abstracting from features of the real steam engine that in his view are not essential to a theoretical understanding of how a steam engine produces motive power by heat (caloric). Accordingly, he abstracts from concrete components such as the furnace and the condenser by asking the reader to “imagine” two bodies, A and B (the temperature of A is higher than B), to which heat (caloric) can be added or from which it can be taken away without effecting any change in their temperature, and which will act as two infinite reservoirs of caloric. Subsequently, he reinterprets his conception of the working of the steam engine represented in the model in terms of relevant parameters (e.g. temperature, pressure, volume, caloric, expansion and compression of the steam) and in terms of three distinct operations. This results in the following reinterpretation of his former description of the steam engine (ibid. pp. 67-68): If we wish to produce motive power by conveying a certain amount of heat from the body A to the body B, we may do this in the following way: (i)

Take some caloric from the body A and use it to form steam. In other words, use the body as if it were the furnace. It is assumed that the steam is produced at precisely the temperature of the body A.

(ii) Pass the steam into a vessel of variable volume, such as a cylinder fitted with a piston, and then increase the volume. When the steam is expanded this way, its temperature will inevitably fall. Suppose that expansion is continued to the point where the temperature becomes exactly that of body B.

Models as Epistemic Tools in Engineering Sciences

719

(iii) Condense the steam by bringing it into contact with B, and, at the same time, subjecting it to a constant pressure, until it is totally liquefied. In this way, B fulfils the role of the injection water in a normal engine. In turn, this developed conception of the steam engine allows further modelling. Here, Carnot makes use of a (preliminary, non-mathematical) notion of ‘reversible’ processes by which he assume that operations (such as i, ii, iii) can also be carried out in the opposite direction. Based on this notion of reversibility, he states that there is no reason “why we should not form steam with caloric from the body B and at the temperature of B, compress it so as to bring it to the temperature of A, continuing the process of compression until complete liquefaction takes place” (ibid. p. 68). Carnot thus conceives of how the steam in the cylinder can brought back to its initial state in order to achieve a closed cycle.7 The introduction of new principles is another aspect of how scientific reasoning by means of models yields knowledge. In Carnot’s modelling this worked as follows. In his description of operations i, ii, and iii Carnot develops a picture that is close to experience (since he used experiential knowledge of how a steam engine works). Subsequently, by introducing the principle that “there is no reason why a process could not be reversed”, he connects knowledge from experience with a completely new principle. This approach results in a description of processes that may not yet be part of one’s experiences; nevertheless, one may believe that they could be brought about by a device. This way of scientific reasoning yields knowledge, not because this process was somehow observed and the description thus represents something external to us, nor because it was deduced from the model or from accepted theories, but because Carnot was able to relate the model at that stage with his conception of reversible processes. At this point, Carnot has developed a model of the heat-engine that goes through a cycle (i, ii, iii, and reverse), but he still needs to find out how this cycle will produce the maximum amount of motive power. With regard to how heat produces motive power, Carnot introduced principle h. By introducing some additional propositions and principles, he explains losses (and avoidance of losses) in the production of motive power by heat. Carnot’s development of these propositions and principles is reconstructed and summarized in the list below, which proceeds from principle h in the former list (ibid. pp. 66-73, selecting, paraphrasing and numbering of principles by the authors). i) Since any process in which the equilibrium of caloric is restored can be made to yield motive power, a process in which the equilibrium is restored without producing power must be regarded as representing a real loss. From reflecting on this latter point, Carnot concludes: 7 Interestingly, by introducing this notion of reversible processes, Carnot also introduces the working of a heat pump (a refrigerator), which is a device that transfers heat from a cooler system to a warmer one by compressing the gas (by exerting an external force).

720

Mieke Boon and Tarja Knuuttila

j) Any change in temperature that is not due to a change in the volume of a body is necessarily one in which the equilibrium of caloric is restored profitlessly. Hence: k) The necessary condition for the achievement of maximum effect is that the bodies used to produce motive power should undergo no change in temperature that is not due to a change in volume. However, l) when a gaseous fluid is rapidly compressed, its temperature rises; and when, on the other hand, it is rapidly expanded, there is a fall in temperature. According to Carnot, some of these principles are ‘obvious’ (e.g. h), whereas others are derived by logical reasoning about the theory of heat (e.g. j). Principle k presents a necessary condition for producing the maximum amount of motive force. At this point, the Carnot model consists of the description of how the steam engine works in terms of a cycle (i, ii, iii and reverse), the epistemic purpose of how this cycle will produce the maximum amount of motive power, and theoretical and empirical principles such as a − l. Again, further modelling proceeds from this model, that is, the model functions again as an epistemic tool in its own further development. From principles h−l, Carnot infers that the cycle avoids any “change in temperature that is not due to a change in volume”. By using principles h − l, he signifies where the problem of achieving the maximum effect lies: If a gas is rapidly compressed, its temperature rises (as stated in l). If we wish to bring this gas back to its original temperature without subjecting it to any further changes in volume, we must withdraw some caloric from it (ibid. p. 74). Hence, the problem is that the gas is brought back to its original temperature while keeping it at constant volume, which, according to principle j, means that “caloric is restored profitlessly”. The model in its current state guides Carnot in constructing an operation that overcomes this problem. Accordingly, he argues that it would be equally possible to withdraw the same caloric during the process of compression in such a way that the temperature of the gas would remain constant. The rise of temperature that would be due to rapid compression is thus avoided. By this solution, Carnot has constructed ‘operation’ (4) of the cycle 3, 4, 5, 6 (described in Section 3.3): 4. The air at TB is placed in contact with the body B; it is then compressed while withdrawing caloric, attaining a decrease in V while T remains constant. It should be noted that this operation could not possibly be derived from mere experience with real steam engines. Likewise, if the gas is rapidly expanded (by which, according to principle l, the temperature would fall), the fall of its temperature can be prevented if we supply to it an appropriate quantity of caloric. Carnot has thus constructed ‘operation’ (6) of the cycle 3, 4, 5, 6:

Models as Epistemic Tools in Engineering Sciences

721

6. The air at TA is placed in contact with the body A; next, the gas expands while supplying caloric, attaining an increase in the volume while the temperature remains constant. From what is given in the model at this point, i.e. by making use of the cycle (i, ii, iii and reverse) and the given principles, ‘operation’ 3 and 5 can be constructed as well: 3. Body A is removed. The gas expands while it is no longer in contact with any body that can act as a source of caloric. Hence, there is a simultaneous increase in the volume of the gas and fall of temperature until it is equal to that of body B. While ‘operation’ (5) is the reverse process: 5. Body B is removed. The gas is compressed, while it is no longer in contact with any body that can withdraw caloric. There is a simultaneous decrease in the volume of the gas and increase in temperature until it is equal to that of body A. At this point, the Carnot model consists of the description of cycle 3, 4, 5, 6 and the theoretical and empirical knowledge represented in a − l. This model meets the epistemic purpose of telling how this cycle will produce the maximum amount of motive power. The model can be used as an epistemic tool in producing knowledge about the behaviour of the real target system (the real steam-engine) because in modelling the hypothetical device (e.g. gas enclosed in a cylinder with a movable piston, and body A and B that represented furnace and cooler) was related to the description of the real steam engine (e.g. boiler, condenser, cooling water and furnace) and to observable occurrences and measurable quantities (such as changes of volume and temperature of a fluid). The model is not a representation of real heat engines. Instead, the knowledge about the real device obtained from this model is confined to its epistemic purpose and thus allows for inferring from it some suggestions as regards the construction of real steam engines. Carnot, for instance, suggested that principles j and k “must constantly be borne in mind in the construction of steam engines. If the principle cannot be strictly observed, any departure from it must be reduced to a minimum.” (ibid. p. 70). Additionally, the Carnot model at this point was used by his successors. They used this model as an epistemic tool in its further development. As was already mentioned, they have built in new representational means such as the differential calculus by which a more refined understanding of the ideal heat-engine was developed, as well as a mathematical description that afforded further development of the Carnot model (and which allowed making calculations such as the maximum theoretical efficiency). Carnot’s successors also used his model in the development of thermodynamics (see footnote 9).

722

Mieke Boon and Tarja Knuuttila

3.8 Model construction and model-based reasoning We have illustrated how Carnot developed his model of the ideal heat-engine, and how at different stages the model guided its own further development. At these stages, the model at that point in the process of modelling was used as an epistemic tool for taking the next modelling step. Strikingly, one can discern a succession of different ways in which the model enabled its own making, thereby also guiding the incorporation of new aspects which will be summarized below. The modelling starts from specifying an intended epistemic function that the model of a device or material has to fulfil, such as finding the theoretical maximum efficiency of heat-engines. Second, in the light of this epistemic function, a phenomenon is identified and conceptualized in terms of which the proper functioning of the device or material can be understood (e.g. the phenomenon of producing work by heat in heat-engines). Third, an idealized device or idealized material that produces the phenomenon of interest is conceptualized. This conception abstracts from several features of the real device or the real material (e.g. from the mechanical working of the real heat-engine) in view of the epistemic purpose. Fourth, the functioning of the idealized device or material is conceptualized in terms of ‘operations’ or physical processes, implying that knowledge of the relevant ‘operations’, physical processes, phenomena or properties is built into the model (e.g. knowledge of physical processes relevant for describing heat flows and the exertion of work that results from exposing the idealized device to certain external conditions was incorporated). Fifth, principles (e.g. how the maximum effect is achieved) and theoretical knowledge (e.g. the gas laws of Boyle, Mariotte and Gay-Lussac) about these ‘operations’ and physical processes in terms of physical variables relevant to the device or material (e.g. P, V, T, and specific heat) are incorporated in the model. Also, experiential principles (e.g. tendency to equilibrium of temperature) were incorporated in the model. Consequently, looking at the Carnot model from the perspective of its construction makes it plausible that the very activity of developing the model (i.e. modelling) guided Carnot in finding the thermodynamic cycle that produces the theoretical maximum efficiency of an ideal heat-engine. We suggest that these different ways of incorporating successive aspects into the model are not particular for the Carnot case, but instead present a more general account of modelling processes in the engineering sciences. We do not claim, though, that we have covered all relevant aspects, nor that all aspects mentioned can be found in every model. Our analysis of this case is at odds with those widely held views on models (stemming from the syntactic view of theories), which assume that models are derived from theories or general principles. Indeed, in modern textbooks, the Carnot engine is usually presented as if it were somehow derived from thermodynamic theory. However, historically it was the Carnot model of the heat-engine that contributed to the theory of thermodynamics and only in retrospect could it be viewed as satisfying the axioms of thermodynamics (cf. [Erlichson, 1999]). With regard to the semantic view of models, we argue that it neglects some crucial questions concerning modelling, particularly how scientists

Models as Epistemic Tools in Engineering Sciences

723

arrive at model systems (such as the ideal heat-engine). In that view, the system model is taken for granted, while we have illustrated that much theoretical work needs to be done in order to arrive at a model system (cf. Hodges, this Volume). Summing up, the Carnot model of a heat-engine is like many other scientific models in that it depicts an ideal entity that can be interpreted in terms of a phenomenon (that of producing motion by heat), which makes the model prone to further scientific examination and explication. Instead of describing a real object (the real heat-engine) the Carnot model actually presents an ideal object, similar to the model of an ideal pendulum, the Lotka-Volterra predator-prey model, or economic models of ideal economies. Also, the process of developing the model (i.e. modelling) has lead to conceptual novelties: the notion of “the efficiency of turning heat into work” did not exist before Carnot’s theoretical model, nor did the idea according to whichany change in temperature that is not due to a change in the volume of a body is necessarily one in which the equilibrium of caloric is restored profitlessly. The model of the ideal heat-engine incorporated various kinds of experiential and theoretical knowledge (e.g. the gas laws) and as it afforded thinking about the behaviour of heat-engines in a novel way it also led for its part to the consequent development of the thermodynamic theory. 4

TOWARDS AN EXPANDED NOTION OF MODELS

We have argued against the generally accepted idea among philosophers that models can be regarded as representations (variously defined) of some real target systems. As an alternative, we have proposed a pragmatic account of models as epistemic tools. We are of course not the first to argue against the representational view of models. Yet, even in those accounts, the notion of representation tends to re-enter the scene when it comes to answering the question why models can be used to knowledge about real target systems. Thus, for instance, after having argued for the importance of building and manipulating models, Morrison and Morgan [1999] claim that we can learn from models because they represent their target systems. We have argued, instead, that conceiving of models as standing in a direct representational relation with some real target systems does not shed light on their epistemic functioning. Philosophically, our proposed notion of models as epistemic tools focuses on the cognitive value of modelling and its different roles in scientific enterprise highlighting the importance of different representational means for model-based reasoning. From the practice point of view, one of the problems of the representational approach is, rather paradoxically, that by concentrating on the relation between the model and its real target system, it abstracts from the actual representational means with which scientists go about building their models. The representational account of models also leads to problems concerning the ontology of models, which has recently attracted quite a lot of interest in the philosophy of science. The question has been whether scientific models should be conceived in terms of the model descriptions (i.e. pictures, diagrams, or mathe-

724

Mieke Boon and Tarja Knuuttila

matical equations) or whether they are abstract or imaginary entities. From the representational perspective it seems crucial to single out the very entity that is supposed to correspond to the real-world systems. Yet both the alternatives proposed, i.e. models as model descriptions and models as abstract or imaginary entities, lead to trouble. Models cannot be identified with model descriptions because mere descriptions in and of themselves signify nothing. On the other hand, it seems difficult to explain how an abstract or imaginary entity, quite apart from its description, succeeds in enabling any reasoning. As opposed to this perspective, our approach to models as epistemic tools invokes the activity of modelling implying thus an extended notion of models as unfolding entities, which are constructed with concrete representational means conveying a hypothetical content. From this perspective, a model reduces neither to an abstract entity nor to the representational means with which it is constructed. The diverse aspects making an irreducible part of modelling include, for instance, (1) the epistemic purpose(s) the model has to fulfill, (2) the phenomenon that determines the function of the device or material of interest, (3) the abstractions and idealizations needed to construct the hypothetical objects, (4) the different types of the representational means used, such as diagrams, pictures or symbols, (5) physical, theoretical, and experiential knowledge or principles that are built into the model, and (6) relevant physical variables and parameters that are either known, measured, or determined otherwise, and which relate the model with what is observable or measurable by means of instruments. It seems to us that without taking these aspects of modelling into account it would be incomprehensible how scientists were able to develop models and reason with the help of them. Often these aspects go without any explicit notice in scientific practice but this does not license philosophers to neglect them in their accounts of models. Last but not least, approaching models as epistemic tools leads us to consider the various epistemic uses of models, such as scientific reasoning, prediction, theory construction, concept formation and design of other artefacts, instruments, or experiments. There is no reason to expect that they draw in the same direction and thus scientists use often different and conflicting models even when considering the same phenomenon, depending on the task at hand. However, finding out more about how the diverse tasks of models perhaps reinforce or alternatively contradict one another seems an interesting direction for further research. indexdesign

ACKNOWLEDGEMENTS This research is supported by a grant from the Netherlands Organisation for Scientific Research (NWO) and the Academy of Finland. Part of this resarch was carried out within the ZIF Bielefeld research project 2007 “Science in the Context of Application”, University of Bielefeld, ZIF, Zentrum f¨ ur Interdisziplinare Forschung, Bielefeld (Germany). We would like to thank Anthonie Meijers, Sjoerd Zwart and the unknown referees for their constructive comments and questions.

Models as Epistemic Tools in Engineering Sciences

725

BIBLIOGRAPHY [Bailer-Jones, 1999] D. M. Bailer-Jones. Tracing the Development of Models in the Philosophy of Science. In Lorenzo Magnani, Nancy J. Nersessian and Paul Thagard (eds.), Model-Based Reasoning in Scientific Discovery. New York: Kluwer, 23-40, 1999. [Bailer-Jones, 2003] D. M. Bailer-Jones. When Scientific Models Represent, International Studies in the Philosophy of Science 17: 59-74, 2003. [Bartels, 2006] A. Bartels. Defending the Structural Concept of Representation, Theoria 55, 7-19, 2006. [Boon, 2006] M. Boon. How Science is applied in Technology, International Studies in the Philosophy of Science 20(1): 27-47, 2006. [Boon, forthcoming] M. Boon. Understanding according to the Engineering Sciences: Interpretative Structures. In Henk De Regt et al. (eds.), Scientific Understanding: Philosophical Perspectives, University of Pittsburgh Press, forthcoming. [Boumans, 1999] M. Boumans. Built-In Justification. In Mary S. Morgan and Margaret Morrison (eds.), Models as Mediators. Perspectives on Natural and Social Science. Cambridge: Cambridge University Press, 66-96, 1999. [Carnot, 1986] S. Carnot. Reflexions on the Motive Power of Fire. Translated and edited by Robert Fox, New York, Manchester University Press, 1986 [1824]. [Clausius, 1865] R. Clausius. The Mechanical Theory of Heat – with its Applications to the Steam Engine and to Physical Properties of Bodies. London: John van Voorst, 1 Paternoster Row, 1865. [da Costa and French, 1990] N. C. A. da Costa and S. French. The Model–Theoretic Approach in the Philosophy of Science, Philosophy of Science 57: 248–265, 1990. [Erlichson, 1999] H. Erlichson. Sadi Carnot, ‘Founder of the Second Law of Thermodynamics’, European Journal of Physics 20: 183-192, 1999. [Clark, 1997] A. Clark. Being There: Putting the Brain, Body, and World Together Again. Cambridge, MA: MIT Press, 1997. [French, 2003] S. French. A Model-Theoretic Account of Representation (Or, I Don’t Know Much About Art. . . but I Know It Involves Isomorphism). Philosophy of Science, 70, 14721483, 2003. [French and Ladyman, 1999] S. French and J. Ladyman. Reinflating the Semantic Approach, International Studies in the Philosophy of Science 13,2: 103-121, 1999. [Frigg, 2002] R. Frigg. Models and Representation: Why Structures Are Not Enough. Measurement in Physics and Economics Discussion Paper Series. London: London School of Economics, 2002. [Frigg, 2006] R. Frigg. Scientific Representation and the Semantic View of Theories, Theoria 55, 49-65, 2006. [Giere, 1988] R. N. Giere. Explaining Science: A Cognitive Approach. Chicago and London: The University of Chicago Press, 1988. [Giere, 1999] R. N. Giere. Science without Laws. Chicago and London: The University of Chicago Press, 1999. [Giere, 2002] R. N. Giere. Scientific Cognition as Distributed Cognition. In Peter Carruthers, S. Stich and M. Siegal (eds.), Cognitive Bases of Science. Cambridge: Cambridge University Press, 2002. [Giere, 2004] R. N. Giere. How Models Are Used to Represent Reality, Philosophy of Science (Symposia) 71: 742-752, 2004. [Godfrey-Smith, 2006] P. Godfrey-Smith. The Strategy of Model-Based Science, Biology and Philosophy 21: 725-740. [Hacking, 1983] I. Hacking. Representing and Intervening. Cambridge: Cambridge University Press, 1983. [Humphreys, 2004] P. Humphreys. Extending Ourselves. Computational Science, Empiricism and Scientific Method, Oxford: Oxford University Press, 2004. [Knuuttila, 2005] T. Knuuttila. Models, Representation, and Mediation, Philosophy of Science 72: 1260-1271, 2005.

726

Mieke Boon and Tarja Knuuttila

[Knuuttila and Merz, forthcoming] T. Knuuttila and M. Merz. An Objectual Approach to Scientific Understanding: The Case of Models. In H. de Regt, S. Leonelli and K. Eigner, eds., Scientific Understanding: Philosophical Perspectives. Pittsburgh: University of Pittsburgh Press, forthcoming. [Knuuttila and Voutilainen, 2003] T. Knuuttila and A. Voutilainen. A Parser as an Epistemic Artefact: A Material View on Models, Philosophy of Science 70: 1484–1495, 2003. [Kroes, 1995] P. Kroes. Technology as Science-Based Heuristics. In J.C. Pitt (ed.): New Directions in the Philosophy of Technology.”. Kluwer Academic Publishers: 17-39, 1995. [Larkin and Simon, 1987] J. Larkin and H. A. Simon. Why a Diagram is (Sometimes) Worth 10,000 Words, Cognitive Science, 11:65-99, 1987. [Lynch, 1990] M. Lynch. The Externalized Retina: Selection and Mathematization in the Visual Documentation of Objects in Life Sciences. In M. Lynch and S. Woolgar (eds.), Representation in Scientific Practice. Cambridge, MA: MIT Press, 153-186, 1990. [Morrison, 1999] M. Morrison. Models as Autonomous Agents, in Mary S. Morgan and Margaret Morrison (eds.), Models as Mediators. Perspectives on Natural and Social Science. Cambridge: Cambridge University Press, 38-65, 1999. [Morrison and Morgan, 1999] M. Morrison and M. S. Morgan. Models as Mediating Instruments. In Mary S. Morgan and Margaret Morrison (eds.), Models as Mediators. Perspectives on Natural and Social Science. Cambridge: Cambridge University Press, 10-37, 1999. [Nersessian, 2008] N. J. Nersessian. Creating Scientific Concepts, Cambridge, MA: MIT Press, 2008. [Portides, 2005] D. P. Portides. Scientific Models and the Semantic View of Scientific Theories, Philosophy of Science 72: 1287-1298, 2005. [Su´ arez, 1999] M. Su´ arez. Theories, Models, and Representations. In L. Magnani, N. J. Nersessian and P. Thagard (eds.), Model-Based Reasoning in Scientific Discovery. New York: Kluwer, 75-83, 1999. [Su´ arez, 2003] M. Su´ arez. Scientific Representation: Against Similarity and Isomorphism, International Studies in the Philosophy of Science 17: 225-244, 2003. [Su´ arez, 2004] M. Su´ arez. An Inferential Conception of Scientific Representation, Philosophy of Science (Symposia) 71: 767-779, 2004. [Suppe, 1974] F. Suppe. The Structure of Scientific Theories. Urbana, University of Illinois Press, 1974. [Suppe, 1989] F. Suppe. The Semantic Conception of Theories and Scientific Realism. Urbana and Chicago, University of Illinois Press, 1989. [Teller, 2001] P. Teller. Twilight of the Perfect Model Model, Erkenntnis 55: 393-415, 2001. [van Fraassen, 1980] B. van Fraassen. The Scientific Image. Oxford: Oxford University Press, 1980. [Weisberg, 2007] M. Weisberg. Who is a Modeler, British Journal for the Philosophy of Science 58: 207-233, 2007. [Zhang, 1997] J. Zhang. The Nature of External Representations in Problem Solving, Cognitive Science 21(2): 179-217, 1997.

MODEL-BASED REASONING IN INTERDISCIPLINARY ENGINEERING Nancy J. Nersessian and Christopher Patton

1

INTRODUCTION

Research in biomedical engineering often confronts the problem that it is both impractical and unethical to carry out experiments directly on animals or human subjects. In our studies of two pioneering biomedical engineering research laboratories we have found a common investigative practice in this interdisciplinary field is to design, build, and redesign in vitro systems, which parallel selected features of in vivo systems. The researchers refer to their in vitro models as “devices.” When biological and engineering components are brought together in an investigation, researchers refer to this as a “model-system.” As one respondent stated: “when everything comes together I would call it a ‘model-system’ [. . . ]I think you would be very safe to use that [notion] as the integrated nature, the biological aspect coming together with an engineering aspect, so it’s a multifaceted modeling system I think that’s very good terminology to describe that.” Another researcher aptly referred to the processes of constructing and manipulating these model-systems as “putting a thought into the bench top and seeing whether it works or not.” The “bench top” refers not to the flat table surface but comprises all the locales where experimentation takes place. These instantiated “thoughts” (mental models) are physical models (devices) that represent what researchers deem to be salient properties and behaviors of biological systems. They are structural, behavioral, or functional analogs of in vivo phenomena of interest. The devices are also systems themselves, with engineering constraints that often impose simplifications and idealizations unrelated to the biological systems they model. In the following analysis we will examine some of these multifaceted systems in the problem-solving practices of the laboratories, especially as they figure in experimental situations. In each case we will examine how manipulating devices and model-systems enables a form of inference — “model-based reasoning” — different from logical inference through manipulating propositional representations. Our analysis derives from a five-year investigation of the research practices of two laboratories; one conducts tissue engineering, the other, neural engineering. These are hybrid engineering and science environments. The hybrid nature of these laboratories is reflected in the bio-engineered model-systems developed by the laboratories and in the characteristics of the researcher-students who are part Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

728

Nancy J. Nersessian and Christopher Patton

of a program aimed explicitly at producing interdisciplinary, integrative thinkers in bio-engineering. The laboratories and learning settings are designed to move beyond the traditional model of collaboration among engineers, biologists, and medical doctors to a new kind of integrative biomedical engineering that will shorten the span between laboratory research and medical application. This is the goal; in reality the research is still not advanced sufficiently for the desired medical applications, and integrative biology and engineering dominates practice. “Lab A” seeks to design off-the-shelf vascular tissue replacements for the human cardio-vascular system. Some intermediate problems that drive the research are: producing “constructs” (models composed of living tissue that mimic properties of natural vessels); examining and enhancing their mechanical properties; and creating endothelial cell sources through mechanical manipulation of stem cells. Although the ultimate objective of the laboratory is to make an artifact, researchers create both new knowledge (e.g., of properties of cells under various conditions) and new know-how (e.g., of tissue engineering techniques) as part of the problem situation. “Lab D” seeks to understand the ways neurons learn in the brain to create aids for neurological disabilities or, more grandly, “to make humans smarter ” (Lab D Director). Its intermediate investigations center on finding evidence of plasticity in a “dish” of multi-electrode neuron arrays, and producing controlled “muscle” activity in robots or in simulated agents (animats), which constitute their model-systems. Since most past research has focused on single neurons, and so little is known about how brains learn, the primary objective of this research is also to create new knowledge and know-how. Significant to our analysis, the frontier nature of both laboratories nature demands that they also design and build novel technologies of investigation. The methods of our analysis are in some respects unusual for philosophy of science and technology, though in accord with a naturalist epistemology. The analysis derives from data we collected in a five-year ethnographic and cognitivehistorical study of the two laboratories. One objective of the study was to extend Nersessian’s previous work on model-based reasoning (see, e.g. Nersessian [2008]) with conceptual models to physical and computational models. Another was to attempt an integrative account of cognitive, social, cultural, and material factors in their development and use [Nersessian, 2005]. As with other ethnographies, that part of the study uses observations and interviews to uncover the activities, tools, and interpretive frameworks that support the research as they are situated in the on-going practices of the community. The cognitive-historical part of the study collects and analyzes data from traditional historical sources (publications, grant proposals, laboratory notebooks, and technological artifacts).1 The aim is to capture the diachronic dimension of the research by tracing the intersecting trajectories of the human and technological components of the laboratory, con1 Although some of the material we quote from comes from published sources, given the regulations governing confidentiality for human subjects research, if the authors are among subjects we are not able to provide citations to that material here. It seems that the possibility of conducting historical research in conjunction with research on human subjects was not anticipated!

Model-based Reasoning in Interdisciplinary Engineering

729

ceived as an evolving cognitive-cultural system, from both historical records and ethnographic data. This novel combination of methods enables developing thick descriptions and analytical insights that capture the dynamics of inquiry characteristic of research laboratories. In approaching the problem of integration, the trick is to create accounts that are neither cognitive with culture tacked on nor the reverse, and this necessitates re-thinking current interpretive categories. In our research we attempt a shift in analytical approach from regarding cognitive and socio-cultural factors as independent variables to regarding cognitive and socio-cultural processes as integral to one another. One way of making the shift towards integration would be to construe cognitive processes as comprising more than what takes place in the head of an individual scientist, and analyze scientific thinking as occurring within complex cognitive-cultural systems comprising humans and artifacts. A central claim of our analysis, then, is that inference is performed through interlocking mental and physical models and that the devices serve as hubs for interlocking cognitive and cultural facets of laboratory research. Our strategy in this paper is to describe modeling practices in each laboratory, and to focus on one example of experimental investigation using a model-system from each lab. We then discuss more generally the nature of the reasoning involved in constructing, manipulating, and revising model-systems. One problem often noted with such “case study” methods is how one generalizes from a specific case or any number of cases. The notion underlying this problem is ‘inductive generalization’ from cases. We subscribe, rather, to the ethnographic notion of ‘transference’ across sites/cases. In ethnography one develops richly detailed descriptions of a site. When conducting comparative investigations of multiple sites, thick descriptions are created of each, and these are examined to see what might possibly transfer across sites — as well as what of importance does not. Although the specifics in each case differ, our insights about model-based reasoning do transfer across laboratories. The main conclusions of our analysis derive from the observation that designing, redesigning, and experimenting with in vitro simulation models (devices) is a signature investigational practice. Physical simulation is an epistemic activity involving hypothesis exploration, testing, and generation, as well as prediction and explanation. The devices serve as hubs for interlocking biological and engineering concepts, methods, and materials, mental and external representations in modelbased inference, design and history, and research and learning. For brevity we refer to this multidimensional notion as interlocking models. 2 TISSUE ENGINEERING LABORATORY: THE “VASCULAR CONSTRUCT MODEL-SYSTEMS” Lab A has as its ultimate objective the development of artificial blood vessels (locally referred to as “constructs”). These are engineered out of living tissue and will need to have the appropriate characteristics to function within the human body,

730

Nancy J. Nersessian and Christopher Patton

such as sufficient strength to withstand the forces of blood flow and endothelial cells lining it that are able to proliferate. An in vivo/in vitro/ex vivo division is a significant component of the cognitive framework guiding practice in Lab A. The test bed environment for developing artificial blood vessels cannot be the human body, so the researchers have to design in vitro facsimiles of the in vivo environment and ex vivo implantation environments (so-called “animal models”) where experiments can occur. In this context, ‘ex vivo’ refers to an animal that has been altered such that experimentation can take place external to its body. The major challenge is to bring together biological and engineered materials with the desired properties so as to perform properly in vivo. As characterized by the Lab Director, the “major barriers” fall into two categories: mechanical properties of the tissue and the influence of mechanical forces on the tissue and cell source strategies that support endothelialization. Many aspects of the in vivo phenomena are known and understood both in biological and mechanical terms, but many are not, such as how cells proliferate. Thus the lab makes contributions (new knowledge and know-how) to basic biology and to engineering applications. The daily research in Lab A is directed towards solving problems that are smaller pieces of the grand objective, such as proliferating endothelial cells within the constructs, which involves, for instance, gene profiling studies, and creating constructs that can withstand the powerful mechanical forces of blood flow in vivo, which involves advancing collagen gel technology. In the next sections we discuss some of the central devices that form the major components of model-systems in these labs and how they interlock in experimental practice.

2.1 The flow channel device Lab A has been in existence since 1986 and was created specifically to move the research, as expressed by the Lab Director, “from animal studies to cell culture.” Early bioengineering experimentation on the vascular system was conducted by the Lab Director and colleagues on blood vessels that were altered while in the living animal. Through surgical interventions blood vessels were made to exhibit pathological conditions consisting in narrowing of native arteries (stenosis). After sacrificing these animals, the morphology of the cells that line the arterial walls at the pathologically altered regions was studied and specific aspects quantified (e.g., elongation and orientation of cell filaments). Simultaneously, arterial flow patterns (velocity profiles) were studied for pathological narrowing of the arteries through creating models that replicated the geometrical dimensions of these observed pathological regions. These structural models were achieved by filling arterial vessels of sacrificed animals with a fluid plastic. After hardening they were used as casts to manufacture replica of the narrowed vessel. These replica models were used in experimental “flow studies” where laser Doppler techniques were used to determine velocity patterns. The results gained from studying cell morphology and from studying velocity patterns in the replica models were correlated to gain

Model-based Reasoning in Interdisciplinary Engineering

731

insights into the relations between variations in wall shear stress due to particular velocity patterns (gradients near the vessel wall) and the morphology of cells lining these vessels. The elaborate material and measurement practices related to the replica models were quickly abandoned, but they launched the director’s program of studying the impact of flow in arterial wall shear stresses in vitro with engineered devices, the research agenda of Lab A. Along with the replica studies and the associated cell morphology studies, the director and other researchers started a line of research with cell cultures of the endothelial cells typically lining the arterial walls. This work provided the basis for the initial configuration of Lab A. Instead of inducing stenosis in living animals, and thus creating particular flow patterns resulting in particular wall shear stresses, they instead exposed the respective cell type in culture to wall shear stresses by “flowing” them in a flow channel device (called the “flow loop” in the Lab). These in vitro experiments on the response of the cells to shear stresses were based on an established fluid dynamic model, specifically, the fluid mechanics of a long channel with rectangular cross-section. Using this device, changes in cell morphology (elongation and orientation) could be related directly to the controlled wall shear stresses. Furthermore, the method of measuring velocity patterns in a replica model was replaced by an engineered model of exact geometrical specification, a flow channel. With the controlled flow channel the correspondence between the mathematical and the physical model became an issue of engineering a channel with the appropriate dimensions (in a physiologically meaningful range), rather than measuring velocity patterns using elaborate laser Doppler technology. The studies using the replica model had in fact dissociated the study of cell morphology from the study of flow patterns, correlating their results after the fact. The flow loop in action is a model-system in which the two foci of study are concentrated into a single system where cultured cells were exposed to flow and thus shear of a well-defined nature. The move to in vitro solved problems related to the fact that it takes twenty four hours to see results of interventions made in animals and during that period many physiological changes take place. However, as a model, it does not represent the diachronic nature of the in vivo environment, even though it is dynamic. Blood flow in vivo changes, for instance, when eating and sleeping. So, eliminating the confounding factors leads to a simulation of the in vivo environment that is “something very abstract because there are many in vivo environments and there are many in vivo conditions within that environment. Things change constantly in our bodies over our lifetimes; including physiological flow rates [....] So I don’t think we are trying to mimic the exact conditions found in vivo.” In the cell-culture line of research, an engineered artifact, a flow channel with the accompanying flow-inducing components serves as an in vitro model paralleling certain in vivo conditions of the blood vessel, including both normal conditions and the pathology that previously was induced in living organisms. The flow loop represents a first-order approximation of shear stresses during blood flow in the artery, “as engineers, we try to emulate that environment, but we also try to

732

Nancy J. Nersessian and Christopher Patton

eliminate as many extraneous variables as possible. So we can focus on the effect of one or perhaps two, so that our conclusions can be drawn from the change of only one variable.” The flow loop provides “a way to impose a very well-defined shear stress across a very large population of cells such that their aggregate response will be due to [it] and we can base our conclusions on the general response of the entire population.” The speed at which the flow loop pump operates reflects knowledge of how fast blood flows in vivo, and the pulse dampener turns the flowing liquid from pulsating flow to a smooth flow that allows control over the constancy of flow. The flow loop allows manipulating the amount of shear stress (speed of flow), duration, and height of the chamber. Using it in flow simulations also enables manipulation of cells or engineered vascular constructs. As the laboratory began to establish itself, flow loop studies on cultured cells were conducted by all members. Over time, new simulation devices have been designed not only to understand mechanical forces creating pathology, but for creating the vascular constructs that will some day repair diseased arteries, as will be discussed below. From the outset, redesign has been a central activity within Lab A. The flow loop provides a major instance of this activity. In working with cultured cells, contamination is a constant problem and this problem was the driving factor in the redesign of the flow channel device. An interview with a former graduate student, now a successful faculty member at another institution, elaborates on this problem and the subsequent redesign: “So, when I got here in 1994 uh, the flow chamber was a mess. It was a benchtop system, it had bulky tubes that looked something like some time machine from the 1950s or something [. . . ]. But anyway it was quite messy and you know culture studies have to be done at 37 degrees so the way that they would do this was you know, incubators were certainly around in 1994, uh, they would wrap these coils, these heating coils around these glass reservoirs and because it had to be a set flow, they would use a hydrostatic pressure difference to derive the flow, and uh, a clamp, a regulated clamp to try and regulate the flow through the chamber and out into the — into the lower reservoir. So you had two reservoirs, one at the top, and one at the bottom, there’d be a hydrostatic difference between them, and then things would flow and then this whole thing would be sitting on the benchtop — big bulky glass reservoirs with bulky tubing [...]. And this was subject to about a 50% success rate.” Interviewer: In terms of contamination? “In terms of contamination. And the reason was because this whole thing had to be assembled outside of the hood [colloquial for ‘the sterile workbench’]. There was no way you could assemble this thing to stand up-this thing was on stands–you have to assemble this part outside of the hood, so basically they we would connect these joints here, and connect them outside of the hood. [. . . ] Doing experiments longer

Model-based Reasoning in Interdisciplinary Engineering

733

than 48 hours was almost impossible, because at experiments longer than 48 hours the incidence of contamination was probably greater than 90%. [.....] I really like compact designs [. . . ]. I instituted a lot of the things I saw over there [referring to the lab at which he interned] in our laboratory, and one of the things was model-revising this design to go into the incubator. And, that was really why we moved from a system that required heating coils and an upper reservoir and a lower reservoir to a system that was just flow driven with a peristaltic pump and a pulse dampener that was — and everything could be done inside the incubator with smaller tubing, little reservoirs as opposed to big reservoirs.” “Model-revising this design,” as the former graduate student described his contribution to this line of research, meant redesigning the physical system that is the flow channel device, its parts (e.g., the reservoirs, the tubing), set-up (e.g. on stands in the lab vs. compact and in the incubator), and the physical principles governing its functional design (e.g. hydrostatic pressure difference vs. integration of a peristaltic pump). The actual flow channel, which is the part where the liquid flows over the cell cultures, was left untouched in this particular redesign. Even though redesign, in this case, did not involve those parts where the cells-in-culture interfaced with the mechanical device, re-engineering this design was central to its function as part of a model-system, which is totally dependent on its resistance to contamination of the cell cultures. The set-up that functions as the model-system is sufficiently decomposable to allow for the independent redesign of its various components. Since then the flow loop has undergone minor redesign, such as related by a current Ph.D. student, because of the introduction of a new device, the construct (described in detail below). In discussing her own redesign, she started with telling of how the researcher just prior to her had modified the flow block to solve some technical problems. The modified device that she inherited had previously been used on cells. She now wanted to use the flow loop to experiment with the vascular constructs seeded with endothelial cells, cut open and placed flat for flowing. These flat constructs are thicker than the muscle cells used before, and bumpy. Because of these features, spacers need to be used between the block and the glass slides in order to improve the flow pattern around the boundary to bring the in vitro model more in accord with flow in the in vivo model. To begin this research, she, together with another new student, had to redesign the device by changing the width of the flow slit to hold the spacers. Most recently another student planned a significant redesign to enable flowing of constructs in tubular shape in order to accommodate implantation in an “animal model ” that will be discussed in the next section. This redesign would mark a significant step in the move towards in vivo implantation in that the constructs would not need to be cut open in order to be flowed.

734

Nancy J. Nersessian and Christopher Patton

2.2 The “construct” The endothelial cells that form the endothelium, a mono-layer of cells that make up the inner lining of the blood vessel are a major target of study in the Lab. In vivo these are in closest proximity to the blood flow. The culturing of cells needs to emulate the naturally occurring conditions of living tissue in an organism to the extent that cells are required to survive and perform in particular ways. This emulation requires such things as the appropriate CO2 levels in incubators and that the incubators keep the cells in the requisite temperature range. Moreover, the types of cells identified for embedding in the vascular substitutes must readily be available and compatible with adjacent tissues. This requires a method for ensuring cell growth and proliferation and cell sources for production. So the introduction of the construct has led also to a line of stem cell research. Until the late 1990s flow loop experiments were done only on cell cultures on slides usually coated with a substance to make them adhere. After flowing they are removed and examined through various instruments such as the coulter counter and the confocal microscope to determine the effects of the mechanical properties of the flow on shape (morphology), alignment, proliferation (reproduction), or migration (locomotion). However, as one researcher observed, “cell culture is not a physiological model; however, it is a model where biologic responses can be observed under carefully designed and well defined laboratory conditions.” So, although many experiments are still done on cell cultures, the problem of constructing a tissue engineered vascular substitute has led to the creation of a new simulation device, the construct. As the lab director recounted, the current research aims “to use this concept of tissue engineering to develop better models to study cells in culture. Putting cells in plastic and exposing them to flow is not a very good simulation of what is actually happening in the body. Endothelial cells, which have been my focus for thirty years, have a natural neighbor called smooth muscle cells. If you look within the vessel wall you have the smooth muscle cells and then the inside lining is the endothelial cells, but these cell types communicate with one another. So we had an idea: let’s try to tissue engineer a better model-system for using cell cultures.” The construct marks a move towards a more physiological model — one whose function is more like the in vivo model along mechanical, physical, and biochemical dimensions. An actual blood vessel is in tubular shape and comprises several layers: the lumen where the blood flows; a first, mono-layer of endothelial cells that sit on collagen, an internal elastic lamina, a second layer of smooth muscle cells, collagen, and elastin, external elastic lamina, and a third layer of loosely connected fibroblasts. In vivo, the cells create an extracellular matrix which is a network of proteins and other molecules, and provides growth factors and mechanical properties. In the in vitro culturing process, the construct is seeded onto a tubular shaped silicon sleeve. Unlike cell cultures on slides, constructs are three-dimensional surfaces in which cells are embedded. The construct is a “wet”

Model-based Reasoning in Interdisciplinary Engineering

735

device — a living “blood vessel wall model ” that simulates in vivo processes. It is a significant component of many current research projects. The hope is that “our construct behaves like a native artery because that’s one step closer to being functional. So we are kind of doing a mimicking thing. So does it respond in the same manner?” When eventually used as replacement systems for the human body, the biological substitutes must replicate the functions of the tissues being replaced. This means that the materials used to grow these substitutes must coalesce in a way that mimics the properties of native tissues. It also means that the cells that are embedded in the scaffolding material must replicate the capabilities and behaviors of native cells so that higher level tissue functions can be achieved. To “respond in the same manner” means, among other things, that it expresses the right proteins and genetic markers. In building a construct, different levels of approximation are used depending on the nature of the experiment. It is possible, for instance, to use only collagen and not add elastin. Most often the cells are not human aortic endothelial and smooth muscle cells. Some experiments are conducted with only a single layer of the blood vessel wall with either endothelial cells or smooth muscle cells. Thus the construct forms a family of models, designed for different experimental purposes. In the experiment we outline below, conducted with baboon cells with a baboon animal model, the third layer was not engineered because the researcher both deemed it unnecessary for her experiment and also surmised that it would grow itself. In addition, she used a teflon scaffolding on the outside of the construct since it is not strong enough to withstand the forces of the baboon’s blood flow. Most significantly, in experiments with the current flow loop, the constructs need to be cut open to lay flat within the flow chamber as designed. The fact that the flowed construct is flat due to the design of the flow loop whereas the blood vessel is curved is an approximation to the tubular surface. However, since the cells are so small with respect to it, the flatness is not an approximation for these main objects of study. Indeed, from the “cell’s perspective” it is not an approximation since “the cell sees basically a flat surface. You know, the curvature is maybe one over a centimeter, whereas the cell is like a micrometer — like 10 micrometers in diameter. It’s like ten thousandth the size, so to the cell it has no idea that there’s actually a curve to it.” That is, flowing liquid over these flat constructs will give accurate enough data on the responses of the endothelial cells that line the arterial wall, because the cells are so small with respect to the arterial wall that the cell’s experience of the wall is as though it lives in a flat world. The medium flowing over the construct lacks a vertical component (flow is two-dimensional) and is unidirectional whereas in vivo “it sloshes around in the blood vessel.” But again the focus is on first-order effects unless there is evidence of a need to consider higher-order effects, such as if they find “that there’s a whole different pattern of genes that are up-regulated in pulsatile shear, or something, maybe then it would be more interesting to use different constructs and stuff,” so as to be able to look at the higher-order effects.

736

Nancy J. Nersessian and Christopher Patton

The introduction of the construct led to a range of new devices for experimenting on it directly, such as the pulsatile bioreactor that was designed for “mimicking the wall motions of the natural artery.” It is used to expose the constructs to mechanical loads in order to improve their overall mechanical properties, what the researchers call “exercising the cells.” Preferably, this is done at an early stage of the formation of the construct, shortly after seeding onto a prepared tubular silicon sleeve. In vivo, arterial wall motion is conditioned upon pulsatile blood flow. With this bioreactor, though, which consists of a rectangular reservoir containing a fluid medium in which the tubular constructs are immersed and connected to inlet and outlet ports off the walls of the reservoir, the fluid does not itself move. Rather, the sleeves are inflated with pressurized culture medium, under pneumatic control (produced by an air pump). The medium functions as an incompressible fluid, similar to blood. By pressurizing the medium within the sleeves, the diameter of the silicon sleeve is changed, producing strain on the cells, similar to that experienced in vivo. However, as a model-system, “the silicon sleeves add the next level of doubt. [...] The construct itself is not actually seeing the pressure that the sleeve does. And because of that you know — it doesn’t actually see a — a pressure — it feels the distention but it doesn’t really feel the pressure. It doesn’t have to withstand the pressure. That’s the whole idea of the sleeve.” These differences between the in vivo and in vitro models arise from the nature of the devices themselves. The construct at present is not strong enough to sustain the actual forces of pulsating blood flow and the bioreactor itself has been designed only as a functional model that achieves a parallel motion through different behavior.

2.3 Preventing platelets: experimenting with a vascular construct model-system In experimental situations models tend to be in interlocking configurations, that is, not as isolated entities but rather as standing in particular relations to other models. The diagram in Figure 1 is a schematic representation of our analysis of the vascular construct model-system for one proposed experiment aimed at solving the problem of platelet formation on constructs, and the resulting thrombosis. The diagram traces the construction, manipulation, and propagation of models within the system that constitutes the experiment. In Figure 1 the models involved are highlighted by thick lines. The experiment is significant because it constitutes a first move in the direction of in vivo research in that the animal model serves as a model for the human system in the context of the experiment. In this experiment an exteriorized shunt connecting the femoral vein and the femoral artery has been placed in a baboon so that a small amount of blood flow can be diverted through a construct during an experiment. The baboon’s blood is injected with iridium so that platelets will be made visible through the gamma camera, a commercially available instrument. Each physical model is constructed to represent and perform as a selected aspect of the cardiovascular system, for example, media and constructs represent and per-

Model-based Reasoning in Interdisciplinary Engineering

     



!       $   



    $ $   %&"'(

$ ) *   

    ! ""

737

      $   

 !! 

 

!!  *

 

 

$ ) *

  *   $ ! 

  

 #

$   

$  +

,  -  

 



  

 

     

. *  

/   /  ! 



   +



Figure 1. A vascular construct model system form as the biological environment of the blood vessel and the flow loop represents and performs as shear stresses do on arterial walls. This experiment initially proposed the redesign of the chamber of flow loop to better approximate the in vivo model in that constructs would be flowed in tubular shape, which is necessary in order to implant them. In the end this turned out to be unnecessary because the researcher had the insight that it should be possible to design an external shunt for the flow loop — on a direct analogy with the shunt used in the baboon model — and attach the construct to it for flowing. During the several re-iterations of the experimental simulation over time, the force with which the media flowed through the construct was adjusted until platelet formation ceased. In general, each experiment with the model-system comprises a number of interlocking models which include at least the following: community models of in vivo phenomena (biological, mathematical, mechanical); engineered in vitro/ex vivo physical models of aspects under investigation; and mental models of in vivo and in vitro phenomena, devices qua in vitro/ex vivo models, and devices qua engineered models. Each mental model is both an individual and a community achievement. Note, too, that models from mathematics, physics, and biology inform the construction of the devices. We return to this notion of interlocking models and to model-based reasoning after considering the model-systems of Lab D.

738

Nancy J. Nersessian and Christopher Patton

3

NEURAL ENGINEERING LABORATORY: THE “EMBODIED DISH MODEL-SYSTEMS”

Lab D is a cutting-edge neuroscience research laboratory whose focus is understanding the fundamental biological nature of learning and memory. Specifically, they are interested in “network level plasticity, learning, and memory” at the cellular level. Historically, neuroscience has been practiced at the two extremes on the spectrum of granularity. At one end there are neurologists who seek a gross, system-wide, psychological understanding of the brain, and at the other end are single-cell electrophysiologists who seek to understand the fundamental properties of a single neuron. This lab, however, sees itself as breaking ground in “mezzo” (meso) level neuroscience. The researchers are interested in the nature of networks of cortical neurons, which are too small and unorganized to be considered a functioning brain, but are at a higher level of organization that the single neuron. Early attempts at investigating the nature of networks of neurons depended on cutting slices of a brain, fixing the neurons (in formaldehyde), and then looking at the physical structure of the neurons underneath a microscope. The Lab D Director was not happy with looking at dead neurons and instead wanted “to look at living things, while the interesting parts would happen.” To this end, he spent many years of postdoctoral research designing and perfecting what has become Lab D’s most prominent model: “The Dish,” and also developing technology for imaging it.

3.1

“The Dish”

Simply put, “The Dish” is a network of cortical neurons living, not as part of a larger brain, but as a small network in a Petri dish. Embedded in the bottom of this Petri dish are 64 electrodes capable of recording and injecting electrical activity in the network. The basic construction of the Dish is as follows: • The MEA: the Microelectrode Array (MEA) is a small, glass, Petri-style dish which has an 8x8 grid of micro-electrodes embedded in the bottom of it. The electrodes poke upwards from the bottom of the dish into the neurons. • Cells: Lab D uses parts of a rat’s cortex as their network. The two basic types of neural cells, neurons and glia (support cells), are harvested from the rat and then cultured in this dish. • Medium: A sugary cocktail of biological chemicals used to feed the cells. • The Lid : A thin film made out of Teflon is stretched over the MEA and held in place by a thick “O” shaped piece of Teflon (like a slice of pipe). Teflon is used specifically because it is non-toxic and allows O2 and CO2 through so that the cells can breathe. However, it is not porous, so it keeps everything else (such as bacteria or fungus) in or out, even when it is stretched.

Model-based Reasoning in Interdisciplinary Engineering

739

There are no established models of neuron communication. In fact, the endeavor of Lab D has been to build these models from the ground, up, as expressed by the Lab Director: “well the level of modeling that I’m very interested in is the network modeling. So there are a lot of people out there who are doing very detailed modeling of single neurons, in fact even [in] this lab X and Y are doing that, but I’m much more interested in what happens when you take a bunch of cells and have them interact with each other, what sort of emergent properties pop up. So, it probably isn’t necessary to include all the details that you would find in a [single neuron] model, with ions and stuff, but it may be, so that’s part of our job is to find out which of the details of biology are important in these sort of network properties and network phenomena and which are sort of incidental.” What is growing in the dish, then, is not a brain, nor is it even a slice of a rat’s brains. Instead, the neurons are disassociated (their connections are broken apart to the point that they free from each other), plated (processed and placed on the MEA), and then cultured so that their connections are allowed to grow back together. As, one researcher described their reasoning for using The Dish as opposed to a brain slice: “Yeah, one single-layer of neurons. We try to get them down in a monolayer. That’s the whole idea. It’s a simpler system to study then. That’s all. And also, I mean, you could do slices that are not disassociated cultures but the problem in slices is it is difficult to maintain them. You have to kind of get the fluid [in] right. Sometimes only the outer cells — you know the medium does not go into the inner layers. The inner layer dies off but the outer layers are fine, stuff like that. So, it’s more difficult to keep them alive. And we want to study over a long term, so we want to keep them alive over months, years.” The Dish as a physical model of the brain of a rat that is constrained by both the state of Lab D’s current understanding of the dish, as well as by technical hurdles having nothing to do with the biological model. In response, Lab D has chosen to simplify their model to a single layer of neurons to reduce the number of possible variables in the system. Doing so gives them a smaller set of questions to ask in order to build their understanding of network processing. However, they believe that this simplified model provides a close enough approximation to yield valid information. The Dish is the lab’s central model-system: the physically constructed embodiment of the Lab’s selective model of the brain. It is an in vitro model of basic in vivo neurological processes. With an in vivo neural system there is specific structure to the pathways and connections in the networks, created by the developmental processes of the animal. Here, though, the neural structure is destroyed

740

Nancy J. Nersessian and Christopher Patton

so that The Dish does not directly reflect any structure within a living rat. Consequently, there is some contention over what The Dish actually models. Some maintain that The Dish is “a model of cortical columns” while others think it “is a model of development [of the brain],” and when pressed, some even admit that “it may just be a model of itself.” However, they all agree that studying The Dish will yield understanding of the basic workings of network-level cortical neural processing, as explained: “First of all, it’s a simplified model, I say that the model is not, it’s artificial, it’s not how it is in the brain, but I still think that the model would answer some basic questions, because the way the neurons interact should be the same whether it’s inside or outside the brain. . . . You know, because we are in an artificial environment, it’s not the same, you know, it’s not the same concentration as it’s in the brain; nothing is the same. I’m growing in an external environment. But, I think that the same rules apply.” The Dish is designed to be a generic model of cortical processing behavior and function. The researchers are not seeking to understand the processing of a specific construction of a particular rat’s brain. Instead, they are interested in how networks of neurons — in general — communicate and process information. There is clearly an intentional lack of specificity in this model in order to be able to build their understanding of neural processing in general. This generic understanding of the networks is not the end goal, however. After building the generic basis of understanding, they plan on building a more refined understanding of neural communication. As stated by the Lab Director: “Clearly it’s missing a lot of other brain parts that are important in what brains do. I’m assuming they are important. And at some point we might be studying cultures with different brains parts mixed together, or specific 3-dimensional pieces that are put together.” The physical construction of the model-system not only frames the inquiry, but also is an integral part in the research progress of the lab as a whole. Only once the researchers understand their current physical model and construct an accurate abstract model (mental, mathematical, etc.) can they progress in the construction of a new model that more closely approximates the in vivo situation. The construction of The Dish itself stretches across biology, chemistry, and electrical engineering, and requires the entire endeavor within Lab D to be interdisciplinary. Though Lab D’s research focuses on a mostly biological entity, it is largely populated by members with backgrounds in electrical engineering and mechanical engineering — and not biology or neuroscience. Unlike Lab A, there are no telltale signs of biology: no flasks, no pipettes, no hazardous waste containers. In fact, aside from a conspicuously covered microscope and an incubator that could be mistaken as a mini-fridge, Lab D’s space looks more like a computer lab than anything else. The most striking features of the Lab are the copious wires

Model-based Reasoning in Interdisciplinary Engineering

741

that crisscross the space carrying electrical signals produced by the neurons to the researchers and their computers. Although all of the researchers use The Dish, there are several different avenues pursued in this lab including pharmacological studies, morphological (imaging) studies, and simulation with both computational models and physical models. Here we highlight the Lab’s electrophysiological investigations. The main goal of the research is to understand how to “communicate” electrically with the network of neurons, and to try to see — indeed to figure out what counts as — evidence of network plasticity (learning). The core of the method involves electrically stimulating a biological neural network and recording the electrical response of the dish. To this end, the electrophysiology technology in Lab D centers on finding new ways to “talk” to The Dish, and then, in turn, trying to understand what the dish is “saying” back to them in response. We will outline three ways in which experiments are conducted with Dish model-systems.

3.2

“Talking” to The Dish: electrophysiology

The objective of the experimentation in Lab D’s electrophysiology research is to understand how information is encoded and processed in The Dish networks. It would be ideal to have direct readings from every single neuron in The Dish. However, with current technology this is impossible. Consequently, access to the Dish is mediated by a comparatively small set of electrodes (the MEA). The signals ultimately received are a representation of neural activity filtered through several models which will be discussed below. It is currently impossible to know the actual neural activity in The Dish. What they study are the “spike” data recorded from the electrodes of the MEA. Historically, the term “spike” refers to the electrical trace left behind when an individual neuron fires. In single-cell electrophysiology, it is possible to read electrical activity directly from the neuron, and as a consequence, the model for neural firing is well known: there is a steep jump in voltage potential as the neuron de-polarizes (fires), and then a proportional drop in potential as the neuron recovers. Multi-cellular neuroscience has borrowed the notion of a ‘spike’, but has modified it to suit their situation. The researchers of Lab D estimate that the electrical activity recorded on a single electrode can come from on average three to five different neurons. When dealing with the electrical traces of many neurons, it is possible (and is often the case) that several neurons around a single electrode will fire simultaneously. As a result, a spike can actually represent the firing of more than one neuron. However due to the fact that they are all on one trace, it is impossible to tell the difference between the firing of a single neuron and the firings of multiple neurons. This altered conception of what a spike represents is shared among the Lab members. Historically, spikes were tagged by hand. However, Lab D has created an automated process of identifying spikes by having developed a piece of software they call the “Spike Detector.” The Spike Detector embodies the Lab’s model of a spike which includes the spike’s “height” (difference

742

Nancy J. Nersessian and Christopher Patton

from the average voltage) relative to the noise on the electrode and the “width” (duration) of the change, along with a few other more subtle characteristics. The Spike Detector checks for jumps in the voltage that match this model of what a spike should look like, tags the spike, and keeps a snapshot of the electrical activity immediately around the spike. The researchers begin their analysis with the filtered spike data. The data that the researchers ultimately use do not come directly from the neural processes in The Dish. There are several pieces of software like the Spike Detector that are used during electrophysiological experiments. These pieces of software are collectively referred to as “filters” and perform a number of different transformations on the raw neural data before the information ever reaches the researcher. Each of these filters embodies a model of an aspect of the neural data. It is possible that the filters miss actual neural firings, or detect a jump in the readings that does not correspond to any action potential (i.e., provide a false positive). Thus the meaning of an individual spike needs to be understood in terms of the filter algorithms that created it, that is to say, in terms of the series of transformations used to produce the data, each one of these transformations being built on a model of the selected electrical signals. The researchers are intimately aware of these processes and perform their analyses in light of this understanding. In the simplest form of analysis, the spike data are transformed into a visualization rendered on a computer screen. There are a number of different types of visualizations the Lab uses for these basic data, and far more for their higherlevel analytical transformations. The simplest visualizations come from the datacapture software they have developed called “MEABench.” As the computer is capturing neural data, MEABench can display the neural activity in real-time. The MEABench screen is divided into 64 rectangles (one for each recording electrode) and arranged to topographically match the layout of the electrodes in the dish. Each rectangle, then, displays a trace of the electrical signals captured on the electrode. When the Spike Detector is turned on, the visualization tool places little red dots at the peaks of the spikes for easy visual identification. The researchers’ ability to “see” the electrical activity of the dish, then, involves a number of interconnected mental, physical, and algorithmic models. The visualization is a representation of the information produced by the filters. The filters are instantiated models of Lab D’s understanding about neural signals. The signals themselves are a model of neural communication (as they abstract away other factors such as chemical signaling). Finally, The Dish itself is a model of neural functioning in rat cortex. This series of interlocking models is the base model-system for any electrophysiological experiment run within Lab D. This experimental set-up, however, is one of the most basic model-systems used in the lab. In the next section we will describe two far more complex “embodied dish” model-systems: the “Animat” and the “Hybrot.”

Model-based Reasoning in Interdisciplinary Engineering

3.3

743

Computational embodiment: animat model-systems “[In] the traditional way to do in vitro physiology . . . the closest thing to behavior is little waves on the oscilloscope screen. [It] has nothing to do with any behavior, other than light on the screen there. And there is not any sensory input other than electrical pulses through a couple of electrodes. You know, it is very disconnected, and one of the things that was really shaping my thinking at that time was this book here. This is the first of a proceedings from this conference — simulation of adaptive behavior, here. And, I think it was 1990 that they had this conference, yeah. [A]ll of the people in that book are simulating animals or what they call “Animats.” The term was coined around that time by the guys who were at this conference. They were simulating these things on the computer or they were building robots that were animal simulations. They were continually reemphasizing the importance of embodiment and being situated.”

Originally, then, the idea for animat model-systems developed in Lab D came from the domain of computational modeling. There the goal was to create a very simple model “world” and a very simple model “animal” and then simulate the activity of the “animal” in the “world.” The Lab Director borrowed this idea of modeling animals, but decided that he could improve on it. While others were using a completely computationally simulated animal, he set out to make a more realistic model of an animal by using The Dish as the “brain.” The term “animat” in Lab D refers to a computationally simulated entity that is controlled by the activity in a Dish. This model-system mimics the fact that learning in vivo is embodied, that is, comes about from sensory stimulation and feedback. An animat consists of The Dish (the “brain” of the computational animal) and two translation programs: one designed to be the simulated sensory input apparatus and one to simulate motor output that exists in a simulated environment. In short, animats are used to simulate the embodiment of the neural networks and are used as a primitive model of an animal functioning in the world. In the description of the experimental set-up we developed in the last section, the electrical signals produced by the neurons were simply recorded and then analyzed. This is typically referred to as “open-loop” electrophysiology. Animats, on the other hand, are part of “closed-loop electrophysiology.” Closing the loop simulates embodied learning. The electrical signals produced by the neurons are not simply recorded. Instead, they are transformed in some meaningful experimental way and then fed back into The Dish. The “sensory” information is then run through a translation program that converts it into a pattern of stimulation to be administered by a stimulation board. The stimulation produces an electrical response in the dish which is recorded and run through a separate translation program which converts the signals into “motor” commands for the animat. The motor commands move the animat in its “world” which, in turn, produces a change

744

Nancy J. Nersessian and Christopher Patton

in the “sensory” information. The change in sensory information is again read from the electrodes in the dish, and the loop continues. “There are lots of different arguments for it. I think probably one of the best arguments is. . . if you look at what neurons do they . . . they learn things. And that’s what we’re trying to figure out. . . how learning works, how memory works. And so if we have neurons in a dish, we want to see them learn something. . . make associations. . . it’s a bit more obvious to see . . . the learning involved in a closed loop [situation].’Cause you define what it is it’s going to learn based on . . . the body you give it and the environment you allow it to work in.” As expressed by the researcher, the animats provide a model-system for studying the relationship between the fundamental nature of information processing in the neurons and the visible behavior of animals. The animat model-system serves to demonstrate network plasticity more convincingly than open-loop experiments, and will, in turn, lead to a better understanding of how to interpret and control the activity in the dish. Just as there are many different types of creatures in the real world, Lab D has created an entire family of model animats, one of which is a simulated “moth.” Here we briefly consider D2’s construction and use of the moth model-system. As conceived by the researcher, “[t]he original model is basically, I have a circle and then the center of the circle, which this is the environment. The center of the circle, where you can find, what you can think of as a light. . . and I wanted, the moth, which would be the animat would move towards the light like moths do.” In the case we analyzed, the screen visualization comprises a circle that delimits the entirety of the moth’s world, a dot in the center represents the simulated “light” and lines across the circle represent the paths that the moth has followed during the duration of the experiment. The moth was given a simulated “eye” to “see” the light and the simulated motor ability to move around in the “world.” The two most interesting parts of any animat experiment are the programs that interpret the neural signals. The sensory translation program enables the moth’s sensory system to “see” where the light was and “see” where it is, itself, and turns this information into a series of neural signals just like an animal sensory system. Once The Dish is stimulated by this sensory input, it responds with its own electrical activity. This activity is recorded and processed by the method described in the electrophysiology section, and then translated into motor commands for the moth. For instance, the neural activity can be treated as a population vector; in other words, each electrode is taken to represent a possible direction of motion. Activity on the left of the dish would indicate that the Moth wanted to move left; activity on the right would mean the Moth wanted to move right; and so on. In essence, the network as a whole determines which direction to go in based on the

Model-based Reasoning in Interdisciplinary Engineering

745

sensory input. Importantly, the translator programs arise from a number of different models. First and foremost, the programs are a model of how D2 understands neural communication, and thus the translation algorithm is an exemplification of his idea about how neurons communicate. Also, put together, the translator programs form a basic approximation of how an animal’s sensory-motor system operates. In sum, an animat, such as the moth model-system is an in vitro simulation of an in vivo model of neural processing that requires “embodiment,” environment, and continual sensory-motor feedback. The sensory translation program derives from a model of the researcher’s understanding of how animal sensory systems translate raw sensory information into meaningful information for the cortex to process. The Dish component, itself, is a model-system that represents basic neural processing. The motor translation program embodies a model of how an animal’s motor system converts neural output into motor function. Finally, taking the model-system together as a whole, the experimental animat is a simplified model of a real-world animal. Running this model-system and analyzing the outcome of the experiment is not the end of the story though. As with most model-systems in these types of cuttingedge communities, the animat system is constantly evolving. Using the results of the first moth, the researcher revised his understanding of neural processing and, in turn, revised his model of moth. After running the moth a number of times, he updated the sensory control system, in particular, he “simplified that even further just to have frequencies only dependent on positive X production. . . Just to make it, easier to analyze in data analysis. . . Just to try and simplify everything. I could just say, ‘Okay, this part of the stimulation does this.’ ” Much like with the Lab’s decision to use monolayer cultures, here the researcher chose to design a less realistic, but more easily understandable model. This limiting case abstraction of an animal gives the researchers less degrees of freedom, but provides for more clearly defined questions and experiments that provide results that are less complex and more easily interpreted.

3.4

Physical embodiment: Hybrot model-systems “So the point is, these cultures are basically the model of the brain, right. The point is to have an embodiment for these cultures which is basically a robot. So, you. . . We have a robot in the other room where you can mount this culture on top of it and you can basically program the robot to do stuff, and see how it behavior changes. So basically, the behavior of the robot controls the input to the dish . . . and how it interacts. So we have the robot running around in some kind of environment doing its stuff according to some algorithm we set, which hopefully does some learning, and then we see how the behavior of the robot changes over time.”

746

Nancy J. Nersessian and Christopher Patton

A “hybrot” (a contraction of “hybrid robots”) provides another kind of embodied experimental model-system for studying neuron learning in. Like an animat, a hybrot provides a generic model of animals. Hybrots take this idea one step further, however, and place the “body” for the network of neurons in the real world as opposed to in a computational simulation. Instead of having a simulated environment with simulated sensory input and simulated motor output, The Dish is given a physical sensory system in the real world with real motor capabilities, and thus provides a physical simulation of in vivo processes. There are several robot “bodies” used in Lab D, some purchased off the shelf such as the “Koala,” and others that are built from scratch in the lab such as a series of “dancing feathers.” The predominant and most interesting hybrot model-system used by the Lab is a mechanical arm that is physically distributed in that part of it resides outside of the physical space of the lab. It is capable of drawing (primitive) pictures that goes by the name “MEArt.” MEArt was designed as both an art project and a research project with the goal of bringing the idea of a neural-mechanical hybrid system to the public. It is “a geographically detached, bio-cybernetic research and development project exploring aspects of creativity and artistry in the age of new biological technologies from both artistic and scientific perspectives.” As an exhibition, MEArt is touted as a “semi-living artist” that uses neurons to control a mechanical arm which draws scribbles on a piece of paper. As a research object, it is a closed-loop electrophysiology model-system which, unlike the animat, has a physical embodiment. The model-system is constructed as follows: a webcam is used as MEArt’s eye, the information from the camera is run through a sensory translation program, the program stimulates The Dish, The Dish responds, those signals are run through a motor translator program, the translator issues motor commands to the arm, and then finally the arm moves and draws a line on a piece of paper. The MEArt model-system serves as a model of an in vivo arm abstracted along a number of dimensions. For instance, the only kinds of information available to MEArt are positional information of its arm and sensory information from its camera. All other sensory dimensions and possible internal states have been abstracted away. Even so, the limited amount of uncertainty and complexity introduced by interaction with the real world along these two dimensions give the researchers a model that more closely approximates real neural input than animat model-systems. As one researcher expressed it, “Yeah. I’ve done it in simulation. That’s usually the first step. . . Just do it in simulation to get an idea. . . I guess one of the points is that once you stick something in the real world. . . and operating. . . you’re adding things that you can’t predict beforehand. And when like. . . Rodney Brooks did that with robots, with AI, he put AI into robots, he found out many different things that would not have been found out with just simulation.”

Model-based Reasoning in Interdisciplinary Engineering

747

There is a MEArt animat that the researchers use first to simulate the likely behavior of the arm. But then the hybrot is created to exist in the real world. The point of the prior computational simulation is to constrain the degrees of freedom tightly in their physical aspects in order to localize tests of the model. Once the general model is built from the animat, the hybrot model serves to explore the validity of the model in the real world. Perhaps, then, the most interesting thing about the hybrots is that they blur the lines between the in vitro and in vivo worlds. Yes, they are clearly in vitro in the sense that the neurons exist outside their natural context and live in an artificial environment. However, the purpose of the hybrot model-system is to bring the neurons closer to their natural environment. The goal of the system as a whole is to more closely model the actual construction of an animal, and consequently the neural input. In this sense, one can view the MEArt model-system as a functioning entity in the in vivo world, as well as an in vitro model of embodied neuron learning. Through the construction of hybrots, Lab D tests not only whether or not they work merely as robots, but by creating physical models that more and more accurately approximate their understanding of a real animal, they physically explore and test their model of animal neural processing. As with Lab A, in Lab D experimental situations models, too, tend to be in interlocking configurations, as standing in particular relations to other models. The diagram in Figure 2 is a schematic representation of our analysis of the MEArt model-system. As can be seen in Figure 2, MEArt is a complex model-system involving several different physical and mental models (models shown with thick black outlines), instruments, and devices which converge and interlock around the MEArt hybrot. Like the vascular construct model system shown in Figure 1, here we trace the steps by which information is generated, manipulated, and propagated through the system. Instead, however, of tracing physical transformations, we trace the information transformations as they are interpreted through each of the instantiated models. Please keep in mind that even though the diagram details the MEArt system, a diagram for the Animat model-system would be identical, save for replacing “Real-World Environment” with “Simulated Environment”, and “Sensory System” with “Simulated Sensory System.” Since this is a closed loop system, there is no actual “start” point, but the best place to begin is with the sensory system (to the right of the figure). Here, raw sensory information about the world is gathered by a video camera and positional sensors for the arm. This information is then sent to the “Sensory to Activity Transform Program.” Here, the raw data are transformed into stimulator commands which feed the information to The Dish. As mentioned earlier, this program serves as an in silico interlock of models. It is both an instantiation of the researcher’s mental model of how sensory data are structured by an in vivo sensory system and an instantiation of the researcher’s mental model of how information is interpreted and processed by The Dish itself. The program then directs a stimulator to inject a pattern of voltage changes into The Dish, giving it its “sensory” input. In response to the stimulation, The Dish

748

Nancy J. Nersessian and Christopher Patton

    !   "  #$   " 

     

   

  



 %



 

           



   



 

   

        

  

 

  

    

   

     

  

  

  

        

    

    

 

Figure 2. MEArt model system

creates a series of electrical impulses as it processes the information. This response is picked up by the data acquisition card, and is fed through the MEABench system. As previously mentioned, the raw data are subjected to a series of filters, each an instantiated model for processing neural activity. The subsequent spike data are then fed into a second “Activity to Motor Transform” program which transforms the data into commands which drive the motor output of the hybrot. Here again we have a convergence of neural communication models instantiated in silico, only this time the models are that of neural response and motor activity. Next, the physical hybrot (the model of an “animal”) produces a physical change in the world by drawing on a sheet of paper. The drawing itself, while possibly aesthetically interesting, also serves as a representation (visualization) of the underlying processes that produced it. Finally, the physical changes in the environment are picked up by the sensory system, and the loop is closed. Significantly, we can see a number of different levels of interlocking models in this model-system. First, at the lowest level there are the fundamental transformations that drive the hybrot. At the highest level, we have the MEArt system in its entirety, as a physically instantiated model of a real-world neural process-

Model-based Reasoning in Interdisciplinary Engineering

749

ing system, and even a selective physical model of a real-world animal. Together, the researchers with the MEArt model system are able to make inferences about the relationship between plasticity (neural learning) and learning (animal behavior). As we can see here, the researchers effectively use MEArt to make inferences through the model system itself. 4

MODEL-BASED REASONING

Biomedical engineering is an interdisciplinary undertaking. The programmatic outcome of biomedical engineering is engineered artifacts built at least in part from biological components. In the cases we examined, these include an actual blood vessel substitute (as in tissue engineering Lab A) and a trained neuron culture for operating a robot (as in neural engineering Lab D). Notably, these are epistemically and ontologically hybrid objects, being simultaneously engineered and biological. These model-systems can be structural, behavioral, or functional (inclusive sense of ‘or’) analogues of real-world biological systems. To a high degree, they reflect the engineer’s design perspective. The following passage from an interview with the Lab A Director is illustrative: LD: “Well, it was clear to me from reading the literature that, and what was really motivating me by 1970-1971, was the fact that these characteristics of blood flow [mechanical forces] actually were influencing the biology of the wall of a blood vessel. And even more than that, the way a blood vessel is designed .” Interviewer: “So, this was influencing the characteristics of the biology” LD: “Yes, right, and influencing biological processes that were leading to disease. The way a blood vessel is designed is, it has an inner lining, called the endothelium. It’s a monolayer; we call it a monolayer of cells because it’s one cell thick. But it’s the cell layer that is in direct contact with flowing blood. So, it made sense to me that if there was this influence of flow [mechanical forces] on the underlying biology of the vessel wall, that somehow that cell type had to be involved, the endothelium.” [stress added] From the outset, he conceptualized the in vivo vessel from the design perspective of an engineer. He went on to note how difficult it was for many years to get biologists studying the cardiovascular system even to be interested in the possibility that mechanical forces play a role in disease. Through processes of designing, constructing, manipulating, and modifying the in vitro model-systems engineers reason about, make hypotheses, and achieve understanding of in vivo biological phenomena. Developing an initial mental model of a blood vessel in terms of mechanical forces, for instance, led to the inference that the response of endothelial cells to the flow was implicated in disease processes.

750

Nancy J. Nersessian and Christopher Patton

Developing the physical models of mechanical forces and of blood vessels, has led, for instance, to inferences about gene expression under specific flow conditions. In this section we address 1) the multidimensional nature of interlocking models; 2) the nature of model-based inference; and 3) how it is that in vitro simulations support inferences about in vivo phenomena.

4.1 Interlocking models Model-systems comprise configurations of hybrid bio-engineered models. In the area of tissue engineering, for instance, some of these models refer to physiological aspects, which can include mental models of the structures or of the functions of the vasculature and physical models that interrelate structure and function at that level. Other models refer to the cellular or to the tissue level, which can again refer to the respective morphology or to aspects of the biochemistry at that level and to their interrelations. Still others point to mathematical models and engineering specifications that can serve to further constrain design options, experimentation or understandings of the biology under investigation. Models are sites of serious, long-term investment of resources, and they define and delimit the research program. One device (an individual model) leads to creating new devices to form various model-systems affording potential experimental situations, such as happened following the introduction of the construct, which in turn led to the new bioreactor or The Dish, which led to building the animats and hybrots. As we can see by all our cases of model-systems, conducting experiments most often requires complex configurations of models. We attempt to capture these kinds of interconnections and more with the multidimensional notion of interlocking models. The devices serve as hubs for interlocking models along many dimensions. Models interlock biological and engineering concepts, methods, and materials (interdisciplinary melding). Models interlock in their design and construction (such as the flow loop modification because of the construct and the construct leading to the pulsatile bioreactor). Models interlock in experimental design. Mental and physical models interlock in model-based inference. Furthermore, models interlock research and learning, and serve as sites of cognitive and socio-cultural integration. Two dimensions of interlocking merit special notice. First, in these engineering labs where design and redesign is an overarching agenda, a significant dimension of interlocking is historical: “history” is to be understood here as intellectually “hands on,” that is meaningfully related to working with devices. Current devices interlock prior and future devices. Redesign of devices in the laboratory is embedded in an understanding of how a certain problem situation has led to the realization of certain design options. In other words, the agenda of redesign characteristic of an engineering research laboratory requires researchers appropriate some of its history as they go about their research. The current design is understood to be conditioned on the problem situation as it existed for the laboratory at a prior time, even if the situation is not fully known to the current researchers.

Model-based Reasoning in Interdisciplinary Engineering

751

Thus, within the laboratory redesign is an agenda, and with it the historicity of the artifacts becomes a resource for novel design options. Second, devices and model-systems are physical models that interlock not only with various community models but also are external representations that interlock with whatever mental representations are used by researchers in reasoning processes. As with other external representations such as diagrams and sketches, physical models constrain and provide affordances that augment the internal representations that participate in problem-solving processes. This is not the place to argue for the point, but Nersessian [2002; 2008] advances the hypothesis that the internal representational structures are also models — mental models (structural, behavioral, or functional analogs representations) — that afford simulative modelbased reasoning. Given that the intended function of physical simulations with model-systems is epistemic, they are integral in the researcher’s mental modeling processes.

4.2

Model-based inference

Here we outline the kinds of reasoning that can be carried out through modelsystems. Characterizing a model, loosely, as a representation of a system with interactive parts with representations of those interactions, an instance of modelbased reasoning: 1) involves the construction or retrieval of a model, 2) derives inferences through manipulation of the model, and 3) those inferences can be specific or generic, that is, they can either apply to the particular model or to the model understood as a model-type, representing members of a class of phenomena. The bio-engineered model-system is both a conceptual and an in vitro physical system representing the in vivo system under investigation. As such it is an abstraction — selective and schematic in nature — that represents dimensions of biological phenomena of interest to the researchers. Models are interpretations of target phenomena (e.g., forces within the human vascular system, learning in the brain) constructed to satisfy constraints drawn from the domain of the target problem (e.g., the biology and physics of the vascular system, the information processing carried out by the brain) and, often, one or more source domain (e.g. the flow loop’s cardio-vascular and engineering domains, the dish’s electro-physiology and engineering domains). Constraints include: spatial, temporal, topological, causal, material, categorical, logical, and mathematical. Performing a simulation can lead to new constraints — or to recognizing previously unnoticed constraints. Inferences made through simulations can provide new data and hypotheses that play a role in evaluating and revising models to comply with constraints or provide potential new understanding into in vivo phenomena. For the model-systems such as we have been studying, models are structural, functional, or behavioral analogs of physical objects, entities, or processes used in experimental situations. Such models represent demonstratively (as opposed to descriptively). The primary evaluation relationship between the in vitro model and what it represents is goodness of fit, that is, what and how well does it cap-

752

Nancy J. Nersessian and Christopher Patton

ture the salient in vivo dimensions. Adequate models need to be of the same kind with respect to salient dimensions of target phenomena (often taking several iterations to achieve this objective). Inferences are derived through manipulations of the model. Operations on models require transformations consistent with the constraints of the domain. Importantly, the kinds of reasoning processes in model-based inference include, though are not limited to (not ordered): abstraction (limiting case, generic, approximation, idealization, generalization); simulation (inferring outcomes or new states via model manipulation (mental, physical, computational)); evaluation (goodness of fit, explanatory power, implications); and adaptation (constraint satisfaction, coherence, plausibility). Selectivity in design of models enables bracketing of (potentially) irrelevant features and focuses attention on those relevant to the problem-solving context. Relevant constraints need to be determined for target and source domains. Abstractive and evaluative processes disregard irrelevant factors. Constructed models can instantiate irrelevant factors, but to reason correctly via a model requires recognizing these as scaffolding for the cognitively germane factors. As noted, different kinds of abstractive processes underlie selectivity and those that figure in model construction, in particular, include: idealization, approximation, limiting case, and generic abstraction. These provide different means of selectively focusing on features relevant to the problem solving while suppressing information that could inhibit the process. Suppression and selective highlighting of features provide ways of representing the problem in a cognitively tractable manner. Most importantly for conceptual innovation, they enable integration of information from different sources. Idealization is a common strategy for relating mathematical representations to phenomena. Thinking of the sides of a triangle as having zero width, from the perspective of a geometrical figure, or the mass of a body as concentrated at a point, from the perspective of determining motion, allows strict application of mathematical formulae. Once mathematized, the idealization provides a point of departure from which to add in information about real-world phenomena as deemed relevant to a problem. Limiting case abstractions involve extrapolation or reduction to a minimum. Approximation provides a means of discounting the relevance of differences. A standard approximation in physics is the “first-order approximation” used when applying a mathematical representation, as with the force of the flow modeled by the flow loop device. Basically it makes the assumption that any higher-order effects are likely to be irrelevant or to be so complex as to make the analysis intractable. A laminar flow — one without currents or eddies — provides a first-order approximation sufficient for solving problems pertaining to many fluid dynamical phenomena. Some abstractions are deliberate, as we saw in choice to use steady laminar flow rather than pulsatile and in the approximation that flow is 2-D and over the flat surface of constructs. Some abstractions occur because the devices are also systems themselves; possessing engineering constraints that often require simplification and idealization not deriving specifically from the biological system they are modeling, as we saw with design and construction of The Dish. As

Model-based Reasoning in Interdisciplinary Engineering

753

mentioned before, The Dish only provides a mono-layer network of neurons instead of the rich 3D connections found in vivo. Here, the researchers have reduced the complexity of the dish to a minimum along one physical dimension. The choice to use only mono-layer networks overcomes two hurdles. First, the recording technology is limited to a mono-layer 8x8 grid embedded in the bottom of the dish. Any neurons that were not in the bottom layer of the dish would not be accessible for recording. Also, from the perspective of data analysis, the mono-layer provides a reasonable reduction of information, while still maintaining the salient qualities of inter-neuron communication. This limiting case makes the interpretation of the electrical response tractable, both the physical and informational. But then, the researchers need always to be aware of how the abstractions might influence the other models with which they interlock in experimental situations. Although abstractive processes often occur in tandem, differentiating them serves to call attention to a kind of abstraction that is especially productive in merging constraints from multiple sources. For instance, in considering the behavior of a physical system, such as a spring, scientists often draw a specific instance, but then reason with it as without specificity as to the number and width of coils. To reason about it as a generic simple harmonic oscillator requires, further, suppressing features of its spring-like structure and behaviors. We call this process abstraction via generic modeling or, simply, generic abstraction. In modelbased reasoning, constraints drawn from different sources need to be understood at a level of generality sufficient for retrieval, integration, and transfer. Further, generic abstraction gives generality to inferences made from the specific models that are constructed. As Berkeley noted, one cannot imagine a triangle-in-general but only some specific instance. However, in considering what it has in common with all triangles, we are able to view the specific triangle as lacking specificity in the lengths of the sides and degrees of the angles. The same concrete representation can be interpreted and understood as generic or specific depending on the demands of the reasoning context. Researchers are able to understand which inferences made from the behavior of a construct apply to blood vessels in vivo and what apply to the specific in vitro construct model. In the experiment within the vascular construct model-system outlined in Figure 1, it is possible to identify inferences that are specific to the flow loop and construct models and inferences made from them that apply to in vivo vascular systems, generally. One is as follows. The hypothesis of the experiment is that conditioning the constructs by first subjecting them to the kinds of forces experienced in vivo will prevent platelet formation (thus reducing the risk of thrombosis). Inferences about the responses of the endothelial cells lining the construct to the forces exerted on them by the fluid flowing in the flow loop are generic to members of a class of cardiovascular systems (to a first-order approximation). But inferences about the strength of the construct based on the collagen scaffold are specific to that model. Another example comes from the physical structure of the network of neurons plated on each dish. The current process for plating neurons onto the MEA calls for breaking apart the connections between the neurons, and placing

754

Nancy J. Nersessian and Christopher Patton

the free-floating cells in a single layer on the bottom of the dish. The cells are then left to re-grow connections to each other. Clearly, the structure of the resulting network has no specific relation to the original cortex. In effect, The Dish is a specific instantiation of a network of neurons that can be used by researchers as and reasoned with as a generic network. In framing The Dish in this manner, it allows the researchers to infer network processing characteristics of a specific network and transfer their findings to other cortical networks. In sum, the kinds of abstractive processes we have been considering provide ways of selectively representing and reasoning with models. Idealization, limiting case, approximation, and generic abstraction provide a means of representing and integrating constraints stemming from different sources into models. To reason correctly by means of the models requires recognizing which features are germane and which are cognitively inert within that problem-solving context. This point requires consideration of how in vitro models represent in vivo phenomena.

4.3 Parallelism and exemplification In vitro models are virtual worlds which bio-engineers design and construct to selectively “parallel ” or “mimic” or “approximate” or “simulate” (to use their words) the in vivo phenomena. In some cases, as with the mechanical forces in the blood vessel and the morphology and alignment of endothelial cells, the phenomena are well-understood. In other cases, the models are the only means they have from which to determine the nature of the in vivo phenomena. They are hypotheses based on incomplete understanding, such as The Dish as a model of learning, and the objective is to develop understanding of the brain or the endothelial cells from the parallel world of the model-system. The physical and computational models are designed and created with the intent of performing simulations with them. Model-systems are sites of systematic experimentation and the nature of the parallelism between in vivo phenomena and their corresponding model-systems is continually evolving, if only in minor ways, as they become better or different kinds of approximations. Thus parallelism is an historical process, both because of the overarching redesign agenda and because of the fact the model-systems are not fast-lived set ups but painstakingly constructed sites of serious investment over considerable time spans. What is the nature of the parallelism or mimicry, that is, of the relationships between the models and what they represent, especially in relation to their status as epistemic artifacts? The goal in creating these virtual worlds is for the models to be of the same kind as the real-world phenomena along particular dimensions, such as first-order processes of blood flow in arteries or type of information processing in the brain. In such respects, the referential relations are built into the design and building of the models. Several accounts of models propose that “similarity” is the basis of the representational relation. Models are similar to and different from what they model along relevant aspects [Cartwright, 1983; Giere, 1988; Nersessian, 1992; 2002]. In the limit, models are isomorphic to relevant

Model-based Reasoning in Interdisciplinary Engineering

755

aspects of the phenomena they model; indeed, if they are successful with the tissue engineered vascular wall model it will be identical to an artery in structure, behavior, and function — that is, itself an artery. In considering how they represent, however, the pertinent word here is “relevant.” Not just any similarities and differences matter. For the model-systems, the relevant aspects are those which exemplify the phenomena being investigated. As introduced by Nelson Goodman in Languages of Art [Goodman, 1968] a representation exemplifies certain features or properties if it “both is and refers to” something which has those feature or properties; that is “[e]xemplification is possession plus reference [p. 53]. One of his examples is that of a tailor’s fabric swatch, which “is a sample of color weave, texture, and pattern; but not of size, shape, or absolute weight or value” [ibid.]. In our cases, the notion of exemplification captures the idea that a model needs to be of the same kind along relevant aspects. The relevant aspects are those which exemplify the objects, entities, or processes under investigation. For instance, the hybrot not only is a model-system used to represent neural processing in vitro, but it is also a system that is, in actuality, doing neural processing. It both refers to selected processing in a brain and is processing in a brain, simultaneously, in the context of the research in Lab D. Catherine Elgin [1996; 2004] has built on Goodman’s notion to address the epistemic problem that science makes extensive use of practices that, if we insist on equating truth and epistemic acceptability, lead clearly to falsehoods, such as limiting case demonstrations, idealizations, curve smoothing, and computational modeling. Yet science accepts the outcomes of such practices as advancing scientific goals, for example, deriving a mathematical representation or predicting future real-world states on the basis of a computational simulation. We can conclude from this either that science is cognitively defective or allow that scientific understanding “is often couched in and conveyed by symbols that are not and do not purport to be true” [Elgin, 2004, p. 116]. She advocates that the epistemic value of modeling, as well as other strategies and symbols (propositional and iconic) used by scientists, lies in their ability to exemplify that is, to “at once refer to and instantiate a feature” [Elgin, 1996, p. 171]. Models thus afford the researcher with epistemic access to selected features of phenomena. As exemplifications, models are devised so as to focus attention on features relevant to epistemic goals. As Elgin points out, physical exemplifications are routinely created for experimental purposes such as when a lump of ore is refined to consist only of iron. The physical abstraction, then, affords epistemic access to the properties of iron in the context of real world experimentation. The bio-engineers we have been studying design and create hybrid exemplifications paralleling the in vivo phenomena they wish to study to the degree of specificity that they believe sufficient or to the degree that lack of knowledge of the phenomena or the nature of the design materials constrains them. The models abstract away irrelevant (or potentially relevant) features, thereby focusing attention on those features salient to the problem-solving context. Understanding what inferences can be made in general about the behavior of endothelial cells in the

756

Nancy J. Nersessian and Christopher Patton

artery or neurons in the brain on the basis of a simulation through a model-system derives not from making inductive generalizations but through understanding how the one model exemplifies selected features of all such phenomena. A properly designed model-system warrants the researchers in pursing where the experimental outcomes of the in vitro world might lead, in lieu of being able to carry these out in the in vivo world. Finally, drawing on the notion of exemplification to explicate how constructed models are selective representations serves as a reminder of the deeply sociocultural nature of representation. As Goodman observed, what a model exemplifies depends on goals, purposes, and context. A paint chip usually exemplifies a color, but might in certain circumstances be taken to exemplify a geometrical shape, such as a rectangle. It exemplifies color within a particular set of social norms surrounding the practice of picking paint for one’s walls or house. In their model-based reasoning practices, the researchers in Lab A and Lab D are drawing on a repertoire of representational practices and the conventions of specific communities surrounding these. 5

CONCLUSION

We have discussed the nature of the model-based reasoning carried out by means of constructing and manipulating physical simulation models in the interdisciplinary context of biomedical engineering. In reasoning processes, models have two faces: mental and physical. That is, inference involves co-constructing and manipulating physical and mental models. The physical models are hybrid designs that merge biological and engineering constraints and represent selective understandings. The mental models, too, interlock biological and engineering concepts and understandings. Problem solving through simulation with model-systems is an epistemic activity that enables inference through selectively creating objects, situations, events, and processes that exemplify those of interest in selective ways. It is also a socio-cultural activity in that, even when considering reasoning activities, the models and the practices of using them are both cognitive and cultural achievements. So, though the model-systems as represented in Figures 1 and 2 convey only a pared-down representations, the model-systems and experiments are to be understood as embedded in rich cognitive-cultural systems distributed in space and time, themselves designed to enable and support such experimentation. Experimentation and inference are conditioned on a fabric of interlocking models — across space, time, people, and artifacts, connecting mental and physical representations. Our meta-analysis has separated some of the threads, but in practice they are intricately woven and inseparable. Devices and model-systems are what socio-cultural studies of science refer to as the “material culture” of the community, but they also function as what cognitive studies of science refer to as “cognitive artifacts” participating in the reasoning, representational, and problem-solving processes of a distributed system. Our point is that within the research of the laboratories, they are both, and it is not possible

Model-based Reasoning in Interdisciplinary Engineering

757

to fathom how they produce knowledge claims by focusing exclusively on one or the other aspect. They are representations of current understandings and thus play a role in model-based reasoning; they are central to social practices related to community membership; they are sites of learning; they provide ties that bind one generation of researchers (around five years) to another; they perform as cultural “ratchets” that enable one generation to build upon the results of the previous, and thus move the problem solving forward. In sum, they are hubs that interlock the various dimensions of the cognitive-cultural fabric in which problem solving takes place in these interdisciplinary engineering research laboratories. ACKNOWLEDGEMENTS We would like to thank other members of our group who have contributed to this research, especially Wendy Newstetter (co-PI), Lisa Osbeck, Elke Kurz-Milcke, Ellie Harmon, and Arvind Venkataramani. We thank the members of Labs A and D for allowing us into their work environment, granting us numerous interviews, and sharing their knowledge with us. We gratefully acknowledge the support of the National Science Foundation ROLE Grants REC0106773 and DRL0411825, and of the Radcliffe Institute for Advanced Study and the National Endowment for the Humanities (Nersessian). BIBLIOGRAPHY [Cartwright, 1983] N. Cartwright. How the Laws of Physics Lie. Oxford: Clarendon Press, 1983. [Elgin, 1996] C. Z. Elgin. Considered Judgment. Princeton: Princeton University Press, 1996. [Elgin, 2004] C. Z. Elgin. “True Enough.” Philosophical Issues 14 (2004): 113-31. [Giere, 1988] R. N. Giere. Explaining Science: A Cognitive Approach. Chicago: University of Chicago Press, 1988. [Nersessian, 1992] N. J. Nersessian. “How Do Scientists Think? Capturing the Dynamics of Conceptual Change in Science.” In Minnesota Studies in the Philosophy of Science, edited by R. Giere, 3-45. Minneapolis: University of Minnesota Press, 1992. [Nersessian, 2002] N. J. Nersessian. “The Cognitive Basis of Model-Based Reasoning in Science.” In The Cognitive Basis of Science, edited by P. Carruthers, S. Stich and M. Siegal, 133-53. Cambridge: Cambridge University Press, 2002. [Nersessian, 2005] N. J. Nersessian. “Interpreting Scientific and Engineering Practices: Integrating the Cognitive, Social, and Cultural Dimensions.” In Scientific and Technological Thinking, edited by M. Gorman, R. D. Tweney, D. Gooding and A. Kincannon, 17-56. Hillsdale, N. J.: Lawrence Erlbaum, 2005. [Nersessian, 2008] N. J. Nersessian. Creating Scientific Concepts. Cambridge, MA: MIT Press, 2008.

This page intentionally left blank

SCALE MODELLING IN ENGINEERING: FROUDE’S CASE Sjoerd D. Zwart

1

INTRODUCTION

Engineers have employed scaling practices since time immemorial, long before scaling methods based on dimensionless numbers were introduced. When building sacred edifices, palaces, vessels and the machinery of war in ever increasing sizes, engineers were compelled to discover how to reliably scale artifacts up or down. An obvious theoretical underpinning of these scaling practices was the Euclidean theory of similar figures and solids contained in Book VI of the Elements. Yet engineers were soon to discover that Euclidean geometry was insufficient when it came to reliably scaling artifacts as it does not consider forces. If a boat is scaled up ten times in the three Euclidean dimensions, then the resulting artifact will be much too weak for practical purposes. Galileo was the first to develop a theory about scaling forces and in the nineteenth century, it was William Froude who developed a theory about scale modelling. Consequently, as theoretical dimensional analysis did not develop until the start of the twentieth century, it is too simple to equate scale modelling with dimensional analysis. This chapter addresses questions about the foundations of scale modelling and related philosophical problems. Let me explain why these questions cannot be considered to have yet been solved. In Froude’s time, the theory of scale modelling was known, in good Euclidean tradition, as the theory of similarity or similitude. Since Froude is the founding father of ‘scientific’ scaling, his endeavours may constitute a good place to start our search for the scientific basis of scale modelling. Traditionally, this basis is taken to be dimensional analysis, but nowhere in his writings does Froude mention or carry out any analysis of dimensions. So in our quest to find the origins of scale modelling we must dig deeper and address the following questions: What is the foundation of Froude’s scaling method if he did not use dimensionless numbers? What is the status of the model laws? Do they have empirical content or are they analytic? And, finally, as scale models are often set apart as different from other models, we will discuss the question whether this special status applies to Froude’s scale models as well. My aim in answering these questions is, in the first place, to shed light on the philosophical foundation of scaling in engineering without reference to dimensional analysis. Furthermore, the outcomes might also have consequences for research Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. © 2009 Elsevier BV. All rights reserved.

760

Sjoerd D. Zwart

into the foundations of scaling methodologies in the modern engineering sciences, which indeed are undoubtedly based on dimensional analysis. Scaling methodologies exhibit an important aspect of engineering knowledge, as they show how engineers can bypass intractable theoretical problems and get their job done by applying practical engineering knowledge. Thus, ultimately, the answers to these questions will serve to show that the philosophy of engineering sciences is an interesting field of research for its own sake. To answer the central questions about the philosophical foundations of scale modelling, we need also to discuss parts of its history and practice; we will do so by digging into one specific example: scale modelling in naval architecture. The outline of this chapter is as follows. First we shall discuss the history of scale modelling, in general, and of scale modelling in naval architecture in particular. Then we will have a thorough look at the way scale models are used today in naval architecture and explain the theory and practice of Froude’s extrapolation method as applied to models in towing tank laboratories. Equipped with this practical laboratory knowledge, we will finally turn to more philosophical questions. First we will consider the origin and reliability of the knowledge produced by scale models and then we will turn to the questions about the foundations of the knowledge produced. At that point we shall consider the special character of scale models and the role and character of similarity laws. As it turns out, it is the empirical assumption that a physical property is ratio scale measurable which guarantees the application of scaling methodologies and superposition of the physical quantities involved. We will close with a summary of the conclusions and predictions for future research.

2

2.1

SOME HIGHLIGHTS IN THE HISTORY OF SCALING

Mechanics

Scaling mechanical devices up and down has been part and parcel of the daily life of engineering from time immemorial. In conjunction with war, commerce and increasing wealth, constructors were required to build and design ever larger artifacts. Common sense tells one that increasing or decreasing the dimensions of a mechanical device will accordingly increase or decrease its capacities. Yet, simply geometrically scaling a working artifact does not always lead to the desired result; one way to proceed, in such a situation, is by systematically varying the different parts of the intended machine and trying to arrive at a result that is as efficient as possible. Let us, for a first historic example of engineering scaling, turn to Alexandrian mechanicians, as they were called, who systematized Greek catapult building. Philo of Byzantium (280–220 BC), also known as Philo Mechanicus, and Hero of Alexandria (10–70 AD) well-known for his experimental ingenuity, both wrote their own Belopoeica, or treatise on artillery. According to Cohen and Drabkin

Scale Modelling in Engineering

761

[1948, p. 318] these treatises are interesting because of the way in which they apply ‘empirical mathematical formulae to the problems of ballistics.’ In his Belopoeica 3, Philo writes (quoted from [Cohen and Drabkin, 1948, p. 318– 319]): Now some of the ancients discovered that the diameter of the bore [receiving the skeins] was the basic element, principle, and measurement in the construction of artillery. But it was necessary to determine this diameter not accidentally or haphazardly but by some definite method by which one could also determine the proportionate measurement for all magnitudes of the instrument. But this could not be done except by increasing or decreasing the diameter of the bore and testing the result. And the ancients did not succeed in determining this magnitude by test, because their trials were [. . . ] conducted [. . . ] merely in connection with the required performance. But the engineers who came later, [. . . ] reduced the principle of construction to a single basic element, viz., the diameter of the circle that receives the twisted skeins. [Philo adds also the reason for the heuristic approach for machines of war design] For it is not possible to arrive at a complete solution of the problems involved merely by reason and by the methods of mechanics; many discoveries can, in fact, be made only as a result of trial. Hacker, reflecting on this passage, describes the way in which the engineers scaled their various machines of war and notices how the classic designers shared their experimental knowledge with those who would need it for the construction of machines of war for the battlefield, [1968, p. 49]. For instance, in the case of catapults, the weight of the stones determined the diameter of the hole holding √ 3 100m where d is measured the skeins by means of the following formula; d = 11 10 in digits (1 digit ≈ 1.9 cm) and m is measured in minas (1 mina ≈ 1 pound). All other parts of the catapult were measured according to the size of the skein hole diameter. But also when arrows were shot, their size determined the skein hole diameter, which was one-ninth of the length of the arrow. The computational outcomes were tabulated and were used by Greek army engineers to construct weapons of war with the desired dimensions, when out at war. The original quotations of Philo and the paraphrases of Hacker clearly show the relevance to engineers of the systematic-scaling method for machines of war. Apart from the need for this method the previous quotations also illustrate the method of what Walter Vincenti calls parameter variation, [1990, p. 139]. According to Vincenti, parameter variation is ‘the procedure of repeatedly determining the performance of some material, process, or device while systematically varying the parameters that define the object of interest or its conditions of operation.’ Again Philo’s quotation clearly shows that twenty-two thousand years ago engineers were already familiar with parameter variation and even considered it necessary to produce reliable engineering knowledge.

762

Sjoerd D. Zwart

The second historical example concerns the introduction to the first day of Galileo’s Dialogue Concerning two New Sciences, [1638]. As is well known, the first new science is about ‘the resistance which solid bodies offer to fracture’, and the second new science is about uniform and naturally accelerated motion. Right at the start, Galileo informs us that his curious imaginary friend Sagredo frequents the shipyard of the Venetian arsenal ‘for the mere pleasure of observing the work of those who, on account of their superiority over other artisans, we call “first rank men”.’ But Sagredo sometimes mistrusts the explanations of the artisans. He says: ‘And notwithstanding the fact that what the old man told us a little while ago is proverbial and commonly accepted, yet it seemed to me altogether false’. Galileo through Salviato, replies SALV: You refer, perhaps, to that last remark of his when we asked the reason why they employed stocks, scaffolding and bracing of larger dimensions for launching a big vessel than they do for a small one; and he answered that they did this in order to avoid the danger of the ship parting under its own heavy weight, a danger to which small boats are not subject? Sagredo agrees and maintains that this ‘current opinion’ in engineering practice must be wrong since it does not follow Euclidean geometry when arguing from small to large machines. He explains: SAGR: Now, since mechanics has its foundation in geometry, where mere size cuts no figure [is of no importance], I do not see that the properties of circles, triangles, cylinders, cones and other solid figures will change with their size. If, therefore, a large machine be constructed in such a way that its parts bear to one another the same ratio as in a smaller one, and if the smaller is sufficiently strong for the purpose for which it was designed, I do not see why the larger also should not be able to withstand any severe and destructive tests to which it may be subjected. In the pages that follow Salviato explains why this argument from geometry fails in the context of mechanical artifacts, leading to the often cited words given on page 130 in the Drake translation: SALV: From what has already been demonstrated, you can plainly see the impossibility of increasing the size of structures to vast dimensions either in art or in nature; likewise the impossibility of building ships, palaces, or temples of enormous size in such a way that their oars, yards, beams, iron-bolts, and, in short, all their other parts will hold together; nor can nature produce trees of extraordinary size because the branches would break down under their own weight; so also it would be impossible to build up the bony structures of men, horses, or other

Scale Modelling in Engineering

763

animals so as to hold together and perform their normal functions if these animals were to be increased enormously in height; for this increase in height can be accomplished only by employing a material which is harder and stronger than usual, or by enlarging the size of the bones, thus changing their shape until the form and appearance of the animals suggest a monstrosity. This is where the theoretical underpinning of what engineers had known for a long time begins, viz. that geometrical scaling is insufficient as a heuristic to increase or decrease mechanical artifacts. Shipbuilders and architects must have known for at least two thousand years that just geometrically scaling of existing devices does not lead to reliable new artifacts. As the examples so far given come from history, the impression may quickly arise that engineers’ use of scale models is mainly a thing of the past, especially if one takes modern computational means into consideration. Nothing could be farther from the truth, though. Just to highlight an example, consider today’s research into the earthquake resistance of complicated building structures. Seismic researchers put scaled-down models on high-tech earthquake shaking tables to find out how designed structures cope with the ground-moving circumstances of earthquakes (see, e.g., [Towhata et al., 2004]). Beijing’s new China Central Television Headquarters (CCTV and TVCC), a Rem Koolhaas and Ole Scheeren (OMA) design, provides an interesting instance. To confirm the seismic performance of this ultramodern structure, a 64-ton copper model was placed on top of a shaking table and an entire team of seismic and structural engineers registered its earthquake performances. Shaking tables are to be found all over the world.1 We may therefore safely conclude that scaling methodologies have been used from the start of recorded history and are still in use today; and if there will be an age in which computer modelling renders all physical modelling obsolete, then that age has not yet begun. Now that we have seen some mechanical scaling examples let us, by way of introduction to Froude’s scaling method, consider some highlights in the history of fluid dynamics.

2.2

Fluid dynamics

Let us start our account of the highlights in the history of fluid dynamics by introducing some terminology. Fluid mechanics is the classical branch of continuum mechanics that studies the relation between forces and movements in fluids, which may be liquids or gases. The study of moving fluid is referred to as fluid dynamics, which is divided into aerodynamics and hydrodynamics depending on whether the relationship is to gases or fluids. Traditionally, the term hydraulics is reserved for more practical engineering knowledge linked to flow in pipes, pumps, turbines, dam design, rivers, channel behaviour, coastal protection and erosion etc. The 1 See [Renda et al., 2004]. Even the most high-tech institutions in the space industry, such as ESTEC, have impressive shaking tables [Hydra].

764

Sjoerd D. Zwart

term stems from the Greek hydraulos which means water organ. Hydrodynamics is generally considered to supply the theoretical background to hydraulics. Let us first have a look at some historical examples in hydraulics, before turning to some highlights in the history of hydrodynamics. 2.2.1

Experimental hydraulics

Experimenting with scaled-down vessels has a long history, which is not always closely connected to the theory of hydrodynamics. It was perhaps the same curiosity that made Sagredo suspicious about the explanation given by the artisans in the Venetian shipyard that left Benjamin Franklin (1706–1790) unsatisfied with the boatman’s explanation about the slow advancement of a track-schuyt in Holland, [Franklin, 1769]. The boatmen explained that the reduced depth of the canal, and therefore an increase in the resistance was the cause of the prolonged traveling time. In a letter, dated May 10, 1768, Franklin reported to John Pringle on his forerunner of modern towing-tank experiments. Franklin had a trough, which was fourteen feet long, six inches deep and wide, and a model boat that was six inches long, two and a quarter inches wide and one and a quarter inches deep. The boat was pulled by the weight of a shilling connected to the boat by a silk thread that passed over a brass pulley at the shilling end. Since Franklin did not have a watch at his disposal, he counted as fast as he could to measure the duration of the model’s journey. He repeated eight experiments three times, one series with the water one and a half inches deep, one series in a depth of two inches, and one series where the water was four and a half inches deep. He tabulated the data, which showed that in a depth of one and a half inches of water, the boat took, on average, a hundred and one ticks, in a depth of two inches it took eighty-nine ticks and in four and a half inches it took only seventy-nine ticks to pass through the trough. Thus, Franklin showed that boats experience more resistance in shallow than in deep waters. Some decades later, Charles Bossut (1730–1814), d’Alembert and the Marquis de Condorcet reached the same conclusion. In 1775, the French government ordered to find out about the resistance that bodies encounter in moving water. To this end, they built a towing tank that was about a hundred feet long, fifty feet wide and seven feet deep. They also concluded that the skin-friction factor was negligible when compared with that of the inertia of a ship. They furthermore considered the friction along the sides and bottom of the boat to be very small and not separable from the resistance of the air. At the end of the eighteenth century (1796–1798), Colonel Mark Beaufoy (1764–1827) carried out towing tests that gave more reliable results. These results were recorded in his Nautical and Hydraulic Experiments, [1834]. Besides doing tests regarding length and surface area, Beaufoy also carried out tests to find out what were the specific effects of the front and back shapes. He found out that the friction experienced did not increase with the square of the increase in velocity but rather with the power equal to a number between 1.71 and 1.81.

Scale Modelling in Engineering

765

The most important period in the history of towing-tank experiments and the underlying theory is the second half of the nineteenth century. At that time, William Froude (1810–1879) carried out his well-known experiments, using scale models such as the Raven and Swan, and later the H.M.S Greyhound, to develop a methodology to predict the resistance of seagoing vessels. In 1852, Ferdinand Reech (1805–1884) formulated what is now known as Froude’s model law, [1852]. ‘One would only have to build one or the other model with linear dimensions√l times as great and multiply all the observed velocities by the quantity u = l for the new system to function similarly to the earlier one, giving rise to forces of which the static intensities would all be augmented in proportion to the cube of the ratio of the linear dimensions’, [Rouse and Ince, 1957, p. 155]. Froude formulated the ‘law of similitude’ thus: ‘the diagram . . . will express equally the resistance of a ship similar to it, but of (n) times the dimension, at various successive velocities, if√in applying the diagram to the case of the ship we interpret all the velocities as ( n) times, and the corresponding resistances as n3 times as great as the diagram,’ [Froude, 1868]. Or, as an equation in which L, V, D are length, velocity and drag and m refers to model and p to prototype: √ (1) (Lp = nLm ∧ Vp = nVm ) → Dp = n3 Dm In the Appendix to his correspondence with The Admiralty [1868], Froude explains how he arrived at his model law. Yet, according to T. Wright, we will probably never trace the original chain of reasoning that led Froude to his final result, [Wright, 1992, p. 244]. Froude seems to have been inspired by the idea that if the model and prototype sail with ‘corresponding velocities’, then the wave patterns around the model and the prototype will have similar forms. Consequently, as the increase in volume proceeds according to the cube of the dimension of length, the work that must be done to create and displace the waves also proceeds by the cube. In the referred to correspondence, Froude formulated his claim straightforwardly, without mentioning the impact of the viscosity of water. As soon as Froude had the Torquay towing tank at his disposal, he immediately started to investigate how the velocity, surface area and configuration of submerged planks influenced their resistance in water. Two years after Froude had published the results of his experiments on surface friction caused by planes, [1872], he changed the application of his ‘law of comparison.’ He distinguished between wave-making resistance and resistance due to other causes of which skin friction was the most important one, and he restricted his ‘law of comparison’ to the wave-making aspect of resistance. This method of extrapolating the scale model data to the prototype is still used today when towing-tank experiments are carried out to predict the resistance of newly designed vessels. As Rouse and Ince succinctly observe, history does not always follow a rational course. F. Reech was the first to formulate Froude’s law of similarity, and nowhere in Froude’s writings can one find the dimensional number that carries his name. Moreover, apart from naval architects hardly anyone would come up

766

Sjoerd D. Zwart

with Froude’s name if asked about the extrapolation method used in today’s towing tank experiments to predict the resistance of new ship hulls [Rouse and Ince, 1957, p. 187]. 2.2.2

Hydrodynamics

At least until the beginning of the twentieth century, experimental hydraulics and theoretical hydrodynamics progressed in relative mutual isolation. This relative isolation is nicely illustrated by the absence, for instance, of the term ‘Froude’ in Neményi’s historical sketch of the main concepts and ideas behind fluid dynamics, [1975]. Rouse and Ince attribute the lack of collaboration between both fields to the absence of a common goal or common problem.2 The purpose of considering the theoretical background of hydraulics is to give some historical context for Newton’s drag law, Bernoulli’s theorem and the NavierStokes equations, which will be referred to in the next sections. Moreover, it shows how practical engineering work such as that of Froude (and also, as we shall see, of Reynolds), although carried out without any link with the underlying theory, ultimately came to be covered by that theory, namely in the form of the NavierStokes equations. Although the number of handbooks on the history of fluid dynamics is not overwhelmingly large, the outlines of this history are well enough documented. For our purposes, we only need to hint at some of its highlights. Two of them are to be found in Newton’s Principia, [1687], Book II, which is about the circular motion of fluids. Newton wanted to prove that Descartes was wrong to assume that all space was filled with matter and that he was furthermore wrong about the planets moving through ‘filled space’ without friction. He therefore embarked upon his investigations into fluid resistance.3 The first highlight is generally accepted to be the start involving the first two propositions, of Section VII, which is about ‘the motion of fluids, and the resistance made to projected bodies.’ The two propositions consider the kinematic and dynamic similarity between two systems with the same respective number of particles. Proposition 32, the first one, addresses the kinematic question of similar motions in systems with the same numbers of particles. First Newton assumes that the two systems are similar in terms of their geometry, density and motion, and that, in modern terms, its collisions are elastic. Newton then claims ‘that the particles of those systems will continue to move among themselves with like motions and in proportional times.’ In the text accompanying the proposition, Newton explains the proposition on the basis of the notions of ‘similar and proportional parts of similar figures.’ The systematic explanation of this kinematic similarity will be the topic of section 3.3. The second proposition of Newton we will consider addresses the dynamic question of the similar motions in the systems of bodies just mentioned. Newton writes: 2 Ibid 3 Ibid

p. 193. p. 83. In [1932, Chap 13], H.J.E. Beth observes this same reactive nature in Book II.

Scale Modelling in Engineering

767

Proposition 33, theorem 27 The same things being supposed, I say, that the greater parts of the systems are resisted in a ratio compounded of the squared ratio of their velocities, and the squared ratio of their diameters, and the simple ratio of the density of the parts of the systems. This quotation turns out to describe exactly the force scale factor (9) used in modern systematic explanations of scaling methodologies. That is the reason why equation (10) is sometimes called Newton’s model law . Moreover, Newton derives a second corollary from Proposition 33 where he already observes that the drag of an object immersed in a fluid varies directly as the square of its velocity. It reads: ‘. . . in the same fluid a projected body that moves swiftly meets a resistance that is as the square of its velocity, nearly.’ The second highlight of the history of fluid dynamics I want to mention concerns the way Newton deals with the viscosity of fluids. Right at the start of section ix, in the hypothesis, Newton formulates his conceptualisation of fluids as continuous media. It is the first adequate formulation of the fundamental property of viscous parallel flows to be found in the literature, namely that the shearing stress is proportional to the velocity gradient. Newton’s hypothesis reads: ‘The resistance arising from the want of lubricity in the parts of a fluid is, other things being equal, proportional to the velocity with which the parts of the fluid are separated from one another.’ Today’s theory about internal fluid friction is based on this description, or rather Stokes’ proper generalization of it. Not all of Newton’s hydrodynamics can, however, be trusted. The reverse might be said to be closer to the truth. Despite the prestige, many of Newton’s claims about fluids do not hold, such as, for instance, theorem 39, which follows in the wake of the hypothesis just quoted [Truesdell, 1953]. Newton’s hypotheses take a prominent place in the history of fluid dynamics as they form the basis to what is now known as the distinction between Newtonian and non-Newtonian fluids. Intuitively, a fluid is Newtonian if it continues to exhibit its ‘fluid properties’ no matter how hard it is stirred and if, for instance, it does not show ‘holes’. More formally, a fluid is Newtonian if at each point the stress is linearly proportional to the strain rate where the stress amounts to applied force per area and the strain rate is the speed at which the deformation takes place. With a Newtonian fluid, the viscosity is the ratio of its shear stress and strain rate. A fluid is more viscous than another if in the case of applied shear stress the rate of deformation is lower than with the other fluid. Water and oil like most other common liquids are Newtonian, just as are all gases. Finally, it is interesting to observe that Newton also introduced the reciprocity principle, which is applied to many modern scaling situations. The principle is mentioned at the start of the Proposition 34 or theorem 28 proof. It reads . . . since the action of the medium upon the body is the same . . . whether the body move in a quiescent medium, or whether the particles of the

768

Sjoerd D. Zwart

medium impinge with the same velocity upon the quiescent body, let us consider the body as if it were quiescent, and see with what force it would be impelled by the moving medium. Newton cannot be said to have developed the similitude theory used in contemporary scale modelling. Nevertheless, the first two propositions mentioned do justify the claim that Newton was the first person in recorded history to contemplate the problem of similitude in fluid dynamics. About half a century after the publication of the Principia, Daniel Bernoulli (1700–1782) published his Hydrodynamica, [1738]. It is worth mentioning not only because he coined the term ‘hydrodynamics’ but also because he described in natural language what was later to become known as ‘Bernoulli’s theorem’ or ‘Bernoulli’s equation.’ Bernoulli based his work on the principle of the ‘conservation of live forces’ to which he also referred when he described the ‘equality between the actual descent and the potential ascent’ of a fluid. From our perspective, the theorem alludes to the conservation of energy, at a time when kinetic energy had only ever been referred to in Leibniz’ notion of ‘live force’, as the product of mass and the square of the velocity. Bernoulli’s theorem asserts that for all fluid points along a streamline the sum of all forms of mechanical energies remains constant. Seventeen years after Hydrodynamica, Leonard Euler (1707–1783) published his Principes Généraux du Mouvement des Fluides in which he produced the simple formula nowadays associated with Bernoulli’s theorem: 1 p + ρv 2 = constant 2

(2)

In this equation p, ρ and v stand for the pressure, density and velocity of the fluid. Formula (2) is well-known today as it is taught in every first-year-degree course in hydrodynamics. Euler derived equation (2) from his famous differential equations for frictionless fluids (3). With frictionless flows, where we can ignore the stresses in the fluid, the gravity force per unit volume (g) plus the pressure force per unit volume (p) divided by the density of the fluid (ρ) equals the total acceleration in the control volume. ∂u ∂u ∂u ∂u 1 ∂p = +u +v +w ρ ∂x ∂t ∂x ∂y ∂z ∂v ∂v ∂v ∂v 1 ∂p = +u +v +w gy − ρ ∂y ∂t ∂x ∂y ∂z ∂w ∂w ∂w ∂w 1 ∂p gz − = +u +v +w ρ ∂z ∂t ∂x ∂y ∂z

gx − (3)

In this equation u, v, w are the three components of the velocity vector field V (x, y, z, t) and the acceleration on the right-hand side is the derivative of this dV , which results from the application of the chain rule. Bernoulli velocity field dt

Scale Modelling in Engineering

769

and Euler modelled fluids as inviscid substances in which no stresses are caused by viscosity. The last highlight in our present brief overview is, of course, the development of the Navier-Stokes equations. This commenced with the work of Louis M.H. Navier (1785–1836) who approached the motion of a fluid in a similar way to Euler. Euler, however, had dismissed the friction between the smallest fluid particles. Navier took as his principle the fact that through ‘the effect of the movement of a fluid the repulsive actions of the molecules are augmented or diminished by an amount proportional to the velocity with which the molecules approach or recede from each other’, [Navier, 1822]. This assumption led Navier to the extra element ε in the Euler equation, which depended on molecular spacing:

(4)

« „ 2 « ∂ u ∂2u ∂2u ∂u ∂u ∂u ∂u −ε + + +u +v +w ∂t ∂x ∂y ∂z ∂x2 ∂y 2 ∂z 2 « „ 2 « „ 2 ∂ v ∂ v ∂2v ∂v ∂v ∂v ∂v −ε + + =ρ +u +v +w ∂t ∂x ∂y ∂z ∂x2 ∂y 2 ∂z 2 « „ 2 « „ 2 ∂ w ∂ w ∂2w ∂w ∂w ∂w ∂w −ε + + =ρ +u +v +w ∂t ∂x ∂y ∂z ∂x2 ∂y 2 ∂z 2

ρgx −

∂p =ρ ∂x

ρgy −

∂p ∂y

ρgz −

∂p ∂z

„

Two decades later George G. Stokes (1819–1903) showed that the ε in the Navier equations represents the effect of internal friction, [1845]. The Navier-Stokes equations like (4) describe the dynamics of incompressible Newtonian fluids. For our purposes, it is interesting to observe that if (4) is reformulated in dimensionless terms it displays both the Froude and the Reynolds number. (5)

1 ∂p − = ∂x (Fr)2

„

∂u ∂u ∂u ∂u + u + v + w ∂t ∂x ∂y  ∂z 

« −

1 Re

„

∂ 2 u ∂ 2 u ∂ 2 u + + ∂x2 ∂y 2 ∂z 2

«

Equation (5) shows the dimensionless Navier-Stokes equation for just the x dimension; the primed variables are dimensionless and a comparison between equations (4) and (5) suggests that where the Froude number is related to gravity, the Reynolds number is related to the viscosity of the fluid. Further information about the relation between equations (4) and (5) can be found in [Szucs, 1980]. It is interesting to conclude that the Navier-Stokes equations very naturally explain the dimensionless numbers of Froude and Reynolds, even though the equation and the numbers were developed in relative mutual isolation. 3

SCALE MODELLING BASED ON PHYSICAL SIMILARITY

The previous sections showed that scaling is an important aspect of many engineering disciplines. In the present section, we will concentrate on a systematic introduction to the theory underlying the scale models used in many towing-tank

770

Sjoerd D. Zwart

laboratories. Before going into detail, we will first concentrate on the relation between direct and indirect research.

3.1 Direct and indirect research Research is conducted amid a web of various constraints and questions such as: What data do we have? What do we know about the subject? What is the kind of answer we are looking for? What improvements do we seek? Engineers have two ways of tackling these difficulties. Either they can do direct research, experimentation or prototype research or they can engage in indirect research. The latter is often model based, incorporating mathematical or scale models. Even prototype research is often so expensive or so demanding that engineers prefer to revert to scale models. 3.1.1

Direct research

In engineering contexts, direct research amounts to artifact or process research that is based on real scale experiments. Its chief advantage lies in the reliability of the data when compared with indirect research. With direct research the object to be investigated is immediately at one’s disposal and can, often, be manipulated at will in a laboratory setting. In many areas of engineering research this advantage is often difficult to achieve. Experimentation involves the systematic manipulation of the values of independent variables and parameters. It may also possibly require many artifacts that are not always available. Examples of such situations include: investigating the behaviour of new designs, or artifacts being too large or too expensive to be reproduced for experimentation in large quantities. Consider, for instance, a completely new design of a specific airplane, ship or rocket. Evidently, in such a case direct research into the characteristics of the new artifact as a whole is impossible. Furthermore, if the artifacts are available in the numbers needed, then the costs involved may become insurmountable so that direct research may become practically impossible. Another complication attached to doing direct research on an artifact produced in series relates to the matter of the object tokens chosen for the experiment being representative of the artifact type. If the sample is biassed, we cannot be certain that the findings will also hold for the other items in the series. A final problem of direct research is that the objects of study may be too large for the laboratory used for experimentation. That is definitely something that holds true, almost by definition, for the largest skyscraper in the world. The value of direct research resides in the fact that it is almost the only way to validate the results of indirect research. Despite all the problems attached to direct research, the results are usually more reliable than those obtained from indirect research. On the rare occasions when direct research is possible, it has the important function of calibrating the results of indirect research. When it comes to the operation or functioning of the artifact in question, engineers usually distinguish between detailed and global testing, which is often a form of direct research. Crash testing cars or testing the drag and lift of aircrafts are

Scale Modelling in Engineering

771

examples of the latter mode, whereas examples of the former are checking fatigue in paperclips or the effects of different windings set of an electric motor. 3.1.2

Indirect research

When physical objects are used to carry out indirect research, such objects are called scale models. The resulting experimental data have to be scaled up or down to become relevant to the intended prototype. To make the model data usable, one has to deploy model laws to describe the similarity between the model and the prototype. The results of the scale model experiments are extrapolated to the situation of the intended prototype. This extrapolation is often based on physical considerations in combination with a mathematical model. Just as with mathematical models, using physical scale models also demands the necessary simplifications. The outcomes of scale-model experiments therefore also need to be validated, and that can be done by carrying out direct experiments. We will see that this is exactly what Froude did with his H.M.S. Greyhound experiments. Indirect research circumvents many of the direct-research problems mentioned above. Since the intended artifact does not need to exist for indirect research to be done, it is often deployed in the very first stages of design. Even if a new steel construction does not yet exist, many computer programs are available that will help to model the construction and determine all the involved stresses and strains. Another advantage linked to greater manageability is the fact that indirect research is often good condition-wise. Carrying out experiments under favourable conditions is easier with indirect research than with direct research partly because such conditions are more easily controlled. Because of all these advantages, experiments carried out during indirect research are often easier to repeat thus resulting in more reliable data about the model. Whether these model data also result in better knowledge about the final artifact depends, of course, on the quality of the validation of these data. Validation remains the weak point of model research. One important branch of indirect research concerns the use of computer or mathematical models. Computer models have the advantage of not having to scale the object of research down to manageable proportions. They are real scale. Consequently, computer models do not have to cope with ‘scale effects’, such as, for instance, the change in the surface-volume ratio when the size of objects is scaled. Mathematical or computer models are always simplifications of the real-life situation. In the process, though, many details are omitted which is why it is often difficult to ensure that such details will not amount to great inaccuracies in the resulting data. Because of the simplification inherent in computer or mathematical models, the validation of these models is very important. This validation may be carried out using direct research or even scale models. Experiments with scale models are often carried out in relation to problems existing in aerodynamics, hydrodynamics and hydraulics, since problems in these areas are often computationally intractable. To do any scale-model experiments

772

Sjoerd D. Zwart

one first needs a physical construct, the scale model m, which is, in relevant respects, first of all geometrically, similar to prototype p. Besides the model one also needs to have model laws, since the experimental data resulting from scale-model experiments have to be scaled up (or down) to be relevant to the intended prototype.4 These model laws determine the relevant physical quantities and prescribe the conditions of similarity for the model and the prototype. To understand the role of the model laws, let us look at some important examples.

3.2

Geometric similarity

In aerodynamics and hydrodynamics the following three kinds of similarity are taken into account: geometric, kinematic and dynamic. The first concerns the similarity in dimensions and form between the model and the prototype. The second, kinematic similarity, pertains to similarities in the velocities, speeds and directions between corresponding points in the model and prototype domains. And thirdly, dynamic similarity, has to do with the similarity between the forces and force components in the fluid domain and bodies. Let us see how engineers proceed when applying the method of scale models, and exploiting the geometric similarity of forms. We start with the correspondence between model and prototype length, both measured using the same unit of length. The dimensionless ratio between these two lengths is called the linear scale factor αL ; thus Lp = αL Lm Within the boundaries of what is practically achievable, the choice of αL is free. Once αL has been chosen, the scale factors of areas and volumes are fixed as well: (6)

2 Am Ap = αL

3 Vp = αL Vm

Notice, incidentally, that the linear scale factor in the x, y and z directions need not be the same. Examples of these distorted scale models (as they are called) are models of river beds where the horizontal scale factor is often much larger than the vertical factor. Equation (6) enables us to determine the volume of large bodies with a complicated shape. We first construct a geometrical similar scale model using αL ; after that we establish Vm by submerging the scale model and measure 3 the volume of the displaced fluid; and by using αL , we can calculate Vp .

3.3

Kinematic and dynamic similarities

Things become more intricate when we consider the kinematic and the dynamic relations between the prototype and the scale model. A typical aero-hydro dynamic problem concerns the relative motion between fluid and a body. In such cases, 4 These rules have different names: model laws, scaling or similarity laws, criterion or laws of similarity or similitude.

Scale Modelling in Engineering

773

variable velocities cause changing pressures that, in turn, give rise to a change in the drag and lift forces, which are the forces parallel and perpendicular to the velocity vector of the moving object. In real life, these phenomena are very complicated and if we want to gain insight into these matters, simplification is inevitable. We consider simplifications in two respects, namely in relation to ideal and real fluids and with or without a free surface. 3.3.1

Ideal fluid and no free surface

We start with the most ‘simple’ case in which the moving object has no free surface which means to say that it is completely submerged in the fluid, which, in the present case, is assumed to be ideal. The latter means that the fluid is inviscid or very ‘thin’ and has no internal ‘fluid friction’; moreover it is assumed to be incompressible. An ideal fluid is furthermore homogeneous, which means that its density is constant everywhere in time and place and the fluid gravity need not be taken into consideration. Consequently, the dominant forces are the inertial forces and the forces of pressure. Ideally, when an object is totally submerged in an ideal fluid, the inertia and pressure forces must be in equilibrium or must cancel each other out at model and prototype levels. Such dynamic similarity thus implies that the scale factor of inertia equals that of the pressure forces. Let us consider this equality. We start with scaling space, and choose a reasonable linear scale factor αL , which is the ratio of the length of the prototype and the length of the scale model: Lp = αL Lm . Next, we may again freely choose the velocity scale factor , αv , (7)

vp = αv vm

and the density scale factor, αρ : ρp = αρ ρm By choosing αv , we also fix the time scale factor . It is the ratio of the linear scale factor and scale factor of velocity, since distance is velocity times time. The time scale factor therefore becomes αL αv Furthermore, the acceleration factor for the scale model is fixed by the linear scale factor and the time factor, as acceleration is distance per time squared. The 2 acceleration scale factor , αL / (αL /αv ) therefore becomes αv2 αL If we want to find out the forces in the prototype by measuring the forces in the scaled experiment and if, in both cases, the inertia forces are in equilibrium with the pressure forces, then it is a necessary condition for the inertia-force scale factor

774

Sjoerd D. Zwart

to be equal to the pressure-force scale factor. Let us start with the inertial forces in the prototype. From Newton’s law of resultant force we know that FpI = mp ap . Since the mass is density times volume, it follows that the inertial-force scale factor 3 2 αv /αL , which equals is αρ αL (8)

2 αρ αv2 αL

If we turn to the pressure force factor, we first observe that FpP = Pp Ap , the pressure force on the prototype equals the pressure times the area of the prototype. According to Bernoulli’s (or Euler’s) law (2), in which v is the relative velocity between the prototype and the fluid, the pressure along a streamline remains constant. We thus substitute Pp = 12 ρv 2 , and the pressure-force scale factor therefore becomes (9)

2 αρ αv2 αL

As (8) equals (9), we may conclude that the necessary conditions for scale methodology have been met. The inertial-force scale factor equals the scale factor of the pressure forces. This equality enables the equilibrium of pressure and inertial forces in the scale model to be scaled up to the prototype scale. We therefore conclude that (10)

2 Fm Fp = αρ αv2 αL

Note, by the way, that Newton’s Proposition 33 on page 766 is the same as (10), which is why it is called Newton’s model law . Formula (10) fits in neatly too, and is partly confirmed by the drag equation, which states that the drag of an object in a fluid is a function of the density ρ of the fluid, the square of its relative velocity v, and some effective cross-sectional area Aref : (11)

FD =

1 2 ρv Aref CD 2

In this law CD is a typical dimensionless drag coefficient. As we have seen in section 2.2.2 Newton did already observe that the drag of a body immersed in a fluid is proportional to the square of its velocity; the second corollary of Proposition 33 reads: ‘. . . in the same fluid a projected body that moves swiftly meets a resistance that is as the square of its velocity, nearly.’ How does (11) confirm the ratio of forces in (10)? The argument hinges on the observation that Newton’s law holds for any ρ and v, and any size of the same geometry with scaling factor αL . Thus (11) holds for the prototype and the model. Consequently, D FpD Fm = = constant 2 L2 ρp vp2 L2p ρm vm m

Scale Modelling in Engineering

from which it follows that FpD =

775

ρp vp2 L2p D Fm 2 L2 ρm vm m

which clearly equals formula (10). We conclude with the observation that, within the boundaries of engineering practice, it is the absence of constraints upon the sizes of the objects to which Newton’s drag law may be applied that grounds the scaling practices in today’s towing tanks. 3.3.2

Real fluid and no free surface

The first step towards a more realistic situation is to consider real fluids where viscosity is not left out of the picture. Viscid fluids are homogeneous and incompressible fluids where the viscosity is taken into account and gravity is left out. The viscosity of a fluid may be seen as the ‘internal friction’ between parallel fluid laminae with different velocities in a non-turbulent fluid stream. If a sheet of metal moves longitudinally in a fluid at rest, we assume that the lamina closest to the sheet will move at a faster speed than the laminae farther away from the sheet. The viscosity of a fluid is thus the friction between the moving laminae. Moving longitudinally in the fluid, the sheet of metal experiences a tangential force caused by the friction between the various fluid layers parallel to the direction of its movement. One can imagine that this friction is larger in pitch than in water. The next picture show a steady laminar flow of a fluid parallel to a plate parallel to this flow. The vectors u indicate the various velocities of the laminae.

As we observed in section 2.2.2, Newton’s viscosity hypothesis states that the shear stress in a small volume of fluid is linearly proportional to its strain rate. This may be expressed as σ = η ∂u ∂z in which η is the (shear) viscosity of the specific fluid in question. Newton’s hypothesis has been confirmed in the case of many gases and simple fluids, in other words, fluids that do not contain very long molecules. These fluids are Newtonian. Now, the tangential force FT that the area of the volume experiences is the shear stress times the area of one of the sides of that volume:

776

(12)

Sjoerd D. Zwart

FT = η

∂u dA ∂z

The question now reads: What is the scaling factor for the frictional forces in 2 . This makes the model and the prototype? From (12) it follows that it is αη ααLv αL the scaling factor of the frictional forces for the model and the prototype equal to αη αv αL , in which αη is unknown. One way to determine αη is by comparing αη αv αL to the inertial or pressure-force scale factor. Consider, for instance, a case where for the prototype the inertial force is equal to the friction force. These forces 2 = αη αv αL , have to be equal in the model as well. It then follows that αρ αv2 αL from which the frictional-force scale factor trivially follows. αη = αρ αv αL If we rewrite this formula in terms of lengths, velocities, densities and viscosities for models and prototypes, it is easy to see that we arrive at: ρp Vp Lp ηp = ηm ρm Vm Lm Since the left and right-hand sides display dimensionless numbers, crosswise multiplication of the numerator by the denominator again results in dimensionless numbers. ρp Vp Lp ρm Vm Lm = ηm ηp The dimensionless number on either side of the equation sign is called the Reynolds number . Scaling experiments are often carried out without changing the fluid. In such circumstances, the Reynolds number is conveniently expressed using ν = η/ρ, since η and ρ are specific characteristics of the fluid under consideration. The Reynolds number then becomes: (13)

Re =

v.L v.L.ρ = η ν

If we want similar turbulent behaviour in the model and the prototype, we must keep Re the same in both situations. The latter constraint is usually called Reynolds’ model law; it warrants similarity regarding the forces of pressure, inertia and viscosity for both model and prototype.5 If we stay with the same fluid, Reynolds’ model law reduces to:

(14)

v m = vp

Lp Lm

5 The status of Newton’s model law and that of Reynolds will be the subject of the fourth section.

Scale Modelling in Engineering

777

If we want to satisfy Reynolds’ model law and the model is smaller than the prototype, the velocity of the model must exceed that of the prototype. Whereas with ideal fluids the empirical law on drag guarantees freedom to choose the speed of the model and the prototype (as long as (10) is obeyed), this freedom of choice disappears when frictional forces are introduced. For the sake of good testing and valid extrapolations, the Reynolds number must be the same in the model and in the prototype case. And as η and ρ remain the same in both cases, according to (14) the choice of αL and the velocity of the prototype determine the velocity of the model. We will see that the constraints implied by (14) contradict those implied by the antecedent of (1). To make things somewhat more concrete, let us consider an example in which the prototype is 100 m long and moves with a velocity of 10 m/s in water with ν = 10−6 . The Reynolds number in this case is 109 . Suppose we choose αL = 10 which means, for a prototype of 100 m that we will work with a model of 10 m. Since ν is 10−6 , as water remains water, the value of Vm following from the fixed value of Re = 109 , will be 100 m/s. This means that we have to test the scale model at a speed of 100 m/s (= 360 km/u). These speeds are only feasible in wind tunnels and circulation tunnels and involve high temperatures, large forces, high power, costly tests, and so forth. Table 1. Velocity of the model in the Reynolds case

Lp vp ν Re

3.3.3

Model 10 m 100 m/s! ← 10−6 109

Prototype 100 m 10 m/s 10−6 109

Inviscid fluid with a free surface

The second possible step when moving from the ideal-fluid case towards a more realistic test situation is to take the force of gravity into account instead of adding viscosity to the ideal-fluid case. In the case of an inviscid fluid with a free surface we are again dealing with a homogeneous and incompressible fluid with gravity but without viscosity. The force of gravity accounts for the surface waves which form substantial friction for, for instance, moving ships. Thus, in the present case, the dominant force alongside the inertia and pressure forces is gravity. The gravitational force on an object is its mass times the acceleration of gravity for the scale model and the prototype. Consequently, the gravitational-force scaling 3 , times the factor of the gravitational force is the scaling factor of mass, αρ αL scaling factor of the acceleration of gravity, αg : (15)

3 αρ αL αg

778

Sjoerd D. Zwart

Again, if the gravitational force in the prototype is equal to, for instance, the pressure force, the same must hold for the scale model, and therefore the respective 3 2 αg = αρ αv2 αL or scaling factors must be the same: αρ αL αv2 = αg αL . With an argument similar to that used in Reynolds’ case one can easily arrive at the Froude number : v Fn = √ gL which, again, must be the same in both situations in order to create similarity between model and prototype. This is, as we will see, exactly the same constraint as stated in the antecedent of formula (1). The identity of the dimensionless Froude number in the model and prototype situation is called the model law of Froude. If Froude’s law is satisfied then the velocity of the model and the prototype will be as follows:  vm = vp

(16)

Lm Lp

In the interests of good testing and valid extrapolation, the calculations must satisfy Froude’s model law. In contrast to the constraint of satisfying Reynolds’ law, if the model is smaller than the prototype, Froude’s model law implies that the model’s velocity has to be smaller rather than larger than the velocity of prototype. Consequently, in the context of maritime scale modelling, constraints (14) and (16) contradict, and cannot be fulfilled at the same time. In the circumstances of the previous example, formula (16) leaves no free choice for the velocity of the model; it has to be 3.2 m/s, which is feasible for wave basins and towing tanks. If, in these circumstances, we take the gravitational acceleration to be 10 m/s2 , the Froude number Fn will be 0.32. Table 2. Velocity model in the Froude case

Lp vp G Fn

Model 10 m 3.2 m/s! ← 10 m/s2 3.2

Prototype 100 m 10 m/s 10 m/s2 3.2

The method of testing in accordance with Froude’s law starts with the choice of the scaling factor of length αL from which the area and volume scaling factors 2 3 and αL , respectively. From (16) it follows that the velocity follow and are αL √ and time scale factors both equal αL . Naturally, the gravitational acceleration

Scale Modelling in Engineering

779

remains the same and if the fluid remains the same, (15) implies that the forces 3 times the model type forces, in accordance with equation in the prototype are αL (1). Thus far, the relevant scaling factor for viscosity-related forces, has however, been completely left out of the picture. Viscosity-related forces are important in Froude’s extrapolation method which we will examine in the next section. 3.3.4

Froude’s extrapolation method

In the foregoing section we saw that when a partly submerged ship moves in a fluid, it experiences resistance in three respects. Firstly, the pressure forces constitute a form drag; secondly, the viscosity of the fluid imposes a frictional force, which is larger when moving in thick syrup than in water; and thirdly, since the ship is party submerged it raises fluid in the form of waves, which will be less in syrup than in water. The second and third forces relate to Reynolds’ and Froude’s model laws. Additionally, we saw that the scaling constraints for extrapolating the wavemaking resistance contradict those for extrapolating the frictional forces created by viscosity. The former requires the model velocity to be lower than that of the prototype, the latter requires the reverse; the model’s velocity must be larger than that of the prototype. It is not uncommon in engineering practice to have to fulfil constraints that contradict each other. William Froude, however, found a practical solution. With hindsight the principle of Froude’s extrapolation method is not complicated. The main idea is that, regarding geometrically similar scale models and prototypes, the total resistance of a vessel to any reasonable velocity is approximately the superposition of the wave-making resistance and the skin friction, for both model and prototype. The skin friction caused by viscosity can be calculated independently for models and prototypes using formula (11), which is intended for totally submerged bodies. The wave-making resistance of the model then equals its total resistance minus the calculated skin friction. This wave-making resistance can be scaled up to the prototype size using the model law (1) laid down by Reech and Froude. And finally the total resistance of the prototype equals the scaled up wave-making resistance plus the calculated skin friction using (1). The skin-friction equation exhibits a drag or skin-friction coefficient CD . This coefficient, which in the towing tank context is referred to as Frictional Resistance Coefficient CF , has been the subject of much research for many years. Finally, in 1957 at the International Towing Tank Conference it was decided to fix this coefficient as follows: 0.075 CF = 2 10 ( log Re − 2) Since the Reynolds number Re = v.L.ρ η , for a hull of any dimension, the frictional resistance coefficient can be calculated using the velocity and length of the wet area of the hull, and the density and viscosity of the fluid.6 6 Interestingly, William Froude never published the skin friction constants that he used to calculate the skin friction of models or the real vessel, although, of course, these values can be

780

Sjoerd D. Zwart

Thus, for all reasonable velocities of the model and the prototype the way to calculate the total resistance is as follows: 1. Take a model that is geometrically similar to the prototype and let the √ velocity of the prototype be αL times the velocity of the model, in calm water, without any interference from air resistance, and measure its total resistance 2. calculate the skin-frictional resistance of the model using equation (11) 3. subtract the skin-frictional resistance of step 2 from the total resistance of the model measured in step 1 to obtain the model’s wave-making resistance 3 , 4. multiply this latter wave-making or residual resistance of the model by αL so that the result is the estimated wave-making resistance of the prototype

5. again, calculate a skin-frictional resistance using equation (11) but now for the prototype 6. add the scaled wave-making resistance of step 4 to the skin frictional resistance of the prototype given in step 5. 7. This last number will be the predicted resistance of the prototype at the velocity stipulated at step 1. Since this method applies to any reasonable pair of model and prototype velocities, it provides us with the total resistance of the prototype for any velocity. This way of predicting the resistance of ship prototypes by extrapolating the data of scaleddown models is known under the name Froude’s extrapolation method. It is still today’s basic scaling method used in towing tank laboratories all over the world. 4

SCALE MODEL KNOWLEDGE PRODUCTION

In the previous sections, we made two observations. In the first place we considered the virtual absence of any systematic connection between the various historic developments in experimental hydraulics and hydrodynamics, although ultimately the Froude and Reynolds numbers did find their way into the Navier-Stokes equations. In the second place we described the rational reconstruction of scaling in naval architecture and Froude’s extrapolation method. Finally, in the present section, we will turn to certain philosophical foundations and the way in which knowledge is produced by applying scaling methodologies. Philosophers of science have not paid much attention to the specifics of scale modelling. Two papers written by Susan Sterrett form a welcome exception to this deduced from the graphs and tables published in [Froude, 1872], and [Froude, 1874]. In [1888], William’s son Edmund published the values of the skin friction constants he used, and only after the publication [Payne, 1936] did William Froude’s skin friction constants become more widely available.

Scale Modelling in Engineering

781

lack of attention, [Sterrett, 2002; 2006]. Interestingly she focuses on scale models because of her interest in model-based reasoning.7 In [2002], Sterrett ‘examine[s] how scale models are used in making inferences’, and in [2006], she concludes that ‘often it is not more detailed models . . . but rather an insightful use of knowledge at hand to determine which similarity principles are appropriate in allowing us to infer what we do not know from what we are able to observe.’ Much of what will be discussed in the present section is in line with, and is sometimes even inspired by, Sterrett’s papers, which form interesting further reading. In this section, we will consider three fundamental aspects related to drawing conclusions from the manipulations of scale models in naval architecture. These subjects are: the special character of scale models, the role and character of model laws in scale modelling; and finally the question about the source of the empirical information originating in only manipulation of units.

4.1

Special character of scale models

Let us turn to the first subject, the special character of scale models. Sterrett correctly points out that in the literature scale models are sometimes set apart from the other kinds of models used in science. Ludwig Boltzmann [1902] and Mary Hesse [1967] serve as examples. In his well-known entry ‘Model’ in the Encyclopedia Britannica in [1902], Boltzmann clearly distinguishes between scientific and ‘experimental’ models, ‘which present on a small scale a machine that is subsequently to be completed on a larger, so as to afford a trial of its capabilities.’ Boltzmann does not mention ‘scale’ or ‘analogue’ models, but he must have had scale models in his mind, since he adds: ‘Here it must be noted that a mere alteration in dimensions is often sufficient to cause a material alteration in the action since the various capabilities depend in various ways on the linear dimensions.’ For Boltzmann scale models were only of secondary importance; he only dedicates the last two paragraphs of his Britannica entry to them. In her classification of models, apart from ‘the senses of “model” that are more central to the structure of “theoretical science” ’ Hesse also distinguishes between logical models and ‘replica’s and analogue machines’ as representing two extreme senses of the word ‘model’. From these two groups, which she considers to be the most important bases for the application of the term ‘model’ in science, she takes the second to be closest to the ordinary language sense of model. Although Hesse mentions wind tunnels, she does not consider specific engineering uses for models. Interestingly, she does mention, however, that analogue machines are often made in situations where we lack the mathematical laws or specifications for the system modelled, or where the mathematics are so complex that drawing reliable practical conclusions is impossible. This often holds in naval architecture situations from which I chose my examples. 7 The chapter by Nersessian and Patton in this Volume provides an interesting example of research into model-based reasoning.

782

4.1.1

Sjoerd D. Zwart

Same physics

Observing the specific character of scale models is one thing, but philosophically the important task is, of course, to analyse the nature of this specific character. One way in which scale models differ from many other kinds of models is from the point of view that they are not simplified, more abstract or reduced models in any way whatsoever. As far as scale modelling in naval architecture is concerned, the physics around the scale models moving in water is of the same level of complexity as that of the prototype. For instance, theoretically, the phenomenon of the scale model’s drag is in no sense simpler or more perspicuous than the same phenomenon when related to a full-sized ship (Froude explicitly mentions this similarity in [1869]). Froude’s extrapolation method illustrates this equivalence of physical complexity between model and prototype; after all, the same formula is used to calculate the viscous resistance or skin friction of the model and the prototype. The difference is, of course, that producing, measuring and adjusting the experimental evidence for a scaled-down model is much easier. In principle, the physical processes are exactly of the same kind as those of the phenomena modelled and the complexity of the physics is not reduced. It is only because of their size that they can be directly studied in a laboratory and that is what makes them epistemologically more tractable. The identity of the physical complexity of any model and prototype shows itself even more clearly if we consider the relationship between scientific theory and the object of study. Just suppose that handbook H describes the relevant physical theory, including the most important heuristics on how to build seagoing vessels. H will then be seen to prescribe how to build vessels of all sorts and dimensions. However, it will probably not be helpful for designing vessels smaller than two meters. The reason for this lower limit is purely a practical one; smaller seagoing vessels have no use. If, however, the scale models are larger than H’s lower limit, H’s theory will be said to describe the model and the prototype and both will be considered to be of equal complexity. Yet, usually, the aim when constructing most models is to leave as many as possible real world complexities out of the picture. Thus, scale models deviate from scientific models in that they have the same level of complexity as their original. Of course, with extreme scaling factors, the physical mechanisms underlying the ship’s behaviour and that of the model may change. An extreme example of such a change would be if the vessel size were comparable to that of colloid particles so that Brownian motions would become relevant. In such a situation, we could not apply Froude’s scale modelling, but even then, nothing would prevent us from studying the relevant physical properties. 4.1.2

Not intermediating

Considerations similar to those of the previous section underpin Sterrett’s observation that scale models do not fit into the familiar picture of models intermediating ‘between something abstract and something concrete’ as advocated in the work

Scale Modelling in Engineering

783

of Morgan and Morrison [1999].8 In this respect they differ significantly from theoretical models in the sense described by Hesse, because the models may be considered to intermediate between abstract theories and the phenomena. Naturally one might object that in some sense scale models do intermediate between the final prototype and the shipbuilding theory. This only holds superficially because it overlooks an important difference. To illustrate this difference, let us consider the well-known ball-and-stick (or Calotte) model of carbon in chemistry. Normally, such a model is constructed out of wooden or plastic balls held with sticks at tetrahedral angles of approximately 109.5°. For the sake of argument, however, we shall consider a ball-and-stick carbon model to be made of ice rather than wood. The carbon model of ice mediates between burned wood in reality and the molecular theory of matter. The difference between the naval architect’s scale model and the ball-and-stick model is not that one represents a concrete situation in the real world while the other does not, since both are concrete. Neither is the difference that one is scaled while the other is not, because in a sense the ball-and-stick model might be considered to be a scaled model as well. The important difference between the two models is that the theory about the drag of the scale model in inviscid fluid with a free surface is precisely the same theory that describes the drag of the prototype. The Navier-Stokes equations predict the drag of both models, and apart from the sheer number of calculations, in theory both situations could be calculated in the same way. The situation is very different, though, with our ball-and-stick model of carbon. There, the theory about chemical bonds does not apply to the balland-stick model of carbon made of ice, because it is all about carbon in the real world and the model does not contain any carbon molecules. Moreover the theory about carbon bonds tells us that if we heat carbon to a hundred degrees Celsius the bonds will remain intact whereas the bonds in our model of carbon made of ice will not. The above example clearly tells the difference between the scale model and the intermediating model. In the former case the theory-model relation is equal to the theory-prototype relation, whereas in the latter case these relations differ substantially. Moreover, the physics of the ball-and-stick model differs significantly from the physics of real life carbon. The ball-and-stick model therefore fulfils a necessary condition for a model to be able to intermediate between the theory and the phenomena, namely the difference between model and phenomenon in terms of theri physical properties. We saw in the preceding section, however, that the physics of the model and the prototype are the same and that scale modelling thus violates this necessary condition. Consequently, scale models fail to intermediate between the theory and the phenomena.

8 This intermediating function is clearly anticipated by the notion of ‘bridging models’ introduced by Bertels and Nauta, [1969, Ch. 6].

784

4.1.3

Sjoerd D. Zwart

No need for fundamental laws

Besides the observation that scale models involve the same kind of physical phenomena as the prototype, Sterrett gives another interesting argument supporting the view that scale models do not intermediate in the way that theoretical models do and so therefore have a special character. She rightly observes: ‘We need not be in possession of laws or equations that, even in principle, would be sufficient to predict what happens in either the model or the object modeled’, [2002, p. 59]. The ‘laws or equations’ in this quotation refer to the abstract theoretical differential equations that ultimately govern the phenomena under study. As was illustrated in the previous section, neither Froude, nor his colleagues in naval architecture who used scale model methodologies used the Navier-Stokes equations to predict the behaviour of the prototype. Even today, with all the computational power we have at our disposal, it is still too complex to solve these equations for seagoing vessels. In the above sections, we saw that Froude approached the problem by using scientific methodology in the same way that John Smeaton (1724–1792) did before him. He formulated his phenomenological model law, endeavoured to confirm it, and after separating skin friction from wavemaking resistance, he decided to only apply his cube law to the latter. As far as scale modelling in towing tanks is concerned, Sterrett’s point about the NavierStokes equations is unconditionally correct. One might comment that Edme Mariotte (1620–1684) and Robert Boyle (1627– 1691) progressed in the same way as Froude did. They did not need to have knowledge of statistical mechanics to apply their laws to predict the pressure of a gas when its volume changed isothermally. There is a difference, however: Boyle’s law does not have anything to do with models. Moreover, Boyle’s law seems to be more of a general empirical description whereas Froude’s law seems to express just a conditional statement. In section 4.2.4, we will elaborate on the similarities and differences between the character of Boyle’s law and the model law created by Froude. If scale modellers, using only rules of similarity, can manage without knowledge about the underlying physical equations, are these equations to be totally eradicated from modern scaling strategies? Sterrett addresses this question in [2002]. Her answer adheres to the traditional line of thought: scale modelling is based on dimension analysis, which uses theoretical considerations to find the relevant physical quantities for the scaling issue in question.9 Her answer reads: ‘The laws [i.e., theoretical equations, SZ] are used to help determine what conditions yield physically similar situations (with respect to a certain phenomenon).’ . . . [T]hey are not used to predict anything directly, but, rather, to figure out what quantities a certain phenomenon depends upon (i.e., viscosity, density, velocity, temperature, length, and so on), [2002, p. 63]. As we have seen, this is emphatically not the way Froude arrived at his extrapolation method. In the first place, he did not use the dimensionless constant that nowadays bears his name, nor did he state 9 Cf.

[Bridgman, 1973], [Langhaar, 1951] and [Palacios, 1964].

Scale Modelling in Engineering

785

that his work on model scaling was founded on the Navier-Stokes equations. We do not know how Froude determined what were the relevant physical variables in the model experiments; probably, it was his engineering ingenuity that brought him his success, since carrying out even thousands of model experiments did not guarantee insights similar to the ones at which Froude arrived.10 On the matter of the context of discovery, Sterrett correctly describes the modern application of dimensional analysis to various fields of research. If researchers apply the method of dimension analysis, the fundamental equations are only employed to find out what the relevant physical quantities are in the model. Yet, to develop a philosophically balanced analysis of scale modelling we must not forget the context of justification. In Froude’s case, it is the confirmation through the Greyhound experiment in combination with the dimensionless’ version of the Navier-Stokes equations, which increases our trust in Froude’s extrapolation method. We must acknowledge that the theoretical equations, perhaps in a computationally intractable way, can still tell us what will happen in the model circumstances. Thus, the underlying equations play an important role in the context of justification. To conclude, scaling models are special since they feature the same physics as the thing modelled; they are not, like theoretical models, intermediating between theory and the phenomena; and, finally, their application does not require any knowledge of the underlying physical laws, as is borne out by Froude’s scaling practices. In the next section we will analyse the nature of the rules of similarity or the model laws.

4.2

Role and character of similarity laws

Usually, the theoretical foundation of scale modelling is taken to be dimensional analysis.11 Dimensional analysis has developed from some plausible observations about the combination of measurements and mathematics in the exact sciences. Let us discuss the three most important ones. Firstly, we may observe that the fundamental laws of physics do not depend on the units being used.12 In [2002, p. 30], Narens formulates this idea as the principle of dimensional invariance: ‘a numerical relation that expresses a valid physical relationship between physical variables has the same mathematical form no matter what proper measurements are used to measure the physical variables.’ Bernoulli’s equation (2) or Newton’s laws of motion apply, regardless whether we measure the variables according to 10 Allegedly,

John Scott Russell, a naval architect of considerable standing, had carried out twenty thousand model experiments [Cooper, 1872, p. 827]. Despite the abundance of his highquality model experiment data, he failed to arrive at reliable conclusions about the resistance of full-scale vessels. That failure led him believe it was impossible to arrive at reliable predictions about prototype resistance by extrapolating data from model experiments. 11 See, for an interesting presentation of dimensional analysis and its foundations Susan Sterrett’s chapter entitled ‘Similarity and Dimensional Analysis’ in this volume. 12 Not everybody considers this to be trivial; see for instance [Luce et al., 1971, Section 10.10], [Narens, 2002] and [Luce, 1996].

786

Sjoerd D. Zwart

the SI, the CGS, the British Mass system or any other metric system with the same choice of fundamental dimensions. Secondly, we do not have as many basic units as there are physical quantities, and we therefore distinguish between fundamental and derived units. The unit of force in the SI system is the ‘newton’, which reduces to kg.m/s2 . The SI system features seven basic units and we could even manage with less. In dimensional analysis, all derived units are reduced to the basic units of the system. In [2002, p. 30], Narens makes the same point about the possible reduction of all derived units. He continues by formulating a fundamental principle about the basic physical quantities: ‘proper physical measurements of a basic physical quantity are related by multiplications by positive reals.’ According to Narens, dimensional analysis is closely related to the two principles just mentioned. He claims: ‘for the purposes of [his] section we may consider dimensional analysis as being a set of techniques that systematically exploit’ the principle of dimensional invariance and the principle about basic physical quantities. Thirdly, with any equation in physics, the left-hand side must display the same dimension or must have the same unit, as the right-hand side; and for any sum or difference between two terms the dimension of the two must be the same. Equations that fulfil this constraint are called ‘dimensionally homogeneous’ equations. This homogeneity is also relevant to constants. If x = cy and the dimension of quantity x differs from that of y, then c is a dimensional constant. In contrast to pure constants, which are just derived from mathematical manipulations, these dimensional quantities fail to be invariant under scale transformations. The principles of homogeneity and of not-being-invariant under scale transformations have developed into the Π-theorem, which states the following. Any equation in n variables and k fundamental or basic units of measurement correctly describing a physical phenomenon can be rewritten as an equation with k − n dimensionless variables. The Froude number is a well-known example of such a variable. 4.2.1

Empirical implications of the ratio scale assumption

Anyone encountering dimensional analysis for the first time will ask himself or herself how analytical manipulations with units that are chosen by convention, can possibly yield empirical knowledge about the world.13 According to Ellis, dimensional analysis is not without its empirical content, which enters as follows: we use our physical experience when we decide which physical factors have to be taken into account for the phenomenon at stake (factors); we do the same when we establish the set of all relevant dimensional constants (constants); and finally, knowledge enters with the ‘standard forms of the basic numerical laws contained in the dimensional formulae themselves’ (form), [1966, Chap ix, sect. F]. Sterrett adds to these considerations that, using a scale model with the same physics as 13 Despite Quine’s passionate attempts to degrade the analytic-synthetic distinction, every account of dimensional analysis addresses the apparent ‘empirical yield from analytical manipulations’ issue.

Scale Modelling in Engineering

787

the prototype, we also add ‘an actual slice of the world’ to the scale modelling arguments, [2002, p. 64]. We will not go into all the details of this fascinating subject now but will concentrate instead on Froude’s case and refer the interested reader to Sterrett’s chapter on dimensional analysis included in the present volume. In the engineering literature of today, Froude’s scaling and extrapolation methods are justified by dimensional analysis, which is understandable from a systematic perspective. Yet, for a more profound philosophical analysis of Froude’s endeavours this approach is not entirely satisfactory for two reasons. In the first place, it is an anachronism as, Froude did not use dimensionless constants in any of his writings; and secondly, but no less importantly, Froude never made reference to the Navier-Stokes equations whereas in modern dimensional analysis, researchers always refer to the underlying theoretical equations to decide which are the relevant physical quantities. It is therefore inappropriate to found our philosophical analysis of Froude’s scaling practices on dimensional analysis. This is especially true if one considers that older engineering scaling practices are well understood from a more fundamental perspective than that of comprehensive dimensional analysis. The assumption in question is the one that a particular empirical characteristic of the physical objects involved is measurable on a ratio scale. This turns out to be a sufficient guarantee for the possibility of scaling. The claim that a phenomenological property is extensive and measurable on a ratio scale is a statement with considerable empirical content. To arrive at an appropriate understanding of this statement, let us turn to the definition of the ratio structure. I will say in reality a property has a ‘ratio-scale structure’ if all its scale values f meet the following rules of a ratio scale [Carnap, 1995, p. 73]:14 1. Rule of equality: e ∼ d ⇐⇒ f (e) = f (d) 2. Rule of additivity: f (e ⊕ d) = f (e) + f (d) 3. Unit rule: for some specific u: f (u) = 1 The rule of equality states that two objects are in equilibrium or can cancel each other out (e ∼ d) regarding the relevant property if and only if their scale values are the same.15 The equality rule partitions the world into classes of objects with the same scale values. According to the additivity rule, the scale value of the concatenation of two objects in reality (e ⊕ d) equals the sum of the scale values of the two objects individually. In physics, the rule of additivity is also sometimes called the principle of superposition. Regarding Froude’s extrapolation method, it implies that the net result of the wave-making resistance and the skin friction is equal to just the sum of both individual forms of resistance. Finally, the unit rule 14 For simplicity’s sake, I will adopt the naive realist point of view that objects and properties exist in the world outside us. 15 Note that the equilibrium notion fits neatly if the properties are forces, thus making them appropriate for Froude’s extrapolation method. With other extensive quantities the term is used in a more figurative sense.

788

Sjoerd D. Zwart

establishes the unit for the scale. In the physical sciences, there is an abundance of ratio-scale quantities like, for instance: spatial length, force, volume, mass, voltage and resistance. Though the ratio rules may seem trivial, when combined they have important consequences. One of these consequences, which is relevant for our purposes is that they allow for the multiplication of a scale value. It follows from the first two rules that: (17)

e ∼ (d1 ⊕ d2 ⊕ . . . ⊕ dc ) ⇐⇒ f (e) = c.f (di ).

Thus, object e is in equilibrium with c concatenations of objects di , all of which have the same scale value, if and only if the scale value of e is c times the scale value of any di . Formula (17), in its turn, allows for the scalability of equilibriums. What does this mean? The equilibrium between objects with an extensive property, say force, can be transferred or scaled to anywhere else on the scale so that they again achieve an equilibrium. Thus, if weight e is in equilibrium with weights d and d  on a balance, then any triple of E, D and D will be in balance again, provided that f (E)/f (e) = f (D)/f (d) = m and D ∼ D . More formally, we may state that for any e, d, E and D satisfying the constraints mentioned formula (17) implies (18)

f (e) = n.f (d) ↔ f (E) = n.f (D)

since both hold if and only if m.f (e) = n.m.f (d). If we apply the rule of equality to (18), we see that it allows for any equilibrium between e and a number of di ’s to be scalable to an equilibrium between E and D’s. Of course, equation (18) does not state that all these equilibria exist in nature; it only indicates where the scaled equilibrium would be located were nature to furnish us with all the necessary building blocks. Statements asserting that a certain physical property has a ratio scale structure carry considerable empirical content. To see this, it suffices to observe that with important physical quantities scientists have failed to find appropriate rules of additivity. Some of the more the well-known examples are the pitch of tones, the hue of colours, the hardness of matter or phenomenological temperature. Most importantly velocity, which in the age of Froude was still believed to be measurable on a ratio scale, turns out to violate the rule of additivity, and therefore fails to have a ratio structure. We will come to velocity’s violation of the Galilean transformations in the next section. Not all these observations are new. As Carnap stated, rules for setting up measurements ‘are not entirely conventional. Factual knowledge is necessary in order to decide which kinds of conventions can be carried out without coming into conflict with the facts of nature’. [Carnap, 1995, p. 68]. In this paper, it is hypothesized that classical dimension analysis and the Π-theorem owe a large part of their significance to the exploitation of the ratio structure of the scales being used. Future research will hopefully reveal how far this hypothesis goes.

Scale Modelling in Engineering

4.2.2

789

Geometric similarity

Let us return to section 3.2 where we unquestioningly accepted that if we increase the side of a cubic aquarium by a factor α, its volume increases by factor α3 . Now we should ask the following question. Are these square and cube laws empirical insights and do we have to apply induction in the way that we had to with Boyle’s law, or are these mathematical truths, which can be determined without empirical investigations? At first sight they are indeed purely mathematical and therefore analytical conclusions not based on any empirical facts. Yet, can we really be sure that if we increase the three sides of the aquarium from one meter, to ten, its volume increases by a factor of one thousand? Of course, if we look at only the mathematics of the question, we see that if the edge of the old aquarium fits ten times the edge of the new one, the ground surface of the latter consists of ten times ten cubes and that the new aquarium allows for ten layers of a hundred cubes, which makes a thousand cubes. But can we be certain that these kinds of scaling will always coincide with what happens in reality? The answer lies in our belief that Euclidean geometry, which is purely analytical, is approximately true for physical space. Here, we have arrived at the heart of the discussion about pure and physical geometry and conventionalism (see, e.g., Carl Hempel’s [1945]). Without entering into the details of that debate, let us consider some observations, which are important for our present discussion. First, the geometric scaling laws just referred to only hold true in terms of pure Euclidean geometry but fail in, for instance, elliptic or Riemannian geometry. If the radius of a sphere cap is doubled, its area on the sphere is not quadrupled as is the case when we double the radius of a circle on a plane. Secondly, if we apply the geometric scaling laws (6) we assume physical space, for engineering purposes, to be Euclidean. Thirdly, the history of non-Euclidean geometry has taught us that finding out the structure of physical space is not an a priori affair and that we are not forced to use Euclidean geometry. Finally, an enormous amount of practical experience through the ages has confirmed that for almost all engineering and other practical purposes we may assume physical space to be Euclidean. This is the reason why we are so confident that in reality, cubic aquaria behave similarly to cubes in our a priori Euclidean thinking. In the light of all these observations, we can only draw a weak conclusion, but it serves our modest goal. The observations show that the geometric scaling laws (6) definitely introduce a considerable amount of empirical content. However, the very moment we assume that Euclidean geometry holds true for physical space, these laws become analytical truths. Interestingly, the ratio-scale assumption is much weaker than the assumption that physical space is Euclidean. Ratio measurability of lengths, areas and volumes only implies the scalability of an equilibrium of, for instance, volumes at one level of magnitude to another level of magnitude. Clearly, the concatenations and the scaling factors used need not satisfy the laws of Euclidean geometry. For in-

790

Sjoerd D. Zwart

stance, we could measure length on Ellis’ dinches scale, where one dinch equals one inch but where the concatenation occurs by putting rods at right angles instead of π rad, [1966, p. 80]. To take another example, we could take the circumference of a square as the measurement of its area; so that quadrupling its area in the Euclidean sense will only be equivalent to doubling it in the circumference sense. If we decide, however, to accept Euclidean geometry to be true for physical space, then lengths, areas and volumes become measurable on a ratio scale. Yet, if, like in modern physics, we abandon the Euclidean assumption, things become less transparent. The curvature of space will possibly be added to the relevant factors and the same holds for the rules of correspondence, which connect the geometrical entities, such as points and lines, to aspects of physical space. As for Lorentz contraction, the conditions under which length and its unit of measurement are compared become relevant.16 These conditions consider kinematic circumstances which brings us neatly to the next section. 4.2.3

Kinematic and dynamic similarity

When we turn to the kinematic and dynamic similarities of section 3.3, we see that scaling theory develops again within classical mechanics and Euclidean physical space. Take for instance the scaling of velocities. According to the theory of special relativity, scaling up velocities by way of formula (7) will lead us astray. The Galilean transformations only apply to low velocities and not to velocities comparable with the speed of light. However much as we might try, we will never be able to measure objects that go faster than the speed of light (assuming that special relativity is true). Consequently, the velocity-scaling law (7) must be false, as it allows for speeds higher than the speed of light. This observation is not intended to criticize engineering scaling theory because, in practice, relativistic effects are very unlikely to occur in fluid dynamics. It is only to show that the velocity scaling law has empirical content and is not analytic or a priori. If one considers again section 3.3, one may wonder whether Newton’s model law (10) is an analytic or an empirical law. Whereas at first sight the geometrical scaling laws seem analytical, Newton’s model law gives the impression of being empirical. This impression is amplified when one reads Newton’s proof. We learned from section 3.3, however, that regarding inertial forces Newton’s model law follows deductively from his law of resultant force (and the law of Archimedes). Thus, the moment Newton’s second law is accepted, his model law follows analytically without any need for further empirical underpinning. In the same section, we saw that the inertial and the pressure-force scale factors are identical, and therefore that a necessary condition for scaling forces was fulfilled. If they had been different an equilibrium between inertial and pressure forces on a model scale would not have been in equilibrium on a prototype scale.17 We may even strengthen this 16 Even in physics, the discussion about the relation between units of measurement and general relativity lingers on. See, for instance, [Pervushin et al., 2004] 17 Interestingly, Bridgman stresses the necessary character of the results of dimension analysis, as well [1973, p. 52].

Scale Modelling in Engineering

791

observation. From the assumption of ratio-scale measurability, it follows that if, at a certain scale level, inertial and pressure forces are in equilibrium, then linear scaling of both forces will be necessary and sufficient to arrive at an equilibrium of forces somewhere else on the scale. Let us then draw our conclusions. The inertial-force scaling factor (8) is based on scaling constraints originating from the three laws of Newtonian mechanics (including the definition of density), on the principles of Euclidean geometry regarding the volumes used, and on the fact that all variables in these laws are measurable on a ratio scale. Newton’s model law gains all its empirical content from his law of acceleration and does not need any independent proof. Whether this way of reasoning forms the foundation to dimensional analysis as well, is a matter for future research. Although the kinematic and dynamic scaling laws definitely have empirical content, clearly, this content derives deductively from classical mechanics, which assumes ratio-scale measurability, and Euclidean geometry. These laws are therefore analytic, and a priori, in the sense that within the paradigm of classical mechanics they cannot be false. In this respect they differ substantially from phenomenological laws such as the laws of Hooke, Boyle, etc. 4.2.4

Froude’s law of similarity

Let us finally turn to the empirical content of Froude’s law of similarity. We will approach the empirical content issue by arguing as strongly as possible that Froude’s model law has much more of an analytical character than an empirical law, like for instance, Boyle’s law. First we will consider the form of these two laws. As for form, pV = k differs from (1), the latter being an implication, whereas the former is an equality. This syntactical difference, however, is unimportant as we may rewrite Boyle’s law in the following way: (19)

p2 V1 p3 V1 = → = for all V1 , V2 , V3 , p1 , p2 , p3 V2 p1 V3 p1

Equation (19) defuses various other arguments about the question of why Boyle’s law would be different from Froude’s model law. The first one is that Froude’s law is used to predict prototype behaviour whereas Boyle’s law is not. Formula (19), however, shows how Boyle’s law can be used to predict the unknown behaviour of V3 and p3 as well. The second argument is that the constant in Boyle’s law is dimensional, whereas Froude used a dimensionless constant. Again, (19) shows that Boyle’s laws can be formulated without a dimensional constant. From a methodological perspective, still another argument favours the similarity between both laws. Let us recall how Froude verified the reliability of his extrapolation method. First, Froude formulated his model law using model experiments without reference to dimensionless numbers to predict the behaviour of the prototype. Afterwards, he verified his predictions by towing the full scale prototype and measuring its drag. These were the successful H.M.S. Greyhound

792

Sjoerd D. Zwart

experiments, which convinced the world of the viability of his approach. Thus, he applied the scientific method of verifying predictions, which were first deduced from the theory. When establishing his well-known law, Boyle applied the same method. This account of Froude’s model law raises a question concerning the usefulness of testing an analytical law. We will return to this question in a moment. First let us return to equation (19), and rewrite Froude’s model law as follows: (20)

Vp2 Lp Dp = → = Vm2 Lm Dm



Lp Lm

3

for all Vp2 , Vm2 , Lp , Lm , Dp , Dm .

In this equation Dp and Dm are the respective wave-making resistance of the prototype and model. Incidentally, although in his writings Froude did not use dimensional numbers, (20) shows that his, or Reech’s, scaling condition is extensively equivalent to the constraint of Froude-number equality at model and scale level. After all, the antecedent of (20) is equivalent to: Vp2 V2 = m gLp gLm Reformulated as in (19) and (20), Boyle’s and Froude’s laws do have similar mathematical forms. The difference in empirical character comes much more clearly to the fore if we consider the formulae in combination with the ceteris paribus clauses. Most importantly, for formula (19) the object must be the same amount of gas (no additions, subtractions or condensations), and the gas must remain at the same temperature. Another assumption is that pressure and volume are extensive physical quantities, measurable on a ratio scale. Furthermore, the volumes must not become too small or the pressures too high, since then the law will start to deviate from reality. Finally, the gas must be protected from all other external disturbing influences. For Froude’s law, the important ceteris paribus clauses prescribes geometrical similarity between model and prototype. A second factor is the assumption that drag caused by wave-making is an extensive physical quantity, a force, which can be in equilibrium with other forces and can be measured on a ratio scale. Another consideration is that the similarity between wave patterns around the model and around the prototype is due to the antecedent of the law and that wave-making is by far the most important part of Dp . Moreover, the scaling of the fluid volumes is assumed to adhere to Euclidean rules, which means to say that they scale volumes in accordance with the cube law. Finally, it is assumed that the model and prototype must be protected from all other external disturbing influences. If we consider these clauses, the difference between the two phenomenological laws becomes apparent. The ceteris paribus clause of Boyle’s law seems, at least in principle, to admit for the possibility of the consequent to be false, while the antecedent is true. It is only if we add extra laws and theories to the ceteris paribus clause, that this possibility seems to disappear. Yet, if the ceteris paribus clauses

Scale Modelling in Engineering

793

of Froude’s model law apply, the truth of the antecedent implies the truth of the consequent. This is mainly due to the law that volumes and therefore weights increase with the cube of the linear scale factor, which deductively follows from Euclidean geometry and from the notion of density. From this perspective it is understandable that Reech arrived at Froude’s model law, before Froude, without having conducted experiments in the same vein as Boyle. Moreover Froude’s model law is often called ‘Froude’s criterion of similitude,’ which also hints at its small amount of empirical content. Still, the reader might ask, why did Froude conduct his H.M.S. Greyhound experiment if his model law was only a tautology? The answer to this question has at least two parts. Firstly, his model law only becomes analytical if all ceteris paribus clauses hold to be true. Secondly, he carried out his experiments to show that his extrapolation method was reliable and not to verify his model law, which is very difficult to deny, once the ceteris paribus clauses hold true. We will not repeat our argument as to why the assumption of drag being measurable on a ratio scale enabled Reech and Froude to transfer an equilibrium from one level on the scale to another. The argument is similar to the one given in the case of Newton’s model law. The assumption that force is ratio-scale measurable, is of the utmost importance to Froude’s extrapolation method. Recall that Froude separated frictional resistance from the other causes of friction, the most important of which was wave-making and that he extrapolated them very differently. The frictional resistance was calculated using the drag formula, the drag coefficient of which, today, is calculated using the Reynolds number. The wave-making resistance cannot, however, be calculated and is thus scaled using Froude’s model law. It is the rule of additivity, the second of the three axioms used to define ratio scales, which allows for the separation of the two forces on the model level and their superposition on the prototype level. Here again we encounter the importance of the ratio scale assumption to the physical characteristics under consideration. To summarise, we may therefore conclude that from a systematic point of view, Froude’s scaling practices were theoretically based on: • Euclidean geometry (if model and prototype are geometrically similar, the increase in water volumes responsible for the wave-making resistance will proceed according to the cube of the linear scaling factor), • Newtonian mechanics (various scaling factors; possible derivation of scaling law, drag law), and • the empirical assumption of ratio-scale measurability (equilibrium of properties in reality is scalable to any level; superposition of skin fraction and wave-making resistance for model and prototype) We showed that these three factors establish an epistemologically adequate account of Froude’s scaling methodology without reference to dimensionless numbers. Overall, the conclusions about the empirical content of the geometrical, the kinematic, the dynamic, and even Froude’s scaling laws are comparable. With-

794

Sjoerd D. Zwart

out any doubt all these laws definitely have empirical content but not in the way that standard phenomenological laws do. The scaling laws derive all their content deductively from classical mechanics, Euclidean geometry and the ratio-scale measurability assumption. These laws are therefore analytic, and a priori, in the sense that, unlike standard phenomenological laws, they cannot be false within the context of classical mechanics. We will end this section by making some remarks on the relation between scaling and measurement theory.

4.3

Relation to measurement theory

The topics in the present section concerning the character of the criteria of similitude are connected to at least three research areas: dimensional analysis, mathematical measurement theory and the methodology or epistemology of engineering. The literature shows that the investigations are not as closely connected as one would expect at first sight.18 Measurement theory researchers have already formulated and addressed some of the issues connecting scaling and measurement theory. R. Luce, for instance, mentions the incompleteness of the presentation of dimensional analysis in the standard representation theory of measurement, [1996, Sect. 4.8]. This incompleteness derives from a lack of a convincing explanation as to why genuine laws of physics should fulfil the principle of dimensional invariance. Narens addressed this problem from a fundamental point of view in [2002]. In [2007, p. 143] Narens distinguishes between the mathematical and epistemological foundations of dimensional analysis. Although many authors have already considered the mathematical foundations of dimensional analysis, a lot still needs to be done regarding the epistemological foundations. According to Narens, for instance, an important aspect of these epistemological foundations concern the relation between definability in science and symmetry invariance, where a symmetry is an isomorphism from a relational structure onto itself. A generally accepted full epistemological foundation of dimensional analysis is, however, still lacking. We encountered, for instance, the question of the relation between dimensional analysis and Euclidean geometry, and the question as to whether ratio-scale measurability is a necessary condition for Π-theorem. The latter issue has already been addressed in measurement theory: ‘dimensional analysis can be extended to incorporate any structure that has a ratio scale representation and that distributes in a suitable conjoint structure’, [Luce, 1996, p. 92]. Yet, the applications of these results and an epistemic analysis of them are still absent. Neither has the case of relativistic velocity gone unnoticed in measurement theory. The standard approach in measurement theory can cope with most structures in classical physics, but it fails to cover, for instance, relativistic velocity. According to Luce, this subject deserves further investigation within measurement theory [1996, p. 92]. Unfortunately, in the present chapter, we had to leave untouched many questions pertaining to the relations between Euclidean geometry, dimensional analysis, 18 The relation between the empirical sciences and measurement theory is still an ‘ongoing dialogue’ as testified by [Luce, 1996].

Scale Modelling in Engineering

795

and ratio scales and their units. Since our topic is a bottom-up investigation into scaling methodology in engineering, we had leave out the mathematical theory of measurement perspective on these questions. An enticing subject on the philosophy of engineering research agenda is therefore: study and bridge the gap between measurement theory, engineering methodology overall and the more specific practices of dimensional analysis. Luce agrees; according to him much work has still to be done before the mathematical knowledge about measurement can be put into scientific practice [1996, p. 79]. 5

CONCLUSIONS AND FUTURE RESEARCH

The scaling of artifacts has been an important problem down the ages and to that end various scaling methodologies have been developed. One such method involves using scale models without explicitly using dimensionless numbers. At the end of the nineteenth century, scientists and engineers started to develop dimensional analysis to adopt dimensionless numbers. The introduction of the Froude and the Reynolds number to the dimensionless version of the Navier-Stokes equation meant that dimension analysis could help to bridge the remarkable gap between theory and practice in fluid dynamics. The combining of scale modelling and dimensional analysis has become one of the most important scaling instruments in modern engineering. In engineering, direct research is distinguished from indirect research, which is based on mathematical and scale models. In many technical contexts direct, or real-scale research, is often difficult to achieve for practical reasons such as cost and safety. Its advantage lies in the validity of the methods since its results are more reliable than those obtained from indirect research. Since indirect research is carried out using scale or computer models, the disadvantage lies in the uncertainty of its validity. Its major asset, however, is that scale models or computer models are often much more manageable and more easily accessible than the actual artifact or process, and the experiments are better controlled. We saw that Sterrett’s observations about the special character of scale models definitely hold for Froude’s scale models. On the matter of the same-physics argument we concluded that the physics of scale models in naval architecture is of the same order of physical complexity as that of the prototype. Moreover, Newton’s drag law may be applied in both situations, because there are no constraints on the sizes of the objects under consideration; this still forms the foundation of skin friction calculations in today’s towing tanks experiments. Neither can Froude’s scale models be seen as intermediating between theory and observations. Unlike a ball-and-stick model of molecules, a scale model is a genuine part of the empirical phenomena to be studied. Finally, Froude did not use the underlying fundamental equations; he did not refer to the Navier-Stokes equations. With regard to the empirical content of the scale-model methods, we saw that Froude worked within the paradigm of classical mechanics. He used Euclidean geometry, Newtonian mechanics and the structure of ratio scale measurements

796

Sjoerd D. Zwart

to arrive at his extrapolation methodology. Overall, the conclusions about the empirical content of the geometrical, the kinematic and the dynamic, and even Froude’s scaling laws are very similar. Without any doubt all these laws have empirical content, but not in the way that phenomenological laws in physics do. The scaling laws derive all their empirical content deductively from classical mechanics including Euclidean geometry and the ratio-scale measurability of the relevant properties. These laws are, therefore, analytic and a priori, in the sense that, unlike standard phenomenological laws, they cannot be false within the context of classical mechanics. Interestingly, Froude’s extrapolation method, when used as an instrument for prediction, is not analytic in the sense that its accuracy for practical purposes has to be tested empirically. Finally, we saw that the ratio-scale measurability assumption carries considerable empirical content. It assumes that any equilibrium in reality can be linearly scaled up or down to any other equilibrium. Froude’s scaling methodology owes part of its virtue to the exploitation of the (implicit) assumption stating that scales under consideration have a ratio-scale structure. An interesting research question is therefore to extend this conclusion and ask whether classical dimension analysis, and therefore the Π-theorem, also owe their strengths to the ratio-scale assumption. It remains to be seen whether the conclusions about Euclidean geometry and classical mechanics drawn in the preceding section can be extended to standard dimension analysis as well. If so, a question arises concerning the philosophical foundations of dimension analysis or, for that matter, scaling when no ratio scales are assumed to exist. Finally, another important question for future research concerns the foundations of dimensional analysis external to classical mechanics. Such analysis could possible take into account, for instance, relativistic effects. ACKNOWLEDGEMENTS The author would like to thank Lex Keuning for his advice and explanations on many ship building and towing tank intricacies. Moreover he would like to thank Peter Kroes and Maarten Franssen for their stimulating discussions and their helpful comments. BIBLIOGRAPHY [Beaufoy, 1834] M. Beaufoy. Nautical and hydraulic experiments. Private Press of Henry Beaufoy, London, 1834. [Bernoulli, 1738] D. Bernoulli. Hydrodynamica, sive de viribus et motibus fluidorum commentarii. Johann Reinhold Dulsseker, Strasbourg, 1738. [Bertels and Nauta, 1969] K. Bertels and D. Nauta. Inleiding tot het modelbegrip. Wetenschappelijke Uitgeverij B.V., Amsterdam, 1969. In Dutch. [Beth, 1932] H. J. E. Beth. Newton’s „Principia”, volume 2 of Historische Bibliotheek voor de Exacte Wetenschappen. P. Noordhoff, Groningen—Batavia, 1932. In Dutch. [Boltzmann, 1902] L. Boltzmann. Model. In Encyclopaedia Britannica, volume XXX, pages 788–791. The Times Printing House, London, 10th edition, 1902. Freely accessible through: http://www.1911encyclopedia.org/Model.

Scale Modelling in Engineering

797

[Bridgman, 1973] P. W. Bridgman. Dimensional Analysis, volume 7, pages 439–449. Encyclopaedia Britannica, 1973. [Carnap, 1995] R. Carnap. An Introduction to the Philosophy of Science. Dover Publications, Inc, New York, 1995. This is an unabridged, corrected republication of the 1974 edition of the work originally published by Basic Books, inc., New York, 1966, under the title ‘Philosophical Foundations of Physics: An introduction to the Philosophy of Science’. [Cohen and Drabkin, 1948] M. R. Cohen and I. E. Drabkin. A source book in Greek Science. Source books in the history of the sciences. McGraw-Hill book company, Inc, 1948. [Cooper, 1872] T. Cooper. Men of Time. A Dictionary of Contemporaries. Routledge, London, 1872. [Ellis, 1966] B. Ellis. Basic Concepts of Measurements. Cambridge University Press, 1966. [Franklin, 1769] B. Franklin. To sir John Pringle. In Experiments and Observations on Electricity: made at Philadelphia in America, pages 492–496. Printed for David Henry, and sold by Francis Newberry, London, 1769. Freely available: http://www.franklinpapers.org/franklin/framedVolumes.jsp?tocvol=10. [Froude, 1868] W. Froude. Observations and suggestions of the subject of determining by experiment the resistance of ships. In The papers of William Froude M.A. LL.D., F.R.S. 1810–1879, pages 120–127. The institution of naval architects, 1868. [Froude, 1869] W. Froude. The state of exiting knowledge on the stability, propulsion and sea-going qualities of ships, and as to the application which it may be desirable to make to her majesty’s government on these subjects. In The papers of William Froude M.A. LL.D., F.R.S. 1810–1879. The institution of naval architects, 1869. [Froude, 1872] W. Froude. Experiments of the surface-friction experienced by a plane moving through water. British Association for the Advancement of Science, 42:118–124, 1872. [Froude, 1874] W. Froude. On experiments with H.M.S. Greyhound. Transactions of the Institution of Naval Architects, XV:37–73, 1874. [Froude, 1888] R. E. Froude. The „constant” system of notation of results of experiments on models used at the admiralty experiment works. Transactions of the Institution of Naval Architects, 29, 1888. [Galilei, 1638] G. Galilei. Two New Sciences, Including Centers of Gravity & Force of Percussion. University of Wisconsin Press, Madison, 1638. Translation: Stillman Drake. [Hacker, 1968] B. C. Hacker. Greek catapults and catapult technology: Science, technology, and war in the ancient world. Technology and Culture, 9(1):34–50, Jan 1968. [Hempel, 1945] C. G. Hempel. Geometry and empirical science. American Mathematical Monthly, 52:7–17, 1945. Reprinted in Readings in Philosophical Analysis, ed. H. Feigl and W. Sellars (New York: Appleton-Century-Crofts, 1949). The World of Mathematics, vol. III, ed. James R. Newman (New York: Simon and Shuster, 1956). [Hesse, 1967] M. Hesse. Models and analogy in science. In The Encyclopedia of Philosophy, volume 5, pages 354–359. MacMillan, New York, 1967. [Hydra, date unknown] Hydra. Hydra multi-axis vibration test facility. http://www.esa.int/techresources/. [Langhaar, 1951] H. L. Langhaar. Dimensional Analysis and Theory of Models. Robert E. Krieger Publishing Company, Huntington, New York, 1951. Edition 1980. [Luce, 1996] R. D. Luce. The ongoing dialog between empirical science and measurement theory. Journal of Mathematical Psychology, 40:78–98, 1996. [Luce et al., 1971] R. D. Luce, P. Suppes, A. Tversky, and D. Krantz. Foundations of Measurements, Additive and Polynomial Representations, volume 1. Academic Press, New York, London, 1971. [Morgan and Morrison, 1999] M. S. Morgan and M. Morrison. Models as Mediators. Cambridge University Press, Cambridge, 1999. [Narens, 2002] L. Narens. Theories of meaningfulness. Scientific Psychology Series. Lawrence Erlbaum Associates, Mahwah, New Jersey, 2002. Dimensional analysis: Sections 1.5, 5.10. [Narens, 2007] L. Narens. Introduction to the Theories of Measurement and Meaningfulness and the Use of Symmetry in Science. Scientific Psychology Series. Lawrence Erlbaum Associates, Mahwah, New Jersey, London, 2007. [Navier, 1822] C. L. M. H. Navier. Mémoire sur les lois du mouvement des fluides. Mémoires de l’Académie Royale des Sciences, 2(6):389–440, 1822. [Neményi, 1975] P. F. Neményi. The main concepts and ideas of fluid dynamics in their historical development. Archive for History of Exact Sciences, 2(1):52–86, January 1975.

798

Sjoerd D. Zwart

[Nersessian and Patton, this volume] N. Nersessian and C. Patton. Model-based reasoning in interdisciplinary engineering: Cases from biomedical engineering research laboratories, this volume. [Newton, 1687] I. Newton. Philosophiae Naturalis Principia Mathematica. S. Pepys, Reg. Soc. Praeses, London, Juli 1687. Mathematical principles of natural philosophy, translated by Andrew Motte and revised by Florian Cajori. Volume 34 of Great Books of the Western World, Encyclopedia Britannica, 1934. [Palacios, 1964] J. Palacios. Dimensional analysis. MacMillan & Co Ltd, 1964. [Payne, 1936] M. P. Payne. Historical note on the derivation of Froude’s skin friction constants. Transactions of the Institution of Naval Architects, 78:93–109, 1936. [Pervushin et al., 2004] V. Pervushin, V. Zinchuk, and A. Zorin. Conformal relativity: Theory and observations, 2004. [Reech, 1852] F. Reech. Cours de Mécanique d’après la Nature Généralement Flexible et Élastique des Corps. Carillian-Goeury et Von Dalmont, Paris, 1852. [Renda et al., 2004] V. Renda, A. Murphy, and P. Sollogoub. Experimental facilities for earthquake engineering simulation worldwide. Technical Report 10, NEA/CSNI/R, 2004. [Rouse and Ince, 1957] H. Rouse and S. Ince. History of hydraulics. Iowa Institute of Hydraulic Research State University of Iowa, 1957. [Sterrett, 2002] S. G. Sterrett. Physical models and fundamental laws: Using one piece of the world to tell about another. Mind & Society, 3(5):51–66, 2002. [Sterrett, 2006] S. G. Sterrett. Models of machines and models of phenomena. International Studies in the Philosophy of Science, 20(1):69–80, March 2006. [Stokes, 1845] G. G. Stokes. On the theories of the internal friction of fluids in motion, and of the equilibrium and motion of elastic solides. Transactions of the Cambridge Philosophical Society, 8:287–319, 1845. [Szucs, 1980] E. Szucs. Similitude and Modeling. Fundamental studies in engineering. Elsevier, Amsterdam, 1980. [Towhata et al., 2004] I. Towhata, S. K. Prasad, G. P. Chandradhara, and P. Nanjundaswamy. Shaking table tests in earthquake geotechnical engineering. Current Science, 87(10):1398– 1404, November 2004. [Truesdell, 1953] C. Truesdell. Notes on the history of the general equations of hydrodynamics. The American Mathematical Monthly, 60:445–458, 1953. [Vincenti, 1990] W. Vincenti. What Engineers Know And How They Know It: Analytical Studies From Aeronautical History. Johns Hopkins University Press, Baltimore, 1990. [Wright, 1992] T. Wright. Scale models, similitude and dimensions: Aspects of mid-nineteenthcentury engineering science. Annals of Science, 49:233–254, 1992.

SIMILARITY AND DIMENSIONAL ANALYSIS Susan G. Sterrett

1 INTRODUCTION AND OVERVIEW The importance of similarity in comprehending things and reasoning about them was recognized before the time of Plato. Similarity continues to be important in philosophy, science, and technology to this day. The historical roots of the concepts of similarity, ratio and proportion will be accorded a brief mention in this article in order to provide a fuller understanding of the relationship between dimensionless parameters and reasoning from similarity. The main topic of this article, however, is the use of dimensional analysis in similarity-based reasoning in current contexts.1

1.1

Ratios and similarity

The use of ratios in reasoning from similarity has a long history. The Pythagoreans, an intellectual community that predated Plato’s Academy, discerned a relationship between observable phenomena and ratios. Certain musical phenomena were correlated with ratios of lengths of unconstrained portions of a lyre string; these ratios were equal to ratios of the first few counting numbers. This discovery seemed to them good reason to expect that all physical phenomena could ultimately be accounted for or described in terms of ratios of whole numbers. The ratios that arose in studying harmony appeared in other important mathematical representations, such as the tetractys, a triangular arrangement of ten marks composed of four rows containing one, two, three and four marks, respectively. Many other numbers had geometrical shapes closely associated with them. This was due to the use of two-dimensional arrays of marks representing whole numbers, which were generated using a certain standard procedure. Thus there was a canonical graphical representation for every number generated using the procedure. The ratios between the sides of the array (i.e., the sides of the square or rectangle formed by the array) were also investigated and associated with particular numbers. Thus the study of geometrical similarity was at first associated with ratios of whole numbers. The simplest example of this was that the arrays for all “square numbers” (the numbers 4, 9, 16, . . ., which had a ratio of sides of 2:2, 3:3, 4:4, . . . respectively) were all square (and so geometrically similar) [McKirahan, 1994, Chapter 9]. 1 For

a discussion of the relationship of dimensional analysis and explanation see [Lange].

Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

800

Susan G. Sterrett

That reasoning from geometrical similarity involved ratio and proportion is later reflected in Euclid’s Geometry, albeit there the ratios are thought of primarily as ratios of line segments per se, rather than as ratios of numbers. Many proofs in Euclid’s Geometry employ the strategy of establishing that two figures are similar in order to draw inferences about other geometrical entities. Such proofs, or portions of proofs, usually proceed by way of identifying certain ratios of lines and then establishing the similarity of figures by means of proportions. Proportions are statements that two ratios are equal. An example is a statement that the ratio of two lengths x and y in a geometrical figure A is equal to the ratio of the corresponding two lengths x and y  in another geometrical figure B, which can be put as: x is to y as x is to y  , or, in a familiar notation, as x : y :: x : y  . (Equality of angles is required for geometric similarity, but inasmuch as establishing the equality of angles is based upon establishing the similarity of right triangles, some of which may be constructed for just this purpose, equality of angles is an instance of establishing similarity of geometrical figures by means of proportion as well.) Once the similarity of two geometrical figures (such as similar triangles or similar parallelograms) has been established, other equalities between ratios are then known to hold, from which conclusions about the quantities that occur in those ratios can then be drawn. There is a more formal conception of geometrical similarity. Geometric similarity in a metric space can be defined in terms of the metric defined on the space and a mapping between corresponding points of the two similar geometrical figures: the distance between any two points m and n in geometrical figure A is equal to some constant times the distance between the corresponding points m and n in geometrical figure B. (Put formally, the condition is: g(m, n) = rg(m , n ), where g is a metric,2 r is some numerical constant, and m and n are points in figure B that are the images of the points m and n in figure A mapped to figure B.) For the special case of Euclidean geometry where g is distance in its usual sense, r would be the scaling factor between two geometrically similar figures A and B. Every pair of figures whose similarity is established using proportions in Euclidean geometry will come out as geometrically similar on this conception as well.

2 A metric is defined on a set. The set along with the metric constitutes a metric space. A metric is defined as follows. If X is an arbitrary set, a function g : X × X → R on X is a metric on X if it satisfies the following conditions for all x, y, z in X: 1. g(x, y) is greater than or equal to 0 (thought of as a distance function, no “distances” are negative); 2. g(x, y) = 0 if and only if x=y (i.e., thought of as a distance function, only the “distance” between an element and itself is zero); 3. the function g(x, y) is symmetric, i.e., g(x, y) = g(y, x) (the “distance” between x and y is the same as the distance between y and x); and 4. the function (x, y) satisfies the inequality g(x, y) + g(y, z) ≥ g(x, z) (the “triangle inequality”).

Similarity and Dimensional Analysis

1.2

801

Physical similarity

Can the notion of similarity developed in geometry be generalized from geometry to natural science? The answer is yes. It is common to regard the notion of physical similarity as just such a generalization; in order to generalize the notion of similarity from geometry to natural science, both the notion of ratio and the notion of shape must be generalized. The correct way to carry out such a generalization of geometrical similarity is to generalize from similarity of geometrical figures to similarity of physical systems.3 The notion of similar systems goes back to at least Newton, who used the phrase himself. Some less formal versions of the idea are found in Galileo. The formal statement of the notion of similar systems as it is used today did not appear in English-language literature until the early twentieth century. Instead of ratios restricted so as to involve only lengths (of line segments), similarity of physical systems is established using ratios that may involve other quantities as well, such as time, mass, and force. As will be explained later, the quantities are organized into a system, and associated with a system of dimensions. There are some constraints on such a system: physical quantities of the same kind all have the same dimension, and any quantity can be expressed in terms of the dimensions of the set of quantities designated as the set of base quantities. The ratios used to establish physical similarity are themselves dimensionless, just as the ratios relevant to establishing geometric similarity are. They may be as simple as a ratio of two quantities of the same dimension (such as Mach number, which is a ratio of two velocities) or significantly more complex (such as Reynolds number, which contains density, velocity, length, and kinematic viscosity). They are known as dimensionless parameters, or dimensionless groups, and many have been given proper names due to their usefulness in science and engineering. The generalization of similarity from geometrical similarity to physical similarity is nontrivial, for the notion of shape is generalized from the shape of a geometrical figure to the configuration of a physical system. The identification of an appropriate physical system may require practical empirical knowledge and the application of fundamental principles, and the physical quantities involved must be identified with an eye to the behavior of interest or phenomenon of interest. There are different species of physical similarity. Often a certain species of similarity is identified with a particular dimensionless parameter. There is a progression of physical similarity from geometric similarity (similarity of the linear relations, i.e., traces similar paths in proportional times) to kinematic similarity (geometric similarity plus similarity of motions, i.e., the same paths and ratios between velocities), to dynamic similarity (kinematic similarity plus the same ratios between forces). All of these arise in hydrodynamics, and are discussed in more detail in Zwart’s chapter in this Volume. There are other kinds of physical similarity 3 Peter Kroes [1989] has urged the point that analogies between systems rather than laws is important. I believe the question has received too little attention among those writing about similarity in philosophy of science.

802

Susan G. Sterrett

used for experimental model testing, though: writing in 1964, Pankhurst4 listed elastic similarity, similarity of mass distribution, electrical similarity, magnetic similarity, and thermal similarity among the kinds of similarity between model and prototype. More recently, dimensionless parameters for physical similarity with respect to phenomena in chemical engineering and fire modeling have been developed [Slokarnik, 2006; Becker, 1976; Hottel, 1961; Sterrett, 2006].

1.3 Physical similarity and dimensionless parameters The generalization from geometrical similarity to physical similarity was not the work of any single thinker, and it took some time for it to be developed into its present form. Now, over two millennia after Pythagoras, the method of physical similarity is a mature, well-established method used in physics, engineering and some applied sciences. The method began in the Renaissance but was not developed and formalized during the late nineteenth and early twentieth centuries. Edgar Buckingham of the National Bureau of Standards set out the method in use as of 1914 explicitly, though the development of the method was due to others, and was later made more mathematically rigorous.5 His 1914 paper “On Physically Similar Systems: Illustrations of the Use of Dimensional Equations” [Buckingham, 1914] has become a landmark reference, and the theorem it contains regarding the reduction of a relation between physical quantities to a relation between dimensionless parameters that can be used to establish physical similarity has consequently become widely known as Buckingham’s Theorem (in spite of his own remarks in later papers ceding credit for priority to Riabouchinsky) [Sterrett 2005, p. 188]. Some of the ideas in Buckingham’s paper appeared elsewhere much earlier, though, notably in France by Vaschy [Vaschy 1892] and in the context of a practical problem in aerodynamics by Helmholtz. Another, less celebrated, point in Buckingham’s 1914 paper is that, even after considering all logical and mathematical constraints, there is often still some leeway left in choosing which dimensionless parameters to use to characterize a given physical system and to establish physical similarity of a model prototype to the system it is intended to model. Here additional conceptualization of the situation and empirical knowledge are involved, which raises philosophical questions about the interrelationship of mathematics, logic, natural science, applied science, and engineering science.6 4 [Pankhurst, 1964, p. 18 and pp. 77 -80]. (Pankhurst was Superintendent of the Aerodynamics Division of the National Physical Laboratory from 1964 until 1970.) 5 Ehrenfest-Afanassjewa [1916] provided more mathematically rigorous concepts (the concepts of homogenous functions and homogeneous equations), which J. Palacios [1964, p. 48 -53] employed to present a more rigorous formalization of the concepts and assumptions involved in deriving a version of Buckingham’s result. Langhaar [1951] also contains a mathematically rigorous treatment of dimensional analysis and similarity. However, Buckingham’s original statement has retained its prominence and is generally the one cited in the engineering literature. 6 The relationship between pure science and engineering regarding ideas involving similarity and dimensional analysis in the years leading up to and immediately following Buckingham’s 1914 paper is discussed in [Sterrett, 2005]

Similarity and Dimensional Analysis

803

In this article we are interested in dimensionless parameters and dimensional analysis insofar as they bear on identifying and reasoning about physically similar systems, but it is noteworthy that dimensional analysis has often been discussed apart from its use in establishing similarity of physical systems. Not only was dimensional analysis discussed as a philosophical topic separately from the topic of similarity during the twentieth century, but similarity was discussed in philosophy and, even, by some in philosophy of science, separately from the use of dimensional equations and dimensionless parameters. Authors of works in philosophy and philosophy of science in the twentienth century who wrote about similarity often did so without mentioning the theory of dimensions or dimensionless ratios. Some formal investigations into the topics of dimensional analysis and physical similarity in philosophy of science were carried out [Krantz, 1971, p. 454-544], but the work was not very well assimilated into philosophy. Often what philosophers mean by similarity is simply sharing a property, such as red; or that the shade of red of one object is very close to the shade of red of another. The philosopher of science Giere gives as an example of “exploiting similarities between a model and that aspect of the world it is being used to represent” the example of a mathematical model of a pendulum in terms of similarities: a scientist picks out “some specific features of the model that are then claimed to be similar to features of the designated real system to some (perhaps fairly loosely indicated) degree of fit” [Giere, 2004, p. 745]. That is not quite what is meant by similarity in this article, unless “feature” is generalized in the manner indicated below, i.e., as the value of one or more of the dimensionless parameters that characterize a system. In this article, similarity refers to a generalized notion of geometric similarity. On such a notion of similarity, similarity is a matter of proportion, rather than a matter of having some specific feature in common. The notion of similarity we use here (physical similarity of systems) might thus be of interest to those who do discuss similarity in philosophy. One often hears it said that “anything is similar to anything else in countless respects” [Giere, 2004, p. 747], and many philosophers thus consider establishing similarity as involving little more than picking and choosing features to suit a pragmatic purpose, rather than being a formal notion capable of providing a rich resource for investigations. Goodman enunciated such a position long ago [Goodman, 1972, p. 446], granting that statements of similarity “are still serviceable in the streets”, but “cannot be trusted in the philosopher’s study.” For, he concluded, “As it occurs in philosophy, similarity tends under analysis either to vanish entirely or to require for its explanation just what it purports to explain”. The prevailing attitude concerning similarity “as it occurs in philosophy” is hence somewhat at odds with that in scientific fields, given that (i) the concept of geometrical similarity is a rich resource for proofs in the discipline of geometry, (ii) the concept of physical similarity, though including more of an empirical element than geometrical similarity and hence requiring more empirical knowledge in order to be used effectively, is nevertheless a widely used, indispensable and

804

Susan G. Sterrett

highly regarded method in many natural and engineering sciences, and (iii) the value of the prospect that the notion of physical similarity could be generalized even further. Contemporary discussions in philosophy and philosophy of science might thus be enriched by taking into account the relationship between physical similarity and dimensional analysis in science. Many treatments of similarity in philosophy of science do not even associate reasoning about similarity with ratio and proportion. It is hoped that, in addition to serving as a reference on the topic, this article might contribute to rectifying that situation. The discussion in this article is centered on the relationship between similarity and dimensional analysis, and some of the philosophical issues involved in understanding and making use of that relationship. Since the connection between similarity and dimensional analysis is via dimensionless parameters, we begin with basics about quantities, units and dimensions. 2

QUANTITIES, UNITS, AND DIMENSIONS

2.1 Quantities, units, and quantity equations A science such as physics involves using equations or proportionality relations to describe physical phenomena. Equations and relations in the physical sciences relate physical quantities to each other. This is so whether they are expressed in words or in symbolic formulae. What exactly is meant by a quantity in this context? Here different approaches have been taken. One approach to the concept of a quantity emphasizes that the physical quantities appearing in physical equations are required to be measurable, and so that the value of a quantity (e.g., “six feet”) consists of two portions: a numerical portion and a unit to which all quantities of that kind can be compared. Maxwell explicitly discussed this conception of a quantity, while recognizing ambiguities in the notation of physical quantities as used in equations in scientific practice. He noted that symbols used as variables in equations of physics lent themselves to two different interpretations: (i) as denoting the lines, masses, times, and so on themselves, and (ii) “as denoting only the numerical value of the corresponding quantity, the concrete unit to which it is referred being tacitly understood” [Maxwell 1890, p. 241]. Each of these interpretations presents a problem, though: The first interpretation doesn’t really apply during the process of performing the numerical calculations. (This is because, on the account of quantities Maxwell was using, arithmetic operations apply to numbers, not to the quantities themselves.) The second interpretation doesn’t satisfy the requirement that “every term [of an equation of physics] has to be interpreted in a physical sense.” Maxwell’s way of resolving the ambiguity he identified was to take a sort of hybrid approach. During the process of calculation, he said, we should regard the written symbols as numerical quantities; they are accordingly governed by the rules of arithmetical operations.

Similarity and Dimensional Analysis

805

However, he could not completely dispense with the first interpretation; he wrote that “. . . in the original equations and the final equations, in which every term has to be interpreted in a physical sense, we must convert every numerical quantity into a concrete quantity . . . ” By “conversion into a concrete quantity,” Maxwell meant multiplying the numerical expression by “the unit for that kind of quantity” [Maxwell 1890, p. 241]. An alternate approach to the concept of quantity is to regard the operations of multiplication and division as applicable to what Maxwell regarded as concrete quantities themselves. On this view, the equations of physics can be regarded as expressing relations between physical quantities, and there is no need for an additional step of interpreting the terms in the equation to yield physical quantities. This alternate approach is the one favored here. Unlike the approach on which a quantity is considered to consist of a numerical portion and a unit, on this alternate approach a quantity can be thought of independently of reference to a unit. To take an example, the definition of the quantity velocity in terms of length and time is thought of as stating a relation between concrete quantities, without involving units or numerical expressions. The relation V = l/t (velocity is proportional to length and inversely proportional to time) is regarded as expressing the relationship between the concrete quantity velocity, a concrete length, and a concrete time, without any hedging regarding the applicability of the arithmetical operations of multiplication and division to concrete quantities. Lodge7 argued for the recognition of fundamental equations of mechanics and physics as quantity equations, i.e., equations that “express relations among quantities” [Lodge 1888, pp. 281–283]. Understood as quantity equations, the fundamental equations of a science “are independent of the mode of measurement of such quantities; much as one may say that two lengths are equal without inquiring whether they are going to be measured in feet or metres; and, indeed, even though one may be measured in feet and the other in metres” [Lodge 1888, p. 281]. Lodge proposed that the arithmetical operations of multiplication and division be applied to concrete quantities of different kinds. In quantity equations, he said, when quantities are represented by numbers or numerical expressions, “that number is the ratio of the quantity to some standard of the same kind [of quantity]”. Lodge made some important points about quantity equations that distinguished them from the numerical equations (which he regarded as derived from quantity equations). Physical (quantity) equations “can only be among quantities of the same kind, or . . . if there are quantities of different kinds in the equation, then the equation is really made up of two or more independent equations which must be separately satisfied, each of these being only among quantities of the same kind.” He also referred to the dimensions of a quantity, noting that the dimensions of a 7 Cornelius [1965a] credits Lodge as the first to endorse a calculus of quantities. Lodge did so an 1888 article in Nature entitled “The Multiplication and Division of Concrete Quantities” [Lodge 1888]. It should be noted that a calculus of quantities for the quantities of length, area, and volume in Euclidean geometry existed before Lodge’s 1888 paper.

806

Susan G. Sterrett

quantity may not always determine the kind of quantity it is. The examples he gave were of two different concrete quantities having the same dimension, such as work and moment of a force. The significance of maintaining a distinction between dimensions and kinds of quantities will be considered in more detail below. The mention of the ratio of a quantity to a standard of the same kind of quantity may seem to bring in the use of units under another guise. Does the alternate approach of a quantity calculus as urged by Lodge really use a notion of quantity that is independent of units after all? The answer is yes, because the need to refer to the particular units used is avoided by an important further step: the requirement that relates units and quantities is a requirement placed on the entire system of units employed in a science, rather than on each unit separately. This requirement is that the system of units be coherent. A system of units is coherent if the relations between the units used for the quantities are the same as the relation between the quantities in the fundamental equations of the science. On this view, of course, the fundamental equations are regarded as quantity equations. Using the example of velocity again, the fundamental equation relating velocity, length and time is that velocity is proportional to length and inversely proportional to time, with a constant of proportionality of unity. The requirement that the system of units used be coherent demands that the relationship between the unit used for velocity, the unit used for length, and the unit used for time is the same as the relationship in the fundamental equation, i.e., the quantity equation V = l/t. As Lodge noted, the constant of proportionality of 1 is a choice we have made in selecting our system of units so that it will be coherent with the fundamental laws of physics. The numerical equation derived from the quantity equation V = l/t will then have the same form as the quantity equation. Thus, if it is known that the system of units is coherent, it follows that the numerical equation has the same form as the fundamental relation. The form of the numerical equation can thus be known independently of actually using units and numerical expressions to express the quantities and then deriving the numerical equation from the quantity equation — so long as the requirement that the system of units is coherent is met. Lodge noted that there is an element of conventionality in our choice of units of length: “If a, b are two lengths, the product ab is always used to represent [the area of] a rectangle whose sides are a, b respectively; though we might have agreed to use it as a representation of a parallelogram with sides a, b containing an angle of (say) 60 [degrees]” [Lodge 1888, p. 282]. The arithmetical operations do not rule out this possibility. The requirement of coherence of a system of units does determine some such choices, however. The step of thinking in terms of requirements placed on a system of units, rather than in terms of requirements placed on the units chosen for each kind of quantity individually, turns out to be crucial to settling one of the most puzzling philosophical disputes on the topic: how to understand the role of quantities, units, and laws when converting from one system of units to another used in electrodynamics. In classical mechanics, conversions from one system of units to

Similarity and Dimensional Analysis

807

another were comparatively straightforward and the questions that arose about the kinds of quantities or form of fundamental equations involved were easily settled by further clarification. This was not so in electrodynamics. The resolution of the issues in electrodynamics is discussed below. The resolution of the issues there contributes to our understanding of the conventional aspects (i.e., the aspects that are a matter of choice) of systems of units and physical quantities in other sciences. We shall see that the requirement that a system of units be coherent turns out to be far more central to understanding the relationship between quantities, dimensions and units than one might have suspected.

2.2

Physical quantities

There have been various philosophical views on the logical relationship between quantities, units, and systems of measurement. Our interest here is the logical priority of physical quantities with respect to units and means of measurement, not historical order of the development of the concepts. Though there is some disagreement among philosophers of science as to what the symbols in a scientific equation represent [Palacios 1964, p. vi], most would at least regard scientific equations as relating quantities only insofar as they are at least in principle measurable. The question then arises as to which is logically prior: scales, units of measurement or the quantities measured? If physical quantities are taken to be logically prior, then the question arises as to what sense it makes to talk of quantities existing independently of the means of measuring them. How, one might ask, can a quantity such as six feet tall be logically prior to a measurement system and specification of a unit? On the other hand, if physical quantities are not logically prior to units and measurement scales, one might ask what basis there could be for claiming that a quantity is anything more than the outcome of a measuring process. One influential view on the relationship of physical quantities and units that is familiar to philosophers of science is Brian Ellis’ view in his early work on the topic of quantities in Basic Concepts of Measurement. His view there is that “the existence of a quantity entails and is entailed by a set of linear ordering relationships” [Ellis, 1966, p. 32].8 Like Percy Bridgman’s view in his Dimensional Analysis [Bridgman, 1963], Ellis thinks that relative magnitude is invariant (so long as the same dimension is associated with the quantities whose magnitudes are being compared) [Ellis, 1966, p. 141]. However, unlike Bridgman, Ellis does not take this point as the starting point of a theory of dimensions. Rather, he uses invariance of relative magnitude of like physical quantities as a criterion for similarity of scales, and then defines dimensions as classes of similar scales [Ellis, 1966, p. 140]. Then, Ellis says, “[w]e may say that two scales X and X  belong to the same dimension if and only if the ratio of any two measurements on X is 8 Ellis distinguishes criteria for the existence of a quantity from criteria for identity of quantities: it is the order, and not the ordering relations, that provides the criteria for identity of quantities [Ellis, 1966, p. 32].

808

Susan G. Sterrett

the same as the ratios of the same measurements on X  .” As for the logical priority of physical quantities and dimensions, many philosophers, including Ellis, are concerned to distinguish their view from what Ellis refers to as a “na¨ıve realist” view. Ellis describes the na¨ıve realist view that most philosophers want to reject as the view that quantities such as yellow or six feet tall are properties of objects, and that such quantities exist prior to specifying a means of measurement. The problem with such a na¨ıve realist view is obvious, as mentioned earlier: how could quantities such as these exist prior to specifying a means of measurement or defining units for measuring it? As numerous philosophers writing on dimensional analysis have noted, dimensional analysis takes place only in the context of the use of scientific equations, many of which can be expressed in such a way that they hold for any coherent set of units (a concept defined below). Ellis’ interest was in identifying the elements of conventionality in means of measurement and units of measurement. He later criticized his own proposal presented in Basic Concepts of Measurement, on which all quantitative properties supervene on quantitative relationships, concluding in 1992 that that account, though it works for spatiotemporal locations, “will not do” for properties such as charge, which he came to think was, unlike length, an intrinsic property [Ellis, 1992, p. 177]. An historical survey of philosophical works on the topic is beyond the scope of this handbook article. Some of the well-known critical studies are: [Ehrenfest-Afanassjewa, 1916; Campbell, 1920; Bridgman, 1931; Langhaar, 1951; Duncan, 1953; Sedov, 1959; Birkhoff, 1960; Palacios, 1964; Pankhurst, 1964; Krantz et al., 1971; Becker, 1976].

2.3 Systems of units In this article, we forgo further discussion of the philosophical questions about the quantities of physical science discussed in the previous section. We now consider the situation faced by someone today making practical use of the theory of physically similar systems. Much of the work that needs to be done in order to identify the quantities of physical science — including not only the empirical work involved in formulating the equations of physical science but the analytical work in interpreting the constants in those equations, determining which equations are to be considered fundamental laws or principles and which relations and equations are to be considered definitional — is done prior to the specification of a particular system of coherent units. Thus if one is already committed to the use of a certain coherent system of units, some things will no longer be in question: the decisions and conventions that affect the number of quantities required in physical science and which ones will be considered basic quantities are no longer in question, so long as the system of units is considered satisfactory. We will use the almost universally-accepted SI system (Le Syst`eme International d’Unit´es) here, not as an authority on these foundational questions, but to illustrate the kinds of foundational issues that can arise in developing a system of units.

Similarity and Dimensional Analysis

809

When new scientific developments suggesting that additional quantities might need to be added arise — as happened with the development of electromagnetism in the late nineteenth century — then these decisions need to be reevaluated and the system may need to be amended or revised. At one time, the CGS system provided a coherent set of units for Newtonian mechanics, yet it was not clear what to say at that time about a coherent set of units for electromagnetism. There was a CGS system for Newtonian mechanics, a CGS-M system of units for magnetism, and a CGS-E system of units for electrical phenomena. It was shown in 19019 by Giorgi that the CGS system could be amended in alternate ways to provide a coherent set of units for the quantities in the equations describing electromagnetic phenomena, so that a choice had to be made. The decision made by the committees governing the SI system was to include a base unit for the physical quantity of electric current; the SI system now provides a coherent set of units with respect to Maxwell’s equations as well as for Newton’s. When used for electromagnetism, the SI system (which follows Giorgi) and the older CGS systems are really different systems of units; for electromagnetic phenomena, unlike for mechanics, switching from one of these systems to the other involves more than a simple change of units, for there will be some equations whose form differs depending on which system one is using.10 Thus some of the units for the older CGS systems are referred to as non-SI units [Taylor and Thompson, 2008, p. 37].11 The calculations needed to switch from CGS units to the Giorgi/SI units were rather complicated. Attempts to explain the source of the complications gave rise to deep-seated disagreements about the nature of quantities, dimensions and units. In 1964, Cornelius summarized the disagreements and presented a resolution that drew heavily on the role of coherence of (a system of) units: “Considering the equations used in electricity to be quantity equations has raised a lengthy dispute. No agreement was reached on the question of whether the dissimilarities between the different systems are caused by a difference in units or in quantities. . . . [T]he real difference lies in a change in the form of the equations and hence in a change in the coherence of units” [Cornelius et al., 1964, p. 1446]. He also provides an explanation of the relationship between quantity and dimension: dimension does not characterize a quantity because dimension is not totally independent of the choice of system of units. Quantities such as charge, length, and current, however, are independent of the specification of a system of units. When a change in system of units involves a change in the form of the quantity equations with respect to 9 See

[Taylor and Thompson, 2008, p. 16]. interdependence of equations of electromagnetism and systems of units is nontrivial. For an entire book devoted to the subject see [Cohen, 2001]. For a more philosophical and comprehensive treatment see the set of four papers by Cornelius et al. on the topic [Cornelius et al., 1964; 1965a; 1965b; 1965c]. 11 As for other points that ought to be noted about the range of applicability of the SI system: The current SI System was developed using equations that do not reflect relativistic effects; this is discussed in [Taylor and Thompson, 2008, p. 13]. One area where revisions or amendations to the SI System could possibly occur in the future is in the development of units for quantities that measure biological effects; the difficulties that would need to be addressed are discussed in [Taylor and Thompson, 2008, p. 14]. Currently such units are recognized, but as non-SI units. 10 The

810

Susan G. Sterrett

which that system of units is coherent, the dimension of the quantity may change. To put it another way: The dimension of a quantity is relative to a system of units inasmuch as a system of units involves selecting the quantity equation(s) that relates the basic quantities of a science. The point is general to all sciences. This resolution of the disagreements in electromagnetism draws on an important point made by Lodge in proposing a quantity calculus: The meanings of products and quotients in a quantity calculus do not have unambiguous meanings. Different choices can be made, even for two different coherent systems. (As explained earlier, this is because coherence is with respect to a particular quantity equation; one could make various choices about interpreting the product ab of two lengths a and b.) According to his explanation, the difference between the Giorgi and various CGS systems of units is really a matter of choosing a different meaning for the ratio of current to length. It is not a matter of a change in [the meaning of a] quantity [Cornelius et al., 1964; 1965a; 1965b; 1965c]. Taking physical quantities to be logically prior to units as described earlier allows Ellis’ point that nothing rules out the possibility that a given physical quantity might be measurable by two dissimilar scales, for choices between two sets of similar scales associated with the physical quantity might be among the items that need to be settled by convention, if one of the organizing principles of a system of units is that each quantity is to have only one dimension associated with it. Thus, although Ellis says that the same physical quantity could be associated with more than one dimension (since, for him, a dimension is a class of similar scales) and the SI System takes the approach that each quantity has only one dimension associated with it, these views are addressing slightly different questions. The statement that “each of the seven base quantities used in the SI system is regarded as having its own dimension” [Taylor and Thompson, 2008, p. 11] is made about a coherent system of units containing some conventional aspects, whereas Ellis’ statement is a statement about the logical possibility of the existence of a physical quantity for which it is possible to construct two dissimilar measurement scales that meet his criteria for a measurement scale.12 The approach taken by the SI system is that it is kinds of physical quantities such as amount of substance or length (rather than quantities or properties such as twenty moles or six feet tall) that are logically prior to units and measuring systems. The physical quantities that the system is committed to providing a means of describing are those that are related by the equations and relations that constitute empirical science, such as masses, volumes, densities, amount of substance, times, distances, velocities, and so on. The SI system is developed within the context of accepted scientific equations, and, as explained below, is a coherent system of units. As remarked in [Krantz, 1971, p. 464], it is true of all well-known physical laws that they “are of such a form that it is unnecessary to

12 In the section on “Dimensional Analysis and Numerical Laws” in the now-classic study Foundations of Measurement [Krantz et al., 1971, p. 454], it is stated at the outset that all the physical measures being discussed will be treated as if they are ratio scales.

Similarity and Dimensional Analysis

811

specify the units in terms of which the several physical quantities are reported, provided only that a fixed system of coherent units is used.” To explain what is meant here in saying that the kinds of quantities used in physical science are logically prior to the units used in physical science, we begin again with geometry and then consider generalizations to the physical sciences. We can make sense of such claims as the claim that a square has sides of equal length or that the ratio of the circumference of a circle to its diameter is always the same, without specifying a system of units, or making any mention of a unit at all. Thus the notion of length can be said to be logically prior to that of units of length. Next, we consider whether we can likewise talk about relationships between physical quantities without specifying units or a means of measurement. Here the empirical is involved as well, as was explained above: the physical quantities are those having a place in a system of quantities used in the equations of an established physical science. The relations between the systems of quantities used in physical science are provided by the fundamental equations and laws of that physical science. Given the relations and equations of empirical science used to identify and define the relations between these quantities, we can make sense of claims that two quantities with which we associate the same dimension (two lengths, two masses, two (local) time intervals, two densities) are equal. In some cases it may turn out that determining the truth or falsity of such a claim requires further investigation, which may involve the use of a measurement systems and units. However, this does not conflict with views on which the dimensions (kinds of quantities) involved can be regarded as logically prior to the specification of a particular system of units. On the approach that it is the relations and equations of science that determine what kinds of quantities there are, kinds of quantities can be regarded as independent of the specification of a system of units. Such a realism is thus not the na¨ıve realism Ellis wished to avoid; it concerns the kinds of quantities that are postulated by the current equations and principles of science, based upon the role they play in a mature science that meets the standards of current scientific practice. It is beyond the scope of this article to deal with questions about the particulars of these standards; what is important to our topic here is that such standards involve checks and balances of various kinds, so that capricious invention of quantities is ruled out. There is no limit in principle to the quantities (and the dimensions associated with them) that can be defined in this way, and many of them can be expressed in terms of others (e.g., the dimension of velocity can be expressed in terms of the dimensions of length and time). The question of which of the many kinds of quantities used in science are to be considered fundamental or basic ones and which are to be considered derived is not completely determined by empirical science, but involves some arbitrary choices.

812

Susan G. Sterrett

2.4 Dimensions and coherent systems of units In developing a system of units, the advantage of taking an approach on which the physical quantities and the dimensions associated with them are considered logically prior to units is that it allows for each unit to be related to the others. The units can be chosen such that they are related to each other by the same equations and relations as the kinds of quantities are. Thus it is possible to define and choose to use what is known as a coherent set of units. As explained in section 2.1, coherence of a set of units is relative to a set of quantity equations of a science that are taken as so basic that they are regarded as fundamental laws of that science. When a system of units is coherent with a set of equations, the equations between the numerical values of quantities have the same form as the equations between quantities themselves.13 This is a crucially important advantage; the aim of a coherent set of units motivated choices about how to handle quantities as they were added in the current SI system, and the fact that other competing alternative systems of units were not coherent counted against choosing them. In the SI system, the basic or fundamental quantities are termed “base quantities.” The base quantities of the SI system are length, mass, time, electric current, thermodynamic temperature, amount of substance, and luminous intensity. All other quantities are considered derived quantities. It is a consequence of the coherence of a system of units that every derived quantity can be written in terms of these base quantities (since these relations are those given by the equations of physics). Each of the base quantities has a dimension associated with it, and the dimension of every derived quantity can be written in terms of the dimensions of the base quantities. Thus, there is a canonical form in which to express the dimension of any derived physical quantity Q. To illustrate, for the SI System, this is stated: dim Q = La Mb Tc Id θε Nf Jg where L is the symbol for the dimension of the base quantity length, M is the symbol for the dimension of the base quantity mass, T is the symbol for the dimension of the base quantity time, I is the symbol for the dimension of the base quantity electric current, θ is the symbol for the dimension of base quantity temperature, N is the symbol for the dimension of the base quantity of amount of substance, and J is the symbol for the dimension of the base quantity of luminous intensity [Taylor and Thompson, 2008, p. 11]. The superscripts a, b, c, d, e, f , and g denote dimensional exponents. (The expression above indicates dimension by the sans serif capital letters shown rather than by the use of brackets.) To put the point in general terms: for any coherent system of units, there is a canonical form in which to express the dimension of any derived physical quantity Q, i.e., using our 13 The official description of the SI System explicitly notes that it is a consequence of using a coherent system of units that “equations between the numerical values of quantities take exactly the same form as the equations between quantities themselves.” [Taylor and Thompson, 2008, p. 12].

Similarity and Dimensional Analysis

813

notation in which brackets indicate the dimension of the quantity: dim Q = [Q1 ]m [Q2 ]n [Q3 ]o [Q4 ]p [Q5 ]r [Q6 ]s [Q7 ]t where Q1 , . . . , Q7 are the symbols for the dimension of the base quantities, whatever they may be, and [Q1 ], . . . , [Q7 ] are symbols for the dimensions of the base quantities. The superscripts m, n, o, p, r, s, and t denote dimensional exponents. The emphasis in the discussion above on the symbols used to represent dimensions is intentional; the proofs deriving consequences from principles of dimensional analysis rely almost entirely on facts about the symbolic representation of dimensions and their manipulation. The dimensions can be manipulated according to the rules of algebraic manipulation, though dimensional equations are of a different sort than the equations of physical science from which they were obtained. The discussion above is meant to provide an appreciation of what is built into the symbolic representation of dimension: it reflects some structural aspects of the empirical content of physics as a whole,14 some simple rules of algebra and standard arithmetic, and conventions of various sorts. It also helps us to see where some of the conventions of the SI system are involved, and in some cases, whether or not they are arbitrary. For example, the choice as to which quantities are taken as base quantities is arbitrary in that there are other sets of quantities that could serve as the set of base quantities. However, once the base quantities are chosen, how the base units are related to each other is not arbitrary, in that the relations between them are determined by the equations of physical science. One of the principles used for the SI System is that the dimension of every quantity, whether base or derived, is unique; that is, there is only one such dimension of canonical form associated with each quantity Q. Since the number of derived quantities is unlimited, and the number of dimensions of canonical form is unlimited, one may ask whether there is a unique quantity associated with each dimension of canonical form. The answer is no: more than one quantity may have a given dimension associated with it, just as more than one quantity may have the same units (heat capacity and entropy are considered physically distinct quantities, though they are both measured in joule/Kelvin; electric current and magnetomotive force are both measured in amperes [Taylor and Thompson, 2008, p. 26]). Thus, one cannot infer from a dimension, the quantity with which that dimension is associated, for the quantity is not uniquely determined. One can make some inferences from dimension, though: if one is using a coherent system of units, one can at least infer what units a quantity has from the dimensions of that quantity. If the dimension is written in canonical form, replacing the dimension by the units used to measure the base quantity with which that dimension is associated will provide the units for that quantity. This is not surprising, for it follows from the fact that every base quantity has a unique dimension and a unique unit associated 14 By “structural aspects of the empirical content of physics as a whole”, I mean the relations between derived and fundamental quantities implied by whatever scientific equations are used in developing a system of units.

814

Susan G. Sterrett

with it, and that every derived quantity has a unique canonical form expressed in terms of the base quantities, that each quantity will have associated with it a unique set of units expressed in terms of the seven base units. 3 DIMENSIONLESS QUANTITIES For some quantities Q, all the dimensional exponents in the expression dim Q = [Q1 ]m [Q2 ]n [Q3 ]o [Q4 ]p [Q5 ]r [Q6 ]s [Q7 ]t will equal zero. Then dim Q = 1, and Q is said to be dimensionless, or of dimension one. The unit associated with it may be considered a derived unit; it is said that such a dimensionless quantity has a coherent derived unit of one. Any ratio of two quantities of the same dimension will be dimensionless, and have dimension one. Thus Mach number and refractive index are dimensionless. The dimension of a plane angle, for which the unit radian is used, is also dimensionless. The SI System also assigns the dimension of one to “counting quantities”, as follows: There are also some quantities that cannot be described in terms of the seven base quantities of the SI at all, but have the nature of a count. Examples are number of molecules, degeneracy in quantum mechanics (the number of independent states of the same energy), and the partition function in statistical thermodynamics (the number of thermally accessible states). Such counting quantities are also usually regarded as dimensionless quantities, or quantities of dimension one, with the unit one, 1. [Taylor and Thompson, 2008, p. 12] The fact that both counting quantities (which take on integral values) and dimensionless parameters used to establish physical similarity (which take on values of real numbers) are of dimension one is not paradoxical. For, in general, although it is true that there is only one canonical dimension per quantity, it is also true that, given a certain dimension, there may be more than one quantity that has the given dimension associated with it. As mentioned in section 1.2 above, similarity of physical systems is established by showing that the dimensionless parameters that characterize the system’s behavior of interest have the same value. The physical meanings of different dimensionless parameters are of course very different, so, if more than one dimensionless parameter is required to characterize the behavior of the system, this is equivalent to saying that the systems need to be similar in both ways in order to be alike with respect to the behavior of interest. The simplest case is the case where the behavior of interest was characterized by a single dimensionless parameter. Some examples of this are that behavior with respect to the existence of shock waves is characterized by Mach number, and behavior with respect to turbulent flow is characterized by Reynolds number. Many dimensionless parameters were in use before the formal basis for the method of physical similarity was developed. These were often conceived of as ratios of two like things, such as two velocities (Mach

Similarity and Dimensional Analysis

815

Number), two lengths (slenderness ratio), or two forces (Froude number, which can be conceived of as the ratio of inertial forces to gravitational forces). Many of the dimensionless ratios in use can be conceived of as the ratio of two like things by rearrangement of the quantities that appear in them (even if they do not appear to be so at first, and even though some algebraic manipulation is required to regroup them into expressions that reveal two like quantities that are physically meaningful). 4 DIMENSIONLESS PARAMETERS AND PHYSICAL SYSTEMS

4.1

Dimensional equations and dimensional homogeneity

Given an equation relating physical quantities, one can write a dimensional equation corresponding to it. This is done by replacing the symbol for the physical quantity with the symbol for dimension associated with it; the standard notation used here is to put square brackets around the symbol for a quantity. If the relation between the quantity and the base quantities of the system is known for each of the physical quantities occurring in the equation, then we have a dimensional equation written solely in terms of the dimensions associated with the basic or fundamental quantities. Thus, for an equation expressing the fact that an elapsed time is equal to the sum of two different distances divided by the two different velocities attained in traversing them that is written: (D1 /v1 ) + (D2 /v2 ) = tT

(i)

(where D1 and D2 stand for designated distances, v1 and v2 stand for designated velocities, and tT stands for the time taken to traverse D1 and D2 ). The dimensional equation corresponding to it, using brackets to indicate dimension of the quantity enclosed in the bracket, does not look that different on first glance: [D1 /v1 ] + [D2 /v2 ] = [tT ]

(ii)

which, when expressed in terms of the dimensions of the basic quantities of the system, becomes: (iii) [L][L]−1 [T ] + [L][L]−1 [T ] = [T ] However, the dimensional equations (ii) and (iii) are not the same type of equation as equation (i) in that they relate dimensions. Here L designates the quantity length and T designates the quantity time. No distinction is made between the two different lengths traversed or the two different velocities attained; what is being related are the dimensions associated with each quantity. Since length and time are taken to be among the basic quantities in a system, we can express the dimensional equation in the basic dimensions [L] and [T ], as we have done in (iii) above. The statement that the equation is dimensionally homogeneous is reflected by a syntactic fact: the exponents of the dimensions [L] and [T ] are the same for the

816

Susan G. Sterrett

terms on each side of the equation. The point is general: if a dimensional equation is stated in terms of a set of dimensions of the base quantities, the criterion of dimensional homogeneity is that the exponents of the dimensions for each basic quantity in a dimensional equation must be the same on both sides of the equation.

4.2 Dimensional homogeneity, dimensionless parameters, and similarity-based inference Much of the practical import of dimensional analysis arises from the powerful and simple logical (or, perhaps, grammatical) principle regarding the requirement of dimensional homogeneity, which is applied to dimensional equations. The principle is known as the principle of dimensional homogeneity. Thus, even if the solution to the associated physical equation is not known, the principle can be employed to obtain information by applying it to dimensional equations. What is somewhat surprising, very striking, and especially significant about the principle of dimensional homogeneity as far as our inquiry into the use of similarity, is that the principle of dimensional homogeneity can produce significant and useful results even in the absence of an equation describing the behavior of interest. Instead, we can apply the principle in a way that makes use of the dimensions of the quantities that must occur in such an equation, even though the equation itself remains unknown. Thus the principle of dimensional homogeneity is a method distinct from the method of transforming equations into dimensionless form. The principle of dimensional homogeneity applied in this way provides useful, significant results by providing similarity criteria from the knowledge of which physical quantities are relevant to the systems and phenomenon of interest. These similarity criteria can then be used to obtain information about one system based on the knowledge of another that is similar to it. The fact that such a significant amount of information can be obtained from a list of the quantities on which a phenomenon depends has often been found somewhat mysterious. Of course, there is a lot of background information that is implicitly being drawn upon in using the principle in this manner: if we are using a coherent system of units, the fact that the dimensions themselves are related by the fundamental equations of a physical science is implicitly involved, so that the information drawn upon in coming up with the similarity criteria certainly involves some scientific equations (e.g., fundamental laws and principles of the science) implicitly, even when an equation describing the behavior or phenomenon of interest in a particular physical system is not at hand. The behavior of interest might be a particular phenomenon, such as the transition to turbulent flow or the generation of shock waves, or it might refer to one aspect of the system’s behavior, such as a temperature distribution, the paths traced out by a particular point or points in the system, a flow velocity profile, or the distribution of forces and stresses within the system. If all that one knows is which physical quantities are the ones upon which a certain phenomenon depends, the principle of dimensional homogeneity can be employed to identify a set of dimensionless ratios upon which the phenomenon

Similarity and Dimensional Analysis

817

depends. The analytical process leading to the identification of the physical quantities upon which the behavior of interest for the system depends when one does not have a governing equation in hand is not entirely a matter of logic and mathematics. A crucial initial step is identifying the physical system whose behavior is to be characterized, and this often involves practical and empirical knowledge required to make the appropriate idealizations and simplifications of the actual system. Then, identifying the physical quantities upon which that system’s behavior depends might involve invoking general scientific principles, such as conservation laws, to identify the quantities involved in force balances or flow balances. For instance, one might use the principle of the conservation of mass and write out mass balances in order to identify all the quantities involved in a system in which mass crosses the system boundary, or one might use principles of equilibrium and draw a free body diagram in order to help identify the physical quantities involved in a mechanical system. The set of dimensionless ratios characterizing the behavior of interest of a certain physical system (and hence, characterizing a class of systems that will be physically similar with respect to the behavior or phenomenon of interest) is not unique. It might seem on first glance that having a set of dimensionless ratios upon which the behavior of interest for a given system depends would not provide much more information, or be of much more use, than merely having a list of the physical quantities upon which the behavior of interest for that system depends. The reason that it is in fact more useful to do so is that, unlike the list of physical quantities, the set of dimensionless ratios that characterizes the behavior of interest for a particular system provides similarity conditions: criteria of similarity of systems with respect to a behavior of interest. The behavior of interest is often indicated by the context and will not always be explicitly identified, but it is important to recognize that although what we are speaking about here is providing formal conditions of similarity, it is always similarity regarding, and hence relative to, some behavior or phenomenon of interest. If all of the dimensionless ratios in such a set have the same value in one particular system as they do in another particular system, then those two systems are physically similar. The set of dimensionless ratios characterizes an unspecified number of physically similar systems. There might be an infinite number of systems whose behavior is characterized by the same set of dimensionless ratios and in which those dimensionless ratios can take on the same values as they take on in the given system. Or, at the other extreme, if the similarity conditions are very restrictive, there might be very few other systems — possibly even none — that are physically similar to a given system. The practical advantage of knowing which ratios (dimensionless parameters) the phenomenon or behavior of interest depends upon, rather than merely knowing which physical quantities the phenomenon or behavior depend upon is that the set of dimensionless ratios that characterizes a certain system’s behavior also provides a way to establish that two different systems are physically similar and, hence, informs a researcher of how to construct a system that will be similar to the given system with respect to the behavior of interest. This is not something that

818

Susan G. Sterrett

can be done directly using a list of physical quantities alone.15 The situation here is just as it is for geometrical similarity: we characterize geometrical shapes by the (dimensionless) ratios relevant to characterizing them, and we establish that two geometrical figures are similar (have the same shape) by showing that all the relevant ratios are the same in one figure as in another. Knowing the dimensionless ratios relevant to geometrically similar figures allows us to construct geometrically similar figures, a strategy used in geometry. Without the knowledge that a certain set of ratios characterizes a certain shape (it need not be a unique set), we could not establish that two geometrical figures are physically similar; whereas, with that knowledge, we are able to do so. Further, the knowledge that certain ratios are equal in the two similar figures or the two similar systems enables the analyst or experimenter to infer the values of a certain physical quantity if the values of all the other physical quantities are known. This step is straightforward: the values can be computed using the fact that the dimensionless ratios are numerically equal, by simple algebraic manipulation. If this manipulation is carried out algebraically, equations relating the value of a given physical quantity in one figure or system in terms of the values of that physical quantity in the second one, to which it is similar. These relations are sometimes referred to as modeling laws, and the relation between corresponding geometrically similar figures or physically similar systems is linear and can be fully described in terms of a scale factor. A common example is the relation expressing the velocities occurring in one system in terms of the velocities in another; these pairs were referred to as corresponding velocities. Often it is desired to know the time interval in one system expressed as a multiple of the time interval in the system modeling it; generally time goes faster in a small scale model than it does in the prototype being modeled. These relations will hold only insofar as the similarity conditions are met, of course.

5

BUCKINGHAM’S THEOREM

The ability to correctly infer the values of quantities in one physical system from knowledge of the values of the quantities in a system that is physically similar to it rests on the ability to correctly establish that the two systems are physically similar. In turn, as we have seen, the ability to correctly establish that two systems are physically similar rests on the ability to identify a set of dimensionless parameters that characterizes the system’s behavior with respect to the behavior or phenomenon of interest. The set of dimensionless parameters is not unique; any such set is sufficient to establish similarity. We now address the question of identifying such sets of dimensionless parameters for a specific problem. One needs not only to find a sufficient set of dimensionless parameters, but also to know that one has found such a sufficient set. 15 The topic of physically similar systems and scientific inference is discussed further in [Sterrett, 2002].

Similarity and Dimensional Analysis

819

One significant point laid out in Buckingham’s paper is that such sets of dimensionless parameters are determined by the principle of dimensional homogeneity, in conjunction with a list of all and only the physical quantities that are relevant to the behavior or phenomenon of interest. Buckingham’s paper also included a result about the minimum number of dimensionless parameters that are required to characterize the system’s behavior and hence the minimum number of dimensionless parameters that are required to establish that two systems are physically similar. This has a practical importance of great value in that it allows one to minimize the number of laboratory experiments that need to be performed in investigations with experimental models, and the paper is often cited for this result. The reasoning in the proof he gave of that result has been criticized for lack of mathematical rigor in some respects. It is, however, quite useful in that it illustrates how to set up dimensional equations and the relations that the exponents must have to each other. The principle of dimensional homogeneity is then applied, and the solution yields the dimensionless parameters in the set. Buckingham’s analysis in setting up the conditions for the theorem proceeds by considering a system undergoing a transformation rather than by considering two distinct physical systems in which the points of one are mapped to the points of the other, reasoning as follows: “Let S be a physical system, and let a relation subsist among a number of quantities Q which pertain to S. Let us imagine S to be transformed into another system S  so that S  ‘corresponds’ to S as regards the essential quantities. There is no point of the transformation at which we can suppose the quantities cease to be dependent on one another; hence we must suppose that some relation will subsist among the quantities Q in S  which correspond to the quantities Q in S . . .. We have to enquire what sort of transformation would lead to [the result that] two systems shall be similar as regards a given physical relation.” [Buckingham, 1914, p. 353] This conception of similarity thus involves the preservation of whatever it is that is required to ensure similarity between the system at any point in the process and the system at any other point in the process (i.e., this might be thought of as the “shape” of the system, which is what sameness of the dimensionless parameters that characterize the system throughout the process ensures). Buckingham presented the application to experimental scale models as a special case of a system undergoing a transformation from one size to another. Some later treatments providing a more mathematically-based proof are: [Pankhurst, 1964; Langhaar, 1951; Duncan, 1953] and [Palacios, 1964]. Here we put aside controversies about the proof of the theorem itself, as we are mainly concerned with understanding the premises of the theorem, the conclusion of the theorem, and the role of the principle of dimensional homogeneity in it. The premises of the theorem are that: (a) There is a dimensionally homogeneous relation between the n quantities p1 , p2 , p3 , . . . pn upon which a certain phenomenon or behavior depends. We denote this by the equation f (p1 , p2 , p3 , . . . pn ) = 0. The quantities p1 , p2 , p3 , . . . pn related by this equation include all the ones upon which the

820

Susan G. Sterrett

phenomenon depends, and the phenomenon depends upon all the quantities on the list. (b) Applying the principle of dimensional homogeneity to the dimensional equation associated with f (p1 , p2 , p3 , . . . pn ) = 0 yields k independent equations relating the exponents in the dimensional equation. In general, there will be one independent equation for each of k base quantities Q1, , Q2 , Q3 , . . . , Qk , where k is the minimum number of base quantities required to express the n quantities p1 , p2 , p3 , . . . , pn in terms of base quantities. The conclusion of the theorem is this: It is a consequence of the principle of dimensional homogeneity that the equation f (p1 , p2 , p3 , . . . pn ) = 0 is reducible to an equation relating (n−k) independent dimensionless parameters π1 , π2 , π3 , . . . , πn−k , where each dimensionless parameter is composed only of products that are among the k base quantities. That is, f (p1 , p2 , p3 , . . . , pn ) = 0 is reducible to an equation of the form f (π1 , π2 , π3 , . . . , πn−k ), i.e., an equation relating (n − k) dimensionless parameters. The set (π1 , π2 , π3 , . . . , πn−k ) will not in general be unique. As explained in section 4.2 above, the advantage of knowing that there exists an equation relating the dimensionless πs is that this knowledge provides similarity criteria for physically similar systems — even if the equation relating the πs is not known. Knowledge of such a set of dimensionless parameters is sufficient for us to determine how to construct a physical system such that it will be physically similar (with respect to a certain behavior or phenomenon of interest) to another given system, if it is possible to do so. Buckingham illustrated the method of physically similar systems with a concrete but fairly involved example: dynamic similarity of a screw propeller. Writing an equation that describes the dynamic behavior of the propeller is difficult, but identifying the physical quantities upon which the behavior of the propeller depends is less so. He identifies the following seven physical quantities as those on which the propeller’s dynamic behavior depends: force, characteristic length and all the length ratios that are required to characterize the shape of the propeller, revolutions per unit time, speed of advance, density and viscosity of the fluid the propeller is immersed in, and gravitational acceleration. Thus there is an equation relating all seven of these quantities; what it is is not specified. There are only three base quantities required (which, in the SI system, would be mass, length, and time) to express the quantities in this list. Thus the minimum number of dimensionless parameters required to characterize the system is 7 − 3 = 4. Here is one of the sets of dimensionless parameters he derives for this problem:   2 2 ρD S ρDi n2 μ2 ρD3 g , , , F F Fρ F where ρ is density, D is diameter (chosen as the characteristic length), S is speed of advance, F is force exerted by the propeller, n is the number of revolutions per unit time, μ is viscosity, and g is gravitational acceleration. These are the

Similarity and Dimensional Analysis

821

πs in the equation mentioned earlier to which the equation in terms of the variables f (ρ, D, S, F, n, μ, g) can be reduced: f (π1 , π2 , π3 , . . . , πn−k ), where f is an undetermined function. 6 SIMILARITY “IN CERTAIN RESPECTS”, COMPLETE SIMILARITY, AND PARTIAL SIMILARITY We now consider more specifically what similarity amounts to. To establish that two physical systems, each of which is characterized by the same set of dimensionless quantities (π1 , π2 , π3 , . . . , πk ) are physically similar, what we have to establish is that the value of π1 in System A is equal to the value of π1 in System B, and that the value of π2 in System A is equal to the value of π2 in System B, and so on for each of the k dimensionless parameter πs. That is what is required to achieve full physical similarity between two systems, with respect to the phenomena and behavior of interest. However, in practice, often only partial similarity is attempted. There are several different reasons why only partial similarity might be attempted. Sometimes the requirement of geometric similarity is deliberately violated in the construction of a model. There are cases in which the benefits of complete similarity, i.e., of keeping the distance relations the same between two systems, are outweighed by some other benefit. Geometrically distorted models (sometimes referred to simply as distorted models) are used extensively in coastal and river models. In order to be able to model large areas with reasonably sized models yet not have the surface tension of the water in the model play an exaggerated role in the model, the vertical dimensions are not scaled down as much as the horizontal dimensions are. Then, there will be more than one scale factor involved for the quantity of length. This is putting things in a somewhat oversimplified manner, as the reasoning about the details of modeling can be quite involved and employ sophisticated mathematics and physics [Hughes, 1993]. Sometimes full similarity is preferred, but it is not possible to achieve it. There are many modeling problems for which the only possible system that is physically similar to the given system is one of the same size, so that, no matter what the resources at hand, it is not possible to build a scale model of a size other than a fullsize one, and still achieve full similarity. This is not a matter of practicality; it is often the case that even in principle it is impossible to simultaneously satisfy all the requirements specified by keeping the value of all the πs on which the phenomena of interest depend the same in the model as in the prototype. So, most models, even geometrically similar ones, will be distorted in the technical sense that not all the dimensionless πs are the same in the model as in the prototype [Hughes, 1993]. The term distorted model is generally reserved for geometrically distorted models, though, and the fact that not all the criteria for physical similarity are met is instead indicated by saying that only partial similarity has been achieved between model and prototype. What is often done is to build one model that aims for similarity with respect to one or only some of the πs, and another model

822

Susan G. Sterrett

that aims for similarity with respect to another one of the πs. Thus, the usual situation is that one has several scale models, none of which has achieved complete physical similarity with the prototype, but each of which is partially similar to it in a prescribed way, and which is used only for inferences with respect to the phenomena or behavior associated with the dimensionless parameters that are the same between model and prototype. The points about the practice of scale modeling hold for the use of physical similarity between systems in general. That is, statements of similarity are by their very nature similarity relative to a certain phenomenon or behavior of the system. Even full similarity is only similarity relative to a certain phenomenon or behavior of the system. Yet there are methods of establishing similarity relative to different phenomena and different system behaviors — as we have seen, we can derive the dimensionless parameters (ratios) that characterize a certain phenomenon or behavior of a system from knowledge about it that does not always entail that we can describe the system behavior in terms of an equation, much less in terms of one for which we know how to obtain solutions. So similarity is always relative similarity, just as in mechanics velocity is always relative velocity. Partial similarity is a different notion than relative similarity, but it, too, is a matter of similarity with respect to a behavior of interest. However, none of these points implies that we cannot say what similarity consists in. Far from it. We can say rather clearly what similarity consists in and what it is based upon, even in the cases in which it is not achievable, and we can say a lot about the consequences that we can reliably draw using it. BIBLIOGRAPHY [Becker, 1976] H. A. Becker. Dimensionless Parameters: Theory and Methodology. Wiley, 1976. [Birkhoff, 1960] G. Birkhoff. Hydrodynamics: A Study in Logic, Fact and Similitude. Princeton University Press, 1960. [Bridgman, 1931] P. Bridgman. Dimensional Analysis. Yale University Press, 1931. [Buckingham, 1914] E. Buckingham. On Physically Similar Systems: Illustrations of the Use of Dimensional Equations, Physical Review 4: 345-376, 1914. [Campbell, 1920] N. R. Campbell. Physics: The Elements. Cambridge, 1920. [Cohen, 2001] D. L. Cohen. Demystifying Electromagnetic Equations: A Complete Explanation of EM Unit Systems and Equation Transformations. SPIE Press, 2001. [Cornelius et al., 1964] P. Cornelius, W. de Groot and R. Vermeulen. Quantity Equations and System Variation in Electricity, Physica 30: 1446-1452, 1964. [Cornelius et al., 1965a] P. Cornelius, W. de Groot and R. Vermeulen. Quantity equations, rationalization and change of number of fundamental quantities I, Applied Scientific Research, Section B, Vol. 12, No. 1. Pp. 1-17, 1965. [Cornelius et al., 1965b] P. Cornelius, W. de Groot and R. Vermeulen. Quantity equations, rationalization and change of number of fundamental quantities II, Applied Scientific Research, Section B, Vol. 12, No. 1. Pp. 235-247, 1965. [Cornelius et al., 1965c] P. Cornelius, W. de Groot and R. Vermeulen. Quantity equations, rationalization and change of number of fundamental quantities III, Applied Scientific Research, Section B, Vol. 12, No. 1, January 1965. Pp. 248-265, 1965. [Duncan, 1953] W. J. Duncan. Physical Similarity and Dimensional Analysis. Edward Arnold and Co., 1953. [Ehrenfest-Afanassjewa, 1916] T. Ehrenfest-Afanassjewa. Der Dimensionsbegriff und der analytische Bau physikalischer Gleichungen, Mathematische Annalen, 1916.

Similarity and Dimensional Analysis

823

[Ellis, 1966] B. D. Ellis. Basic Concepts of Measurement. Cambridge University Press, 1966. [Ellis, 1992] B. D. Ellis. Conventialism in Measurement Theory. In C. W. Savage and P. Ehrlich, eds., Philosophical and Foundational Issues in Measurement Theory. Lawrence Erlbaum, pp. 167–180, 1992. [Giere, 2004] R. Giere. How Models Are Used To Represent Reality, Philosophy of Science, Vol. 71, December 2004. [Goodman, 1972] N. Goodman. Problems and Projects. Bobbs-Merrill, 1972. [Hottel, 1961] H. C. Hottel. Fire Modeling. In The Use of Models in Fire Research, Publication 786, National Academy of Sciences, National Research Council, Washington, D. C., 1961. [Hughes, 1993] S. A. Hughes. Physical Models and Laboratory Techniques in Coastal Engineering. World Scientific, 1993. [Krantz et al., 1971/2007] D. H. Krantz et al. Foundations of Measurement, Vol. I Additive and Polynomial Representations. Academic Press, 1971. Republished by Dover Publications, 2007. [Kroes, 1989] P. Kroes. Structural Analogies between Physical Systems. British Journal for the Philosophy of Science 40, 145-154, 1989. [Lange, forthcoming] M. Lange. Dimensional explanations, Noˆ us, forthcoming. [Langhaar, 1951] H. L. Langhaar. Dimensional Analysis and Theory of Models. John Wiley, 1951. [Lodge, 1888] A. Lodge. The Multiplication and Division of Concrete Quantities, Nature 38: 281-283. 1888. [Maxwell, 1890] J. C. Maxwell. Dimensions. In The Encyclopaedia Britannica: A Dictionary of Arts, Sciences, and General Literature, Ninth Edition, Volume VII. Henry G. Allen Company, Publishers, 1890. Pp. 240-242. [McKirahan, 1994] R. McKirahan. Philosophy Before Socrates: An Introduction with Texts and Commentaries. Hackett Publishing, 1994. [Palacios, 1964] J. Palacios. Dimensional Analysis. (Translated from the Spanish by P. Lee and L. Roth), Macmillan, 1964. [Pankhurst, 1964] R. C. Pankhurst. Dimensional Analysis and Scale Factors. Butler and Tanner, 1964. [Sedov, 1959] L. I. Sedov. Similarity and Dimensional Methods in Mechanics. Academic Press, 1959. [Slokarnik, 2006] M. Slokarnik. Scale-Up in Chemical Engineering. Wiley, 2006. [Sterrett, 2006] S. G. Sterrett. Models of Machines and Models of Phenomena, International Studies in Philosophy of Science. Vol. 20 2006, p. 69-80. [Sterrett, 2005] S. G. Sterrett. Wittgenstein Flies A Kite: A Story of Models of Wings and Models of the World. Pi Press (Penguin Books), 2005. [Sterrett, 2002] S. G. Sterrett. Physical Models and Fundamental Laws: Using One Piece of the World to Tell About Another, Mind and Society, Vol. 3, March 2002, p. 51-66. [Taylor and Thompson, 2008] B. N. Taylor and A. Thompson, eds. The International System of Units (SI ). NIST Special Publication 330, 2008 Edition. ( United States version of the English text of the eighth edition (2006) of the International Bureau of Weights and Measures publication Le Syst` eme International d’ Unit´ es (SI )). National Institute of Standards and Technology. Issued March 2008. http://physics.nist.gov/Pubs/SP330/sp330.pdf downloaded June 23, 2008. [Vaschy, 1892] A. Vaschy. Sur les lois de similitude en physique. Annales T´ el´ egraphiques, 1892. Pp. 25-28.

This page intentionally left blank

MEASUREMENT THEORY AND ENGINEERING Patrick Suppes

1

INTRODUCTION

A conceptual analysis of measurement can properly begin by formulating the two fundamental problems of any measurement procedure. The first problem is that of representation, justifying the assignment of numbers to objects or phenomena. What we must show is that the structure of a set of phenomena under certain empirical operations and relations is the same as the structure of some set of numbers under corresponding arithmetical operations and relations. Solution of the representation problem for a theory of measurement does not completely lay bare the structure of the theory, for there is often a formal difference between the kind of assignment of numbers arising from different procedures of measurement. This is the second fundamental problem, determining the scale type of a given procedure. The scale type is based on the proof of an invariance theorem for the representation. This is another way of stating the second fundamental problem. In this introduction we expand on this general viewpoint. Section 2 then focuses on four types of measurement that exemplify, in the simple mathematical setting of finite equally-spaced models, the detailed nature of the representation and invariance theorem. Section 3 considers in greater detail the important case of extensive measurement, with special attention to qualitative approximation with upper and lower measures. Section 4 develops a fairly general qualitative theory of the probability of error. Section 5 contains brief remarks on several topics of general interest that could not be covered in detail.

1.1

Numerical representation

A representation of something is an image, model, or reproduction of that thing. References to representations are familiar and frequent in ordinary discourse. Some typical instances are these: Sleep is a certain image and representation of death. The Play is a representation of a world I once knew well. Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. © 2009 Elsevier BV. All rights reserved.

826

Patrick Suppes

In some cases we can think of a representation as improving our understanding of the object represented. Many of us certainly understand the proportions of a building better — especially the layout of the interior — after examining its architectural drawings. The formal or mathematical theory of representation has as its primary goal such an enrichment of the understanding, although there are other goals of representation of nearly as great importance — for instance, the use of numerical representations of measurement procedures to make computations more efficient. Before turning to isomorphism of models of a theory as the general formal approach to representation, it is important to characterize two models or structures of a theory being isomorphic. Roughly speaking, two models of a theory are isomorphic when they exhibit the same structure from the standpoint of the basic concepts of the theory. The point of the formal definition of isomorphism for a particular theory is to make this notion of same structure precise. It is to be emphasized, however, that the definition of isomorphism of models of a theory is not dependent on the detailed nature of the theory, but is in fact sufficiently independent often to be termed ‘axiom-free’. The use of the phrase ‘axiom-free’ indicates that the definition of isomorphism depends only on the set-theoretical character of models of a theory. Thus two theories whose models have the same set-theoretical character, but whose substantive axioms are quite different, would use the same definition of isomorphism. These ideas may be made more definite by giving the definition of isomorphism for algebras. Here a structure (A, ◦,−1 , e) is an algebra if A is a nonempty set, ◦ is a binary operation from A × A to A, e is an element of A, and −1 is a unary operation from A to A. DEFINITION 1. An algebra A = (A, ◦,−1 , e) is isomorphic to an algebra A =  (A , ◦ ,−1 , e ) if and only if there is a function f such that (i) the domain of f is A and the range of f is A , (ii) f is a one-one function, (iii) if a and b are in A, then f (a ◦ b) = f (a) ◦ f (b), 

(iv) if a is in A, then f (a−1 ) = f (a)−1 , (v) f (e) = e . When we ask ourselves whether or not two distinct objects have the same structure, we obviously ask relative to some set of concepts under which the objects fall. It is an easy matter to show that the relation of isomorphism just defined is an equivalence relation among algebras, i.e., it is reflexive, symmetric, and transitive. As a rather interesting example, we might consider two distinct but isomorphic groups which have application in the theory of measurement. Let one group be

Measurement Theory and Engineering

827

the additive group of integers. In this case, the set A is the set of all integers, the operation ◦ is the operation of addition, the inverse operation −1 is the negative operation, and the identity element e is 0. As the second group, isomorphic to the first, consider the multiplicative group of all integer powers of 2. In this case, the set A is the set of all numbers that are equal to 2 to some integer power, the operation ◦ is the operation of multiplication, the inverse operation is the standard reciprocal operation, i.e., the inverse of x is 1/x, and the identity element is the integer 1. To establish the isomorphism of the two groups A = (A, +, −, 0) and A = (A , ·,−1 , 1), we may use the function f such that for every integer n in the set A f (n) = 2n . Then it is easy to check that the range of f is A , that f is one-one, and f (m ◦ n) = f (m + n) = 2m+n = 2m · 2n = f (m) · f (n) = f (m) ◦ f (n) f (n−1 ) = f (−n) = 2−n = and

 1 = f (n)−1 , 2n

f (0) = 20 = 1.

It should be apparent that the same isomorphism between additive and multiplicative groups is possible if we let the set of objects of the additive group be the set of all real numbers, positive or negative, and the set of objects of the multiplicative group be the set of all positive real numbers. From the standpoint of the theory of measurement, this isomorphism is of interest primarily because it means that there is no mathematical basis for choosing between additive and multiplicative representations. In attempting to characterize the nature of the models of a theory, the notion of isomorphism enters in a central way. Perhaps the best and strongest characterization of the models of a theory is expressed in terms of a significant representation theorem. As outlined informally earlier, by a representation theorem for a theory the following is meant. A certain class of models of a theory, distinguished for some intuitively clear conceptual reason, is shown to exemplify within isomorphism every model of the theory. More precisely, let M be the set of all models of a theory, and let B be some distinguished subset of M. A representation theorem for M with respect to B would consist of the assertion that given any model M in M there exists a model in B isomorphic to M. In other words, from the standpoint of the theory every possible variation of model is exemplified within the restricted set B. It should be apparent that a trivial representation theorem can always be proved by taking B = M. A representation theorem is just as interesting as the intuitive significance of the class B of models and no more so. Homomorphism of models. In many cases within pure mathematics a representation theorem in terms of isomorphism of models turns out to be less interesting than a representation theorem in terms of the weaker notion of homomorphism.

828

Patrick Suppes

A good example of this sort is provided by theories of measurement, and the generalization from isomorphism to homomorphism can be illustrated in this context. When we consider general practices of measurement it is evident that in terms of the structural notion of isomorphism we would, roughly speaking, think of the isomorphism as being established between an empirical model of the theory of measurement and a numerical model. By an empirical model we mean a model in which the basic set is a set of empirical objects and by a numerical model one in which the basic set is a set of numbers. However, a slightly more detailed examination of the question indicates that difficulties about isomorphism quickly arise. In all too many cases of measurement, distinct physical objects are assigned the same number, and thus the one-one relationship required for isomorphism of models does not hold. Fortunately, this is the only respect in which we must change the general notion, in order to obtain an adequate account for theories of measurement of the relation between empirical and numerical models. The general notion of homomorphism is designed to accommodate exactly this situation. To obtain the formal definition of homomorphism for two algebras as previously defined, we need only drop the requirement that the function establishing the isomorphism be one-one. When this function is many-one but not one-one, we have a homomorphism that is not an isomorphism.

1.2 Invariance in theories of measurement In connection with any measured property of an object, or set of objects, it may be asked how unique is the number assigned to measure the property. For example, the mass of a pebble may be measured in grams or pounds. The number assigned to measure mass is unique once a unit has been chosen. The measurement of temperature in ◦ C or ◦ F has different characteristics. Here an origin as well as a unit is arbitrarily chosen. Other formally different kinds of measurement are exemplified by (1) the measurement of probability, which is absolutely unique, and (2) the ordinal measurement of such physical properties as hardness of minerals, or such psychological properties as intelligence and racial prejudice. Use of these different kinds of transformations is basic to the main idea of this chapter. An empirical hypothesis, or any statement in fact, which uses numerical quantities is empirically meaningful only if its truth value is invariant under the appropriate transformations of the numerical quantities involved. As an example, suppose a psychologist has an ordinal measure of I.Q., and he thinks that scores S(a) on a certain new test T have ordinal significance in ranking the intellectual ability of people. Suppose further that he is able to obtain the ages A(a) of his subjects. The question then is: Should he regard the following hypothesis as empirically meaningful? HYPOTHESIS 1. For any subjects a and b, if S(a)/A(a) < S(b)/A(b), then I.Q.(a) < I.Q.(b).

Measurement Theory and Engineering

829

From the standpoint of the invariance characterization of empirical meaning, the answer is negative. To see this, let I.Q.(a) ≥ I.Q.(b), let A(a) = 7, A(b) = 12, S(a) = 3, S(b) = 7. Make no transformations on the I.Q. data, and make no transformations on the age data. But let φ be an increasing transformation which carries 3 into 6 and 7 into itself. Then we have 7 3 < , 7 12 but 7 6 ≥ , 7 12 and the truth value of Hypothesis 1 is not invariant under φ. The empirically significant thing about the transformation characteristic of a quantity is that it expresses in precise form how unique the structural isomorphism is between the empirical operations used to obtain a given measurement and the corresponding arithmetical operations or relations. If, for example, the empirical operation is simply that of ordering a set of objects according to some characteristic, then the corresponding arithmetical relation is that of less than (or greater than), and any two functions which map the objects into numbers in a manner preserving the empirical ordering are adequate. It is then easy to show that, if f1 and f2 are adequate in this sense, then they are related by a monotone-increasing transformation. Only those arithmetical operations and relations which are invariant under monotone-increasing transformations have any empirical significance in this situation. A representation theorem should ordinarily be accompanied by a matching invariance theorem stating the degree to which a representation of a structure is unique. In the mathematically simple and direct cases it is easy to identify the group as some well-known group of transformations. For more complicated structures, for example, structures that satisfy the axioms of a scientific theory, it may be necessary to introduce more complex apparatus, but the objective is the same, to wit, to characterize meaningful concepts in terms of invariance. One note to avoid confusion: it is when the concepts are given in terms of the representation, for example, a numerical representation in the case of measurement, or representation in terms of Cartesian coordinates in the case of geometry, that the test for invariance is needed. When purely qualitative relations are given, which are defined in terms of the qualitative primitives of a theory, for example, those of Euclidean geometry, then it follows at once that the defined relations are invariant and therefore meaningful. On the other hand, the great importance of the representations is the reduction in computations and notation they achieve, as well as understanding of structure. This makes it imperative that we have a clear conception of invariance and meaningfulness for representations which may be, in appearance, rather far removed from the qualitative structures that constitute models of the theory.

830

1.3

Patrick Suppes

Scales of measurement

Counting is an example of an absolute scale. The number of members of a given collection of objects is determined uniquely. There is no arbitrary choice of a unit or zero to be made. In contrast, the measurement of mass or weight is an example of a ratio scale. Any empirical procedure for measuring mass does not determine the unit of mass. The choice of a unit is an empirically arbitrary decision made by an individual or group of individuals. Of course, once a unit of measurement has been chosen, such as the gram or pound, the numerical mass of every other object in the universe is uniquely determined. Another way of stating this is to say that the measurement of mass is unique up to multiplication by a positive constant. (The technical use of ‘up to’ will become clear later.) The measurement of distance is a second example of measurement of this sort. The ratio of the distance between Palo Alto and San Francisco to the distance between Washington and New York is the same whether the measurement is made in miles or kilometers. To avoid certain common misunderstandings, amplification is needed of the claim that no empirical procedure for measuring mass determines the unit of mass. A chemist, measuring a sample of a certain ferric salt on an equal-arm balance with a standard series of metric weights, might find this claim surprising, for he might suppose that the selection of a standard series of metric weights had fixed as part of his empirical procedure one gram as the unit of measurement. There are at least two lines of argument that should prove effective in convincing the chemist that his supposition was incorrect. In the first place, it would be pertinent to point out that exactly the same information could be expressed by converting this final measurement into a measurement in pounds or ounces. To express the measurement in this latter form, no further empirical operations with the equal-arm balance would have to be performed. But, it must be admitted, the chemist could well reply that, although no further empirical operations had to be performed with the present balance, the use of the conversion factor from grams to pounds entails appealing to empirical operations previously performed by some Bureau of Standards in determining the conversion factor. An analysis of his retort takes us to the second line of argument, which goes deeper and is more fundamental. To begin with, his appeal to previous empirical operations on other balances may be turned back on him, for to justify labeling his measurement with his given standard series as a measurement in grams, appeal must also be made to previously performed empirical operations, namely, those which constituted the calibration of his series of weights as a standard metric series. The important point is that the empirical operations performed by the chemist himself establish no more or less than the ratio of the mass of the ferric salt sample to a subset of weights in his standard series. And the same kind of ratio statement may be made about the empirical operations that led to the calibration of his standard series by the technicians of the firm which produced the series. In other words, the statement by the chemist:

Measurement Theory and Engineering

831

(1) This sample of ferric salt weighs 1.679 grams may be replaced by the statement: (2) The ratio of the mass of this sample of ferric salt to the gram weight of my standard series is 1.679, and the manufacturer of my series has certified that the ratio of my gram weight to the standard kilogram mass of platinumiridium alloy at the International Bureau of Weights and Measures, near Paris, is .0010000. The measurement of temperature is an example of a third formally distinct kind of measurement mentioned earlier. An empirical procedure for measuring temperature by use of a thermometer determines neither a unit nor an origin. In this sort of measurement the ratio of any two intervals is independent of the unit and zero point of measurement. For obvious reasons, measurements of this kind are called interval scales. Examples other than measurement of temperature are measurements of temporal dates, linear position, or cardinal utility. In terms of the notion of absolute, ratio, and interval scales we may formulate the second fundamental problem for any exact analysis of a procedure of measurement: determine the scale type of the measurements resulting from applying the procedure. The general notion of a scale is not something we need to define in an exact manner for subsequent developments. For the purpose of systematizing some of the discussion of the uniqueness problem, we may define a scale as a class of measurement procedures having the same transformation properties. Examples of three different scales have already been mentioned, namely, counting as an absolute scale, measurement of mass as a ratio scale, and measurement of temperature as an interval scale. Moh’s hardness scale, according to which minerals are ranked in regard to hardness as determined by a scratch test, and the Beaufort wind scale, whereby the strength of a wind is classified as calm, light air, light breeze, etc., are examples of a fourth, weak kind of scale. The social sciences abound with such ordinal scales, the significance of which is discussed below. Numbers are also sometimes used for pure classification. For example, in some states the first number on an automobile license indicates the county in which the owner lives. The assignment of numbers in accordance with such a scale is arbitrary except for the assignment of the same number to people in the same county, and distinct numbers to people in distinct counties. This weakest scale is one where numbers are used simply to name an object or person. The assignment is arbitrary. Draft numbers and the numbers of football or soccer players are examples. Such scales are usually called nominal scales. We have distinguished, in an intuitive fashion, five types of scales. We will in the following pages primarily confine our discussion to an analysis of the first four. They are not, however, the only scales of measurement, and we shall upon occasion mention others. We now want to characterize each of these five scales in terms of their transformation properties. For future reference each of the five transformations mentioned

832

Patrick Suppes

is formally defined. To begin with, an absolute scale is unique up to the identity transformation. DEFINITION 2. The identity transformation on the set of real numbers is the function f such that for every real number x, f (x) = x. So far, counting has been our only example of an absolute scale. A ratio scale is unique up to a similarity transformation. DEFINITION 3. Let f be a real-valued function whose domain is the set of real numbers. Then f is a similarity transformation if and only if there exists a positive real number α such that for every real number x f (x) = αx. Measurement of mass and of distance have been our two examples of ratio scales. An interval scale is unique up to a linear transformation. DEFINITION 4. Let f be a real-valued function whose domain is the set of real numbers. Then f is a linear transformation if and only if there is a positive number α and a number β such that for every number x f (x) = αx + β. If in the measurement of temperature we wish to convert x in degrees Fahrenheit to Celsius we use the linear transformation defined by α = 59 and β = − 160 9 . That is, 5 160 5 . y = (x − 32) = x − 9 9 9 Obviously every similarity transformation is a linear transformation with β = 0. An ordinal scale is unique up to a monotone transformation. DEFINITION 5. Let f be a real-valued function whose domain is some set of real numbers. Then f is a monotone transformation if and only if f is a monotoneincreasing transformation or a monotone-decreasing transformation, i.e., if x < y then f (x) < f (y) – monotone increasing, and if x < y then f (x) > f (y) – monotone decreasing. The classificatory analysis of this section is summarized in Table 1. 2

FINITE EQUALLY SPACED MODELS OF MEASUREMENT

In this section some of the simplest nontrivial examples of measurement structures are considered first. The basic sets of objects or stimuli will in all cases be finite, and the adequacy of the elementary axioms for various structures depends heavily on this finiteness. In addition to their finiteness, the distinguishing characteristic of the structures considered is that the objects are equally spaced in an appropriate sense along the continuum, so to speak, of the property being measured. The

Measurement Theory and Engineering

833

Table 1. Classification of scales and measurement Scale Absolute Ratio

Interval

Ordinal

Nominal

Transformation Properties Identity Similarity (Multiplication by a positive number) Linear (multiplication by a positive number and addition of an arbitrary number) Monotone or monotone increasing 1-1

Examples Counting, relative frequency Mass, distance, electric current, voltage Temperature, potential energy, linear position, cardinal utility Moh’s scale, Beaufort wind scale, qualitative preferences Phone numbers, Social Security numbers

restrictions of finiteness and equal spacing enormously simplify the mathematics of measurement, but it is fortunately not the case that the simplification is accompanied by a total separation from realistic empirical applications. Finiteness and equal spacing are characteristic properties of many standard scales, for example, the ordinary ruler, the set of standard weights used with an equal-arm balance in the laboratory, or almost any of the familiar gauges for measuring pressure, temperature, or electric current.

2.1

Extensive measurement

The axioms of extensive measurement are developed in this section with three specific interpretations in mind. One is for the measurement of mass on an equalarm balance, one is for the measurement of length of rigid rods, and one is for the measurement of subjective probabilities. Other interpretations are certainly possible, but detailed remarks are restricted to these three. From a formal standpoint the basic structures are triples Ω, F,  where Ω is a nonempty set, F is a family of subsets of Ω and the relation  is a binary relation on F. By using subsets of Ω as objects, we avoid the need for a separate primitive concept of concatenation. As a general structural condition, it shall be required that F be an algebra of sets on Ω, which is just to require that F be nonempty and be closed under union and complementation of sets, i.e., if A and B are in F then A ∪ B and −A are also in F. The intended interpretation of the primitive concepts for the three cases mentioned is fairly obvious. In the case of mass, Ω is a set of physical objects, and for two subsets A and B, A  B if and only if the set A of objects is judged at least as

834

Patrick Suppes

heavy as the set B. It is probably worth emphasizing that several different uses of the equal-arm balance are appropriate for reaching a judgment of comparison. For example, if A = {a, b} and B = {a, c} it will not be possible literally to put A on one pan of the balance and simultaneously B on the other, because the object a is a member of both sets. But we can make the comparison in at least two different ways. One is just to compare the nonoverlapping parts of the two subsets, which in the present case just comes down to the comparison of b and c. A rather different empirical procedure that even eliminates the need for the balance to be equal arm is to first just balance A with sand on the other pan (or possibly water; but in either case, sand or water in small containers), and then to compare B with this fixed amount of sand. Given the standard meaning of the set-theoretical operations of intersection, union, and complementation, no additional interpretation of these operations is required, even of union of sets, which serves as the operation of concatenation. In the case of the rigid rods, the set Ω is just the collection of rods, and A  B if and only if the set A of rods, when laid end to end in a straight line, is judged longer than the set B of rods also so laid out. Variations on exactly how this qualitative comparison of length is to be made can easily be supplied. In the case of subjective probabilities or objective propensities, the set Ω is the set of possible outcomes of the experiment or empirical situation being considered. The subsets of Ω in F are just events in the ordinary sense of probability concepts, and A  B if and only if A is judged at least as probable as B. Axioms for extensive measurement, subject to the two restrictions of finitude and equal spacing, are given in the following definition. In the definition, and subsequently, we use the standard definitions for equivalence ≈ in terms of a weak ordering and also of a strict ordering. The definitions are just these: A ≈ B if and only if A  B and B  A; A B if and only if A  B, and not B  A. Also ∅ is the symbol for the empty set. DEFINITION 6. A structure Ω = (Ω, F, ) is a finite, equally spaced extensive structure if and only if Ω is a finite set, F is an algebra of sets on Ω, and the following axioms are satisfied for every A, B, and C in F: 1. The relation  is a weak ordering of F; 2. If A ∩ C = ∅ and B ∩ C = ∅, then A  B if and only if A ∪ C  B ∪ C; 3. A  ∅; 4. Not ∅  Ω; 5. If A  B then there is a C in F such that A ≈ B ∪ C. From the standpoint of the standard ideas about the measurement of mass or length, it would be natural to strengthen Axiom 3 to assert that if A = ∅, then A ∅, but because this is not required for the representation theorem and is unduly restrictive in the case of probabilities, the weaker axiom seems more appropriate.

Measurement Theory and Engineering

835

In stating the representation theorem, we use the notion of an additive measure μ from F to the real numbers, i.e., a function μ such that, for any A and B in F, i.

μ(∅) = 0,

ii.

μ(A) ≥ 0,

iii.

if A ∩ B = ∅ then μ(A ∪ B) = μ(A) + μ(B).

THEOREM 1. Let Ω = (Ω, F, ) be a finite, equally spaced extensive structure. Then there exists an additive measure μ such that for every A and B in F μ(A) ≥ μ(B) if and only if A  B. Moreover, all singleton sets fall into at most two equivalence classes in F ; if there are two, one of these equivalence classes contains the empty set. The elementary proof of this theorem and some of the preceding discussion is to be found in Suppes [1969, pp. 4–8]. THEOREM 2. Let A = (A, F, ) be a finite, equally spaced extensive structure. Then a numerical representation of A is invariant under the group of similarity transformations of Table 1. It is important to point out that, for all three of the empirical interpretations of the primitive concepts used in the theory of extensive measurement characterized in Definition 1, many important empirical procedures required for each interpretation are missing. Students in elementary engineering or physics laboratories are taught many details about measuring mass, or weight if you will, with equal-arm balances, or length with measuring rods or tapes. But even these details are inadequate as an account of fundamental measurement. For such measurement must be used at an earlier stage to validate the standard weights or rods used by the students. The empirical procedures are much more elaborate for creating a set of standard weights, for example. The axioms of Definition 1 come into play in creating ‘equal’ standard weights or checking that a given set of standard weights does seem satisfactory. I say ‘catedseem’ deliberately, for there is no natural end to the empirical investigation of standard measurement procedures or instrumentation in physics or other sciences. The art of creating good measuring instruments is a complicated and subtle one, certainly so at the frontier of many sciences. Such art, I have argued repeatedly in previous publications, cannot be adequately described in a manual, but requires an apprenticeship of training, many components of which are nonverbal. Can you imagine, for example, becoming a good tennis player by just reading a book about it or listening to a few lectures? The same is true of any laboratory or manufacturing set of skills required to produce good measurement instruments.

836

Patrick Suppes

2.2 Difference measurement Referring to the distinction between extensive and intensive properties discussed earlier, I could easily make a case for entitling this subsection intensive measurement, for it is characteristic of difference measurement that no operation corresponding to addition is present, and no meaningful combination of objects or stimuli is postulated for the difference structures. As before, the basic set of objects or stimuli will be nonempty and finite, but in the case of difference structures the relation on the set will be a quaternary one. I denote the basic set of objects by A and the quaternary relation by . The idea behind the quaternary relation  is that ab  cd holds when and only when the qualitative (algebraic) difference between a and b is at least as great as that between c and d. In the case of similarity judgments, for example, the relation  would hold when the subject in a psychological experiment judged that the similarity between a and b was at least as great as the similarity between c and d, due account being taken of the algebraic sign of the difference. The inclusion of the algebraic difference requires some care in interpretation; for example, in many similarity experiments a natural algebraic sign is not attached to the similarity. Instances that satisfy the present requirement are judgments of utility or of pitch or of intensity of sound; in fact, any kind of judgments in which the subject will recognize and accept that the judgments naturally lie along a one-dimensional continuum. We define for the quaternary relation  just as for a binary relation, and ≈: ab cd if and only if not cd  ab, ab ≈ cd if and only if ab  cd and cd  ab. It is also convenient to have at hand certain elementary definitions of the binary relation of strict precedence or preference and the relation ≈ of indifference or indistinguishability. These definitions are the following. a b if and only if ab aa.

(1)

a ≈ b if and only if ab ≈ ba.

(2)

In order to express the equal-spacing part of our assumptions, we need one additional definition, namely, the definition that requires that adjacent objects in the ordering be equally spaced. For this purpose we introduce the definition of the binary relation J. This binary relation is just that of immediate predecessor. aJb if and only if a b and for all c in A if a c, then either b ≈ c or b c. (3) I now turn to the definition of finite equal-difference structures. The axioms given follow those given by Suppes and Zinnes [1963].

Measurement Theory and Engineering

837

DEFINITION 7. A quaternary structure A = (A, ) is a finite, equally spaced difference structure if and only if A is a finite set and the following axioms are satisfied for every a, b, c and d in A: 1. The relation  is a weak ordering of A × A; 2. If ab  cd, then ac  bd; 3. If ab  cd, then dc  ba; 4. If aJb and cJd, then ab ≈ cd. Keeping in mind the empirical interpretations mentioned already, it is easy to grasp the intuitive interpretation of each axiom. The first axiom just requires that the quaternary relation  be a weak ordering in terms of the qualitative difference between objects or stimuli. Axiom 2 is the most powerful and fundamental axiom in many ways. It expresses a simple necessary property of the intended interpretation of the relation . Axiom 3 just expresses a necessary algebraic fact about the differences. Notice that Axioms 1–3 are necessary axioms. Only Axiom 4 is sufficient but not necessary; it relates J to the quaternary relation ≈. The intuitive idea of this axiom is that if a stands in the relation J to b, and c stands in the relation J to d, then the difference between a and b is judged to be the same as the difference between c and d, due account being taken of algebraic sign. From these four axioms we can prove the following representation theorem. (Proofs of Theorems 3–8 in this section are given in [Suppes, 2002, Ch. 3–4].) THEOREM 3. Let A = (A, ) be a finite, equally spaced difference structure. Then there exists a real-valued function ϕ on A such that for every a, b, c, and d in A ϕ(a) − ϕ(b) ≥ ϕ(c) − ϕ(d) if and only if ab  cd. The matching invariance theorem is the following. THEOREM 4. Let A = (A, ) be a finite, equally spaced difference structure. Then a numerical representation of A is invariant under the group of linear transformations of Table 1. Upon casual inspection it might be supposed that the first three axioms of Definition 7 would characterize all finite-difference structures for which a numerical representation could be found. However, Scott and Suppes [1958] showed that the theory of all representable finite difference structures is not characterized by these three axioms and indeed cannot be characterized by any simple finite list of axioms. On the other hand, it might be thought that with the addition of the nonneccessary Axiom 4 it would be difficult to satisfy the axioms, because an arbitrary collection of stimuli or objects would not. However, if the stimuli being studied lie on a continuum, then it will be possible to select a standard sequence that will satisfy the axioms, just as is done in the case of selecting a standard set of weights for use on an equal-arm balance.

838

Patrick Suppes

2.3 Bisection measurement Relational structures closely related to the finite difference structures are bisection structures A = (A, B) where B is a ternary relation on the finite set A with the interpretation that B(a, b, c) if and only if b is the midpoint of the interval between a and c. The method of bisection has a long history in psychophysics, but it is important to emphasize that satisfaction of the axioms given below requires no assumptions of an underlying physical measurement. All we need is the intuitive idea of a qualitative continuum, and even that is not needed for formal purposes. After the fundamental psychological measurement in terms of the method of bisection has been made, it is desirable when possible to find a computationally simple psychophysical function relating physical measurements of the same magnitude to the psychological measurements. The axioms given below for the method of bisection imply a number of checks that should be satisfied before it is asserted that a numerical representing function exists, but these checks have often been ignored in the experimental literature that reports use of the method of bisection. For the simplest set of axioms and definitions, we take both the bisection relation B and the ordering relation  as primitive, but it is easy to eliminate  by definition. We use the binary relation J as defined earlier (Equation 3). DEFINITION 8. A structure A = (A, , B) is a finite, equally spaced bisection structure if and only if the set A is finite and the following axioms are satisfied for every a, a , b, c, and c in A: 1. The relation  is a weak ordering of A; 2. If B(abc) and B(abc ) then c ≈ c ; 3. If B(abc) and B(a bc) then a ≈ a ; 4. If B(abc) then a b and b c; 5. If aJb and bJc then B(abc); 6. If B(abc) and a Ja and cJc then B(a bc ). The intuitive interpretation of the axioms is relatively transparent. The first axiom is already familiar. Axioms 2 and 3 require uniqueness of the endpoints up to equivalence, which clearly separates bisection from betweenness. Axiom 4 relates the ternary bisection relation and the binary ordering relation in a natural way, although it imposes a formal constraint on the bisection relation which would often be omitted. Inclusion of this order property as part of the relation B simplifies the axioms. Axiom 5 is a strong assumption of equal spacing, and Axiom 6 expresses an additional feature of this equal spacing. In view of the axioms given earlier for difference structures, it is somewhat surprising that Axiom 6 can be shown to be

Measurement Theory and Engineering

839

independent of Axiom 5, but it is easy to give a model of Axioms 1–5 to show that this is the case. For we can take a model with B(abc) if and only if aJb and bJc and satisfy all of the first five axioms. The representation theorem assumes the following form. THEOREM 5. Let A = (A, , B) be a finite equally spaced bisection structure. Then there exists a real-valued function ϕ defined on A such that for every a, b, and c in A (i) ϕ(a) ≥ ϕ(b) if and only if a  b, (ii) 2ϕ(b) = ϕ(a) + ϕ(c) and ϕ(a) > ϕ(b) > ϕ(c) if and only if B(a, b, c). The matching invariance theorem is the following: THEOREM 6. Let A = (A, , B) be a finite bisection structure. Then a numerical representation of A is invariant under the group of linear transformations of Table 1.

2.4

Conjoint measurement

In many kinds of experimental or observational environments, it turns out to be the case that the measurement of a single magnitude or property is not feasible or theoretically interesting. What is of interest is the joint measurement of several properties simultaneously. In this case we consider axioms for additive conjoint measurement. The intended representation is that we use ordered pairs of objects or stimuli. The first members of the pairs are drawn from one set and consequently represent one kind of property or magnitude. The second members of the pairs are objects drawn from a second set representing a different magnitude or property. Given the ordered-pair structure, we shall only require judgments of whether or not one pair jointly has more of the ‘conjoined’ attribute than a second pair. It is easy to give examples of interpretations for which this way of looking at ordered pairs is natural. Suppose we are asked to judge the capabilities of individuals to assume a position of leadership in an organization. What we are given about the individuals are evaluations of their technical knowledge on an ordinal scale and a charisma measure on an ordinal scale. Thus for each individual we can say how he compares on each scale with any other individual. The problem is to make judgments as between the individuals in terms of their overall capabilities. The axioms given below indicate the kind of further conjoint ordinal conditions that are sufficient to guarantee finite equally spaced conjoint measurement, where in this case the equal spacing is along each dimension. As a second example, a pair (a, p) can represent a tone with intensity a and frequency p, and the problem is to judge which of two tones sounds louder. Thus the subject judges (a, p)  (b, q) if and only if tone (a, p) seems at least as loud

840

Patrick Suppes

as (b, q). Other examples from disciplines as widely separated as economics and physics are easily given, and are discussed in considerable detail in Krantz et al. [1971, Ch. 6]. It is to be stressed that the additivity representation sought in this section is a special case. Generalizations of additivity are discussed in the reference just cited. It is also to be noted that the restriction in this section to ordered pairs rather than ordered n-tuples is not essential. Before turning to the axioms of (additive) conjoint measurement, we need a couple of elementary definitions that permit us to define ordering relations on the individual components. On the basis of axioms on the ordering relations between pairs, we shall be able to prove that these ordering relations on the components are also weak orderings. In the following elementary definitions A1 is the set of first components and A2 the set of second components. Thus, when reference is made to an ordered pair (a, p), it is understood that a is in A1 and p is in A2 . a  b if and only if for all p in A2 , (a, p)  (b, p).

(4)

In terms of this relation we define a b and a ≈ b in the usual fashion. Also, a similar definition is needed for the second component. p  q if and only if for all a in A1 , (a, p)  (a, q).

(5)

We also use the notation already introduced for the relation  on A1 ×A2 , namely, (a, p) (b, q) if and only if not (b, q)  (a, p), and (a, p) ≈ (b, q) if and only if (a, p)  (b, q) and (b, q)  (a, p). The axioms for additive conjoint measurement in the finite, equally spaced case are embodied in the following definition. DEFINITION 9. A structure (A1 , A2 , ) is a finite, equally spaced additive conjoint structure if and only if the sets A1 and A2 are finite and the following axioms are satisfied for every a and b in A1 and every p and q in A2 : 1. The relation  is a weak ordering on A1 × A2 ; 2. If (a, p)  (b, p) then (a, q)  (b, q); 3. If (a, p)  (a, q) then (b, p)  (b, q); 4. If aJb and pJq then (a, q) ≈ (b, p). The intuitive content of the four axioms of Definition 9 is apparent, but requires some discussion. Axiom 1, of course, is the familiar requirement of a weak ordering. Axioms 2 and 3 express an independence condition of one component from the other. Thus Axiom 2 says that if the pair (a, p) is at least as great as the pair (b, p) then the same relationship holds when p is replaced by any other members q of A2 , and Axiom 3 says the same thing about the second component. Axiom 4 is sufficient but not necessary. It states the equal-spacing assumption, and corresponds closely to the corresponding axiom for finite, equally spaced difference structures.

Measurement Theory and Engineering

841

It might be thought the monotonicity assumption that if (a, p) ≈ (b, q) and a b, then q p, also needs to be assumed as an axiom, but this additional assumption is not necessary: it can be proved from the first four axioms alone. The statement of the representation theorem, to which we now turn, assumes exactly the expected form. The only thing to note is that the two real-valued functions on each component are welded together by the same unit. This is reflected by the common change of unit α in the theorem, but a different origin is permitted. THEOREM 7. Let (A1 , A2 , ) be a finite, equally spaced additive conjoint structure. Then there exist real-valued funtions ϕ1 and ϕ2 on A1 and A2 respectively such that for a and b in A1 and p and q in A2 ϕ1 (a) + ϕ2 (q) ≥ ϕ1 (b) + ϕ2 (p) if and only if (a, q)  (b, p). It is worth noting that the next invariance theorem (Theorem 8) has a natural geometrical interpretation. If we think of the functions ϕ1 and ϕ2 mapping pairs into the Cartesian plane, then the uniqueness theorem says that in the standard geometrical sense, any change of scale must be uniform in every direction, but the origin can be translated by a different distance along the different axes. THEOREM 8. Let (A1 , A2 , ) be a nontrivial finite equally spaced additive conjoint structure. Then a numerical representation (ϕ1 , ϕ2 ) is invariant up to the group of pairs (f1 , f2 ) of linear tranformations of (ϕ1 , ϕ2 ), but with the restriction that f1 and f2 have a common multiplicative constant, i.e., there are numbers α, β, and γ such that f1 ◦ ϕ1 = αϕ1 + β and f2 ◦ ϕ2 = αϕ2 + γ. (The requirement in the theorem that the structure be nontrivial is just that both A1 and A2 each have two distinct elements that are not equivalent under the ordering.) 3 EXTENSION AND APPROXIMATION OF EXTENSIVE MEASUREMENT

3.1

Extension of the finite case

We start with three primitives: a nonempty set A, a binary relation  on A, and a closed binary operation ◦ that maps A × A into A. The interpretation is: A is a set of objects or entities that exhibit the attribute in question; a  b holds if and only if a exhibits, in some prescribed qualitative fashion, at least as much of the attribute as b; and a ◦ b is an object in A that is obtained by concatenating (or composing) a and b in some prescribed, systematic fashion. An example is weight measurement (see Kisch [1965] for a detailed history of methods and equipment) in which the elements of A are material objects; a  b is established by placing

842

Patrick Suppes

a and b on the two pans of an equal-arm pan balance and observing which pan drops; and a◦b means that a and b are both placed in the same pan with a beneath b. As previously, we write a ≈ b if and only if a  b and b  a; and a b if and only if a  b and not (b  a). DEFINITION 10. Let A be a nonempty set,  a binary relation on A, and ◦ a closed binary operation on A. The triple A, , ◦ is a closed extensive structure iff the following four axioms are satisfied for all a, b, c, d ∈ A: (i) Weak order: A,  is a weak order, i.e.,  is a reflexive, transitive, and connected relation. (ii) Weak associativity: a ◦ (b ◦ c) ≈ (a ◦ b) ◦ c. (iii) Monotonicity: a  b iff a ◦ c  b ◦ c iff c ◦ a  c ◦ b. (iv) Archimedean: If a b, then for any c, d ∈ A, there exists a positive integer n such that na ◦ c  nb ◦ d, where na is defined inductively as: 1a = a, (n + 1)a = na ◦ a. The structure is called positive if, in addition, it satisfies (v) Positivity: a ◦ b a. THEOREM 9. Let A be a nonempty set,  a binary relation on A, and ◦ a closed binary operation on A. Then A, , ◦ is a closed extensive structure iff there exists a real-valued function φ on A(φ : A → Re) such that for all a, b, ∈ A (i) a  b iff φ(a)  φ(b); (ii) φ(a ◦ b) = φ(a) + φ(b). Another function φ satisfies (i) and (ii) iff there exists α > 0 such that φ = αφ. The structure is positive iff for all a ∈ A, φ(a) > 0. The axioms of Definition 10 and the proof of Theorem 9 are taken from Krantz et al. [1971, Ch. 3].

3.2

Qualitative approximation with upper and lower measures and transitive indistinguishability

The psychological consideration of thresholds below, for which perceptual or other comparative judgments are difficult, if not impossible, was initiated by Fechner [1860/1966]. An important early mathematical analysis was given by Wiener [1921]. Much of the modern literature begins with Luce’s definition [1956] of a semiorder, which was axiomatized as a single binary relation in the finite case by

Measurement Theory and Engineering

843

Scott and Suppes [1958]. Some of the most significant contributions have been by Falmagne [1971; 1974; 1977; 1985]. The probabilistic analysis of thresholds dates at least from works of Thurstone [1927a; 1927b]. Falmagne [1976; 1978] has also been a central contributor to this approach, with a number of other papers written with colleagues: Falmagne and Iverson [1979], Falmagne et al. [1979], and Iverson and Falmagne [1985]. An extensive review of all this literature is given in Suppes et al. [1989, Ch. 16]. Almost all of the work referred to assumes that the indistinguishability of similar events, objects, or stimuli is a nontransitive relation. The implicit assumption is that with many different discriminating observations, many initially indistinguishable events may be separated. Here the opposite is the starting point and the reason for the use of the word “transitive” in the title. It is a consequence of the axioms introduced that indistinguishability is an equivalence relation, and so, transitive. The remainder of this section draws heavily on Suppes [2006]. In the previous section I reviewed briefly extensive measurement focused on the construction of a finite standard ratio-scale representation. The basis for transitive indistinguishability is now easy to explain. An object weighed is assigned to a unique minimal interval, for example, one between 1.9 g and 2.0 g. The binary relation of two objects, a and b, not part of the standard sequence, being equivalent in weight, a ≈ b, is that they be assigned to the same minimal interval in the standard sequence. This relation is obviously an equivalence relation, i.e., reflexive, symmetric, and transitive, but in the system of approximation developed, these properties are not directly testable, but rather consequences of weighing operations with standard already “calibrated” sets of weights. So, in the notation used later, an object assigned to the minimal interval (1.9 g, 2.0 g) is said to have, as an approximation, upper measure (of weight) w∗ (a) = 2.0 g and lower measure w∗ (a) = 1.9 g. In practice, for all but the most refined procedures of measurement, no statistical analysis of having weight in such a minimal interval is given. In the cases when the minimal interval of the standard sequence is just on the borderline of instrument performance a statistical analysis can be given for repeated measurements. The ordinary practice is not completely in accord with my use of a minimal interval and thereby the assignment of an upper and a lower bound as the appropriate approximate measurement. But what is done is closely and simply related. As taught in elementary physics courses, to express a measurement as “accurate to 0.1 g”, for example, the measurement is written as 1.9 ± 0.1 g. What is usually recommended in practice is to use two adjacent minimal intervals to reduce uncertainty and to express the measurement itself as a single number. The axioms given in Section 3 could easily be changed to accommodate this use of two adjacent rather than one minimal interval. This same ± notation is also widely used to express the statistical standard error of repeated measurements. It is conceptually important here to retain both upper and lower measures, for the foundational view formalized in the axioms is that no finer measurement than that of a minimal interval is available in the given

844

Patrick Suppes

circumstances. And no theoretical construction of a probability distribution for location within the minimal interval makes much scientific sense. The point being emphasized is that the formalization given is meant to be a step closer to much, but certainly not all, actual practice of measurement when a fixed standard-scale representation is available. As a matter of terminology, what I have called a finite equally spaced extensive structure, could just as well be called a finite standard-sequence extensive structure. The terminology of standard sequences is familiar in the literature on the foundations of measurement. This language suggests the useful term standard sets for the sets of weights forming a standard sequence. For subsequent use it is important to note that for two sets of standard weights A and B, if they are not equivalent in weight then the minimum difference possible between them is the weight of one atomic set. More exactly, the ordered pair of sets (A, B) is a minimal pair of standard sets if μ(A) − μ(B) = μ(one atomic set), i.e., their difference is actually the minimum for nonequivalent standard sets. Note that if (A, B) is a minimal pair, A B. The equivalence of such pairs is a   useful notion to define. Two minimal pairs (A, B) and (A , B ) are equivalent if   μ(A) = μ(A ) and μ(B) = μ(B ). Here are three observations that are pertinent to later discussions. (1) If (A, B) and (C, D) are minimal pairs, then μ(A) − μ(B) = μ(C) − μ(D). (2) Obviously the ordering relation  can be extended to minimal pairs (A, B) and (C, D): (A, B)  (C, D) iff A  C, which we could have used earlier to define equivalent minimal pairs. (3) The empty set ∅ is a standard set. Assuming now a finite equally spaced extensive structure (also referred to as a finite standard sequence), additional axioms are given for measuring approximately any physical object in the range of the standard sequence. The primitive concepts are now (i) a set Ω of objects, (ii) a nonempty family F of subsets of Ω, (iii) a subset S of Ω, whose elements form a finite standard sequence, (iv) a subset W of objects to be measured, i.e., W = F|W − {∅} is the family of all nonempty subsets of W . (The notation F|W means that the family F of subsets is restricted to subsets of W .) (v) a binary relation  on F, but not assumed to be a weak ordering of W . This is proved later. As before, we define: W1 W2 iff W1  W2 and not W2  W1 . Also, W1 ≈ W2 iff W1  W2 and W2  W1 .

Measurement Theory and Engineering

845

If (S1 , S2 ) is a minimal pair and S1 W1 S2 , then (S1 , S2 ) is said to be a minimal pair for W1 , and also W1 is said to have a minimal pair. DEFINITION 11. A structure Ω = (Ω, F, S, W, ) is an approximate extensive structure with a finite standard sequence if and only if W is a nonempty finite set, W ⊆ F|W is the family of all nonempty subsets of W , and the following axioms are satisfied for all S1 , S2 , S3 and S4 in F|S and all W1 and W2 in W: 1. (S, F|S, ) is a finite equally spaced extensive structure; 2. S ∩ W = ∅ and S ∪ W = Ω; 3. W1  W2 or W2  W1 ; 4. If W1 S1 W2 then W1 W2 ; 5. If S1 W1 S2 then S1 S2 ; 6. W1 S1 or S1 W1 ; 7. If (S1 , ∅) is a minimal pair then W1 S1 ; 8. If W1 ∩ W2 = ∅, S1 W1 S2 , S3 W2 S4 and S1 ∩ S3 = ∅, then S1 ∪ S3 W1 ∪ W2 S2 ∪ S4 ; 9. If W1 ∩ W2 = ∅ then there are standard sets S1 and S2 such that S1 ∩ S2 = ∅, S1 W1 and S2 W2 ; 10. If W1 W2 then there is a standard set S1 such that W1 S1 W2 ; 11. W1 has a minimal pair of standard sets. Some comments on these axioms are appropriate. Axiom 1 just brings the structure of standard sets within the approximation framework. Axiom 2 requires no overlap of objects between those in S, calibrated for standard sets, and those in W , objects to be weighed. Axiom 3 is the only axiom expressed purely in terms of weighed objects, without tests using standard weights. Its requirement of connectedness of  for W is familiar. Axioms 4–11 then formulate testable assumptions that are sufficient to justify approximate measurement of weights falling within the range of standard sets. Because both sets S and W are finite, each axiom can be directly tested on an equal-arm balance. Axiom 4 provides the test for W1 being strictly heavier than W2 , namely find an S1 such that W1 S1 and S1 W2 . Axiom 5 states a transitivity condition, so to speak, on the relationship between standard sets and weighed sets or objects. If S1 is heavier than W1 and W1 is heavier than S2 , then it must be the case that S1 is heavier than S2 . Axiom 6 excludes any weighed object W1 being of exactly the same weight as any standard set. Weaker forms of this axiom are possible, but with

846

Patrick Suppes

attending complications of the test conditions. The axiom is similar to familiar “forced-choice” axioms in the measurement of beliefs or actions. Axiom 7 requires that any weighed object W1 be heavier than any minimal positive standard set S1 . This axiom permits an equal-arm balance or comparable device to not be sensitive to any positive weight smaller than a minimal standard set. Axiom 8 is obviously the generalization to approximate measurement of the usual qualitative axiom of addition exemplified in Axiom 2 of Definition 1. Axiom 9 guarantees that given disjoint sets W1 and W2 to be weighed, disjoint standard sets that are least upper bounds, S1 for W1 and S2 for W2 , can be found that are also disjoint. This does not follow from other axioms, because if W1 ∪ W2 = W , the union of the disjoint least upper bounds, S1 ∪ S2 can be one atomic standard set larger than a least upper bound of W itself, so S must be enlarged to cover this case. The possibilities are made explicit in Theorem 12. Axiom 10 is a test for W1 to be strictly heavier than W2 , and the test is, of course, relative to the coarseness of the standard sets. Axiom 11 guarantees that any objects, or set of objects, to be weighed fall within the range of the standard sets by having a minimal pair of standard sets, i.e., a discrete least upper bound and a discrete greatest lower bound among standard sets. A sample of elementary theorems are stated first, with a focus on the transitivity of relations and ≈ between sets of objects to be weighed. THEOREM 10. If W1 W2 and W2 W3 then W1 W3 . The next theorem shows that the equivalence relation ≈ for standard sets has the congruence property for on the set S × W . THEOREM 11. If S1 ≈ S2 and S1 W1 then S2 W1 . The next theorem asserts the testable criterion for W1 and W2 being indistinguishable. THEOREM 12. W1 ≈ W2 if and only if W1 and W2 have equivalent minimal pairs. By similar methods, we may prove a closely related result. THEOREM 13. Let (S1 , S2 ) be a minimal pair for W1 , and (S3 , S4 ) be such a pair for W2 . Then W1 W2 iff S1 S3 . We are now in a position to assert the transitivity of the indistinguishability of weights. THEOREM 14. If W1 ≈ W2 and W2 ≈ W3 then W1 ≈ W3 . The importance of the next theorem for determining the approximation that holds under addition of two disjoint sets W1 and W2 of objects to be weighed is brought out in the discussion following the theorem. THEOREM 15. If W1 ∩ W2 = ∅, then there exist standard sets S1 , S1 , S2 and S2 such that S1 ∩ S2 = S1 ∩ S2 = S1 ∩ S2 = S1 ∩ S2 = ∅, and

Measurement Theory and Engineering

847

(i) (S1 , S1 ) is a minimal pair for W1 , (ii) (S2 , S2 ) is a minimal pair for W2 , (iii) (S1 ∪ S2 , S1 ∪ S2 ) and (S1 ∪ S2 , S1 ∪ S2 ) are equivalent minimal pairs for W1 ∪ W2 , or (S1 ∪ S2 , S1 ∪ S2 ) and (S1 ∪ S2 , S1 ∪ S2 ) are equivalent minimal pairs for W1 ∪ W2 . In adding the approximate weight of two collections of physical objects, from weighing them individually, the approximate result does not enable us to infer which one of the two disjuncts formulated in Theorem 15 holds. These two disjuncts describe two adjacent but different minimal intervals. But there is one important feature to note. Addition does not increase the interval of approximation after addition. So, in Theorem 15, when we are given W1 and W2 , without further information we do not know in which minimal interval W1 ∪ W2 lies, but, as the disjunctive conclusion of the axiom asserts, it is just one of two adjacent minimal intervals, and by making the comparison empirically, we can determine which. The disjunctive clause (iii) of Theorem 15 and the assumption of exactness, i.e., no approximation, in the measurement of the standard sequence itself, mark a difference from the discussions and results on approximation in several different places in Foundations of Measurement [Krantz et al., 1971, Vol.I, Sections 2.2.2, 3.10.3, 4.4.4, and 5.4.3; Suppes et al., 1989, Vol. II, Section 16.6.2; Luce et al., 1990, Vol. III, Sections 19.5.4 and 21.8.2]. In fact, the standard concept of a pair (μ∗ , μ∗ ) of upper and lower measures, useful as measures of approximation, is not introduced anywhere in the three volumes of Foundations of Measurement. The definition of such a pair (μ∗ , μ∗ ) follows in form that given earlier for a measure μ. DEFINITION 12. Let Ω be a nonempty set and F a nonempty family of subsets of Ω closed under intersection and union, and let (μ∗ , μ∗ ) be a pair of real-valued functions defined on F. Then the structure (Ω, F, (μ∗ , μ∗ )) is an upper-lower measure structure if and only if the following axioms are satisfied for every A and B in F: 1. μ∗ (∅) = μ∗ (∅) = 0; 2. μ∗ (A) ≥ μ∗ (A) ≥ 0; 3. If A ⊇ B then μ∗ (A) ≥ μ∗ (B) and μ∗ (A) ≥ μ∗ (B); 4. If A∩B = ∅, then μ∗ (A)+μ∗ (B) ≤ μ∗ (A∪B) ≤ μ∗ (A∪B) ≤ μ∗ (A)+μ∗ (B). The concept of a pair (μ∗ , μ∗ ) of upper and lower measures is not new. It goes back at least to the use of inner and outer measures in analysis in the latter part of the nineteenth century by Carathedory [1917/1948] and others. Use in probability goes back at least to Koopman [1940a; 1940b].

848

Patrick Suppes

The representation of approximate measurement is explicitly given in terms of upper and lower measures. Theorem 15, or something roughly equivalent, is needed to establish the subadditive and superadditive properties of the upper and lower measures. These properties are formulated explicitly in part (v) of the next theorem. THEOREM 16. (Representation Theorem) Let Ω = (Ω, F, S, W, ) be an approximate extensive structure with a finite standard sequence. Then there is a measure μ on F|S satisfying Theorem 1, and an upper-and-lower measure pair (μ∗ , μ∗ ) on F|S ∪ W such that for any S1 and S2 in F|S and W1 and W2 in W: (i) μ∗ (S1 ) = μ(S1 ) = μ∗ (S1 ); (ii) If (S1 , S1 ) is a minimal pair for W1 , then μ(S1 ) = μ∗ (W1 ) > μ∗ (W1 ) = μ(S1 ); (iii) μ∗ (W1 ) > 0; (iv) if W1 ⊇ W2 , then μ∗ (W1 ) ≥ μ∗ (W2 ) and μ∗ (W1 ) ≥ μ∗ (W2 ); (v) if W1 ∩ W2 = ∅ then μ∗ (W1 ) + μ∗ (W2 ) ≤ μ∗ (W1 ∪ W2 ) ≤ μ∗ (W1 ∪ W2 ) ≤ μ∗ (W1 ) + μ∗ (W2 ). Comparison of the inequalities of clause (v) of the theorem just proved with the two disjunctive qualitative possibilities expressed in Theorem 15 suggest that a tighter bound can be proved, and this is the case. The inequalities in clause (v) can be tightened to (v’) by the insertion of the term μ∗ (W1 ) + μ∗ (W2 ) which is justified by Theorem 15. COROLLARY 1. (v ) μ∗ (W1 ) + μ∗ (W2 ) ≤ μ∗ (W1 ∪ W2 ) ≤ μ∗ (W1 ) + μ∗ (W2 ) ≤ μ∗ (W1 ∪ W2 ) ≤ μ∗ (W1 ) + μ∗ (W2 ). I have not stated an invariance result for Theorem 16, for the obvious one follows from this part of Theorem 1. But there is a different related consideration of greater interest. The minimal interval of the finite standard sequence S = (S, F, ) that is a part of any structure of approximate extensive measurement, as characterized by Definition 11, fixes the qualitative empirical precision of the empirical measurements. Now consider a second finite standard sequence T for measuring the same property of the subsets of W , and let (T1 , T1 ) be the minimal interval of T . Then, unlike the conventional acceptance of a unit of extensive measurement, in the case of approximate measurement, we have a directly qualitative comparison of precision given by the empirical ratio of (S1 , S1 ) to (T1 , T1 ). For example, the “scale” I regularly use to weigh myself has a minimum interval of 0.25 lb, but another one I use less often has a minimum interval of 0.1 kg. Since 1 kg = 2.20 lb, the ratio of 0.25 lb to 0.1 kg is .25/.22, which is, to two decimals, 1.14. So the standard sequence calibrated in the metric system is slightly more precise,

Measurement Theory and Engineering

849

although both “scales” provide minimum intervals beyond the precision ordinarily observed or recorded for most purposes. Any further refinement of either one is of little or no interest for the purpose of measuring body weight. Similar examples are easily given for measurement of length using different finite standard sequences. Moreover, the approximate theory developed here in terms of upper and lower measures can easily be extended by the same methods to difference measurement, bisection measurement, and conjoint measurement, and with somewhat more difficulty to several dimensions, e.g., affine or Euclidean geometry. Not surprisingly, applications of upper and lower measures have been most applied to approximate measurement of subjective probability. A comprehensive review and analysis is given by Walley [1991]. My own earlier contribution, Suppes [1974], uses upper and lower probabilities, but with nontransitive indistinguishability. The focus here has been on approximate measurement, but a very different theory of upper and lower probabilities can be derived from a direct set-theoretical generalization from random variables as random functions to random relations. An indication of the theoretical difference is that the upper and lower measures derived from random relations by Suppes and Zanotti [1977] are capacities of infinite order in the sense of Choquet [1953]. In contrast, the upper and lower measures considered here for approximate measurement are not even capacities of order two. Clearly, the sense of approximation introduced here and in Suppes [1974] is in no sense the only possibility. 4 QUALITATIVE PROBABILITY DISTRIBUTIONS OF ERROR In the standard theory of fundamental extensive measurement, qualitative axioms are formulated that lead to a numerical assignment unique up to a positive similarity transformation. The central idea of the theory of random quantities is to replace the numerical assignment by a random-variable assignment. This means that each object is assigned a random variable. In the case of extensive quantities, the expectation of the random variable replaces the usual numerical assignment, and the distribution of the random variable reflects the variability of the property in question, which could be intrinsic to the object or environment, or due to instrumental errors of observation. In any case, the existence and use of distributions with positive variances is almost universal in the actual practice of measurement in most domains of science. It is a widespread complaint about the standard foundations-of-measurement literature that too little has been written that combines the qualitative structural analysis of measurement procedures and the analysis of variability or error. In view of the extraordinarily large number of papers that have been written about the foundations of the theory of error, which go back to the eighteenth century with fundamental work already by Simpson, Lagrange, and Laplace, followed by the important contributions of Gauss, it is surprising that the two kinds of analysis have not often been combined. Part of the reason is the fact that, in all of this long history, the literature on the theory of errors has been intrinsically quantitative

850

Patrick Suppes

in character. Specific distributional results have usually been the objective of the analysis, and the assumptions leading to such results have been formulated in quantitative probabilistic terms. This quantitative framework is also assumed in the important series of papers by Falmagne and his collaborators on randomvariable representations for interval, conjoint, and extensive measurement (see Falmagne [1976; 1978; 1979; 1980; 1985]; Falmagne and Iverson [1979]; Iverson and Falmagne [1985]). The work reported here combines in one analysis the qualitative structures characteristic of the foundations of measurement and the probabilistic structures characteristic of the theory of error or the theory of variability. Details are given in Suppes and Zanotti [1992], and Suppes et al. [1989, Ch. 16]. The approach to the distribution of the representing random variables of an object consists of developing, in the usual style of the theory of measurement, qualitative axioms concerning the moments of the distribution, which are represented as expectations of powers of the representing numerical random variable. The first natural question is whether or not there can be a well-defined qualitative procedure for measuring the moments. This is discussed in Section 4.1, Section 4.2 presents the qualitative primitive concepts and Section 4.3 the axiom system and the representation theorem.

4.1 Variability as measured by moments Before giving the formal developments, we address the measurement of moments from a qualitative standpoint. We outline here one approach without any claim that it is the only way to conceive of the problem. In fact, we believe that the pluralism of approaches to measuring probability is matched by that for measuring moments, for reasons that are obvious given the close connection between the two. The one approach we outline here corresponds to the limiting relative-frequency characterization of probability, which we formulate here somewhat informally. Let s be an infinite sequence of independent trials with the outcome on each trial being heads or tails. Let H(i) be the number of heads on the first i trials of s, P (heads) = lim H(i)/i, i→∞

with the provision that the limit exists and that the sequence s satisfies certain conditions of randomness that need not be analyzed here. In practice, of course, only a finite initial segment of any such sequence is realized as a statistical sample. However, ordinarily in the case of probability, the empirical procedure encompasses several steps. In the approach given here, the first step is to use the limiting relative-frequency characterization. The second step is to produce and analyze a finite sample with appropriate statistical methods. Our approach to empirical measurement of qualitative moments covers the first step but not the second of giving detailed statistical methods. Thus, let a0 be an object of small mass of which we have many accurate replicas — so we are assuming here that the variability in a0 and its replicas, aj0 , j = 1, 2, . . . are

Measurement Theory and Engineering

851

negligible. Then we use replicas of a0 to qualitatively weigh an object a. On each trial, we force an equivalence, as is customary in classical physics. Thus, on each trial i, we have  (1) (2) (m )  a ∼ a0 , a0 , . . . , a0 i . The set shown on the right we symbolize as mi a0 . Then, as in the case of probability, we characterize an , the nth qualitative raw moment of a, by 1 n mi a0 , j→∞ j i=1 j

an ∼ lim

but, in practice, we use a finite number of trials and use the estimate a ˆn : 1 n m a0 , j i=1 i j

a ˆn ∼

and so also only estimate a finite number of moments. It is not to the point here to spell out the statistical procedures for estimating an . Our objective is only to outline how one can approach empirical determination of the qualitative raw moments. There is one important observation to deal with. The observed data, summarized in the integers m1 , m2 , . . . , mj , on which the computation of the moments is based, also constitute a histogram of the distribution. Why not estimate the distribution directly? When a distribution of a particular form is postulated, there need be no conflict in the two methods, and the histogram can be of further use in testing goodness of fit. The reason for working with the raw moments is theoretical rather than empirical or statistical. Various distributions can be qualitatively characterized in terms of their raw moments in a relatively simply way, as the examples to be considered show. Furthermore, general qualitative conditions on the moments are given in the Representation Theorem. Alternative qualitative approaches to characterizing distributions undoubtedly exist and as they are developed may well supersede the one used here. We now turn to the formal developments. In proving the representation theorem for random extensive quantities we apply a well-known theorem of Hausdorff [1923] on the one-dimensional moment problem for a finite interval. HAUSDORFF’S THEOREM. Let μ0 , μ1 , μ2 , . . . be a sequence of real numbers. Then a necessary and sufficient condition that there exist a unique probability distribution F on [0, 1] such that μn is the nth raw moment of the distribution F , that is to say, 1 μn = tn dF, n = 0, 1, 2, . . . , 0

is that μ0 = 1 and all the following inequalities hold:

(6)

852

Patrick Suppes

    k k μν+1 + μν+2 + · · · + (−1)k μν+k  0 f or k, ν = 0, 1, 2, . . . . (7) μν − 1 2 A standard terminology is that a sequence of numbers μn , n = 0, 1, 2, . . . is completely monotonic iff Inequalities (2) are satisfied, in more compact binomial notation μν (1 − μ)k  0, for k, ν = 0, 2, . . . (for detailed analysis of many related results on the problem of moments, see Shohat and Tamarkin [1943]). It is important to note that we do not need an additional separate specification of the domain of definition of the probability distribution in Hausdorff’s theorem. The necessary and sufficient conditions expressed in the Inequalities (2) guarantee that all the moments lie in the interval [0, 1], and so this may be taken to be the domain of the probability distribution without further assumption.

4.2 Qualitative primitives for moments The idea, then, is to provide a qualitative axiomatization of the moments for which a qualitative analogue of Inequalities (2) obtains and then to show that the qualitative moments have a numerical representation that permits one to invoke Hausdorff’s theorem. Thus, the qualitative structure begins first with a set G of objects. These are the physical objects or entities to which we expect ultimately to associate random variables. More precisely, we expect to represent the selected extensive attribute of each object by a random variable. However, in order to get at the random variables, we must generate from G a set of entities that we can think of as corresponding to the raw moments and mixed moments of the objects in G. To do that, we must suppose that there is an operation · of combining so that we can generate elements an = an−1 · a, which from a qualitative point of view, will be thought of as corresponding to the raw moments of a. It is appropriate to think of this operation as an operation of multiplication, but it corresponds to multiplication of random variables, not to multiplication of real numbers. We shall assume as axioms that the operation is associative and commutative, but that is should not be assumed to be distributive with respect to disjoint union (which corresponds to numerical addition) can be seen from a random-variable counterexample, given in Gruzewska [1954]. We turn now to the explicit definition of a semigroup that contains the associative and commutative axioms of multiplication. DEFINITION 13. Let A be a nonempty set, G a nonempty set, · a binary operation on A and 1 an element of G. Then A = (A, G, ·, 1) is a commutative semigroup with identity 1 generated by G iff the following axioms are satisfied for every a, b, and c in G and s and t in A: 1. If a ∈ G, then a ∈ A. 2. If s, t ∈ A, then (s · t) ∈ A.

Measurement Theory and Engineering

853

3. Any member of A can be generated by a finite number of applications of Axioms 1 − 2 from elements of G. (A is the set of finite strings with alphabet G.) 4. a · (b · c) = (a · b) · c. 5. a · b = b · a. 6. 1 · a = a. Note that, because of the associativity axiom, we omit parentheses from here on. Note, further, that, on the basis of Axiom 3 we think of elements of A as finite strings of elements of G. Intuitively the elements of A are qualitative mixed moments. Furthermore, because the product operation · is associative and commutative, we can always write the mixed moments in a standard form involving powers of the generators. For example, a · a · a · c · a · b · c = a4 · b · c2 . This expression is interpreted as the qualitative mixed moment consisting of the fourth raw moment of a times the first one of b times the second one of c. We denote this semigroup by A. Our last primitive is a qualitative ordering of moments. As usual, we will denote it by . The first question concerns the domain of this relation. For purposes of extensive measurement, it is useful to assume that the domain is all finite subsets from the elements of the semigroup A. We may state this as a formal definition: DEFINITION 14. Let A be a nonempty set and  a binary relation on F, the family of all finite subsets of A. Then A = (A, F, ) is a weak extensive structure iff the following axioms are satisfied for every B, C, and D in F: 1. The relation  is a weak ordering of F. 2. If B ∩ D = C ∩ D = , then B  C iff B ∪ D  C ∪ D. 3. If B = , then B . Superficially the structure just defined looks like a familiar structure of qualitative probability, but in fact it is not. The reason is that because A is an infinite set, we cannot assume F is closed under complementation, because that would violate the assumption that the subsets in F are finite. An important conceptual point is that we do require the ordering in magnitude of different raw moments. One standard empirical interpretation of what it means to say that the second raw moment, a2 , is less than the first, a1 , was outlined previously. A formal point, appropriate to make at this stage, is to contrast the uniqueness result we anticipate for the representation theorem with the usual uniqueness up to a similarity (i.e., multiplication by a positive constant) for extensive measurement. We have, in the present setup, not only the extensive operation but also the semigroup multiplication for forming moments; therefore, the uniqueness result is absolute (i.e., uniqueness in the sense of the identity function).

854

Patrick Suppes

Given this strict uniqueness, the magnitude comparison of am and an for any natural numbers m and n is not a theoretical problem. It is of course apparent that any procedure for measurement of moments, fundamental or derived, will need to satisfy such strict uniqueness requirements in order to apply Hausdorff’s or other related theorems in the theory of moments. Within F, we may define what it means to have n disjoint copies of B ∈ F: 1B = B (n + 1)B ∼ nB ∪ B  , where nB ∩B  = , and B  ∼ B and ∼ is the equivalence relation defined in terms of the basic ordering  on F. Axiom 3, of Definition 15, which follows, will simply be the assumption that such a B  always exists, and so nB is defined for each n. It is essential to note that this standard extensive or additive recursive definition is quite distinct from the one for moments an given earlier.

4.3 Axiom system for qualitative moments The goal is to provide axioms on the qualitative raw moments such that we can prove that object a can be represented by a random variable Xa , and the nth raw moment an is represented by the nth power of Xa (i.e., by Xna ). For convenience, we shall assume the structures we are dealing with are bounded in two senses. First, the set G of objects will have a largest element 1, which intuitively means that the expectation of the random variables associated with the elements of a will not exceed that of 1. Moreover, we will normalize things so that the expectation associated with X1 is 1. This normalization shows up in the axiomatization as 1 acting as the identity element of the semigroup. Second, because of the condition arising from the Hausdorff theorem, this choice means that all of the raw moments are decreasing in powers of n (i.e., if m  n, then an  am ). Obviously the theory can be developed so that the masses are greater than 1, and the moments become larger with increasing n. This is the natural theory when the probability distribution is defined on the positive real line. Here, the exponential notation for qualitative moments an is intuitively clear, but it is desirable to have the following formal recursive definition: a0 an

=

1,

=

an−1 · a,

in order to have a clear interpretation of a0 . Before giving the axiom system, we must discuss more fully the issue of what will constitute a qualitative analogue of Hausdorff’s condition, Inequality (2). We have only an operation corresponding to addition and not to subtraction in the qualitative system; thus, for k, an even number, we rewrite this inequality solely in terms of addition as follows:

Measurement Theory and Engineering

    k k μν+2 + · · · + μν+k  μν+1 + · · · + kμν+(k−1) , μν + 2 1

855

(8)

and a corresponding inequality for the case in which k is odd. In the qualitative system, the analogue to Inequality (3) must be written in terms of union of sets as follows for k even:     k ν+2 k ν+1 ν+k a ∪ ··· ∪ a  a ∪ · · · ∪ kaν+(k−1) . (9) aν ∪ 2 1 When k is odd, aν ∪

    k ν+2 k ν+1 a a ∪ · · · ∪ kaν+(k−1)  ∪ · · · ∪ aν+k . 2 1

(10)

There are several remarks to be made about this pair of inequalities. First of all, we can infer that, for a ≺ 1, as opposed to a ∼ 1, the moments are a strictly decreasing sequence (i.e., aν aν+1 ). Second, the meaning of such terms as

k ν+2 was recursively defined earlier, with the recursion justified by Axiom 3 2 a below. It is easy then to see that the unions indicated in Inequalities (2) and (5) are of disjoint sets. On the basis of the earlier terminology, we can then introduce the following definition. A qualitative sequence a0 , a1 , a2 , a3 , . . . is qualitatively completely monotonic iff Inequalities (4) and (5) are satisfied. DEFINITION 15. A structure A = (A, F, G, , ·, 1) is a random extensive structure with independent objects — the elements of G — iff the following axioms are satisfied for a in G, s and t in A, k, m, m , n, and n natural numbers, and B and C in F: 1. The structure (A, F, ) is a weak extensive structure. 2. The structure (A, G, ·, 1) is a commutative semigroup with identity 1 generated by G. 3. There is a C  in F such that C  ∼ C and C  ∩ B =  . 4. Archimedean. If B C, then for any D in F there is an n such that nB  nC ∪ D. 5. Independence. Let mixed moments s and t have no common objects: a. If m1  ns and m 1  n t, then mm 1  nn (s · t). b. If m1  ns and m 1  n t, then mm 1  nn (s · t). 6. The sequence a0 , a1 , a2 , . . . of qualitative raw moments is qualitatively completely monotonic. The content of Axiom 1 is familiar. What is new here is, first of all, Axiom 2, in which the commutative semigroup, as mentioned earlier, is used to represent

856

Patrick Suppes

the mixed moments of a collection of objects. Axiom 3 is needed in order to make the recursive definition of (n + 1)B well defined as given earlier. The special form of the Archimedean axiom is the one needed when there is no solvability axiom, as discussed in Section 3.2.1 of Krantz et al. [1971]. The dual form of Axiom 5 is just what is needed to prove the independence of the moments of different objects, which means that the mixed moments factor in terms of expectation. Note that it is symmetric in  and . The notation used in Axiom 5 involves both disjoint unions, as in m1, and the product notation for mixed moments, as in (s · t). Axiom 6 formulates the qualitative analogue of Hausdorff’s necessary and sufficient condition as discussed above. THEOREM 17. (Representation Theorem) Let A = (A, F, G, , ·, 1) be a random extensive structure with independent objects. Then there exists a family {XB , B ∈ F} of real-valued random variables such that: (i) every object a in G is represented by a random variable Xa whose distribution is on [0, 1] and is uniquely determined by its moments; (ii) the random variables {Xa , a ∈ G} are independent; (iii) for each a and b in G with probability one Xa·b = Xa · Xb ; (iv) E(XB )  E(XC ), iff B  C; (v) if B ∩ C = , then XB∪C = XB + XC ; (vi) if B = 0, then E(XB ) > 0; (vii) E(Xn1 ) = 1 for every n. Moreover, any function φ from Re to Re such that {φ(XB ), B ∈ F}, satisfies (i)–(vii) is the identity function. If we specialize the axioms of Definition 15 to qualitative assertions about distributions of a particular form, we can replace Axiom 6 on the complete monotonicity of the sequence of qualitative moments of an object by much simpler conditions. In fact, we know of no simpler qualitative way of characterizing distributions of a given form than by such qualitative axioms on the moments. The following corollary concerns such a characterization of the uniform, Bernoulli, and beta distributions on [0, 1], where the beta distribution is restricted to integer-valued parameters α and β. COROLLARY 2. (Corollary to Representation Theorem) Let A = (A, F, G, , ·, 1) be a structure satisfying Axioms 1–5 of Definition 15, and, for any a in G, assume a  1.

Measurement Theory and Engineering

857

I. If the moments of an object a for n  1 satisfy (n + 1)an ∼ 2a, then Xa is uniformly distributed on [0, 1]. II. If the moments of an object a for n  1 satisfy an ∼ a, then Xa has a Bernoulli distribution on [0, 1]. III. If the moments of an object a for n  1 satisfy (α + β + n)an+1 ∼ (a + n)an , where α and β are positive integers, then Xa has a beta distribution on [0, 1]. Note that a Bernoulli distribution on Xa implies that all the probability weight is attached to the end points of the interval, so that, if p is the parameter of the distribution, as in standard notation, then E(Xa ) = (1 − p) · 0 + p · 1 = p. We remark informally that some other standard distributions with different domains may also be characterized qualitatively in terms of moments. For example, the normal distribution on (−∞, ∞) with mean equal to zero and variance equal to one is characterized as follows: a0 ∼ 1, a1 ∼ , a2 ∼ 1, a2(n+1) ∼ (2n + 1)a2n for n  1. 5

CONCLUDING REMARKS

There are several topics not adequately covered in the preceding pages that I would like to mention briefly. Geometric generalizations. The approximate extensive measurement methods of Section 3, in fact, even the finite equally-spaced measurement models of Section 2, have natural geometric generalizations. And not just to Euclidean geometry. Equal-spacing models can easily be given for spherical geometry, or the rather different example of affine geometry. The approximations of extensive measurement by upper and lower measurement can fit in well as a framework for spatial measurements in geometry and physics. The details are somewhat more complicated, but the conceptual viewpoint is easily generalized to n-dimensional geometries of various types, especially those used in scientific and engineering data analysis.

858

Patrick Suppes

More on approximation. Section 3 was mainly concerned with a particular sort of approximation, but the subject is much broader than what could be developed in detail here. Perhaps most missing is any attempt to survey the large and complex literature in statistics on measurement errors of all kinds. Matters of importance range from the standard introduction in linear regression models of normally distributed errors with zero mean to the estimation of prediction errors in nonlinear models. In much of this work, the framework is quantitative and the theory of measuring the errors themselves is not developed, but assumed without comment. Of course, the limits of accurate measurement are stressed, especially in quantum mechanics. The 19th century ideal of ever better approximations of the measurement of physical quantities has been replaced by such theoretical results as the Heisenberg uncertainty relations, and, more generally, by the realization that measurement imvolves physical interaction between instrument and object being measured. The disturbances of such interactions block the road to ideally exact measurements of any continuous quantity. Sections 3 and 4 display the tension that naturally arises if too unified an approach to approximation is taken. The results of Section 3 are in terms of upper and lower measures. In contrast the rather elaborate qualitative theory of distributions of error in Section 4 is focused entirely on results stated in terms of standard probability measures. How are these two approaches to be reconciled? There are several ways of answering. First, the upper and lower measures of Section 3 are, in a sense already mentioned there, pre-statistical. The upper and lower measures introduced are not standard at all, for example, in the elementary introductions of errors and their measurement in student lab manuals in physics, chemistry or most engineering subjects. The intention here is to provide a conceptual foundation for what is often unspoken in these matters. A naive student may ask, “What does it mean to report a weight measurement as 4.7 ± ·1 gm?” The answer can be simple, or as complicated as the axiomatic analysis of Section 3. And Section 3 offers only part of the complicated answer. Manufacturers’ reports of the accuracy of the instruments they sell, psychologists’ study of human errors, and always the disturbing effects of interactions with the environment, which may not even be labeled as errors, all these and still other sources as well, add to the complexity of the total picture. Section 4 does provide a different take on errors. It addresses the problem of providing a foundation, based on qualitative axioms, of various probability distributions widely used in statistical analysis of data. This is one way, important for the systematic development of measurement theory, to give an axiomatic underpinning for these familiar distributions, and to see how they naturally arise from qualitative principles. References to the literature. The numerous articles and books on the theory of measurement published in the 20th century are spread over many different scientific and engineering fields. A large collection of references focused on foundational

Measurement Theory and Engineering

859

questions can be found in the three-volume work Foundations of Measurement (referenced here as Krantz et al., [1971], Vol. I; Suppes et al., [1989], Vol. II; Luce et al., [1990], Vol. III). The list of references in these three volumes occupies many pages at the end of each volume. (This work was reprinted in paperback by Dover Publications, New York, in 2007.) BIBLIOGRAPHY [Carathedory, 1917/1948] C. Carathedory. Vorlesungen u ¨ ber Reelle Funktionen. New York: Chelsea Publishing Company, 1948. (First edition 1917). [Choquet, 1953] G. Choquet. Theory of capacities, Annales de l’InstitutFourier, 5, 131—295, 1953. [Falmagne, 1971] J. C. Falmagne. The generalized Fechner problem and discrimination. Journal of Mathematical Psychology, 8, 22—43, 1971. [Falmagne, 1974] J. C. Falmagne. Foundations of Fechnerian psychophysics. In Contemporary Developments in Mathematical Psychology Vol. 2, D.H. Krantz, R.C. Atkinson, R.D. Luce, and P. Suppes, eds., pp. 127—159. San Francisco, CA: Freeman, 1974. [Falmagne, 1976] J. C. Falmagne. Random conjoint measurement and loudness summation. Psychological Review, 83, 65—79, 1976. [Falmagne, 1977] J. C. Falmagne. Weber’s inequality and Fechner’s problem. Journal of Mathematical Psychology, 16, 267—271, 1977. [Falmagne, 1978] J. C. Falmagne. A representation theorem for finite random scale systems. Journal of Mathematical Psychology, 18, 52—72, 1978. [Falmagne, 1979] J. C. Falmagne. On a class of probabilistic conjoint measurement models: Some diagnostics properties. Journal of Mathematical Psychology, 19, 73—88, 1979. [Falmagne, 1980] J. C. Falmagne. A probabilistic theory of extensive measurement. Philosophy of Science, 47, 277—296, 1980. [Falmagne, 1985] J. C. Falmagne. Elements of Psychophysical Theory. London and New York: Oxford University Press, 1985. [Falmagne and Iverson, 1979] J. C. Falmagne and G. Iverson. Conjoint Weber laws and additivity. Journal of Mathematical Psychology, 20, 164—183, 1979. [Falmagne et al,, 1979] J. C. Falmagne, G. Iverson, and S. Marcovici. Binaural “loudness” summation: Probabilistic theory and data. Psychological Review, 86, 25—43, 1979. [Fechner, 1860/1966] G. T. Fechner. Elemente der Psychophysik. Leipzig: Druck and Verlag von Breitkopfs H¨ artel. Translated by Helmut E. Adler. Elements of Psychophysics (Vol. 1). New York: Holt, Rinehart, and Winston, 1860/1966. [Gruzewska, 1954] H. M. Gruzewska. L’arithm´etique des variables al´eatories (The arithmetic of random variables). Cahiers Rhodaniens, 6, 9—56, 1954. [Hausdorff, 1923] F. Hausdorff. Momentprobleme f¨ ur ein endliches Intervall (Moment problem for a finite interval). Mathematische Zeitschrift, 16, 220—248, 1923. [Iverson and Falmagne, 1985] G. Iverson and J. C. Falmagne. Statistical issues in measurement. Mathematical Social Sciences, 10, 131—153, 1985. [Kisch, 1965] B. Kisch. Scales and Weights. New Haven, Connecticut: Yale University Press, 1965. [Koopman, 1940a] B. O. Koopman. The axioms and algebra of intuitive probability. Annals of Mathematics 41(2): 269—292, 1940. [Koopman, 1940b] B. O. Koopman. The bases of probability. Bulletin of the American Mathematical Society 46: 763—774, 1940. [Krantz et al., 1971] D. H. Krantz, R. D. Luce, P. Suppes, and A. Tversky. Foundations of Measurement, Vol. I: Additive and Polynomial Representations. New York: Academic Press, 1971. [Luce, 1956] R. D. Luce. Semiorders and a theory of utility discrimination. Econometrica, 24, 178—191, 1956. [Luce et al., 1990] R. D. Luce, D. H. Krantz, P. Suppes, and A. Tversky. Foundations of Measurement, Vol. III: Representation, Axiomatization, and Invariance. New York: Academic Press, 1990.

860

Patrick Suppes

[Scott and Suppes, 1958] D. Scott and P. Suppes. Foundational aspects of theories of measurement. Journal of Symbolic Logic, 23, 113—128, 1958. [Shohat and Tamarkin, 1943] J. A. Shohat and J. D. Tamarkin. The Problem of Moments. New York: American Mathematical Society, 1943. [Suppes, 1969] P. Suppes. Studies in the Methodology and Foundations of Science: Selected Papers from 1951 to 1969 (pp. 4—8). Dordrecht: Reidel, 1969. [Suppes, 1974] P. Suppes. The measurement of belief. Journal of the Royal Statistical Society (Series B), 36, 160—191, 1974. [Suppes, 2002] P. Suppes. Representation and Invariance of Scientific Structures. Stanford, CA: Center for the Study of Language and Information, 2002. [Suppes, 2006] P. Suppes. Transitive indistinguishability and approximate measurement with standard finite ratio-scale representations. Journal of Mathematical Psychology, 50, 329— 336, 2006. [Suppes et al., 1989] P. Suppes, D. H. Krantz, R. D. Luce, and A. Tversky. Foundations of Measurement, Vol. II: Geometrical, Threshold and Probabilistic Representations. New York: Academic Press, 1989. [Suppes and Zanotti, 1977] P. Suppes and M. Zanotti. On using random relations to generate upper and lower probabilities. Synthese, 36, 427—440, 1977. [Suppes and Zanotti, 1992] P. Suppes and M. Zanotti. Qualitative axioms for random-variable representation of extensive quantities. In Philosophical and Foundational Issues in Measurement Theory, C.W. Savage and P. Ehrlich, eds., pp. 39—52. Hillsdale, NJ: Lawrence Erlbaum, 1992. [Suppes and Zinnes, 1963] P. Suppes and J. L. Zinnes. Basic measurement theory. In Handbook of Mathematical Psychology, Vol. 1, R.D. Luce, R.R. Bush and E.H. Galanter, eds., New York: Wiley, pp. 3—76, 1963. [Thurstone, 1927a] L. L. Thurstone. A law of comparative judgment. Psychological Review, 34, 273—286, 1927. [Thurstone, 1927b] L. L. Thurstone. Psychophysical analysis. American Journal of Psychology, 38, 368—389, 1927. [Walley, 1991] P. Walley. Statistical Reasoning with Imprecise Probabilities. London: Chapman and Hall, 1991. [Wiener, 1921] N. Wiener. A new theory of measurement: A study in the logic of mathematics. Proceedings of the London Mathematical Society, 19, 181—205, 1921.

TECHNOLOGICAL EXPLANATION

Joseph C. Pitt

1

INTRODUCTION

The purpose of this chapter is to provide an account of technological explanation. The topic is relatively unexplored. Therefore, in many respects this is as much an attempt to lay out the territory that needs to be covered as it is a fully adequate theory of technological explanation. The structure of the chapter is as follows: after a discussion of the need for a theory of technological explanation, I differentiate technological explanation from physical, teleological, psychological, and social explanation. Attention is then directed to answering questions as a means of providing technological explanations. A distinction between internal and external audiences is also introduced to provide a means for characterizing different kinds of explanations in terms of the audiences to which they are directed and the kinds of questions which when answered provide the appropriate explanation. Next the concept of system is introduced. The idea to be developed is that a crucial component of a technological explanation is placing the artifact/mechanism/activity/function to be explained in a relation to other parts of the system in which it is embedded. The strong position that there is no explanation without relating the thing to be explained to something else is laid out. This idea is elaborated by showing how artifact specific issues such as the design, function or structure of an artifact can only be adequately explained by reference to the system in which they have a role. Inevitably talk of systems will bring us to social systems as technological artifacts and the degree to which an explanation of some aspect of a technical artifact requires appeal to some aspect of a social system in which it functions.1 Finally, following a discussion of some examples, there is a discussion of the lack of symmetry between explanations of technological successes and technological failures and the importance of that lack of symmetry. 1 A theme that follows from this line of thought and also further develops a position laid out in the last chapter of [Pitt, 2000] is that since mature sciences are embedded in a technological infrastructure, any adequate theory of scientific explanation requires a theory of technological explanation. Unfortunately a full development of this relationship would take us far afield from the topic at hand.

Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

862

Joseph C. Pitt

2

RELEVANCE

Why do we need a theory of technological explanation? The standard answer to a similar question, “Why do we need a theory of scientific explanation?” has a fairly straightforward answer: “Because science is supposed to explain the world and we need to know when such explanations are good ones.” But clearly a similarly phrased response to the question about technological explanation won’t do. The task of justifying the need for a theory of technological explanation is that, unlike for scientific explanation, which only has to account for why things in the natural world do the things they do, there are lots of different kinds of questions we can ask about our technologies. That suggests that either there are many types of technological explanations or that the account we develop will be unique. While that doesn’t answer the primary question, it does suggest a strategy: first, identify what technological explanations are about. I will not talk about Technology, but rather about technologies, specific technologies.2 They are everywhere. And that is the very point. We are surrounded by, embedded in, dependent on, supported by, amused through our technologies. They make the way we live possible. They also have positive and negative impacts on us and on the ecology of the planet. They are the expression of the creative, inventive, and perhaps malevolent aspects of our collective character. In short, our technologies and how we use them are what marks us out as human.3 That means that if we want to know what we are and how we got to this point we need to explain how we created our technologies and how they assisted us and/or restrained us. This means that a theory of technological explanation is relevant to all forms of human activity, since they all involve dealing with technologies, including science. What a theory of technological explanation will provide is the means to explain how an artifact came to be what it is. This can be a causal story but it will also be partially an appeal to a variety of social factors. A theory of technological explanation will also provide the means to explain the role of the artifact in our lives and the impact introducing the artifact had on our social structures, goals, and values. It will, finally, also provide the means to explain technological failures and to distinguish questions concerning system failure from issues of assessing blame and responsibility. 3 TECHNOLOGICAL VERSUS SCIENTIFIC EXPLANATION As noted above, the key to developing an account of technological explanation is answering the question: what are we explaining? Unlike in science, where in the past it has been assumed, incorrectly, that the answer is fairly straightforward, as noted above the focus in technological explanations can be multifold. The traditional view has it that the purpose of a scientific explanation is to help us understand why the world works the way it does in specific circumstances. The 2 See 3 Not

[Pitt, 2000, Chapter 1]. everyone agrees with this claim, especially Ashley Shew. See her [2007].

Technological Explanation

863

history of the development of theories of scientific explanation reveals consistent attempts to find a general theoretical account of what makes for a good explanation that applies across all the sciences.4 From the first major modern efforts in this direction by Hempel in [1948], there have been numerous such attempts to construct a general theory of explanation. These efforts may now decline as work in the histories and philosophies of the individual sciences reveals major differences among them in terms of methods, that in turn seem to require the development of individual theories of explanation for physics, biology, chemistry, geology, etc.5 But even if universal theories of scientific explanation decline in usefulness in the natural sciences, it is not clear that they could have been of use for the technical sciences when considered apart from the natural sciences, that is, when technologies are not seen as mere applied science and the technical sciences, like the various disciplines of engineering, are considered in their own right. It is the very universal ambition of these philosophical theories that renders them inapplicable in the technical sciences. Consider, for example, Hempel’s account.

3.1

The DN theory

Known as the Deductive-Nomological Theory, Hempel’s Covering Law Theory (DN) requires that in the premises of the deductive argument that constitutes the explanation there must appear the expression of at least one law of nature. If we were to apply this account to technological issues needing explanation, this would require that there exist natural laws governing technologies. However, there are no natural technological laws, except maybe the Law of Unintended Consequences. In the absence of laws of nature for technologies, this sort of theory will not work. As noted, a significant component of a DN explanation is the requirement of the presence of a law. However, neither the model, nor the theory of explanation that supports it provides a decision procedure for selecting which law to choose. Laws, in this context, are formulated in the context of theories. Hempel’s unarticulated assumption seems to be that in science there is only one theory in play at a time and hence there is no need for a decision procedure. But this assumes that even when there is only one theory in play that we know which of its many laws and generalizations to employ in this explanation. Further, since recent historical and sociological considerations have begun impinging on philosophical ruminations, we now know that scientific work rarely takes place in such a clean environment. In the context of doing science, there are generally multiple accounts vying for supremacy and much of the excitement in the sciences comes from the clash of theoretical explanations. In scientific, as in technological, contexts, to forget that these are very human activities, subject to the strengths and weaknesses of any 4 The literature on explanation is vast. It is far too large to discuss here. Fortunately, there are two major narrative histories of the debates. The first is by Wesley Salmon [1989] in his introduction to the edited volume Scientific Explanation. The second, more recent discussion, is by Jeroen de Ridder [2007]. 5 Robert Cummins also alludes to this possibility in his [1975] article.

864

Joseph C. Pitt

human epistemological endeavor, is to set goals that are unattainable. To do so is to ignore what can actually be accomplished. The DN model and responses to it are only one type of explanation and they were devised with physics as the model science to do the explaining.6 It is a way of explaining why natural things in the world do the things they do by appeal to the structure of nature. And that is one reason they seem inappropriate for matters technological. Technological explanations are unique insofar as they concern technological issues, issues that emerge because of things human beings have built. Some of these items require knowledge of how nature works in order to be constructed, e.g., optics for telescopes, chemistry for drugs. But the request for a technological explanation will not be exhaustively satisfied by an appeal to the physics or chemistry of the matter. Why is this? One response, to be developed below, is that an adequate technological explanation must consider the audience to which it is addressed. But a second reason is that technologies are made by humans and at some point it is always appropriate to consider the impact on human living of a given technology. So a physical explanation is never an exhaustive technological explanation. Finally, while the design and production of technological artifacts is a complicated set of interlocking and overlapping processes, it is not the case that there is only one way to proceed in developing technologies. The shape, function, components of an artifact take the form they have because of a variety of contingent circumstances, i.e., there is no one way to do this job. It is, perhaps, capturing that sense of contingency that is the most important and most difficult part of a theory of technological explanation.

3.2 Other theories of explanation Other theories of explanation have been devised which try to account for the phenomena to be explained by appeals to other factors than the natural world. There are for instance, teleological explanations, social explanations, and psychological explanations. While each of these can provide good explanations for some things and/or events, they are all individually inadequate for the purposes of technological explanation because, like physical explanations, they tell only part of the story. Thus a teleological explanation accounts for the behaviors of a given phenomenon in terms of the final end for which it was constructed. However, that tells us nothing about why it has the design it does or why it is constructed out of these materials rather than these others nor how its parts work together. Likewise, social explanations ignore the physical world in which artifacts are embedded. This is not to say that there are not social components of an adequate technological explanation, it is just that more is needed than just the social. Thus, if I want an explanation of why this dam was built here, it will surely not be enough to appeal to the politics involved. The geology of the site plays a role as well as the availability of materials, the appropriateness of its location etc. 6 See

[Pitt, 1988].

Technological Explanation

865

Regarding the technological the task of constructing a general theory of explanation has not even been an objective, with one exception to be discussed below, since it is not clear that a general account of explanation in this domain is possible given the kinds of things that can be in need of explanation. It may also be the case that since there was a common misconception that since, it is alleged, technologies are merely applications of some science or other,7 then the search for a technological explanation will naturally revert to a scientific explanation. In fact, it is not clear what a theory of technological explanation is supposed to explain. If we focus only on technological artifacts (leaving systems and social technologies aside for the moment), the design and function of the artifact can each be in need of explanation [Kroes, 1998] . Let us refer to these points as artifact specific issues. Attending to artifact specific issues is important, but we need also to explain the artifact’s social impact and the values or value structure associated with its development and evaluation [Winner, 1986]. When you introduce the topic of values, things get very sticky very quickly. Some will argue that technological artifacts are foils of ideological systems [Winner 1986], others will claim that they are valueladen in other ways (Kroes, personal communication) and others continue to claim that artifacts are value neutral [Pitt, 2000]. In each of these scenarios it may be the case that it is the decision-making structures behind the development of the artifact that need to be explained in order to understand how an object came to be what it is and do what it does. To develop a theory of explanation for these latter subjects will take us further and further away from the task of explaining specific features of specific artifacts. In the long run this is what might be necessary; to explain some feature of an artifact may require that ultimately we have to explain the motivation for its coming into being, which will in turn require an explanation of the social and economic system from which it emerged, etc. However, there is a danger in taking only this direction. To move into this mode runs the risk of falling under the seductive spell of social constructivism and its mantra, “it is social all the way down”. However, it is not the case that all is social, not, at least in some non-trivial fashion. Nevertheless, acknowledging the social allows for a distinction between artifact specific explanations and social explanations. This is a distinction that will prove useful as we explore the kind of explanation we seek with respect to the technological. One non-constructivist area that has attempted to explain technical or technological development in general terms is economics. Like most economic accounts, the appeal to rational self-interest, market forces, evolutionary scenarios or class conflict, presents a narrow vision of the factors involved in our complicated technical world, (see [Elster, 1983]). For reasons explained below, technical explanation cannot rest on what amounts to ideological critique couched in economic terminology. More to the point, economists of various stripes have been concerned to explain technological change in economic terms, which is not the same as offering a technological explanation. 7 See

[Pitt, 2000] for an argument against the technology as the handmaiden of science view.

866

Joseph C. Pitt

4 QUESTIONS AND INTERNAL AND EXTERNAL AUDIENCES Returning to artifact specific issues, we find that a close examination of their explanatory demands takes us beyond the specific to a level of greater generality, even if short of universality. The issue of explanatory demand is crucial to unraveling the problem of technological explanation. If, as writers from Hempel to Achinstein agree,8 scientific explanations are answers to why-questions, to form an adequate answer depends to a very large degree on who is asking the question and to whom the answer is directed; in other words, an adequate answer depends on the audience. But explanations are more than answers to why-questions. They also answer how-questions. Further, there are at least two very different audiences asking questions about technologies: internal and external and it is not obvious from the start which type of question they are asking and what is the best way to answer it. The internal audience consists primarily of workers within a specific technological context, i.e., engineers, designers, etc. These are the individuals who are involved in developing the technologies in question. Their questions concern issues surrounding the design, the materials employed, the nature of the system into which the technology fits (more on systems below), meeting the design specifications and so forth. In short, they ask a lot of how-questions. The external audience consists of technology users, entrepreneurs, developers, politicians, critics, etc. Further, with an external audience there will be differing demands of generality. The same question can be asked by different audiences but it can be answered by appealing to more or fewer specifics. For example, if the question is “Why did that light bulb turn on?” one specific answer could be that I flipped the switch, thereby construing the question as a how-question. That might be all that is needed. However, the simple question may mask a deeper one such as “where does the electricity to power the light come from?” which is a more complicated how-question. The answer to that deeper question might appeal to the concept of an electrical grid and how distinct places, like houses, get connected to the grid and how electricity is dispersed throughout a local site such as a house through a wiring system. In this context a why-question might be of the form “Why is the switch placed at that height?” The answer appeals to building codes and opens the door to social factors. A full explanation of why the light bulb turned on, therefore, requires a lot of ground to be covered, from the wiring of the house, the electrical grid, building codes.9 It involves a little bit of science, a lot about design and function and an acknowledgment of the role of the social. So it is beginning 8 See

[Pitt, 1988]. some respects, this emphasis on the different scopes of the issues depending on the question asked, is reminiscent of Larry Bucciarelli’s discussion in Chapter 1 of his [1994] classic when he ruminates on the question “Do you know how your phone works?” He concludes, “I conjectured that there could be no unique criterion for judging responses; there could be as many legitimate, that is to say accurate, ways to describe how the telephone works as there are respondents” (p. 4). 9 In

Technological Explanation

867

to look like a technological explanation is going to be a complicated thing. But the fact that it is complicated does not mean it is unstructured. The structure is supplied by invoking the concept of a system. 5 TERMINOLOGY A word on terminology — technical versus technological. “Technological” is used here to discuss systems, both mechanical and social, as means of controlling and manipulating the environment, writ large and small. “Technical” is used to refer primarily to artifacts and mechanical systems, but there is nothing crucial in using these terms in these ways. What is important is the realization that explanations in the realm of the technological/technical require appeals to systems of varying complexity. The relevant system or systems constitutes an explanatory matrix that must be demarcated with care.10 Depending on which system is invoked, a different explanation may be required. In this sense the construction of a technical/technological explanation may be more of an art than a science. Indeed, it is a function of the skill of the person asked to give the explanation in determining what will satisfy the questioner. 6 SYSTEMS Common to adequate answers for internal and external audiences, however, is the notion of a system. A system is a set of relations among other things, places, artifacts, social institutions, and individuals. “System” is a broader notion than a similar concept, “context”. A context is a specific set of relations in a specific space and at a specific time, even if that time is actually an extended period, such as the Scientific Revolution. “System,” however, denotes a more general and abstract set of relations which can be represented as schema’s, for example, as line drawings showing various types of connections without there actually being such a system in physical existence. Thus, every context can be seen as a system, but not every system is a context. The system may be a simple one, such as the relationship between a person, the light switch, and the light bulb, or it may be more complicated involving systems within systems such as an electrical grid and the wiring of a house. Upon reflection we find that no aspect of an artifact is ever explained in isolation, it is always with respect to its relationship to something else and wherever there is such a relationship, real or conceptually imposed, there is a system. At its most basic, a system is a structured relationship between two or more parts. Further, I will argue that understanding that some artifact is itself a system or embedded in a system is essential to being able to offer or understand a technological explanation. 10 In some respects this idea that we offer an explanation in the context of a system resonates with Cummins’ [1975] proposal that functional explanations are offered against a set of background assumptions and tacit knowledge.

868

Joseph C. Pitt

That is, a good technological explanation relies on the idea that objects, person and systems are related to one another in differing ways and proper placement in the appropriate system of the thing (broadly construed) to be explained is crucial to being able to understand the explanation as well as to formulate one. This is also why understanding who the audience is for an explanation is so important. The person offering the explanation must be able to refer to a system that will be understood by the person to whom he or she is offering the explanation. In this respect then and for our purposes, “system” will be considered a fundamental concept and the factors bearing on adequate technological explanations will rely on considerations of systems. However, it is not enough simply to appeal to a system, for there are two major problems. The first concerns individuating systems. The second concerns the kind of information provided in the answer. As alluded to above, when it comes to individuating systems the question is how do we determine which is the most appropriate system for our purposes in this specific instance? One way to approach this problem is to return to the idea of differing audiences and why- and how-questions. If we also introduce the idea of a feed-back loop we can begin to see how the appropriate system can be identified. The appropriate system will be a result of the audience asking for the explanation. To determine which audience, and how complicated a system we need to appeal to, we ask questions. Let us return to the light switch example. If the initial question is “Why did the light come on?” and the initial answer is “Because I flipped the switch” and if there is no further question forthcoming — then the explanation, for that audience, at that time, is complete. At this point we can, if we wish, assume further that the questioner is a member of an external audience not interested in the further workings of switches and electric grids.11 But, if the questioning continues, we must begin to explore the kind of answer that would satisfy the questioner and reassess our initial assumption as to what kind of answer would suffice. It might come to mind that the person asking the question might not be a member of the external audience, he or she might be a member of the internal audience and wishes to know the source of the electricity and how the grid is constructed. In short, it will take some effort to determine what kind of an answer will satisfy the person seeking an explanation. However, at some point the questioning has to end — this process is not the same one as the four year old child asking why the sky is blue, unwilling to settle for any answer. The practical end point in a technical explanation is marked by an indication of satisfaction by the individual asking for the explanation. But, it might be asked, does this not end in a tight circle: the explanation is sufficient when the audience is satisfied and the audience is satisfied when the explanation is sufficient. No, it is not circular for the explanation is 11 Of course individuals can be members of both internal and external audiences. Thus an electrical contractor could be satisfied by the “I flipped the switch” answer when that is really all he wanted to know at that time in that place. On the other hand, when taking a busman’s holiday he might pursue further questions just because he wants to see if the folks who wired this house did anything different from what he would have done.

Technological Explanation

869

adequate when the audience is satisfied and the audience is satisfied when they have no more questions. However, it may be the case that the audience is satisfied for the wrong reasons. This is the second problem. But before we turn to it, we need to look at a similar proposal to the one just proposed. In his 1975 paper “Functional Analysis”, Robert Cummins develops an account of a functional explanation in contradistinction to Hempel’s [1965] DN theory. Although developed independently, there are many similarities between Cummins’ account and some of what is being proposed here. Cummins, for example, wants to explain a biological function in terms of its contribution to the proper working of a biological system, where that is accounted for in terms of an organism’s disposition to do A in circumstances C. He employs what he calls an “analytical strategy”, by which he means, roughly, accounting for a given function by way of analyzing it into sub-functions. He also talks about the organization of the organism in terms of a “program”. Thus, you can explain the function of a sub-function in terms of its contribution to the program of the organism, where the program is a description of what the organism is supposed to do or how it is supposed to behave. Now, technological explanations have to account for more than functions, but Cummins’ view supports, in many respects, the intuitions behind the systems account of technological explanation being developed here. It is especially interesting to see how he handles the tension between seeking ever-finer sub-functions and more comprehensive programs. He offers no systematic answer to the problem of when to stop seeking finer details and when to rely on the “sophistication” of the program. It seems to be a balancing act that depends on the degree of acceptance of the given explanation, a process very similar to the situation in systems explanation for technologies, that is how to arrive at the stopping point in asking questions [Cummins, 1975, pp.760-762]. How an answer is phrased also can make a difference with respect to its degree of acceptance. We are all familiar with the advertisers’ gimmick of making their product appeal more desirable by using such phrases as “scientifically proven” or “as shown in a scientific study at a major university”. According to Dennis Carlat in the June 2008 issue of Wired, a recent study at Yale University reported that spurious explanations were deemed more satisfactory when preceded by the phrase “Brain scans indicate.” That suggests that appealing to the audience’s valued set of experts or expertise can bring about satisfaction with an explanation without actually achieving a state of genuine intellectual understanding; satisfaction here is a psychological as opposed to an epistemic state. However, this is no more a problem than that we face when investigators use false statistics or appeal to made-up data. Sometimes researchers lie. Sometimes fathers, tiring of the constant “why?” of their child, will simply make up an answer that the father knows will satisfy them. Sometimes a person attempting to provide a technological explanation will throw in an appeal to something he thinks might satisfy his audience, knowing that it is actually misleading. These things happen — but, as in detecting scientific fraud, we must constantly be on guard. That these things happen does not

870

Joseph C. Pitt

discredit the proposed account any more than that lying scientists discredit all scientific investigations.

6.1 System and design Returning to the issue of explaining artifact specific issues we can see further the importance of system. Consider some of the factors involved in explaining the design of an artifact. Artifacts do not take the shapes they have by accident; they are designed with specific factors in mind. Some such factors include how the artifact is to be used — so one factor in design concerns how the artifact relates to a broader system of use — in its use it will interact with other artifacts, or the natural world, itself a system of systems. Another factor involves the marketability of a product — attractive and user-friendly designs sell — here part of the system is actually outside the design process itself insofar as it is the system of sales and consumers, an Aristotelian final cause as it were (see [Bucciarelli, 1994]). Another consideration involves the cost, availability, and reliability of materials, factors again involving appeal to a broader system and may in turn relate to marketability and use, which themselves require appeal to systems. At this point we are in a position to answer a possible objection to the account being proposed. It has been suggested here that the determination of the adequacy of an explanation, by way of answering why- and how-questions, is a direct response to the audience asking the questions and our ability to relate the answer to appropriate systems. That, however, it might be argued, fails to distinguish an explanation from an adequate explanation. But we have already noted that satisfying answers to why-questions direct the questioner to the manner in which the artifact functions in a system, relating to other artifacts and other components of the system. So, if the question is “how do I get the light to come on?” directing the questioner to begin by lighting a votive candle before throwing the switch would not count as an adequate answer since there is no way to relate such an action, lighting a candle, to the electrical system in an satisfactorily explanatory fashion.

6.2 System and function Function is another system-intrinsic feature of an artifact.12 The function of an artifact can only be fully explained in terms of how it fits into a system.13 It does not matter if we are talking about function as use or function as in how it works, we cannot escape appealing to a broader context. Even if we are only concerned with 12 The literature on functions is almost as enormous as that on explanation. I am relying here on a common sense appreciation of what a function is, recognizing that the circumstances in which we appeal to “function” may in fact change the meaning of the term. Thus, asking for the function of a turn indicator on the steering column of an automobile is not same as asking how well a device functions. 13 Although the account here was developed independently of [Cummins, 1975], the two approaches agree strongly on this point.

Technological Explanation

871

the mechanics of the artifact we still are employing a system for the explanation since the artifact itself is a system if it has more than one part. The larger question of why it does what it does must involve an appeal to something beyond the artifact itself. Consider an example, the internal combustion engine (ICE). We can ask many questions regarding the ICE. What does it do? If the answer is “it transforms energy,” we have a clearly unsatisfactory answer. We can pursue satisfaction by moving in one of two directions or both. If we are a member of the internal audience, we can ask how it transforms energy, seeking a description of the mechanics of the artifact. If we are a member of the external audience, “Transforming energy” by itself doesn’t tell us anything of value. So, let’s change the question. What is the ICE used for? “It is used to propel a tractor.” Now we are getting somewhere, but we have also appealed to an incipient system or two. The first system is that created by the relation between the ICE and the tractor. The second is the relationship between the tractor and something else. The tractor is a system of parts that are brought together to produce an artifact that can, among other things, plow a field. If you didn’t know anything else you wouldn’t see that as an explanation. But if you do in fact know something about agricultural production, and if you can begin to see the utility of the artifact in terms of producing a crop , harvesting the crop, getting the crop to market (system of transportation), selling the crop (economic system), having it transformed into something usable as food, etc., then the explanation of the function of the ICE as a mechanism for propelling a tractor begins to make sense. A fully adequate answer requires also knowing how the ICE produces energy and what it is used for, both of which require appeals to systems.

6.3

System and structure

Structure, the final artifact-specific issue to be discussed here, is also closely aligned with the design of an artifact, and an explanation of why an artifact has the structure it does also requires an appeal to systems. A design is a design of some structure or other. We can speak of the internal or external structure of an artifact. For example, in designing a modern skyscraper, it is common to configure the building around a central service utility column containing elevators, power and water lines, etc. What the external structure looks like is a function of many factors, including location, building codes, materials, cost, aesthetics and, last but not least, the architect’s ego and the desires of the individual or group commissioning the design. Each of the factors listed are themselves systems – different kinds of systems, to be sure, but systems nonetheless. In seeing how the structural and functional designs are related we can see a number of things. First, if we simply looked at the external structure of the building, we would not learn very much about it. Looking at the external structure only reveals the aesthetics of its design and how the building relates to its surroundings, if we are provided with a site plan. If we took a political approach we might speculate on the kind of power statement the building makes given its size. However, to appreciate the

872

Joseph C. Pitt

design fully, we need to know more than how the building looks and whether it fits in its location, we need to know how the structural and functional components work together. We need to know, for example, how people who work on the top floors are going to get there. This is where we appeal to the internal structure of the building — the role of the central core in proving the elevator conduit, along with the means for getting electricity to the rest of the building. We can now turn to an account of what it means to say that the three artifactspecific issues we have been considering, design, structure and function of an artifact, can be explained in terms of systems. And here we can clearly differentiate between conceiving of the building as a system and conceiving of it in its context. From a systemic point of view, we don’t need to think about a particular building, we can do this in generic terms. We can, for instance, draw a large three-dimensional rectangle on a piece of paper. Then we can add some stick figure trees, to give some sense of its size, and perhaps dot in some other buildings to show how it its supposed to fit in a location. Then we turn to the building itself and pencil in the central core — indicating where the electric lines will go, how the plumbing can work — schematic floor plans. But if it is a specific building I am working on, then we need to know exactly how high it will be, where in the city it will be located and what putting it there will do to the area. We also need to know about the location infrastructure, does it have enough water for the building, how will the electricity get to the building, parking. When we turn to the internal structure of the building, it is not enough to say that each floor will contain offices, elevators, and rest rooms, we need to show where these will be located, how large they will be, traffic patterns, and so forth. Here some social factors come into play as we have to decide, for example, between an open floor plan or individual offices. In that case, how many corner offices and how many windows? It might be the case that the external design of the building does not allow for corner offices as the corners are structural. Here the architect needs to be sensitive to office politics and human psychology. As already noted, an explanation is a response to, among other questions, howand why-questions. Thus, why does this artifact have this design? The answer will be in terms of the systemic factors noted above. If we are talking about automobile designs, one possible answer could appeal to aesthetic fads, or the interchangeability of parts in different models — always to something external to the artifact itself. Likewise for function: an artifact does what it does in order to contribute to the successful operation of a system in which it has a role to play. To explain the structure of an artifact, we appeal to a variety of systemic factors that contributed to the artifact having the structure it does. The explanation consists in positioning the artifact with respect to the system in such a way as to answer the why-question satisfactorily. Thus, if we are explaining why all (so it seems) so-called cross-over vehicles tend not only to look alike, but are equally ugly, we might refer to the attitude of social superiority owners of such vehicles cultivate. Thus: “we don’t have to drive big gas-consuming monsters like Chevrolet Suburbans or Ford Expeditions or Land Rovers to have four wheel drive — but to let

Technological Explanation

873

you know that, we purchase cars so ugly you can’t ignore them” — perverse, but once placed in a social environment, the ugliness of these vehicles can be explained in a satisfactory fashion. So far we have been concerned with the explanations of what, for lack of a better term, we can call “hard” artifacts, automobiles, buildings, hammers. We have argued that explanations of the design, structure, and function of hard artifacts involve appeals to systems in which they are embedded in some form or other. These explanations are, minimally, answers to how- and why-questions. But, there are other kinds of questions that can be asked regarding hard artifacts. One disarmingly simple one is: “What does it do?” To answer this question we also need to place the artifact in a system. Explaining what the artifact does involves relating one part of the system to another. Thus, flipping on an ordinary light switch connects the power grid to the light bulb. When explaining what an automobile does we immediately appeal to the broader system of the social and natural world when we say that it is a means of transporting people and goods from one place (which must be defined in the context of a system) to another. When we explain what a skyscraper does, we turn the building itself into a system in which people work, work which may involve communicating (via different systems) with people in other places (system required) in order to move goods and services around the world, thereby appealing to the broader social system of transnational commerce. 7 THE SOCIAL The explanation of hard artifacts increasingly involves appeal to the social domain as we get farther and farther away from explaining the mechanics of the artifact, i.e., how it does what it does, to an internal audience of, for example, engineers and respond to the concerns of an external audience asking about its impact on the society. But to understand the social in its explanatory mode is to understand it as a system, or a set of systems. Examples of social systems used to explain features of hard technical artifacts include economic markets, communication systems, legal systems, building codes, and standardized metrics. But there are also non-social systems we appeal to by way of explanation such as the environment. Thus the new movement called green architecture is a response to increasing awareness of the effects of the built environment on the ecology of the planet. Such explanations only make sense in the context of thinking of the ecology of the planet as itself a system, something we seem increasingly to do. In addition to hard technical artifacts we must consider the nature of explanations in the context of various social technologies, i.e., so-called soft artifacts. Social technologies help human beings arrange their affairs. We often explain what people do by appeal to some feature of a social technology. Thus, a response to the question “Why did he slow down?” could be “Because the speed limit was lowered.” Implicitly this is an explanation by way of appeal to the regulative powers of the law — one’s behavior is often shaped by legal constraints. Legal systems are developed in order to provide orderly means for adjudicating conflict

874

Joseph C. Pitt

and for shaping social behavior — they are deliberate constructions that have a function.14 To explain these functions we must appeal to the needs of a different system: society. To call society a system is obviously problematic. Which society do we mean? Sometimes society is synonymous with nation state, sometimes with a broader historical/cultural group, e.g., Western society, sometimes with more specific groups such as religions, sometimes with geographic locations such as Southeast Asia. Then there are even smaller groupings such as the Mafia, a city, a sport. As international trade becomes increasing global in its structure and interdependence, we just may end up at some point meaning the world, which tells us very little. Thus, explanations that appeal to society must be society-specific or run the risk of being empty. Not all forms of societal control of behavior are the result of deliberate constructions such as a legal system. There are cultural constraints that may not be obviously derived from a system. For example, it is often difficult to understand economic factors, economic theories not withstanding, that influence behavior as occurring within a system — for there may be conflicting systems at work, such as, at least in the United States, economic security versus public service. This is where there is a need for careful work identifying conditions that can be used to isolate systematic factors to explain how social technologies themselves can play an explanatory role.

8 EXPLAINING FAILURES We now turn to a different issue that calls for technological explanation: failure. Technical failures occur at all levels, from O-rings on a space launch vehicle to voting machines to power grids to social services systems to education systems. The analysis and explanation of failure comes under the general heading of forensics. Explanations of hard technical artifacts involve engineering forensics. Engineering Forensics can be seen as a form of reverse engineering, during which process a mechanism is taken apart to see what makes it work. In the specific case of engineering forensics, an investigation into an incident in which some artifact, mechanism, or system failed is undertaken with the objective of determining if an artifact or mechanism was the cause of failure and what specifically went wrong. Failures of social systems and social technologies require social forensics. Social Forensics go beyond engineering forensics in that these investigations examine the failure in the context of a social situation, looking for human failures and sys14 There is an air of paradox to the claim that legal systems are deliberate constructions when one considers what are often referred to as “common law” systems, for these have developed over time and generally in a piecemeal fashion. However, it is enough for our purposes to point out that they couldn’t have developed at all if the idea of a law was not generally understood. Once in place, the society can add laws as it sees need. Granted this is different from constructing a legal system and then imposing it on a society, but common to both is the acceptance of the idea of law.

Technological Explanation

875

tem failures. The point of engineering and social forensics is to explain why the artifact or the social system failed to do what it was designed or evolved to do. Hard artifact failure usually comes as a surprise and sometimes is accompanied by social misery or even disaster. The example of the failure of the O-rings in the launch vehicle of the shuttle Challenger gives us a good case of an explanation that deals with a particular part of a complex artifact system that didn’t do the job it was expected to do despite the fact that the demands placed on it exceeded its specifications. It might be argued that the failure occurred not so much because of the failure of the artifact system, but because of the social system in which it functioned. The warnings of the engineers that the O-rings would fail were overruled for other, some say political [Vaughan, 1996], reasons. This is a valuable example because it exhibits the complexity of the relationship between hard artifact systems and social systems and between engineering forensics and social forensics. The Rogers Commission Report, the official report of the U.S. government’s investigation into the incident, identified the specific cause of the disaster as the failure of the O-rings, but also noted that there were other systemic failures to be considered as well. The failure of the O-rings was a structural failure in the sense that this part did not work in this structure under certain circumstances, circumstances outside the specifications for the part in question. It might be argued that this is a functional explanation. But that is not clear. For when we appeal to the circumstances outside the specifications for the O-ring, we are appealing to more than the function of the O-ring. The failure of the shuttle mission as a result of the failure of the O-ring can also be explained in terms of the failure of the social system that developed, managed and used the shuttle system. What we need to guard against here is making the social system the explanation for everything that went wrong — that way lies the rhetoric of social criticism, but it does little to explain the failure in a way that leads to corrections that actually make a difference. In short, appealing only to the social has the effect of black-boxing the engineering.

8.1

The Challenger example

In the discussion of the Challenger disaster, the question of the responsibilities of the engineers involved is often raised. Is it enough for engineers to design artifacts that meet the specifications of the client or do they have further responsibilities to determine if those specifications are reasonable given the functions of the ultimate product? To raise such issues also allows a distinction between seeking causes for failure and assessing blame. On the one hand we want to know what broke and why. On the other hand, we want to know who was responsible for the situation. In the first case we are looking for technological explanations, in the second, for judicial or even moral ones. It may be the case that an ultimate resolution of the entire situation requires assessing blame and putting constraints in place to correct for whatever actions or inactions occurred leading to the problem. But it is not clear that doing so is a necessary part of a technological explanation,

876

Joseph C. Pitt

construed as an outcome of engineering forensics. One might argue, however, that it is a necessary end point for social forensics, for it is not enough to know why the system failed, but what needs to be done to fix it and sometimes that means identifying individuals as failing to be responsible, instituting review processes, laying down what amounts to moral directives, etc.

8.2 The 2000 US presidential election example It can also be argued that we can explain the results of the 2000 United States Presidential election in terms of both a hard artifact technical failure and a set of failures of the social technologies in which it was embedded. The so-called hanging chads were a result of the failure of the voting machines in Florida to operate correctly. The resulting decisions by state officials and finally the United States Supreme Court can be seen as failures of the social and political systems. Laws were broken but no action was taken by state officials or even federal officials to identify responsible individuals and have them prosecuted. In the second case, we have the politics of judicial appointment at the highest level trumping legal precedent and the procedures specified by the U.S. Constitution.

8.3

The Ladbroke Grove railroad crash example

Finally we should look at a different kind of failure case, also leading to disaster. This is one in which political agendas did not overtly appear to play major roles, but compounded instances of human negligence led to a sad result and, it can be argued, there was no specific technical failure. The case in question is the Ladbroke Grove rail crash outside of London, England on 5 October 1999. At approximately 8 am two trains traveling at high speeds collided, resulting in 31 deaths and 523 other casualties. The immediate cause was identified by an official investigation as the failure of the engineer of one of the trains to obey a stop signal. Subsequent study, however, revealed a more complicated story. The operator of the train that missed the red signal had only been on the job for two months. The signal itself was non-standard — structured as a reversed “L.” The red signal was located to the left of the other lights, rather than at the bottom of a standard three light array. Further, the signal was obscured from view by overhead roads and bright sun. Finally, those responsible for the maintenance of the track and its signals had failed to take necessary action to correct the problems at this site, despite the fact that there had been eight instances of trains missing the red signal there in the previous six years — luckily all those trains managed to stop before an accident occurred. The frequency of such occurrences should have alerted someone to a problem. While the signal had worked properly, there apparently had been a series of events that contributed to the final tragedy that no one had bothered to connect. Yes, there was system failure, but not out of blind ambition or greed or stupidity. In this case it seems that a series of small changes over time resulted in a situation that no one had anticipated, despite a number of warning signs over the years. However, in this case, the official investigation lead by Lord Cullen

Technological Explanation

877

resulted in record fines, suggesting that there was a clear determination of blame. 9

9.1

CONCLUSIONS AND OBJECTIONS

A final defense of systems

It may be objected that this account of technological explanation is inadequate; it is simply too soft, resting as it does on interpersonal skills rather than rigorous logical connections. As we have seen, the use of iterated why-questions provides a method for both individuating systems and locating the appropriate explanation. That is, by finding out what kind of framework will provide understanding on behalf of an inquirer, we insure that the technological/technical explanation actually answers the question and, as we have suggested, answering how- and why-questions is what explanation is all about. This also helps with regard to another, so far unexplored, issue: how to determine what exactly the question is. Earlier it was noted that the questioning develops in the manner of a feedback loop. The point here is as much to find out what the questioner is actually asking as it is to arrive at a satisfactory answer. In fact, until both the questioner and the person seeking to provide the explanation know what the question is no satisfactory explanation will be forthcoming.

9.2

Failure, success and symmetry

There is yet one final objection (for the time being): the apparent asymmetry between the technological explanation of why something works and why something fails. On the surface it appears as if the explanation for why something works is that it accords with general processes that we understand. In the case of why something fails, we appeal to the particulars of the case. There are two different issues here. Again we see the unwelcome influences of older discussions on our expectations for present concerns. In particular, we see the continuing influence of Carl Hempel. Hempel laid down the condition of symmetry for the logical structures of both explanation and prediction. That is, in DN the same logical form characterizes both explanations and predictions; they are both deductive arguments with their premises containing a law and statements of initial conditions, leading to a statement of fact x, (if one is explaining single facts) or the statement of a prediction that x will occur. If one takes, by way of analogy, prediction to apply to why things work and explanation to apply to failures, there might be something to work with. That is, it might appear that the objection to the asymmetry in technological explanations has some ground. This assumes that the objections raised earlier to the DN model don’t apply to this question of symmetry. But they do — for in DN explanations we need to know which laws to use, and we cannot assume that there is only one theory in use at a time. If the problems produced by issues of individuating laws and competing theories are real problems, then DN fails in deeply serious ways, and the call for symmetry

878

Joseph C. Pitt

of logical structures would seem to fail as well. In short, it is not clear that the explanation of why something works must have the same logical structure as an explanation of why something fails to work. Further, it is rarely the case that an explanation of why something works needs to appeal to the expertise of an operator or the coherence of a system. Those are taken for granted when we do a walk-through of an artifact. Likewise, the presumption from the start is that the parts are properly designed and manufactured. Thus when we explain why the light goes on in the house when I flip the switch, it is in the context of a very big ceteris paribus clause. All things being equal, if the objects are properly designed and manufactured and if the operators are properly trained and competent and conscientious, and if the system as a whole works well, then when I do X, I can expect Y . The more difficult explanations regard the failures, for explaining failure requires that we figure out which components of the ceteris paribus we ought to have thought twice about and what it was in those components that needs fixing. In short, there is a built-in asymmetry in technological explanations and that is not necessarily a bad thing; consider what would happen if this were not the case. If we failed to operate with the very large ceteris paribus clause bracketing our explanations of why things work, we would have to build in all the constraints it is designed to wash over and the end result would be stasis. Thus, let us consider our light switch. The light comes on in the room when I flip the switch only when the switch is properly manufactured and wired, but we know electricians make mistakes in wiring things, and we know that materials used in manufacturing can be flawed, so the switch, even if properly constructed, may not work properly. But we know that we are unjustified in assuming that the switch is properly constructed because the operators at the manufacturing facility that make the switch also make mistakes, sometimes because they are improperly trained, or because they had a fight with their husbands or wives, or because they had two beers at lunch. Knowing all that can go wrong just with the switch, I will be unable to explain why the light comes on because there is too much that can go wrong in the system as a whole. An explanation with all those caveats is no explanation. On the other hand, digging into those assumptions the ceteris paribus clause hides when we turn to explaining failure is just what makes such explanations so difficult to construct but also so valuable. Here we have to uncover the very things we needed to ignore before. Not because they are necessarily part of the final explanation, but because they might be, and to ignore those possibilities is to offer only a surface explanation, which, again, is no explanation at all. In conclusion, the asymmetry between explaining how things work and why they fail is essential to providing technological explanations. In the case of explaining success we need to simplify, and in the case of explaining failure we must get very complicated. This is not to say that explanations of why things work can’t be complicated — they certainly can be. The point here is that the kinds of complications are of a different order when we try to explain why things didn’t work as we expected them to. The complications have to do with the human

Technological Explanation

879

factors, individual or aggregated. This also explains why we don’t blame the artifacts when things go wrong, we blame the people who use or misuse or abuse them. Thus, it is true that guns don’t kill, people do. BIBLIOGRAPHY [Bucciarelli, 1994] L. L. Bucciarelli. Designing Engineers. Cambridge, MA: MIT Press, 1994. [Carlat, 2008] D. Carlat. Mind Readers. Wired, June, pp. 120-128, 2008. [Cummins, 1975] R. Cummins. Functional Analysis. The Journal of Philosophy 72, 741-765, 1975. [Elster, 1983] J. Elster. Explaining Technical Change, Cambridge, UK: Cambridge University Press, 1983. [Hempel, 1948] C. Hempel. Studies in the Logic of Explanation. Philosophy of Science,15, 1 135-175, 1948. [Hempel, 1965] C. Hempel. Studies in the Logic of Functional Explanation. In Aspects of Scientific Explanation and other essays in the philosophy of science. New York: Wiley, 1965. [Hughes, 1983] T. Hughes. Networks of Power: Electrification in Western Society, 1880-1930. Baltimore: Johns Hopkins University Press, 1983. [Kitcher and Salmon, 1989] P. Kitcher and W. C. Salmon, eds. Scientific Explanation, University of Minnesota Press, 1989. [Kroes, 1998] P. Kroes. Technological Explanations: The relation between structure and function of technological objects. Society for Philosophy and Technology, Vol. 3, http:scholar.lib.vt.edu/ejournals/SPT/v3n3/KROES.html, 1998. [Kroes, 2001] P. Kroes. Technical Functions as Dispositions: A Critical Assessment. Techn´ e, Journal of the Society for Philosophy and Technology, Vol. 5, http:scholar.lib.vt.edu/ejournals/SPT/v5n3/KROES.html, 2001. [Pitt, 1988] J. C. Pitt. Theories of Explanation, New York: Oxford University Press, 1988. [Pitt, 2000] J. C. Pitt. Thinking About Technology, (Originally published by Seven Bridges Press, New York), now at www.phil.vt.edu/HTML/people/pittjoseph.htm, 2000. [de Ridder, 2007] J. de Ridder. Reconstructing Design, Explaining Artifacts: Philosophical Reflections on the Design and Explanation of Technical Artifacts. Simon Stevin Series in the Philosophy of Technology Vol. 4 (Ph.D. Dissertation), Delft University of Technology, Department of Philosophy, 2007. [Rogers Commission, 1986] Rogers Commission report, Report of the Presidential Commission on the Space Shuttle Challenger Accident, 1986. [Shew, 2007] A. Shew. Beaver Dams, Spider Webs, and the Sticky Wicket: An Investigation into What Counts as Technology and What Counts as Knowledge. MS Thesis, Science and Technology Studies Graduate Program, Virginia Polytechnic Institute and State University, 2007. [Vaughan, 1996] D. Vaughan. The Challenger Launch Decision: Risky Technology, Culture and Deviance at NASA. Chicago: University of Chicago Press, 1996. [Vincenti, 1990] W. Vincenti. What Engineers Know and How they Know it: Analytical Studies from Aeronautical History. Baltimore: Johns Hopkins University Press, 1990. [Wimsatt, 1980] W. C. Wimsatt. Teleology and the Logical Structure of Function Statements. Studies in the History and Philosophy of Science, 3: 1-80, 1980. [Wimsatt, 2002] W. C. Wimsatt. Functional Organization, Analogy, and Inference. In Ariew A. (ed.), Functions: New Essays in the Philosophy of Psychology and Biology, Oxford, New York: Oxford University Press, pp. 173-221, 2002. [Winner, 1986] L. Winner. The Whale and the Reactor. Chicago: University of Chicago Press, 1986.

This page intentionally left blank

Part V

Norms and Values in Technology and Engineering

This page intentionally left blank

INTRODUCTION TO PART V Ibo van de Poel, associate editor At first sight the notion that technology and engineering are value-laden hardly seems controversial. After all, engineering pertains to the creation of useful technological artifacts. Technological artifacts have certain, intentionally designed, functions and are therefore good for serving certain ends. This basic insight lends, so it seems, a normative dimension to technology.1 Given this basic normativity of technology, one might wonder why certain authors (e.g. [Florman, 1987; Pitt, 2000]) have contended that technology is valueneutral (the so-called neutrality thesis). Upon closer inspection, however, the neutrality thesis does not deny that technologies are useful nor does it maintain that technology is without value. It asserts instead that technology is a mere means to an end. Technologies may be valuable as means to ends — and thus have what may be called instrumental value or utility value — but their ultimate value depends on their particular uses and such choice of ends depends more on the actual users than on the technology itself or its designers. As the American National Rifle Association once put it: “Guns do not kill people, people kill people”. The thesis that technology is value-neutral then rests on the assumption that instrumental value — i.e. the value of being a means to an end — is ultimately no value at all. This assumption has indeed been defended by some moral philosophers (e.g. [Dancy, 2000]). If one wants to maintain, against the neutrality thesis, that norms and values are in some respects inherent to technology and engineering, one could adhere to a variety of argumentative strategies, which are not mutually exclusive. Firstly, one could argue that the neutrality thesis is wrong and that technology is inherently normative. This strategy is employed in the chapter by Radder, which critically discusses four different existing approaches to the normativity of technology. On the basis of a detailed account of the notions of technology and normativity, Radder concludes that norms are inherent to technology. Similarly, the chapter by Van de Poel (in particular Section 3) argues that some types of values are inherent to technology. Secondly, one could argue that the instrumentality of technological artifacts does at least support some normative statements. Thus, Franssen discusses the normativity that is involved in statements like “This is a good hammer”, “This coffee-vending machine malfunctions” and “This car ought to start”. Thirdly, one could argue that as a practice engineering is not value-neutral (even if as an instrument technology perhaps is). Engineering is not value-neutral as a 1 I use the notion “normative” to refer both to the evaluative domain (in which values prevail) and to the prescriptive or deontic domain (in which norms prevail).

Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

884

Ibo van de Poel, associate editor

practice because it is guided — like any practice (cf. [MacIntyre, 1984]) — by a range of normative standards. One might think, for instance, of the normative standards that are implied in technical codes and standards but also of the fundamental canon of many American engineering codes of ethics: “Engineers shall hold paramount the safety, health and welfare of the public”. Pritchard discusses the professional normative standards that are typical of engineering. As the above points already illustrate, the normative dimensions of technology and engineering can be analyzed from at least two perspectives. One perspective is that of the normative standards inherent in engineering as a practice. The other relates to the normativity inherent in technological artifacts and systems. The second perspective differs from the first from the point of view that the normativity inherent in technological artifacts and systems is not only shaped by, and affects, engineers, and other technology developers, but also by users and various stakeholders. The two perspectives are not of course unrelated. This is underscored by the canon in engineering codes of ethics that requires engineers to hold paramount the safety, health, and welfare of the public. One of the places where the two perspectives meet is in the design phase. Design is obviously steered by professional engineering standards, but it is also the terrain where the expected value of to-be-designed technical artifacts is contemplated and created. The chapter by Van de Poel focuses on how the expectation of value plays a role in the design process. It focuses in particular on the plurality of values that affect design and on how engineers can, and according to some people, should deal with conflicts between those values in the design process. Given the normativity of technology and engineering, a range of more specific values can be distinguished that play a part in engineering and technology. One such value is efficiency. In broad outline, efficiency can be described as the achieving of an end with as few as possible resources. This matches the instrumental notion of technical artifacts as means to ends. In striving for efficiency engineers likewise see the ends as given and try to devise efficient means to achieve such ends. This fits the self-image that at least some engineers have of themselves as neutral problem solvers and it helps to explain the primacy of efficiency in engineering, at least rhetorically; its role in actual design practice is sometimes less obvious [see the chapter by Alexander, Section 5]. Efficiency is, however, only seemingly neutral, because the notion crucially depends on what is adopted as an end and what is defined as the input that needs to be minimized; issues that regularly lead to considerable controversy. Moreover, as is analyzed in detail by Alexander, the history and application of the notion of efficiency is much broader than engineering, even if engineering is a main area where the value is applied. A second category of values that has a role to play in engineering and technology is aesthetic value. Apart from areas such as product design and architecture, little attention has so far been paid to aesthetic values in engineering. The chapter by Schummer, MacLennan and Taylor embarks on a more general analysis of the role of aesthetic values in technology and engineering design by looking at three dif-

Introduction to Part V

885

ferent areas: urban landscape planning and architecture, chemistry and software engineering. A third category of relevant values are moral values. One moral value that is recognized as crucial by almost all engineers is safety. Hansson thus discusses risk and safety in technology and, among other matters, the ethical issues of technological risk. Safety is obviously not the only relevant moral value in engineering. Other relevant moral values include sustainability, health, privacy, justice and democracy. Some of these moral values are discussed in general terms in the chapters by Van de Poel and by Mitcham and Briggle. The chapters in Part VI, which deal with the philosophical issues of specific engineering disciplines, discuss the more specific moral concerns raised by specific technologies or engineering domains such as biotechnology and computing and information technology. A fourth category of values that is relevant to engineering is epistemic value. If one assumes that science and engineering have different aims (knowledge versus useful products), then the thesis that science and engineering are characterized by different epistemic values has at least some prima facie plausibility. Engineering science does not, however, fit neatly into this framework because it usually aims at obtaining knowledge to design useful artifacts. A way of putting engineering science into the picture might be by looking at the suggestion that the research system is currently going through a transformation from what is sometimes called Mode 1 knowledge production to Mode 2 knowledge production [Gibbons et al., 1994]. Mode 2 is typically assumed to be more application-driven, more interdisciplinary, more undertaken by actors other than universities and research institutes and more heterogeneous. Ziman has suggested that the Mertonian norms of communalism, universalism, disinterestedness, originality, and organized scepticism (CUDOS) that characterize Mode 1 are replaced by what he calls the PLACE (proprietary, local, authoritarian, commissioned, and expert) norms in Mode 2 [Ziman, 2000].2 Much engineering research seems to fit the description provided for Mode 2 better than the Mode 1 description, and Ziman’s thesis could be seen to express the expectation that engineering science is characterized by different epistemic norms than those of traditional science. The Mode 1 – Mode 2 thesis is, however, usually viewed as a transformation of the research system rather than a description of the two modes of research that exist side by side. Ultimately it is hard to find in the literature a systematic exploration of the relevant epistemic differences between science and engineering science which explains why there is no separate chapter on epistemic values in engineering in this part. Obviously, there is still a lot of interesting work to be done on the subject of epistemic values in engineering. The final two contributions discuss two subdisciplines that have made major contributions to the analysis of normative issues in technology and engineering. Grunwald gives an overview of the area of technology assessment (TA). Technology assessment began as an attempt to predict and assess the impacts that technology 2 Ziman uses the terms academic science and post-academic science instead of Mode 1 and Mode 2.

886

Ibo van de Poel, associate editor

has on society. It therefore initially focused on the normativity inherent in technologies rather than in engineering practice; some even believed that TA was value neutral because it only predicted technological consequences while the evaluations had to be done by politicians. In the course of its history, TA become more and more aware that it was itself a value-laden enterprise. Moreover, attempts were made to feed back TA insights into the development phase, for example through Constructive Technology Assessment [Schot and Rip, 1997], so also integrating the normativity inherent in engineering practice. The chapter by Mitcham and Briggle discusses ethics of technology. Ethics of technology and engineering ethics are relatively recent subdisciplines of philosophy. The interaction between ethics and technology is, however, much older and, as they demonstrate, dates back to antiquity. Not only have moral concerns and ethical theorizing influenced engineering practices and the development of technology over the course of time but technology has also influenced ethics, according to some to the extent that we even need new ethical theories if we are to fully grasp technology (e.g. [Jonas, 1984]). BIBLIOGRAPHY [Dancy, 2000] J. Dancy. Should we pass the buck? In The Good, the True and the Beautiful, A. O’Hear ed. pp. 159-73, Cambridge University Press, 2000. [Florman, 1987] S. C. Florman. The Civilized Engineer. St. Martin’s Press, 1987. [Gibbons et al., 1994] M. Gibbons, C. Limoges, H. Nowotny, S. Schwartzman, P. Scott and M. Trow. The New Production of Knowledge: The Dynamics of Science and Research in Contemporary Societies. Sage, 1994. [Jonas, 1984] H. Jonas. The Imperative of Responsibility: In Search of an Ethics for the Technological Age. University of Chicago Press, 1984. [MacIntyre, 1984] A. MacIntyre. After Virtue. University of Notre Dame, 1984. [Pitt, 1997] J. C. Pitt. Thinking about Technology: Foundations of the Philosophy of Technology. Seven Bridges Press, 2000. [Schot and Rip, 1997] J. Schot and A. Rip. The past and future of constructive technology assessment. Technological Forecasting and Social Change, 54, 251-68, 1997. [Ziman, 2000] J. Ziman. Real science. What It Is, and What It Means. Cambridge University Press, 2000.

WHY TECHNOLOGIES ARE INHERENTLY NORMATIVE Hans Radder

1

INTRODUCTION

The title of this chapter implies that technologies are inherently normative. Explaining and defending this claim requires, first of all, a plausible account of the notions of ‘technology’ and ‘normativity’. For this purpose, Section 2 characterizes a (type of) technology as a ‘(type of) artifactual, functional system with a certain degree of stability and reproducibility’, and it addresses the crucial question of how we may successfully realize such technologies. Next, Section 3 explains the notion of a norm as ‘a socially embedded directive concerning what people should (or should not) say or do’, and it examines several important aspects of how norms function in actual practices. Section 4, then, introduces the distinction between contingently and inherently normative technologies and it provides a detailed discussion of the question of why technologies are inherently normative. After a general outline of the argument for this claim, Sections 4.1 through 4.4 address the central aspects of the inherent normativity of technologies on the basis of four different approaches, arranged in chronological order. These sections discuss the normativity of technology in the case of Langdon Winner’s account of the political nature of artifacts, in my own analysis of the material/social control needed to realize stable and reproducible technologies, in the script theory of technological design advocated by Madeleine Akrich, and in the notion of use plans proposed by Wybo Houkes and Pieter Vermaas. The merits and problems of these approaches are assessed and their implications for the main question of this chapter examined. Section 4.5 uses the analyses and examples from the preceding sections to illustrate several important aspects of the role of normativity in technological practice. The fifth and final section summarizes the argument for the claim that technologies are inherently normative and it briefly explores the question of whether there are other ways in which technologies can be said to be inherently normative. It will be clear that the focus of this chapter is on the philosophy of technology. However, since a central aim of the engineering or technological sciences is to contribute to the design and implementation of technologies, this chapter is essential to engineering or technological scientists as well. The basic point of relevance is that, if they aspire to anticipate the successful implementation of their designs, Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

888

Hans Radder

engineering and technological scientists need to be sharply aware of the inherent normativity of technologies.

2

TECHNOLOGY

Any systematic philosophical discussion of the normativity of technology should be based on a plausible account of the very idea of ‘technology’. This is of course a huge subject, that has been treated in detail in a large number of studies (for extensive reviews, see Van der Pot [1994]; Mitcham [1994]; Mitcham and Schatzberg [this volume, Part I]). It will be clear that all those studies cannot be addressed here. What I will do in this section is to outline a theoretical account of technology that, I hope, is both plausible and suitable as a starting-point for an in-depth discussion on the subject of this chapter. In part, this account employs notions that are quite common in the philosophy of technology; in part, it builds on and develops several ideas taken from my own work.1 The resulting theory of technology has two interrelated parts. I start with a characterization of a (type of) technology as a (type of ) artifactual, functional system with a certain degree of stability and reproducibility, and I explain the key notions of this characterization. In the second part, I examine the questions of how, where and when we may successfully realize and maintain such technologies, which will reveal a number of equally important, additional characteristics of technologies.

2.1 A first characterization of technologies The distinct elements of this characterization of technologies can be, briefly and rather formally, described as follows. A system is any aggregate (or collective) of mutually interacting material entities within a certain region of space and time. Thus, a system possesses not only a spatial but also a temporal dimension, which allows us to see processes as systems. Although the notion of a system may also be used in more substantial ways,2 this unassuming definition is appropriate in the present context. It may also be phrased by saying that technologies have a systemic character, because they result from bringing, and keeping, together two or more material entities. By definition, everything that is not included in the system is its environment. The relevant part of the environment comprises the conditions for the successful functioning of the technology in question. Thus defined, the notion of a system includes the qualification ‘within a certain region of space and time’. This important qualification entails that, in thinking about actual technologies, we should always take into account their spatiotemporal location(s). 1 See in particular [Radder, 1986; 1988/1984; 1996]. This section draws on, but includes some slight revisions of, the presentation in Radder [2008]. 2 Such strong notions often imply the idea of goal orientation (see, e.g., [Feenberg, 1999, pp. 117-119]).

Why Technologies Are Inherently Normative

889

An artifactual system is a system that is the result of human intervention in the material and social world. A minimal interpretation of the phrase ‘is the result of’ is that such artifactual systems would simply not exist without this intervention by human beings. In technological practice such interventions will usually be intentional, but this is not necessary on the minimal interpretation. An artifactual system is functional if its potential to play some role intended by one or more human agents can be realized by embedding the system in a suitable environment. Thus, the notion of functionality refers at once to the potentialities of a technological system, to the realizability of the system and the relevant parts of its environment, and to the intentions of one or more human agents (for example, policy makers, financiers, designers, manufacturers or users). Consequently, this notion does not presuppose a principled distinction between ‘proper’ functions (those intended by designers) and ‘improper’ functions (those intended by other people). The phrase ‘can be realized’ means that the technology in question can be actually realized in some region of space and time (perhaps still in the future). This requirement of the actual realizability of the function of a technology excludes realization conditions that are purely fictional or merely ‘in principle’ and it implies that ‘being functional’ is, in part, dependent on specific characteristics of the environment. Furthermore, the system should possess a certain degree of stability. That is to say, it should be able to perform its function across a variety of situations and during a substantial period of time. Moreover, different systems of the same type should be able to exhibit the same function, that is, the system should be reproducible. Thus, the notion of reproducibility also shows the significance of distinguishing between technological systems as tokens and technological systems of the same type or of different types. As we will see, the notions of stability and reproducibility are crucial to an appropriate understanding of technologies. Yet, in philosophical theories of technology, they are often taken for granted or not taken into account at all. By way of an example, consider a token of a washing machine. This particular device constitutes a system in the sense of an aggregate of mutually interacting parts. Thus, its sides contain the water while some of the heat of this water will warm up these sides. Given the minimal interpretation of the notion of system, where to draw the boundary between system and environment is a matter of pragmatic focus. That is to say, depending on our purposes we may define the technological system as more or less inclusive. Thus, apart from the washing machine, this system might include the plug socket or even the electric power station, the washing powder factory, and so on. Of course, these different definitions of the system will go together with different delineations of the corresponding environment.3 The washing machine is obviously artifactual, since this machine would not exist without human intervention. Its most common function is to clean the 3 Thus, the notions of a system and (the relevant part of) its environment include the idea of composition but point beyond the usual sense of that idea toward a more comprehensive kind of ‘composition’ that is necessary for the stable and reproducible functioning of a technology.

890

Hans Radder

laundry of the people who own or use it, but it might also be primarily used as a table. The designers, manufacturers, owners and users expect that the device will not break down permanently after its first run, but that it will work in a stable manner for some period of time. How long exactly cannot be fixed in advance, because it depends on the quality of the build of the machine and on the specific environment in which it is supposed to function. Thus far, we have considered a particular washing machine, a token of this technology. Usually, many similar machines of the same type are being produced, and both producers and users expect that each of these washing machines of the same type works in the same, stable way. That is to say, they assume the reproducibility of the tokens of this technology. How general is this account of technology? Does it cover each and every technology that we may think of? For instance, is every technology a system in the sense of an aggregate of mutually interacting parts? In principle, we can imagine limit cases of ‘technologies’ that are of one piece, that is, without parts. We should also realize, though, that it is often the case that technologies that look homogeneous at first sight are in fact the result of a process of artifactual composition. Thus, the coffee mug and spoon on my desk may look at first to be homogeneous, but the mug has been constructed out of a number of different synthetic materials and the spoon out of a number of different metals. Keeping the mug as a stable configuration requires a continuing interaction between its components, and the same applies to the spoon. Furthermore, is a technology always artifactual in the sense of being artfully made by some human being or beings? Again, there might be rare examples of ‘technologies’ which are not artifactual. Think of a stone that is picked up to be used as a weapon. Finally, from a purely logical point of view, one might question the necessity of including the notion of reproducibility in our characterization of technologies. Yet, although it is not logically necessary, in actual practice it is generally seen as desirable to have at least the possibility of producing more than one token of the same type of technology. To sum up, while functionality and stability are necessary features of any technology,4 the features of system, artifact and reproducibility apply ‘almost universally’, and they certainly cover the interesting and important cases. Hence, the notions of system, artifact, functionality, stability and reproducibility are crucial to an appropriate understanding and assessment of real technologies.

2.2 How to realize and maintain technologies So far, this characterization of technologies is probably quite plausible, but it is also quite limited. Hence, it should be significantly expanded by posing and answering the questions of how, where and when we may realize and maintain such (types of) artifactual, functional systems with a certain degree of stability and reproducibility. There are two basic problems that must be solved. First, as the 4 Note that even throw-away devices are supposed to function stably during a brief period of time.

Why Technologies Are Inherently Normative

891

word ‘artifactual’ implies, well-functioning technologies need to be artfully made. Having them available in the first place requires active human intervention. Second, since the world may, and often does, change in substantial ways in the course of time, keeping a technological system stable and reproducible also requires active human intervention. Sometimes this intervention is small-scale and routine, more often it is elaborate and demands continuing attention. That is why technologies, if they are expected to keep functioning, cannot be left to themselves. How to tackle these two basic problems? Firstly, we need the capability to, literally, put together a technological system that has the potential of performing the required function. This means that we need to have at our disposal the materials, resources, skills and knowledge that are required for designing, constructing and using the technological system in the first place. Secondly, since we want the technology to be stable and reproducible, we need to exercise such control so as to keep a once functioning system steadily functioning inside its intended spatial domain and during the required period of operation. A crucial feature of technologies is that this control should not merely be applied to the system in question but should extend to the relevant aspects of its environment. As we will see in more detail in Section 4, in this context ‘to control’ primarily means ‘to be sure of and, if needed, to make sure’ rather than ‘to check explicitly’. indent Let us go back to the washing machine to illustrate this sketch of how we may realize and maintain a certain (type of) technology. Clearly, putting together a washing machine requires a lot of knowledge about a variety of processes needed to realize the function. How to make the device waterproof and strong enough, how to heat the water, how to construct the spin-drier, how to program the entire washing cycle, and so on? In technological practice, sometimes we merely have know-how about realizing these processes, sometimes we possess more detailed, explicit knowledge of such processes. Of course, knowledge as such is not enough. We also need to have available to us the materials for constructing a functioning washing machine. This includes not just the parts of the device itself but also the water, the electric power and the washing powder. Furthermore, we need the resources (such as tools) and the skills to put the device together. Finally, in order to have a stable and reproducible washing machine, a number of further conditions should be met. First, the supply of clean water, safe electric power, effective washing powder, and the like, as well as the draining of dirty water needs to be guaranteed. Second, the users of the machine should have the required skills to operate it. This may merely require the capability to follow the instructions from the user manual. However, it may also include adjusting these instructions to local circumstances (for instance, harder or softer water areas) or even inventing new ‘instructions’ in order to cope with novel local problems. Third, a certain measure of stability and reproducibility can only be guaranteed if enough skilled maintenance and repair facilities are available. We see that it is the requirement of stability and reproducibility in particular that demands a connection between our technological system and its wider environment. Because all kinds of psychological, social and cultural factors may

892

Hans Radder

influence the stability and reproducibility of a (type of) system, this environment is not just a physical environment. If some users become unemployed, cannot pay their electricity bill any more and are cut off by the utility company, their (token) washing machine will not function any more. Or, if a major and lasting stagnation of energy supplies arises because the resources have been exhausted and hence entire power stations have to shut down, most or all current types of washing machines will be rendered useless in terms of their cleaning function. A further important connection between a technological system and its wider environment derives from the notion of function. In actual practice, ‘to function’ is not an all or nothing matter, but a matter of degree. A technology may function perfectly, well enough, just OK, badly and so on. But whether or not a technology functions perfectly, well enough, just OK or badly is also dependent on the characteristics and requirements of the user (and hence on the environment of the technology). A simple portable radio may function well when it provides background music for a not especially musical person, but the same artifact functions badly for a connoisseur who really intends to hear the details of a complex symphony. In addition, this example shows that how well a technology functions also depends upon the availability of alternatives, that is, better or worse existing options to realize the same function. In a situation in which a hi-fi stereo set is also available, the functioning of a portable radio will be assessed differently from a situation in which this radio is the best example of stated function there is.

2.3 Some philosophical implications The theoretical and practical implications of this theory regarding the issue of why technologies are inherently normative will be discussed in Section 4. Here I will briefly mention some other philosophical implications. First, technological systems are, by definition, material. This excludes systems of nonmaterial entities (such as institutions), which are sometimes called ‘social technologies’. More generally, I think that we should resist the tendency to call any goal-oriented procedure (such as, for instance, ‘giving a command’ or even ‘interpreting a poem’) a ‘technology’, for the simple reason that this would make almost anything a technology and thus it would undermine the usefulness of the very notion of technology.5 Next, the focus on artifactual systems also differentiates the theory from accounts of technology in terms of knowledge, that is, technology as the logos of techne (see [Layton, 1974; Mitcham, 1994, Chap. 8]). The theory of technology presented here explicitly acknowledges the significance of technological knowledge (both know-how and know-that), yet its primary focus is on the realization of artifactual, functional systems with a certain degree of stability and reproducibility. Furthermore, the structural connections between system and environment, including the social environment, imply that there is no fundamental contrast between so-called ‘stand-alone technologies’ (say, a bicycle) and ‘large-scale infras5 In the same spirit, Koningsveld [2006, pp. 216-227] distinguishes nonmaterial from material practices and administrative or management sciences from technological sciences.

Why Technologies Are Inherently Normative

893

tructural systems’ (say, the car transport system), in the sense that only the latter would be a ‘socio-technical’ system, while the former would be a ‘nonsocial’ artifact. Yet, the emphasis on the significance of the social conditions for the successful functioning of technologies does not mean that ‘technology’ and ‘society’ are inextricably interwoven in a ‘seamless web’. In this respect, I follow the forceful critique of the alleged inextricability of this web in [Gingras, 1995]. The point of the proposed theory of technology is, after all, to make useful distinctions between particular material systems and the relevant parts of their material environment on the one hand, and their specific social conditions and effects on the other. Finally, the focus on realization and its social conditions does not imply an anti-realist interpretation of the working of technologies, which is the usual view in social constructivist approaches to technology (e.g., [Bijker, 1995]). As in the case of experimental science, the potentialities that enable the stable and reproducible realization of a technology can be explained in a realist sense as being humanindependent, that is, as not being constructed by human intervention and social practices (see [Radder, 1996, Chap. 4]). As will become clearer in the course of the chapter, the proposed theory of technology aims to integrate, in Carl Mitcham’s terms, elements from both the engineering and the humanities tradition in the philosophy of technology (see [Mitcham, 1994]). On the one hand, it explains what it means to design and make working technologies. On the other, it reflects on the kind of world that is implied by, and required for, having these technologies work.

3

NORMATIVITY

The second key notion that has to be explained, given the subject of this chapter, is the notion of normativity. Although the term normativity is often used in scholarly literature, its meaning is not very clear. As a kind of jargon term, it roughly denotes something like ‘the subject of norms’ or ‘the role of norms’ or, even more vaguely, ‘issues having to do with norms’. For reasons of convenience, I will also use the term normativity in this, admittedly vague, sense. In addition, however, we need a more precise characterization of the more basic notion of ‘norm’ and the corresponding adjective ‘normative’. In this section, I will first discuss, in a general way, the question of the nature and role of norms. The next section, then, will focus on technological normativity and address our main question of why technologies are inherently normative. At the most basic level, a norm is a socially embedded directive concerning what people should (or should not) say or do. A somewhat weaker phrasing of the same idea is that norms pertain to those actions and assertions which are considered desirable (or undesirable). The actions and assertions to which norms may apply are taken to be publicly accessible. The social embedding entails that

894

Hans Radder

not following accepted norms may lead to some form of sanction(s).6 The adjective ‘normative’ may then be defined as ‘implying one or more norms’.7 By definition, accepted norms presuppose (positive) value judgments of the (goals of the) actions we should perform and of the (meaning of the) assertions we should make. Thus, the norm that we should not tell lies clearly presupposes that ‘speaking the truth’ is considered to be a positive value. As to the reverse relationship, it is important to realize that in practice the same value may lead to rather different normative recommendations. Most people will endorse the value of scientific integrity but, without further specification, this does not tell us which norms this implies for the direction of scientific practice (for more about values, see Van de Poel’s chapter in this volume, Part V).

3.1 Normativity in practice In addition to these basic definitions, we need some further explanation of the nature of norms and the role they play in actual practice. First, from a theoretical perspective we may distinguish between different kinds of norms: epistemic norms (‘good knowledge should be explanatory’); methodological norms (‘medical research should use the double-blind approach’); technological norms (‘a good device should be efficient’); social norms (‘when introducing yourself, you should shake hands’); political norms (‘all governments should be democratically controlled’) and moral norms (‘nobody should kill animals for no particular reason’). In practice, however, classifying norms into kinds may be less straightforward. Why is the requirement of double-blind research methodological and not epistemic? Is the norm of democracy purely political, or can it just as well be classified as social or moral? Moreover, in practice, different kinds of norm will often be intertwined. Medical scientists who follow the methodology of double-blind research at the same time act according to the social norms of their profession. Thus, distinguishing 6 Although the theory of technology presented in this chapter differs from J¨ urgen Habermas’ views in important respects, the primacy of publicly accessible actions and assertions and the inclusion of the notion of social embedding are in line with his approach to the issue of normativity. For instance, Habermas [1971, p. 92] states that norms ‘define reciprocal expectations about behavior and . . . must be understood and recognized by at least two acting subjects. Social norms are enforced through sanctions. Their meaning is objectified in ordinary language communication.’ Some philosophers apply norms to private beliefs and desires, in addition to publicly expressed beliefs and desires. Because of its rationalistic or moralistic presuppositions, this extension of the normative from the public domain to the private realm is bound to be contestable. 7 Thus, normative statements are statements about what one should (not) do or say. Another option is to call normative statements in the above sense ‘prescriptive’ and then include both prescriptive and evaluative statements in the class of normative statements. This is the approach chosen by Maarten Franssen in his contribution to this Handbook [Franssen, this volume, Part V]. Although I prefer to keep both a conceptual and an empirical distinction between norms and evaluations, terminologically both options are in principle possible. Since Franssen’s approach assigns primacy to evaluative statements about artifacts (the prescriptive statements, it is claimed, follow on and follow from evaluative statements), the two chapters are in this respect complementary.

Why Technologies Are Inherently Normative

895

different kinds of norms may be useful from a theoretical perspective, but it will not always reflect the ways in which norms are being employed in practice. A second qualification pertains to the scope of norms. Suppose we have the norm that ‘people ought to do or say x’. Then the scope of this norm may vary for two reasons. The number of people that are, or can be expected to be, confronted with x may vary, or the number of times x occurs, or can be expected to occur, may vary.8 Thus, a norm may have a large, moderate or small scope. ‘One should respect the dignity of other people’ is meant to apply to all human beings in all situations where they meet their fellow humans. In contrast, the scope of ‘in situation s a social scientist should use statistical method m’ may be quite limited, in particular when situation s does not arise very frequently. Corresponding to the variety of scope, the significance ascribed to a particular norm and the sanctions that may be applied to people who do not live up to it, may also differ substantially. Third, norms may be explicit or implicit. In the above examples all norms were formulated explicitly. Often, however, we phrase things in a more implicit way. Thus, in certain contexts one might say: ‘in The Netherlands we do shake hands when introducing ourselves’ or ‘your design is not very efficient’. Usually, in such contexts it will be clear enough that such phrases are meant to be normative. However, norms may also be implicit in a deeper sense. Important illustrations are the norms that are implicit in a prevailing vocabulary or in a standard practice. Thus, political debates may be framed in such a way that ‘representative democracy’ is the only legitimate notion of democracy. Or medical scientists may implicitly endorse the double-blind method in their practice as the one and only correct method for testing hypotheses. In both cases, a normative claim is operative, which also has the effect of excluding alternative claims or actions. An important critical task of philosophy has always been to unearth such deeper, implicit norms which are hidden in dominant vocabularies and routine practices.9 Fourth, it is obvious that norms are not always obeyed. This also happens in the case of norms that are widely accepted within a certain practice or culture. However, although norms may be contested, changed or abolished, they also possess a regulative character and are not immediately rejected in the face of norm-breaking behavior. What is more, the possibility (and probably even the actuality) of going against a norm seems to be one of its essential features. Furthermore, norms that are not followed do have empirical consequences. After all, the burglar prefers to work at night and in any case tries to hide or avoid all traces that might lead to his or her exposure. Hence, it would be wrong to conclude that norms that are not being followed do not exist or are inconsequential in practice, because they are not being followed. Finally, even if norms are generally agreed upon, they do not determine what real people will say or do in real situations. This is true for (at least) two reasons. 8 Described in this (empirical) way, the scope of a norm should not be confused with its validity. 9 In this respect the term ‘prescription’ may be less appropriate, since it seems to be biased toward explicit normative practices.

896

Hans Radder

The first is that a norm itself does not tell us if and when it applies. Consider the norm that when introducing oneself, one should shake hands. Whether or not this norm applies depends strongly on the context: on the size of the group, on the nature of the social interaction, on the country one happens to live in, and so on. A similar story holds for ‘thou shalt not kill’. In cases of war and selfdefense, can we speak of killing in the sense of the norm? The general argument is that establishing whether or not a norm applies requires a contextual judgment that takes into account all relevant aspects of the situation (see also Pritchard’s chapter in this volume, Part V). A further reason why norms do not determine behavior is this: in practice, it is often the case that different norms are operative, which recommend actions or assertions that cannot be simultaneously done or made. Not lying would require always speaking the truth, but not offending people unnecessarily implies that speaking the truth is not always appropriate. Again, which norm will be followed in such cases is highly context dependent. A general conclusion then, in particular from the last two points, is that norms do not determine each and every detail of what people in fact say and do. Nevertheless, accepted norms are clearly operative in the sense of guiding a practice or culture in a specific way, to a smaller or greater extent. Without these norms, the practice or culture would be different from what it in fact is.

4 WHY TECHNOLOGIES ARE INHERENTLY NORMATIVE Having explained the basic features of technology and normativity I will now address their relationship. In view of their broad range and huge impact, technologies can be expected to be related to normative issues in myriad ways.10 It is not the aim of this chapter to cover all these different normative issues. Instead, the focus is on the inherent normativity of technologies and the purpose of this section is to explain why all technologies are inherently normative. Given the account of technology, norms and normative claims in the previous sections, a technology is inherently normative if its realization implies one or more norms or normative claims about what to say or do. Technologies that are normative, but not inherently so, will be called contingently normative. Although I will occasionally point to the normativity of assertions, in what follows the focus will be on normative claims about what to do. In abstract terms, the argument for the claim that technologies are inherently normative is relatively straightforward. In Section 2 we have seen that technologies can be characterized as artifactual, functional systems with a certain degree of stability and reproducibility, and we have discussed the issues at stake in the realization of such technologies. In addition, we have established the crucial role of the (material, psychological, social or cultural) environment in realizing technologies in a stable and reproducible manner. An important conclusion of this theoretical 10 See, for example, the chapters by Franssen, Van de Poel, Mitcham and Briggle, all in this Volume, Part V.

Why Technologies Are Inherently Normative

897

analysis is that the successful design, manufacture, use and maintenance of a technology requires a specific intervention in and control of the material, psychological, social or cultural environment in which the technology is supposed to function. More particularly, it follows that technologies are inherently normative, because their stable and reproducible realization in some region of space and time requires that the people in that region should behave in such a way as to enable, and not disturb, the intended functioning of the technology. As we have seen in Section 2, the relevant part of the environment comprises the conditions for the successful functioning of a technology. This implies that certain behaviors by all the people who are, or might be, present in that part of the environment are required, while other behaviors are forbidden. The point is that we cannot simply assume that those behaviors will be, and will remain to be, displayed. We have to reckon with the fact that the relevant part of the world, including the behaviors of human beings, is, or may be, changing at any moment. For this reason, a successful functioning of a technology requires a control of those behaviors. The claim that technologies are inherently normative is in part theoretical, that is, derived from the theoretical characterization of technologies as artifactual, functional systems with a certain degree of stability and reproducibility. In part, the claim is empirical, because the inherent normativity of technologies is necessitated by the actual changeability of the behavior of human beings in the relevant part of the environment of the technological system. Thus, the conditions for the successful realization of a technology ought to be satisfied. It is important to see that this ‘ought’ is really a normative ought, for two different reasons. First, as we have seen, norms presuppose values and the aim of following norms is to contribute to the achievement of these values. Hence realizing the conditions for technological success is normative because it contributes to the value of having a well-functioning technology available. Second, it may also be the case that realizing these conditions is seen as normatively undesirable, because it clashes with other, more weighty, social or moral values. I will come back to these issues in Section 4.2. In addition to this explanation of the argument for the inherent normativity of technologies, it is crucial to see what is not implied in this argument. As can be expected in the case of an argument for the inherent normativity of technologies, the argument is of a general nature: it exploits a general characterization of what it means to have a functioning technology, it is based on a general empirical understanding of the complexity and variability of our material and social world, and its conclusion is a general normative requirement on the behavior of groups of people. However, whether or not a particular technology can, or will, be (wholly or partly) successfully realized is a contingent matter. The same applies to the issue of which specific behavior is judged to be normatively required for this technology. A similar point pertains to the question of who exercises the required control: it may be explicitly incorporated in the overall realization of the technology by the designers, or it may be a matter of self-control by its users, or it may exploit the control that has already been realized for some other reason, and so on. To sum up,

898

Hans Radder

in any particular case it is a contingent matter whether or not the people involved will follow the required norms, what the specific contents of these norms will be, and how the required behavior will be realized in practice. These practical issues are not fixed by the general argument and they have to be studied empirically by focusing on the relevant technological practices. In the following parts of this section, I will focus the discussion on four more specific views that are directly relevant to the issue of the inherent normativity of technologies. This focus serves several purposes. First, it fleshes out and develops the rather abstract argumentation presented thus far: what does it mean, in concrete terms, to say that ‘the people in the relevant part of the environment of a technological system should behave in such a way as to enable, and not disturb, the intended functioning of the technology’ ? In addition, it illustrates the different aspects of the role of norms, discussed in Section 3, for the case of technological practice. At the same time, this focus fits the idea of the Handbook by providing an exposition and discussion of four different approaches to the philosophy of technology and their relevance to the issues of technology and normativity. The four approaches, which have mostly been developed independently of each other, are arranged in chronological order. In each case, I will first introduce these approaches in their own terms. Next, I’ll briefly discuss some of their merits and problems. And finally I will investigate what they imply for the question of the inherent normativity of technology.

4.1 Contingently and inherently political artifacts The first view is Langdon Winner’s. In a well-known paper, he sets out and defends the claim that technological artifacts have politics.11 Later on in this section, I will explain the relationship between politics and normativity. But first I should introduce and discuss Winner’s claim in detail. At the start of his paper, Winner explains his basic point as follows. At issue is the claim that the machines, structures and systems of modern material culture can be accurately judged not only for their contributions to efficiency and productivity and their positive and negative environmental side effects, but also for the ways in which they can embody specific forms of power and authority. [Winner, 1986, p. 19] He emphasizes that it is technological artifacts in themselves, and not merely their social or economic environments, that can have political properties [Winner, 1986, p. 20]. While everybody could agree on the latter, it is the former which is the critical point. Winner distinguishes two ways in which technological artifacts can be political. Technologies of the first type may be called contingently political, in contrast to the inherently political technologies. Because the phrasing of the 11 See [Winner, 1986, Chap. 2]. This chapter, entitled ‘Do Artifacts Have Politics?’, was originally published in 1980.

Why Technologies Are Inherently Normative

899

claims and the distinction between the two types of political technologies is quite important, I quote the relevant passage in full. First are instances in which the invention, design, or arrangement of a specific technical device or system becomes a way of settling an issue in the affairs of a particular community. Seen in the proper light, examples of this kind are fairly straightforward and easily understood. Second are cases of what can be called ‘inherently political technologies’, man-made systems that appear to require or to be strongly compatible with particular kinds of political relationships. Arguments about cases of this kind are much more troublesome and closer to the heart of the matter. [Winner, 1986, p. 22] The political nature of technologies of the first type is due to their consequences. Once realized, these technologies may have specific and enduring implications for the ways in which different groups of people live their lives. Here Winner introduces a further distinction. In some cases, the technologies may be intentionally designed with the political effects in mind. Winner emphasizes, however, that many of the most important cases of political technologies are rooted in an ongoing social process that is not the result of a conscious policy of individual people. The most cited case is Winner’s example of intentional political implications. He observes that many of the bridges over the New York parkways on Long Island are unusually low. These bridges were designed in the 1930s by Robert Moses, one of New York’s most influential town planners and builders from the 1920s to the 1970s. Winner’s explanation is that the bridges were intentionally built that low by Moses in order to exclude buses, which are too high to fit underneath the bridges, from these Long Island parkways. The political point is that these low bridges limited access by poor and black people—who did not own cars and usually traveled by public bus transport—to Jones beach, a favorite public park ‘meant for’ the white, upper and upper-middle classes. Thus, the ‘innocent’ concrete overpasses embodied a ‘social class bias and racial prejudice in physical form’ [Winner, 1986, pp. 22-23]. The mechanical tomato harvester exemplifies a case of a technology in which the political effects were not consciously planned by its designers. Nevertheless, this device — or better, an increasingly improved series of versions of it — entailed significant political consequences that seamlessly fit in a larger social pattern. Thus, in rural California, the large-scale introduction of mechanical tomato harvesters led to an increase in production and profits for a strongly decreased number of growers. At the same time, it brought about the bankruptcy of a large number of smaller growers and the loss of many jobs in the tomato industry. Winner emphasizes that, in the cases discussed so far, the political consequences were contingent, that is to say, they were not dictated by the technological requirements of these (kinds of) artifacts as such. A different designer could have built a higher kind of bridge (a contingency in design); and in a situation without small growers and without a substantial tomato industry (a contingency in pro-

900

Hans Radder

duction conditions), the political effects of the introduction of mechanical tomato harvesters would have been different. In the case of the second type of political technologies the situation is different. Here, the technological artifacts are claimed to be inherently political. Again, Winner distinguishes two versions of this type. One version claims that the adoption of a given technical system actually requires the creation and maintenance of a particular set of social conditions as the operating environment of the system. [. . . ] A second, somewhat weaker, version of the argument holds that a given kind of technology is strongly compatible with, but does not strictly require, social and political relationships of a particular stripe. [Winner, 1986, p. 32] The “requires” of the first version is a matter of judgments of practical necessity. The effective, efficient and safe operating of large sailing vessels, railways, nuclear power stations and the like, is generally judged to be dependent on a concomitant centralized and authoritarian social organization. Similarly, the ‘strong compatibility’ between a particular technology and a specific political environment is again a matter of practical judgment in a given social context. Thus, many people see solar energy as being much more easily compatible with democratic and local forms of social organization than other forms of energy production. 4.1.1

Some STS criticisms and their rebuttal

Both in philosophy of technology and in science and technology studies (briefly, STS) Winner’s claim that artifacts have politics has often been discussed, sometimes approvingly and sometimes critically. In some cases, the link with Winner’s argumentation is rather weak. Thus, Joseph Pitt criticizes Winner’s views and claims that “tools and technical systems are inherently ideologically neutral” [2000, p. 72], but he does not address Winner’s specific arguments for the inherently political nature of technological artifacts. Jane Summerton’s article “Do Electrons Have Politics? Constructing User Identities in Swedish Electricity” [Summerton, 2004] does not discuss Winner’s account either, in spite of what is being suggested by the title of her paper. In particular, this paper leaves unclear the sense in which the socially constructed user identities (the politics) are required by the specific configurations of electrons that run through the Swedish electricity cables and electrical devices (the artifacts). Some other STS authors, however, have addressed Winner’s views more head-on, and have drawn a number of quite critical conclusions. In this subsection, I briefly describe and assess two sharp attacks on Winner’s claim that artifacts have politics: one by Bernward Joerges and a closely related one by Steve Woolgar and Geoff Cooper. Taken together, these criticisms can be summarized in four claims, some pertaining to the examples discussed by Winner and some pertaining to his general line of argument. The focus of the critical comments is on Winner’s account of Moses’ bridges.

Why Technologies Are Inherently Normative

901

The first and principal point is that the low bridges did not at all prevent black and poor people from visiting the Long Island beaches. As a matter of fact, it is claimed, these beaches were accessible both via alternative car routes and, more importantly, through alternative public transport by bus and by train. Joerges concludes that the bridge story is “counterfactual” [1999a, p. 411] and “a bit of a scam” [1999b, p. 451]; and according to Woolgar and Cooper [1999, pp. 442-443], their ‘counterevidence’ has shown that “the example is simply wrong”.12 A second point of contention concerns the intentions Moses had in building his parkways and bridges. Did he intentionally design these parkways and bridges out of social class bias and racial prejudice? Joerges [1999a, p. 418] asserts that Moses “never pursued explicitly racist schemes” and he suggests alternative interpretations for the case at hand, which he claims to be more plausible. The last two points are more general. The third point questions the notion of “the consequences of a technology”. As the title of their paper indicates, Woolgar and Cooper strongly emphasize the ambivalence of technological artifacts. In particular, they claim that technologies do not have “definite consequences’ about which a “definitive story” (such as Winner’s) could be told. The ’extitfourth and final point is a critique of Winner’s general claim that artifacts have politics. Thus, Joerges claims to have shown that “what Winner asserts about technical artifacts is doubtful for any technical artifact, not just for Moses’ low bridges” [Joerges, 1999b, p. 450]. Similarly, Woolgar and Cooper state that, depending on which story is being told, “technology ... does and does not have politics” [Woolgar and Cooper, 1999, p. 443]. Let me now assess these criticisms on the basis of the account of Winner’s views presented in the first part of this section. The core of the first point of criticism is that the low bridges did not prevent poor and black people from accessing the beaches, because of the availability of alternative ways to get there. This fact as such, however, is fully compatible with Winner’s statements, as an accurate reading of the relevant passages shows. Thus, Winner claims that the low bridges limit access to the beaches or, in general, that technologies influence the form of life of the people involved [Winner, 1986, p. 23 and p. 28]. Given these claims, a sensible (primarily empirical) debate could focus on the degree of limitation and the kind of influence that, given the specific setting, comes with the technologies under discussion. Winner’s critics, in contrast, base their attack on the misinterpretation that the bridges fully prevented access to the beaches, that one had to use the low bridges routes.13 The point is important because — as Winner’s careful phrasing shows — the world of technology is not a world of universal regularities. The second point of criticism questions Winner’s account of Moses’ intentions. 12 These authors then go on to discuss the question of how to explain what they call “the uncritical acceptance and potential endurance” of the low bridges story. Since my focus is on the substance of Winner’s claims, I will not enter into this debate. 13 See [Woolgar and Cooper, 1999, pp. 434-435] and [Joerges, 1999a, p. 417]. This kind of reasoning can be found more frequently in STS. From an all-or-nothing perspective, there is no space for nonlocal patterns, which are not universal but still pattern a variety of situations in normatively significant ways (see [Radder, 1996, Chaps. 5 and 8]).

902

Hans Radder

Of course, it is not always easy to establish someone’s intentions, certainly not in the case of historical figures. Yet, it is possible to make plausibility arguments. In this respect, my feeling is that the case needs further investigation. On the one hand, there is some evidence for Moses’ racist intentions; on the other, Joerges’ alternative interpretation of Moses as a promoter of the automobile society or even as an early ecologist [1999a, pp. 416-420] is worth exploring, although it is not (yet) plausible as it stands. But even if Joerges’ interpretation should prove to be plausible on closer inspection, it is important to see that the scope of his criticism is rather limited. First, since their interpretations do not seem to be incompatible, Joerges and Winner may both be right. Second, and more importantly, Winner repeatedly emphasizes that, in the case of contingently political artifacts, the most significant and most frequently occurring political consequences are not the result of the intentions of one or a few individual designers. The third point put forward, in particular by Woolgar and Cooper, is that technologies do not have definite consequences. In so far as this point is meant to follow Winner’s usage of ‘political consequences’, and hence applies to contingently political technologies, it is fully compatible with Winner’s view, and even stated explicitly by himself. In all the cases cited above [including the low bridges case] the technologies are relatively flexible in design and arrangement and variable in their effects. Although one can imagine a particular result produced in a particular setting, one can also easily imagine how a roughly similar device or system might have been built or situated with very much different political consequences. [Winner, 1986, p. 29] Furthermore, at no point in Winner’s text can one find even a suggestion that he sees his story as ‘definitive’. In as far as Woolgar and Cooper’s criticism is meant to apply to any technology, it leads us to the final point. This final point is a general critique of the overall claim that technological artifacts have politics. Here, however, the perplexing fact is that none of the critics has addressed, let alone refuted, Winner’s central claim. They have focused on what Winner designates as the less significant version of the less interesting type of his claims. That is to say, they have hardly mentioned the case of non-intended contingently political artifacts and they have completely ignored the strong cases of inherently political artifacts. Hence, their general criticisms are unsubstantiated. 4.1.2

The inherent normativity of political artifacts

The conclusion is that the STS criticisms of Winner’s position are superficial and, for the most part, inconsequential. Hence, it makes sense to proceed and examine the connection between Langdon Winner’s account of the political nature of artifacts and the issue of the inherent normativity of technologies. In which sense can political artifacts be said to be (inherently) normative? The answer is different for the two types of political technologies. In the case of the first version of inherently

Why Technologies Are Inherently Normative

903

political technologies, the connection to normativity is straightforward. Given the choice for a particular technology (for instance, a large-scale railway system), the required social organization as the operating environment of the system (a centralized organizational structure) should also be created and maintained. In the case of the second version, the relation is less direct. ‘Strong compatibility’ between a type of artifact and some kind of socio-political arrangement does not strictly require the realization of this arrangement. In a weaker sense, one may say that such artifacts suggest or reinforce the arrangement. The case of the contingently political artifacts is different. These artifacts do have particular political effects, but their normative consequences will depend on the specifics of the context. Hence, these technologies are clearly not inherently normative. The low New York bridges are contingently related to the norm that poor and black people should not be allowed on the beaches of Long Island. Similarly, the tomato harvester is contingently related to the norm that the economic value of increased and more efficient production should prevail over the social value of maintaining employment.

4.2

Closed systems and their normative implications

In section 2, I characterized a technology as an artifactual, functional system with a certain degree of stability and reproducibility. A working technology needs to possess some degree of stability and reproducibility. The aim of this section is to analyze this requirement in more detail and to explain its normative implications. As we have seen, keeping a technological system stable and reproducible requires control of both the system and the relevant aspects of its environment. The nature and implications of this control may be analyzed in more detail with the help of the notion of the ‘closedness’ of the technological system as an important necessary condition for achieving stability and reproducibility. Hence, the focus of this section is on the normative implications of technological closedness.14 This leads to an approach that is akin in spirit but more detailed than (and distinct in content from) Winner’s account. Given the variability of our material and social world, stability and reproducibility do not come automatically, but require us to close the technological system through controlling the relevant interactions between this system and its environment.15 In my chapter in Part I of this Handbook, I analyzed these interactions in order to illuminate the conceptual-theoretical and empirical relations between (experimental) science and technology. Here I exploit this analysis for the purpose of addressing the issue of the inherent normativity of technology. As we have seen, we may distinguish three basic types of interactions between the technological system and its environment: required, forbidden and allowed interactions. The required 14 See [Radder, 1988/1984, Chap. 3] and [Radder, 1986]; a slightly revised version of the latter article has been published in [Radder, 1996, Chap. 6]. 15 The notion of inter action allows a comprehensive analysis of both the impacts of the environment on the system and the ways in which the system influences the environment. The present chapter, however, focuses primarily on the impacts of the environment on the system.

904

Hans Radder

interactions need to be actively produced and maintained to enable the stability and reproducibility of the technology. Forbidden interactions are those interactions that would disturb the stable and reproducible working of the technology in question, and hence such interactions need to be removed or prevented from taking place. Finally, allowed interactions do, or may, occur but they do not have any (enabling or disturbing) influence on the stable and reproducible functioning of the technology. Thus, realizing and maintaining a working technology requires the power and control to bring about the required interactions and to eliminate or prevent the forbidden interactions. We may summarize this account by defining a closed technological system as a system for which the required interactions have been realized and maintained, and for which the forbidden interactions have been removed or prevented. Since closedness is necessary for the stable and reproducible functioning of a technology (in a certain region of space and time), if we would like to have a working technology, we need to be able to close the technological system in that region.16 Thus far, this is a theoretical analysis. In line with the general explanation in the introduction to this section, nothing so far is implied about the realizability of particular, closed technological systems, or about the ‘we’ who would like to see the technology realized and the ‘us’ who are required to exercise the relevant control. Moreover, what is needed to close a particular system will also depend on how we have specified the required functioning of the technology in the first place. Hence, analogous to the approach in Section 2, the theoretical analysis should be complemented by addressing the question of whether, and if so, how particular, closed technological systems may be realized in actual practice. I discuss this question with the help of two concrete examples. Consider first the example of a contact lens. Contact lens technology constitutes a system but, as I explained in Section 2, how to delineate this system is a matter of analytic focus. We may focus on relatively small systems (the lens alone, lens and eye, lens and cleansing liquids and user procedures and the like) but we may also include much larger systems (factories for producing cleansing liquids, facilities for opticians for doing check-ups and the like). Concomitant social arrangements include health care provisions, insurance systems and so on. Let us focus on the eye-contact lens system. To close this system we have to examine the interactions between this system and its environment. An important required interaction is the supply of sufficient moisture to the eye and lens. In part, this is done through our tear ducts, in part by applying artificial liquid. The 16 In

theory, one might think of the possibility of closing a system by transforming all forbidden interactions into allowed ones. In practice this is not possible, however, because of the fundamental complexity and variability of the material and social processes that take place in the environments in which a technology is supposed to function. That is to say, strictly fool-proof technologies do not exist. Of course, this does not exclude the possibility of transforming a particular technological system into a more stable one by changing some forbidden interactions into allowed ones. An interesting example is the recent research on ‘self-healing materials’. In the case of reinforced concrete, for instance, the intrusion of water may be changed from a forbidden into an allowed interaction by adding certain bacteria to the concrete structures.

Why Technologies Are Inherently Normative

905

latter requires, at the very least, production facilities for these liquids and quite a bit of social regulation around it. In some cases, these days it may even require successfully passing the special detection devices at airports; if your cleansing liquid is confiscated, after some time your lens will not function anymore so long as no new liquid is available. An obvious forbidden interaction is the intrusion of dust and dirt into the lens and eye. This may be avoided by observing proper hygienic procedures, by wearing sun glasses in hot and windy weather, and so on. It implies that for small children, Alzheimer patients and workers in unclean circumstances the technology cannot be expected to work in a stable and reproducible manner. Finally, in the case of the eye-lens system, there is a host of allowed interactions. The sound of talking people and radios playing music induce allowed interactions. People smoking and wet weather are also allowed but only within certain limits. What this example shows is that the realization and maintenance of the required interactions and the elimination or prevention of the forbidden interactions requires a substantial intervention in, and control of, the environment in which the technology is supposed to function. This intervention and control is at the same time material and social. In human affairs, the material realization of a working technology always involves a concomitant social realization of the relevant conditions needed for having the technology work in a stable and reproducible manner (see also Sørensen’s chapter in this Volume, Part I). In the case of contact lenses, users are required to exercise self-discipline by following the relevant hygienic procedures, children are warned not to touch the cleansing liquids, a certificate is awarded to those opticians who have successfully passed the training courses and governmental offices oversee the environmental aspects of the cleansing liquid factories. The typical situation is indeed one in which a variety of actors (instead of one single actor) is involved in creating and maintaining the required intervention and control of the relevant aspects of the environment of the technological system.17 As we have seen in Section 2, the functioning of a technology may be a question of degree. A particular technology may function optimally, reasonably well, just OK, poorly or not at all. Of course, in the last case no control needs to be exercised. In the other cases, the higher the standards for the functioning of the technology are set, the stronger the normative constraints must be. Thus, in the case of the contact lens, poor hygiene might still lead to just OK functioning (at least for a limited time), but following appropriate cleansing procedures is really necessary for optimal functioning. As a second illustration, consider the case of nuclear energy technology. Again, different arrangements may be chosen as our focal system. We may focus on the reactor, on the reactor plus other apparatus and operators present in a power plant, on the mining and transport of nuclear fuel, on the storage of radioactive nuclear waste, and so on. And again there are the concomitant social arrangements, such as the regulations for choosing suitable locations for the plants, the safety measures 17 This compares with Michel Foucault’s account of the decentered exercise of power (see, e.g., [Foucault 1982]).

906

Hans Radder

for workers and people living in the neighborhood of the plants, treaties against the proliferation of nuclear technology, the Greenpeace donation system for critics of nuclear energy technology. Clearly, a comprehensive account of this technology would include a very large number of material and social arrangements. By way of example, let us focus on the nuclear reactor itself as our technological system. For the purpose of closing this system, we need to control its interactions with its environment. In this case, the required interactions include securing a sufficient supply of nuclear fuel and, rather different but equally important, securing the employment of skilled, disciplined and reliable operators and technicians. As for the forbidden interactions, intrusion into the reactor buildings by terrorist saboteurs needs to be prevented. Finally, although the number of allowed interactions will be much smaller than in the case of the eye-contact lens system,18 they are not fully absent. Presumably, playing music and the presence of a bit of dust and dirt in the reactor building do no harm, while smoking might be a borderline case in view of the risk of fire. Closing the overall system of nuclear energy technology demands an effective control of the required and forbidden interactions during the entire period of operation for all the components of this technological system. Again, this control cannot be limited to the material arrangements, but it also pertains to the behavior of all the people who interact, or might interact, with the technological system. In particular in the case of the safe disposal of highly radioactive waste, this is an unresolved and, in view of the extremely long period during which effective control should be exercised, unsolvable problem. The cases of contact lens and nuclear energy technology also show quite clearly that closedness is not sufficient for the working of a technology. The effective control of the relevant interactions with the environment is not enough to get a well-functioning technological system. In addition, this system needs to possess the appropriate potentialities which enable it to play its intended role when it is embedded in the right environment. 4.2.1

Some points of debate

How plausible is this account of closedness as a necessary condition for the stability and reproducibility of technologies? Brian Wynne claims that the approach is valuable but incomplete, and he suggests adding the following three points. (a) Closure of the system as the ideal pursued by . . . technologists can never be complete, and is more problematic the more socially and physically extended is a technological system. . . . (b) As science becomes increasingly an economic resource in industrial competition, the arbitrary properties of technologies as covert and authoritarian social experiments are amplified. This is because the rush to exploit scientific knowledge as commercial technology allows less time and social access 18 See also Charles Perrow’s [1984] analysis of nuclear energy as a complex and tightly coupled technological system.

Why Technologies Are Inherently Normative

907

for wider system-problems and interferences to be anticipated. . . . (c) Technologies can therefore be regarded as the arbitrary imposition of power, and their discourses, including scientific risk analytic discourses, as the ideological easement of the social experiment of power. [Wynne, 1988, pp. 163-164] However, apart from some aspects of point (c), most of these ‘additions’ are already implied in the approach sketched thus far. Point (a) is taken into account by emphasizing the conditional nature of the theoretical analysis: if we would like to realize a technology, we should be able to close the technological systems in question. In practice, successfully closing technological systems is possible, but it is by no means guaranteed. Moreover, the larger the spatiotemporal extension of a technological system, the harder it will be to close it. Thus, whether or not closing a particular system may be expected to be feasible and desirable is a crucial empirical and normative issue, which cannot be settled on the basis of a theoretical analysis alone. In fact, making these specific distinctions and connections between theoretical, empirical and normative questions has been a characteristic and persistent feature of this approach to the philosophy of technology. Similarly, point (b) restates conclusions I drew from an analysis of the Dutch debate on the feasibility and desirability of nuclear energy during the 1980s [Radder, 1986, pp. 674-678]. Also, Wynne’s last point can be fully endorsed in as far as it refers to the use of risk analysis in closing the system of nuclear energy production. Taken as a general claim we have to be more careful, though. The basic idea of the above analysis is that technologies can do something for us (to wit, function in a stable and reproducible way), if we are prepared to pay a price through complying with the necessary material and social control and discipline. However, the question of whether or not this control and discipline constitute an ‘arbitrary imposition of power’ cannot be answered on the basis of a theoretical analysis, since this also requires sustained empirical research and normative debate. In fact, one of the points of the above analysis is to stimulate more democratic debate on the pros and cons of specific technologies. It cannot be denied that practical decisions are often strongly influenced by prevailing power relationships. Yet, cases of a free and democratic acceptance of the required control and discipline occur as well. Under the present circumstances, I am forced to accept the price that comes with the realization of nuclear energy, but I am choosing to pay the price of wearing contact lenses. On this point, there seems to be a significant contrast to Winner’s approach. Winner [1986, p. 22] defines politics as ‘the arrangements of power and authority in human associations as well as the activities that take place within those arrangements’. Both his linking of power and authority, and his examples of the exercise of power, strongly suggest that he sees politics as intrinsically repressive and coercive, especially regarding underprivileged social groups. In contrast, I prefer to define politics as (aiming at) the governance of society through shaping our material and social world in specific ways, and thus excluding the realization

908

Hans Radder

of alternative worlds [Radder, 1996, Chap. 7]. In modern societies this essentially includes the realization of specific technological systems. This governance may be performed through the authoritarian exercise of coercion and repression, but also through democratic deliberation and decision-making. What will happen in concrete circumstances is a practical, normative issue that cannot be decided by an a priori, theoretical definition of the notion of power. A further question regarding the plausibility of the above analysis pertains to the appropriateness of the notion of closedness. Because closedness does not exclude the presence of required and allowed interactions, it should not be confused with full isolation. Instead, it may be usefully compared to the closedness of a door. The door of a room should not fully isolate it. Its required interactions with the environment may include letting in the appropriate people as well as enough fresh air. Its forbidden interactions may include letting through too much noise or unwanted people like burglars. Finally, its allowed interactions may include soft noises or conduction of small amounts of heat. In spite of this explanation, the term ‘closedness’ might still be seen to be confusing, in particular because of the important role of the required interactions.19 For this reason, it may be advisable to drop this term entirely and focus on the idea of (attempts at) controlling the required and forbidden interactions rather than on the term ‘closedness’ as such. After all, it is the account of the interactions between technological system and environment that forms the core of the above analysis. 4.2.2

The inherent normativity of ‘closed’ systems

Finally, there is the question of the normativity of technologies, and especially of their being inherently normative. The analysis in this section implies that technologies are normative in the following sense: the requirement of stability and reproducibility demands that the people who are, or might be, involved in the realization of the technology in question should co-operate in realizing the required interactions and should not obstruct the prevention or elimination of the forbidden interactions. Especially in the case of comprehensive technological systems, this normativity is significant and the question of how the people involved can be made to follow the norms is of crucial importance. Moreover, if my theory of technology and the above analysis is right, technologies are inherently normative. This implication is not merely conceptual, because it is also based on the empirical fact of the variability of the material and social worlds in which the technologies are supposed to function. It is this variability that necessitates the intervention, control and discipline needed to obtain stable and reproducible technologies.

19 See, e.g., [Ducheyne, 2005, pp. 324-326], who does admit that in practice systems are only relatively closed, but still defines a closed system as ‘hermetically isolated and independent from its environment’, and accordingly sees such an absolute isolation as the ideal. Similarly, the start of the above quotation suggests that something like this is also presupposed by Wynne.

Why Technologies Are Inherently Normative

909

Thus, in an ontological sense, technologies are inherently normative. Answering the epistemological question of which specific norms should be followed in order to realize and maintain a particular technology is a matter of practical inference. Suppose that x is the result of a certain technology, which may be achieved by exercising the control c. On the basis of the premises: ‘one wants to accomplish x’ and ‘unless c is done, x will not be accomplished’, one concludes that ‘c should be done’ (see also Kroes’ chapter in this Volume, Part III]). Epistemically, the second premise is the crucial one. Depending on the degree of plausibility of this premise, the normative claim ‘that c should be done’ may be seen as really necessary, as strongly advisable, as maybe a good thing to do or as not at all obvious. Normatively, the first premise is at least as crucial. An important question is who is the ‘one’ that wants to accomplish x? Critical approaches to technology (e.g., [Feenberg, 1999]) insist that this question should be answered in a democratic manner. On the basis of the analyses in this section, we may add that the issues of both the feasibility and the desirability of realizing working technologies should play an essential role in democratic decision-making concerning the acceptability of these technologies [Radder, 2008].

4.3

The script of technological artifacts

Another approach that is directly relevant to the theme of this chapter focuses on the relation between the design and the use of technological artifacts. This is the so-called script approach, proposed by Madeleine Akrich. A script is a scenario that anticipates a specific kind of users and use of a particular technology. The claim is that designers of a particular artifact possess or develop a specific representation of the intended users and use. The script, then, is the result of ‘inscribing’ this representation in the physical design of the technology. Akrich develops this idea as follows. Designers thus define actors with specific tastes, competences, motives, aspirations, political prejudices, and the rest, and they assume that morality, technology, science, and economy will evolve in particular ways.20 However, she immediately goes on to emphasize that this definition may fail and that the assumption may prove to be wrong. What will actually happen to a technological design is essentially contingent and, instead of a successful technological object, the designers may end up producing a chimera. To be sure, it may be that no actors will come forward to play the roles envisaged by the designer. Or users may define quite different roles of their own. . . . Thus, if we are interested in technical objects and not in chimerae, we cannot be satisfied methodologically with the 20 [Akrich, 1992, p. 208]. For some other discussions of the script approach, see [Latour, 1992; Oudshoorn and Pinch, 2003; Van Oost, 2003].

910

Hans Radder

designer’s or user’s point of view alone. We have to go back and forth continually between the designer and the user, between the designer’s projected user and the real user, between the world inscribed in the object and the world described by its displacement. [Akrich, 1992, pp. 208-209] Obviously, failure is more probable when the (physical or socio-cultural) distance between designers and users is large. This is the case, for instance, with processes of technology transfer between designers in developed countries and users in developing countries. By way of example, Akrich discusses the attempted transfer of a photoelectric lighting kit from France to certain African countries. The kit consisted of a photoelectric panel to convert solar energy into electricity, a battery to store the electric energy, and a lamp to produce the light. In an attempt to make the kit ‘foolproof’, the designers had tried to exclude any interference (as a potential source of damage) by local users and local conditions. For this purpose, the design included direct current circuitry, connecting wires of fixed length, nonstandard plugs, and waterproof batteries. Hence, the script defined the users as unskilled, and prescribed them not to interfere in any way with the device. In fact, however, this script proved to be non-viable since it prevented a smooth adjustment to the local circumstances: repairs could not be done by local electricians, replacement of broken components was not possible and so on. The result was that local installation and maintenance technicians ‘were confronted with considerable difficulties’ and that users ‘lost control over the installation’.21 The script approach has been taken up in other studies of technology and its basic idea has been applied to quite a few other cases. A quick example is the road bump. Its script prescribes that the users of the road, in particular car drivers, have to slow down on pain of inflicting substantial damage to their cars or to themselves. Ellen van Oost’s study of the development of electric shavers by the Philips company since the end of the 1930s constitutes a more extended case. In the course of this period, a specific gender differentiation emerged between electric shavers for men and for women. Designers drew on stereotypical representations of the male and the female user, and inscribed these representations in the different designs of electric shavers. In doing so, they both reproduced and reinforced these gender stereotypes. Van Oost [2003, p. 206] summarizes her conclusions as follows: Masking the technology was a systematic element of the gender script of the Ladyshave. The methods used included using perfume to mask the smell of oil, linking the shaver to lipstick, transforming the shaver into a beauty set, and eliminating visible screws. The script of the Ladyshaves hides the technology for its users both in a symbolic way (by presenting itself as a beauty set) and in a physical way (by not 21 [Akrich, 1992, p. 210]. Akrich’s text leaves it unclear whether these difficulties were temporary or led to a more permanent failure of the technology.

Why Technologies Are Inherently Normative

911

having screws that would allow the device to be opened). The Ladyshave’s design trajectory was based on a representation of female users as technofobic. The design of men’s shavers, in contrast, exhibited these devices as the latest hightech products and suggested the technological competence of their users. Again, it is emphasized that the impact of these design scripts on the users should not be seen in a deterministic way, but rather as more or less compelling given the nature of the wider socio-cultural context. 4.3.1

The scope and limitations of the script approach 22

In some respects, the script approach is similar to the views discussed in Sections 4.1 and 4.2. For instance, the case of the transfer of the photoelectric lighting kit could be easily reconstructed to fit Winner’s analysis of inherently political artifacts or my analysis of the material/social conditions needed to realize stable and reproducible technologies. In some other respects, however, there are smaller and larger differences between these analyses and the script approach. First, design and use constitute only a part of the realization of a technology as an artifactual, functional system with a certain degree of stability and reproducibility. Next to design there is production, and next to use there is social regulation, to mention just two further aspects of technologies. And although the notions of users and use may be employed in a relatively broad way, as is done by Akrich in her account of the transfer of the photoelectric lighting kit, quite often they are taken in a much narrower way, as in the case of the users of electric shavers discussed by Van Oost. Furthermore, it is sometimes suggested and sometimes explicitly stated that the anticipated behavior of the users is required for the working of the artifacts (see, e.g., [Van Oost, 2003, p. 195]). However, it is not always clear what this ‘working’ involves. The case of the photoelectric lighting kit is relatively straightforward, in that this technology does not work if the lamp cannot be made to shine because of deficiencies in its context of use. In the case of the Ladyshave, however, the relation between the inscribed representation and the working of the shaver is much looser. Presumably, in this case, working primarily amounts to being economically successful in terms of sales.23 Finally, in line with the descriptivism of the STS approach, script analyses are limited to descriptions of actual technological practices. There is no (relatively independent) theoretical level at which it is possible to distance oneself from the actual course of technological developments. For this reason, there is no normative level that would enable a normative assessment of technologies. Normative assessments — for instance of the stereotyping of women as technofobic — are lacking 22 For

a more detailed assessment of this approach, see [Oudshoorn and Pinch, 2003, pp. 7-16]. same ambiguity regarding the notions of ‘success’ and ‘working’ can be found in social constructivist claims that the successful working of a technological artifact can be fully explained in social terms. 23 The

912

Hans Radder

from the script approach. Furthermore, Akrich [1992, p. 222, my emphasis] holds that ‘we are able to say that technical objects changed, stabilized, naturalized, or depoliticized social relations only with the benefit of hindsight’. Apparently, the world is assumed to be so radically contingent that no plausible claims at all can be made about the future. The implication is that any deliberate, future-oriented technology policy is pointless. Within the script approach, the forward-looking problem of how to shape future technologies cannot be sensibly posed.24 In this respect, this approach strongly contrasts with the other accounts discussed in this section. Yes, we have no certainty about what will happen in the future. What this implies, however, is that deliberation concerning more desirable or less desirable futures is essentially fallible (but not pointless). 4.3.2

Scripts and (inherent) normativity

What about the inherent normativity of the inscribed representations of use and users? Do scripts of technological artifacts imply norms? In answering these questions we can make use of the preceding points of comment. As we have seen, some of the examples of the script approach, such as the photoelectric lighting kit, can be shown to exhibit an inherent connection between the particular technologies and specific norms that should be followed to have the technology work in a stable and reproducible way. Similarly, in the case of the technological system consisting of a road, a road bump, a car and a driver, stable and reproducible working requires that the driver respects the norm of driving slowly when approaching the bump. The script approach, however, does not include a general characterization of technologies. As is frequently the case in empirical approaches, it is assumed either that we already know what a technology is, or that a technology is what the actors that are being studied call a technology. For this reason, the general question of whether or not technologies are inherently normative cannot be answered on the basis of the script approach. Moreover, in quite a few cases ‘working’ is also defined in terms of economic success. In these cases, such as the design and use of the Ladyshave, the possible links between the functioning and the normativity of the artifact are clearly contingent.

4.4 Artifacts, use plans and functions A second approach that focuses on the relationship between design and use is the use-plan approach advocated by Wybo Houkes and Pieter Vermaas. It is comparable to the script approach in its subject matter, but quite different in its philosophical method and style. The basic idea is that the design of a technological artifact essentially includes, or should include, a use plan. Design processes are seen as involving not only the description of a type of artifact, but also, or even primarily, the construction of an appropriate use plan. 24 For

a more detailed exposition of this argument, see [Radder, 1998a; 1998b].

Why Technologies Are Inherently Normative

913

Central to our action-theoretical analysis is the notion of a use plan. Defining a plan somewhat loosely as a goal-directed series of considered actions, a use plan of object x is a series of such actions in which manipulations of x are included as contributions to realizing the given goal. [Houkes and Vermaas, 2004, p. 57] In addition, use plans need to be communicated to prospective users. For this purpose, a variety of means are available: user manuals, explicit instructions or demonstrations, television ads, use cues added to the artifact, and so on. In constructing novel use plans, designers may draw on their own creativity, on the resources of socio-cultural traditions or on available scientific and technological knowledge. In developing this account, two further distinctions are introduced: between standard and nonstandard use, and between rational and irrational use. Use plans based on socio-cultural traditions and scientific or engineering design practices are claimed to give rise to standard use. Nonstandard use may arise from (innovative or idiosyncratic) individual users. For example, the standard use of a screw driver employs it to fasten screws; using it to open a paint-can may be called innovative nonstandard use, while its employment to kill mosquitoes would be seen as idiosyncratic nonstandard use. Furthermore, rational, or proper, use is distinguished from irrational, or improper, use. Uses that are appropriate, effective and efficient are called ‘rational’. Of the mentioned uses of a screw driver, the first two exemplify rational use (which implies that rational and standard use do not coincide), while the third is presumably a case of irrational use. In actual practice, the distinction between standard and nonstandard use is said to be a matter of degree, and the same seems to apply to the contrast between rational and irrational use. Moreover, in the course of time particular artifact uses may be classified differently. Standard or rational uses may become nonstandard or irrational and vice versa. Finally, Houkes and Vermaas emphasize the contrast of their action-theoretic philosophy of artifacts with artifact theories that focus on functions. In their view, it is not the function of an artifact which is the primary notion but its use plan and use. In short, our theory of function features use plans, justification, and communication as central concepts. Roughly speaking, an artifact function is any role played by an artifact in a use plan that is justified and communicated to prospective users. Hence, functions are plan-relative: it makes no sense to ascribe functions to an object that is not, metaphorically speaking, embedded in a use plan. [Houkes and Vermaas, 2004, p. 66] The mentioned justification involves giving good reasons for the belief that the artifact is able to play the relevant role and for the belief that execution of the use plan contributes to realizing its goal. In the case of a professional designer, the justification of the use plan will mostly be based on scientific or technological

914

Hans Radder

knowledge; in the case of an innovative user, it will usually be based on his or her experiential knowledge. 4.4.1

Merits and problems of the use-plan approach 25

The use-plan approach, as briefly summarized above, possesses several merits. The relational conception of technological functions is vastly superior to objectivistic accounts that try to explain functions exclusively on the basis of the features of the technological objects. This conception is compatible with the definition of functionality presented in Section 2. Furthermore, since the notion of a use plan is akin to the idea of a script, it shares the valuable features of the script approach. As we have seen, a script is a specific representation of prospective users that is inscribed in the technical content of an artifact. If we interpret the notion of ‘technical content’ in a broad way, a script may include not just the material use cues of the artifact itself but also user manuals and explicit instructions about, or demonstrations of, the use of the artifact (compare with [Houkes and Vermaas, 2003, p. 40]). However, the similarity between the script and the use-plan approach also implies that the latter shares some of the limitations and problems of the former. For a start, the notions of use and users in the use-plan approach seem to be even narrower than those in the script approach. Both the examples and the general explanation strongly suggest that users are identified with individual consumers or end-users of simple and familiar artifacts. From this perspective, a factory worker who uses an artifact made in another factory as part of the production process for another artifact is excluded from consideration. Furthermore, the approach seems to be restricted to isolated artifacts and their immediate users. This point is illustrated by comparing the limited account of a use plan for a washing machine proposed by Houkes and Vermaas [2003, pp. 36-37] with the comprehensive analysis of this technology provided in Section 2 of this chapter. The general point is that the ‘relevant aspects of the environment of a technology’ include much more than just its immediate users. Furthermore, the distinction between proper or rational use and improper or irrational use entails a certain a priori bias, since the requirement that a philosophy of technology should ‘underwrite’ the proper-improper distinction [Houkes and Vermaas, 2004, p. 53] tends to privilege the designers’ or engineers’ viewpoints. The problem is that terms like ‘proper’ or ‘rational’ and ‘improper’ or ‘irrational’ possess considerable pejorative connotations. But what is the point of denouncing somebody who kills mosquitoes with a screw driver as behaving improperly or irrationally? Or, to offer a real-life example, why should philosophers a priori condemn as improper or irrational any attempt at ‘democratic rationalization’,

25 For

further discussion of this approach, see Preston’s chapter in this Volume, Part II.

Why Technologies Are Inherently Normative

915

such as the ‘hacking’ of the French Minitel system in the early 1980s?26 In this respect, the script approach is more open-minded. Related to this is the fact that the advocated notion of the rationality of a use plan is limited to its utility for an individual user. The implication is that a plan for designing and using gas chambers for mass killings is fully rational. Houkes and Vermaas [2004, p. 70, note 16] acknowledge such problems, but think they can be solved through supplementing their basic, designer-centered assumptions. It seems more plausible, however, that this ‘supplement’ will have to subvert their distinction between rational and irrational uses, because of the possibility that a use plan that is rational (respectively, irrational) according to the definitions of Houkes and Vermaas may be irrational (respectively, rational) according to a broader notion of appropriate use.27 4.4.2

Use plans and inherent normativity

The final question is again: are technologies, according to this use-plan approach, inherently normative and, if so, in what sense? Although the authors of this approach do not, as far as I know, provide an explicit definition, we may infer that ‘a technology’ is understood to be ‘an artifact with a justified function and use plan’. Significantly, the justification pertains to the beliefs that the artifact can function in the relevant way, and that the envisaged use plan will facilitate the realization of this function. If this reading is correct, technologies are inherently normative because they imply that there should be good reasons for the beliefs in question. This is clearly a type of epistemic normativity. Indeed, as the authors emphasize [Houkes and Vermaas, 2004, pp. 66-67], their theory is primarily an (action-theoretic) theory of function ascriptions. It does not tell us what it means for an artifact to have a function. For this reason, the theory does not answer the question of whether technologies are inherently normative in an ontological sense, or so it would seem. However, by relating use plans to the account of technology presented in Section 2, we may be able to say something more. After all, in order to have an actually functioning artifact, it is not enough to envisage some arbitrary use plan that will, in principle, facilitate the realization of this function. What is also crucial is that this use plan is feasible, that it can be actually realized in the relevant region of space and time in a stable and reproducible manner. Similarly to the script approach, in the use-plan approach technologies can be said to be inherently normative because they imply the norm that the (anticipated or actual) users should follow the use plans in the prescribed way in order to have the artifacts perform their intended function. But again, given the above-mentioned limitations to use and users, a variety of other inherent norms remain outside the scope of the use-plan approach. 26 For this example, see [Feenberg, 1995, Chap. 7]; for the general notion of democratic rationalization, see [Feenberg, 1999, Chaps. 4–6]. 27 On these points, see the more inclusive discussion of use plans and responsibility in [Brumsen, 2003]. For a broader account of appropriate technology, see [Radder, 1996, Chap. 7].

916

Hans Radder

4.5 Technological Normativity in Practice In Section 3 we have seen that, when studying the role of norms in actual practice, a variety of aspects need to be taken into account. In this section, I briefly illustrate these qualifications of normativity in the case of technological practice. I use the examples given in the previous sections, but to avoid repetition, I will not again address the issue of the inherent or contingent normativity of these examples. Different kinds of norms can be found easily in the mentioned examples. Consider the case of washing machines. Clearly, the design of these technologies needs to conform to technological norms (for instance, efficiency), social norms (such as safety), political norms (say, sustainability) and so on. It is also obvious that these distinctions are not always clear-cut. Is efficiency a technological or an economic requirement? Similarly, does sustainability count as a political or as a moral norm? The scope of norms that are implied in, or contingently related to, technological practice may vary considerably. Assuming that the use-plan approach implies that people should not behave irrationally, one should not use a screw driver for killing mosquitoes. This is obviously a norm with a relatively small scope: although it applies to all human beings, the number of occasions for a potential application of this norm is limited. In contrast, the norm that women ought to have no interest in technology (exemplified through the design of the Ladyshave) clearly has a much larger scope. Many technologies (for instance, a washing machine or a contact lens) include a user manual or other kinds of instructions for users. These instructions make explicit the norms concerning how people should behave in order to have the technologies work as intended by their designers, manufacturers or sellers. When norms are physically inscribed in the technological artifacts, they are clearly implicit and, perhaps, intentionally left implicit. Thus, the norm that Jones beach should not be accessible to the poor people of New York City was implicitly inscribed in the physical design of the bridges over the Long Island parkways. As we have seen, the question of whether or not this implicit normative inscription was made intentionally by Moses is answered differently by Winner and his critics. Of course, also in technological practice, norms can and will be broken. In the case of nuclear energy, especially as regards the aspect of storing highly radio-active nuclear waste, this can be anticipated as being highly probable. In the example of the African uses of the photoelectric lighting kits, it proved to be the case with hindsight that the people involved did not behave as they should according to the script of the technology. In the case of contact lenses, a user might not be able to exercise the required hygiene on a particular day. Thus, depending on the case in hand, the consequences of not conforming to such norms may be disastrous (as in nuclear energy), a substantial problem (as with the lighting kits), or less consequential (for instance, when the occasional lack of hygiene causes a minor eye infection that can be cured by not wearing the contact lenses for one or two days).

Why Technologies Are Inherently Normative

917

For the sake of argument, let us suppose (with the proponents of the use-plan approach) that people should not act in irrational ways. Even so, some people might question the applicability of this norm to the case of killing mosquitoes with a screw driver. After all, they might judge that the norm of rationality does not apply to cases where the use of the artifact does not do any harm (even if this use looks inefficient or ineffective). Finally, the issue of trade-offs between opposing norms can be illustrated by the cases of contact lens and nuclear energy technologies. In the contact lens case, many people allow their aesthetic norms to overrule the norms that come with the required self-discipline and control. In contrast, in the case of nuclear energy, many people prefer to stick to norms of long-term safety over the norms of possible short-term economic gains. 5 CONCLUSION Technologies are inherently normative. That is to say, the (actual or anticipated) realization of a technology in a certain region of space and time implies one or more norms about what the people who (might) interact with the technological system should, or should not, do. These people should behave in such a way that they enable, and do not disturb, the working of the technology in question. This argument draws both on the theoretical characterization of technologies provided in Section 2 and on the fact that the appropriate realization of the enabling and non-disturbing conditions for having the technology work as intended cannot be taken for granted. To flesh out this argument, Sections 4.1 through 4.4 provide detailed discussions of several important aspects of the inherent normativity of technologies, while Section 4.5 addresses some of the complexities of the issue of technological normativity in actual practice. It is obvious that technologies can be contingently normative, or normatively relevant, in innumerable ways. This chapter has focused on the more specific question of the inherent normativity of technologies. An interesting further question is whether the argument provided at the beginning of Section 4 captures the only way in which technologies can be inherently normative. Of course, it is always difficult to claim to have a definitive answer to such a question. But what can be done is to exclude some of the other ways in which technologies may be alleged to be inherently normative. In concluding the chapter, I will briefly look at two other possible arguments for inherent normativity. First, in addition to, and largely independently of, having functions, technologies may also carry symbolic meanings. Specific types of cars may count as social status symbols and certain designs may be seen as especially fashionable or aesthetically pleasing. Such symbolic meanings will entail certain normative claims, such as the social norm that ‘as a top soccer player, I ought to drive a Porsche’. Clearly, symbolic meanings and associated norms play important roles in the case of many technological artifacts. It should be equally clear, though, that there are many technological products which do not possess a symbolic meaning. Thus, even if a particular type of car may carry symbolic meaning, this will most prob-

918

Hans Radder

ably not apply to the types of bolts employed in its construction. Yet, these bolts are also artifactual, functional systems with a certain degree of stability and reproducibility, and hence are also technologies. More generally, quite a few, or probably most, technologies are simply taken for granted and used without any notice.28 Hence, the norms derived from the symbolic meanings of technologies are context dependent and contingent: they are not inherent in the technologies themselves.29 A second claim about technology and normativity says that ‘artifacts ought to perform their function’. This claim is explicitly meant to capture a type of inherent normativity of technological artifacts. However, as Maarten Franssen (in his chapter this Volume, Part V) shows, the ground of this alleged artifactual normativity cannot be found in the artifacts themselves. (Indeed, which sanction would be applied to a certain artifact if it would violate this norm?). Instead, these norms derive from a justified belief in the well-functioning of those artifacts. To make this point more concrete, consider this example. If I buy a dirt-cheap artifact at the flea market or if I get an artifact made by a friend (whom I know is not particularly handy), these artifacts may work but there is no expectation that these artifacts ‘ought to work’. In these cases the saying applies that ‘one should not look a gift horse in the mouth’. This shows that, in as far as normative expectations are present, they are based on some kind of (formal or informal) social contract or agreement, usually between the manufacturer or seller and the buyer of the artifact. Where such a contract or such an agreement is absent, there is no justified expectation and hence no ‘artifactual normativity’. Thus, neither symbolic meanings nor justified expectations entail that technologies are inherently normative. My discussion should not be taken to imply, however, that contingent norms are by definition less significant than inherent norms. Given a particular technology, its inherent norms may well be judged less important that some of the related contingent norms. More generally, whether inherent or contingent norms are seen to be more (or less) significant will be strongly dependent on the specific context of the technological systems and their material and social environment.

28 In

this respect, compare with Albert Borgmann’s notion of the ‘device character’ of modern technology [Borgmann, 1984]. 29 In addition, overemphasizing symbolic meanings is questionable for two further reasons. First, it implies a one-sided focus on separate artifacts (‘the car as a status symbol’) and hence may lead to a neglect of the broader systems in which these artifacts are incorporated; second, it contributes to the persistence of the current myth that we live in a ‘postindustrial society’ in which material production has become unimportant. See also [Henwood et al., 2001] and [Wajcman, 2004, pp. 121-122], who rightly criticize the claim that what counts in our ‘digital world’ is not material production but only information and communication.

Why Technologies Are Inherently Normative

919

ACKNOWLEDGMENTS In presenting earlier drafts of this chapter I received useful feedback from several audiences. I would like to thank the participants at the preparatory Eindhoven workshop on the Handbook project, the audience at the 2007 Delft workshop on ‘Philosophy and Engineering’ (in particular, Maarten Franssen), and the members of the research group ‘Philosophy of Science and Technology’ at VU University Amsterdam (in particular, Edwin Koster). The detailed comments by Ibo van de Poel stimulated me to further clarify the main claims of this chapter, while the thoughtful copy editing by Neil Milne provided the finishing touch.

920

Hans Radder

BIBLIOGRAPHY [Akrich, 1992] M. Akrich. The De-Scription of Technical Objects. In Shaping Technology/Building Society, W.E. Bijker and J. Law, eds., pp. 205-224. MIT Press, 1992. [Bijker, 1995] W. E. Bijker. Of Bicycles, Bakelites, and Bulbs. Toward a Theory of Sociotechnical Change. MIT Press, 1995. [Borgmann, 1984] A. Borgmann. Technology and the Character of Contemporary Life. University of Chicago Press, 1984. [Brumsen, 2003] M. Brumsen. Gebruiksplannen en Verantwoordelijkheid. In Handelingsontwerpers. Een Wijsgerige Visie op Ingenieurswerk, M. Scheele and P. Vermaas, eds., pp. 111-120. Damon, 2003. [Ducheyne, 2005] S. Ducheyne. Joan Baptista Van Helmont and the Question of Experimental Modernism. Physis, 2, 305-332, 2005. [Feenberg, 1995] A. Feenberg. Alternative Modernity. The Technical Turn in Philosophy and Social Theory. University of California Press, 1995. [Feenberg, 1999] A. Feenberg. Questioning Technology. Routledge, 1999. [Foucault, 1982] M. Foucault. The Subject and Power. In Michel Foucault: Beyond Structuralism and Hermeneutics, H.L. Dreyfus and P. Rabinow, pp. 208-226. University of Chicago Press, 1982. [Gingras, 1995] Y. Gingras. Following Scientists through Society? Yes, but at Arm’s Length! In Scientific Practice. Theories and Stories of Doing Physics, J.Z. Buchwald, ed., pp. 123-148. University of Chicago Press, 1995. [Joerges, 1999a] B. Joerges. Do Politics Have Artefacts? Social Studies of Science, 29, 411-431, 1999. [Joerges, 1999b] B. Joerges. Scams Cannot Be Busted, Social Studies of Science, 29, 450-457, 1999. [Habermas, 1971] J. Habermas. Technology and Science as ‘Ideology’. In Toward a Rational Society, J. Habermas, pp. 81-127. Heinemann, 1971. [Henwood et al., 2001] F. Henwood, S. Wyatt, N. Miller, and P. Senker. Critical Perspectives on Technology, In/Equalities and the Information Society. In Technology and In/Equality, S. Wyatt, F. Henwood, N. Miller and P. Senker, eds., pp. 1-18. Routledge, 2001. [Houkes and Vermaas, 2003] W. Houkes and P. Vermaas. Gebruiksplannen. In Handelingsontwerpers. Een Wijsgerige Visie op Ingenieurswerk, M. Scheele and P. Vermaas, eds., pp. 29-43. Damon, 2003. [Houkes and Vermaas, 2004] W. Houkes and P. Vermaas. Actions Versus Functions: A Plea for an Alternative Metaphysics of Artifacts. The Monist, 87(1), 52-71, 2004. [Koningsveld, 2006] H. Koningsveld. Het Verschijnsel Wetenschap, revised and enlarged edition. Boom, 2006. [Latour, 1992] B. Latour. Where Are the Missing Masses? The Sociology of a Few Mundane Artifacts. In Shaping Technology/Building Society, W.E. Bijker and J. Law, eds., pp. 225-258. MIT Press, 1992. [Layton, 1974] E. T. Layton, Jr. Technology as Knowledge. Technology and Culture, 15, 31-41, 1974. [Mitcham, 1994] C. Mitcham. Thinking through Technology. University of Chicago Press, 1994. [Oudshoorn and Pinch, 2003] N. Oudshoorn and T. Pinch. Introduction. How Users and NonUsers Matter. In How Users Matter. The Co-Construction of Users and Technology, N. Oudshoorn and T. Pinch, eds., pp. 1-25. MIT Press, 2003. [Perrow, 1984] C. Perrow. Normal Accidents. Basic Books, 1984. [Pitt, 2000] J. C. Pitt. Thinking about Technology: Foundations of the Philosophy of Technology. Seven Bridges Press, 2000. [Radder, 1986] H. Radder. Experiment, Technology and the Intrinsic Connection between Knowledge and Power. Social Studies of Science, 16, 663-683, 1986. [Radder, 1988/1984] H. Radder. The Material Realization of Science. Van Gorcum, 1988. (Originally published as H. Radder. De Materi¨ ele Realisering van Wetenschap. VU Uitgeverij, 1984). [Radder, 1996] H. Radder. In and about the World. Philosophical Studies of Science and Technology. State University of New York Press, 1996. [Radder, 1998a] H. Radder. The Politics of STS. Social Studies of Science, 28, 325-331, 1998.

Why Technologies Are Inherently Normative

921

[Radder, 1998b] H. Radder. Second Thoughts on the Politics of STS. Social Studies of Science, 28, 344-348, 1998. [Radder, 2008] H. Radder. Critical Philosophy of Technology: The Basic Issues. Social Epistemology, 22, 51-70, 2008. [Summerton, 2004] J. Summerton. Do Electrons Have Politics? Constructing User Identities in Swedish Electricity. Science, Technology, and Human Values, 29, 486-511, 2004. [Van der Pot, 1994] J. H. J. Van der Pot. Steward or Sorcerer’s Apprentice? The Evaluation of Technical Progress, Volume I and II. Eburon, 1994. [Van Oost, 2003] E. van Oost. Materialized Gender: How Shavers Configure the Users’ Femininity and Masculinity. In How Users Matter. The Co-Construction of Users and Technology, N. Oudshoorn and T. Pinch, eds., pp. 193-208. MIT Press, 2003. [Wajcman, 2004] J. Wajcman. TechnoFeminism. Polity Press, 2004. [Winner, 1986] L. Winner. The Whale and the Reactor. University of Chicago Press, 1986. [Woolgar and Cooper, 1999] S. Woolgar and G. Cooper. Do Artefacts Have Ambivalence? Moses’ Bridges, Winner’s Bridges and Other Urban Legends in S&TS. Social Studies of Science, 29, 433-449, 1999. [Wynne, 1988] B. Wynne. Unruly Technology: Practical Rules, Impractical Discourses and Public Understanding. Social Studies of Science, 18, 147-167, 1988.

This page intentionally left blank

ARTEFACTS AND NORMATIVITY Maarten Franssen

1

INTRODUCTION: ARTEFACTS, ARTEFACT FUNCTIONS, AND THEIR NORMATIVE DIMENSION

The basic defining characteristic that distinguishes artefacts from the other tangible objects with which our environment is filled is a relational and historical one. Artefacts are objects that have been made by people for a particular purpose. Two objects may be identical in all of their physical properties, yet one may be a natural object while the other is an artefact. This may seem a very implausible thing to occur for the ordinary consumer products that are the prototypical artefact for most people, but examples are far from imaginary: artificial diamonds or rubies, for instance, are much used in various branches of industry. The defining characteristic limits artefacts to a subclass of the class of all things that are made by humans, in the sense of resulting from an intentional act that consists in the modification and reshaping of material bodies. Other elements in this class, for instance waste products, such as exhaust fumes and sawdust, and other things that are the by-products or side-effects of human action, such as footprints and fingerprints, are not artefacts, because, although the process from which they result is (presumably) done for a purpose and the maker is aware of producing them, the fact that they are made is accepted rather than intended and the maker has no purpose for them. Usually, that is, since for each of these examples a case can be imagined where they are made for a purpose and therefore artefacts: sawdust to fill a doll, exhaust fumes to commit suicide, footprints or fingerprints to lead someone astray. Arguably, everything that can be made can be made to serve a purpose and can therefore classify as an artefact. This above definition of an artefact is ambiguous concerning the relation between ‘made’ and ‘for a purpose’. The purpose can be associated with the process of making the object or with the resulting object itself. This ambiguity is closely related to an important distinction between two sorts of artefacts: on the one hand, artworks, and on the other hand, useful artefacts such as tools, instruments and consumer products, in short, as I will call them, technical artefacts. It has been argued that in the case of artworks, the purpose for which they are made is the making of the work of art itself.1 Technical artefacts, in contrast, are made to 1 Artworks with a particularly problematic ontological status, such as symphonies or novels, must perhaps be seen as instructions to make the work again and again, where the actual work of art is the performed symphony or the unfolding story rather than the score or the text.

Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

924

Maarten Franssen

serve a purpose that is still in the future while they are made. They are made to be used, and generally the sole purpose of making them is to make them available for future use. Of course it may happen that someone makes an artefact in order to serve a purpose he or she has at precisely that moment, in order to use it right away when finished and discard it afterwards, but even then the purpose of the act of making the artefact is different from the purpose for which the artefact is being made. To be used in a particular way and thereby enable or support its user to realize a particular outcome, is what is called an artefact’s function. This cannot serve as a sharp contrast to artworks, however. It seems to me undeniable that artworks — that is, paintings, sculptures, stories, musical performances — also serve purposes: to create certain mental states or conditions in people.2 Still, of we refer to the achievement of such a purpose through a work of art using the work, there remains a clear difference in the form use takes for both categories: the use of artworks is exhausted by the perception of them, while using a technical artefact involves a much more complicated, bidirectional causal interaction between the user and the artefact, which can be referred to as manipulation in a broad sense.3 Although technical artefacts are made to be used for a purpose, we need not have recourse to an artefact when we seek the use of an object as an intermediary between us and the achievement of some goal we have. The usefulness of an object lies in its physical properties, and we can use any object that is available as long as it has, roughly, the required characteristics. We can use a rock to hammer a pole into the ground, to cut a rope, to keep some cloth from being blown away, and so forth. Indeed, historically, the class of artefacts originated in the first modifications of such natural objects, in order to make them (slightly) more fit for the purpose for which they were being used. In using ‘found objects’ for a purpose, we are of course not restricted to natural objects; we can also use artefacts for purposes that have no relation to the purpose they were made for serving, as happens when we use, for example, a hammer as a counterweight. In doing this we in a sense ignore their status of artefact; we take into consideration only those of their physical properties that help us to achieve our goals, just as we would do with a natural object. The mere using of an object, be it a natural or an artefactual object, for a purpose is to be distinguished from the making of an artefact. The making of an artefact is the intentional modification of one or more of the properties of a natural object (where ‘object’ includes lumps of stuff), resulting in the creation of an artefact.4 The notion of modification is very broad, however, and if this is all 2 These people are not necessarily the ‘users’ of the artwork. Even if works of art serve a purpose, to say that they are used for that purpose, or to specify who uses them when, is much more difficult for artworks than for most technical artefacts, although for some artefacts, such as components or large-scale systems, this is also problematic. 3 For an extensive treatment of the works of art in relation to the general notion of artefact, see [Dipert, 1993]. 4 Here I am ignoring all associated ontological issues, for instance, in what sense the artefact is identical to the modified object.

Artefacts and Normativity

925

that separates the class of artefacts from the class of natural objects used, the two classes become, so to speak, contiguous. If, during my weekly walk through the local wood, at one time I pick up a dead branch and strip it of all awkward twigs to use it as a walking stick, and at another time I pick up another dead branch and after close inspection decide that no awkward twigs need to be removed, it seems pointless to insist that I am using tokens of two quite different sorts of things as a walking stick on these two occasions, one a natural object pure and simple and the other an artefact.5 However, we need not settle this matter in order to have an adequate grasp of the difference between natural objects used for a purpose and artefacts designed and, presumably, used for a purpose. If, in contrast, the object we use for a purpose — any purpose — is already an artefact, we do not create it as an artefact but only as a specimen of a new type of artefact. It should also be noticed that many technical artefacts are not used directly, or hands-on, so to speak. Many artefacts are components, performing a function at the inside of another artefact, which is the one that is, properly speaking, used. What we do with components is install them, not use them. If someone uses an assembled artefact, he or she does not at the same time use its components if only because the user often enough has no inkling that most of these components are there at all. In this chapter I will speak of all artefacts as if they are used, but ‘used’ should then be interpreted in a broad sense, as ‘used or installed’. The above considerations may serve as an introduction to the class of artefacts. What I wish to discuss in this contribution is a property that all entities in this class seem to have, and that entities not belonging to it seem to lack: they support some specific sorts of judgements. First of all, they support evaluative statements like ‘This is a good drill’ and ‘Pumps of trademark X are not very good’, or ‘This camera usually functions well’ and ‘The radio in the spare room is a poorly functioning one’. With respect to artefacts such statements are generally considered to make perfect sense, whereas they are not in order as far as natural objects are concerned. There is no distinction between good ones and poor ones among stones, stars, electrons, oak trees, and so forth. It seems, however, that certain biological items are an exception to this: we do say that someone has a bad heart, or a good pair of lungs, and we also talk of good watchdogs and poor milk-cows. Any explanation of why artefacts, unlike natural objects per se, support evaluative statements must be able to explain why specific biological items are thought to support them as well. An example of a second sort of judgement supported by artefacts is ‘This coffeevending machine malfunctions’. We generally use the expression ‘malfunctioning’ in case a technical artefact fails to perform its function entirely, not just poorly. It is tempting to interpret a malfunction statement as representing a boundary case, shorthand for ‘This is an extremely poor coffee-vending machine’, by which it simply becomes an example of the sort already discussed. This may be moving 5 Cf. the view of Richard Sorabji [1964] that a natural object of which some property has merely been contemplated for modification in order to make it usable for a particular purpose thereby already becomes an object-with-a-function and, perhaps, an artefact.

926

Maarten Franssen

too fast, however. Even a poor artefact of a particular kind can still pass discontinuously from functioning to malfunctioning. By this we do not mean that it drops a bit more on the scale of good to poor. A malfunctioning coffee-vending machine certainly is, at that precise moment, a very poor coffee-vending machine, but once it has been discovered that a fuse has blown and once that fuse has been replaced, the machine suddenly becomes again the excellent or not-so-excellent coffee-vending machine it was before its fuse blew. There is more than one source or mechanism at work behind these qualifications.6 Finally, there is a third type of statement concerning artefacts, which is related to malfunction statements but has a form that separates them clearly from either malfunction judgements or evaluative judgements. Being confronted with a malfunctioning coffee-vending machine, we may say that, although the thing does not make us a cup of coffee, that is what it ought to do, or what it is supposed to do. That is exactly why we call the machine malfunctioning: although it does not perform the function of making coffee, this does not change the fact that its function is to make coffee, and this is, therefore, what it should do. Compare our different treatment of natural objects: if we have found, day after day, that samples of fluid, when heated, expand, but finally, while heating a sample of water from 0◦ C to 4◦ C, discover that it does not expand, it is not correct to say that, although this particular sample does not expand, it still ought to expand. Instead we retract our claim that all samples of fluid expand when heated. Any of the three sorts of statements introduced above can be found mentioned as evidence that artefacts have, so to speak, a normative dimension. In the remainder of this chapter I will investigate whether these judgements indeed represent a normative dimension inherent in artefacts, and if so, in what exact sense. But before I start out doing that, I should point out that a discussion of the normative dimension of technical artefacts does not exhaust by far the normative dimension of the practice of designing and making such artefacts — that is, the practice of engineering — nor the normative dimension of the practice of using them or deciding about them — in short, the practice of technology as a whole. As practices, these activities have a normative dimension in being rule-governed, from a relative micro-level of individual design decisions to a relative macro-level of business and government policies concerning the implementation and regulation of technologies. The conception of normativity associated with technological practice is in many respects a social one: norms and rules are conceived as institutionalized, as requiring matching expectations in people throughout society and as implying the existence of sanctions on norm violation. In this chapter I focus on the normative dimension of technical artefacts, rather than the normative dimension of technological practice generally, and I will do so using a different conception of normativity, focussing on the rational individual. For a discussion of the normative dimension of technology as a practice, see the chapters by Radder and by 6 Hansson [2006], however, sees malfunctioning as a form of functioning poorly, not as discontinuous.

Artefacts and Normativity

927

Pritchard in this volume. In section 5, however, I say a few things that have a bearing on technology as a social practice. So, given the observation that various sorts of judgements are applied to technical artefacts, and that these judgements are widely seen as revealing that artefacts have a normative dimension, or that the notion of function is inherently normative, two important questions emerge. The first is: is the interpretation of these judgements supported by artefacts correct? In particular: do the various examples of normative judgements articulate the same ‘dimension’, and is this dimension correctly identified as normative? The second question is: what does it mean? In particular: why is it deemed important that this alleged normative dimension is identified, and in what way does it increase our understanding of artefacts? In this chapter I will primarily deal with the first question, but I will also address the second. 2 THE NOTION OF NORMATIVITY Is there indeed something that unites the judgements ‘This is a poor camera’, ‘This is a malfunctioning camera’, and ‘Making photos is what this camera is supposed to do’, and is this something correctly identified as normativity? To answer this question, what we need is a characterization of normativity. Traditionally, normativity has first of all to do with what is right. Rightness and its opposite wrongness concern our actions: a particular act by a particular person can be the right thing to do or the wrong thing to do in this person’s circumstances, that is, the thing this person ought to or ought not to do. Most authors take the domain of the normative to be broader than being about what we ought or ought not to do. Issues concerning the rightness or wrongness of actions belong to the deontic part of the normative. Another part of this domain concerns what is good rather than right. Goodness and badness pertain not to our actions but to the situations in which we act and that make us act, or to the situations that arise from our actions. So what is good or bad, or what has value, are first of all states of affairs. Apart from states of affairs, however, also other things that are in some way related to actions are said to have value: persons as the originators of actions, motives as the ‘drives’ of such actions, and objects as far as they are the product of such action: artefacts. Issues concerning value belong to the evaluative part of the normative. If the normative is taken to be a unified realm, evaluative and deontic statements must be related. There has been little consensus on how this relation must be seen, however, in particular not concerning basic issues like whether the relation between what is good and what one ought to do is analytic or synthetic, necessary or contingent, or whether what one ought to do is dependent upon or reducible to what is good, or whether what is good is dependent upon or reducible to what one ought to do, or whether they are independent. Following a recent proposal by Dancy [2005], I will characterize the normative in general as being about the difference that facts about the world make to the

928

Maarten Franssen

question what to do or believe or aim for.7 A normative fact, more in particular, is the second-order fact that a particular fact or set of facts about the world is relevant to someone’s deliberations concerning his or her actions, beliefs or goals. This ‘definition’ does not characterize the normative exclusively in non-normative terms; instead it brings to light the basis underlying all normativity, namely the relevance relation between the question what to do, believe or aim for and the facts about what the world is like. These facts that make a normative difference need not be ‘purely’ natural facts as they are commonly understood. They include intentional facts, that is, facts stating what certain people do, feel, expect and the like. Such facts matter just as much or even more to what motivates our own behaviour as natural facts; someone’s grief, for instance, surely is a reason to comfort her. Even normative facts, if one accepts their existence, could perhaps be taken to make a normative difference, but since these normative fact can themselves be analyzed as second-order facts that state the normative significance of still other, lower-level facts, the definition ultimately analyzes normativity in terms of a normative relevance relation linking non-normative (natural and/or intentional) facts to the actions, beliefs and goals of persons.8 In talking of normative facts, I may be seen to take a stand in the debate between cognitivism and non-cognitivism in ethics. Cognitivist take it that statements concerning what we ought to do can be true or false and that there are normative facts ‘out there’ that serve as the truth-makers of normative statements. Noncognitivists hold that normative statements lack truth-values; they are a different sort of thing compared to descriptive statements, which do have truth-values. Instead, normative statements, for example, express an attitude the speaker has toward someone’s actions or beliefs. In my opinion, however, my analysis is independent of whether cognitivism or non-cognitivism is ‘correct’. A non-cognitivist reading of my definition would interpret it as saying that a normative statement is one that expresses that a particular fact is taken to be relevant to a person’s question what to do, believe or aim for. On the general characterization of the normative adopted here, deontic statements — statements that say that one ought or ought not to do something — are, of course, straightforwardly abbreviated normative statements. They settle the question what to do by saying what one ought to do, but they do this without bringing in the features of the world that ground this claim, or this fact, if you believe that the normative claim reports a fact. Note that on the adopted characterization of normativity, the relation between a prescription and the ground for it is basic, not the prescription itself. The broad characterization of normativity I adopt here also allows an interpretation of evaluative statements as abbreviated complete normative statements, but the abbreviation now concerns more the issue 7 These are the three ingredients involved in human action distinguished by most theories; see the chapter by Kroes, Franssen and Bucciarelli on rationality in engineering design in this Volume. 8 Many philosophers, however, would claim that intentionality is an inherently normative notion. In will not discuss here the complication of this position for the question how normativity is distributed over the various ingredients in the definition.

Artefacts and Normativity

929

what to do than the grounds for doing it. An evaluative statement, saying of an object or a situation that it is good or bad, indicates that a particular feature of the object or the situation is of relevance to the question what to do, without, however, specifying this relevance any further. Indeed, the relevance relation between first-order facts and the question what to do or believe or aim for has still to be spelled out. I will take it to be a reasongiving relation. Since the work of Joseph Raz [1975], the notion of a reason is seen as the basic normative concept. The normative concerns the reasons we have, or take to have, for doing, believe, or aiming for something.9 If we have a reason to do something, however, this does not make it the case that we ought to do it. If some fact within the totality of facts give us a reason for doing something, this does not preclude that another fact gives us a reason against doing this thing Whether or not we ought to do something depends on the balance of all reasons we have for or against doing it.10 Although we can distinguish between reasons to do a particular thing, reasons to believe something and reasons to aim for something, these three ingredients of action are not on a par. Beliefs and aims or desires are much less open to voluntary control than actions. Many beliefs and desires occur to us spontaneously, not as a result of any deliberation. If you wonder whether you should form the belief that there is a red tomato on the table in front of you on the basis of this red tomato lying on the table in front of you, you already believe that there is a red tomato on the table in front of you. If you wonder whether there is any reason to aim to have a vanilla ice cream just now, you (plausibly) already possess the longing or desire for vanilla ice cream just now. Basic beliefs and basic desires or preferences (in the everyday sense of the word rather than the technical sense from decision theory) are prior to reasoning and rationality. When reviewing or weighing reasons for or against beliefs and desires, this usually concerns reasons either to retain or to drop a belief formed previously on the interpretation of sensory evidence or to adopt a belief, as a result of reasoning rather than direct sensory evidence, or it concerns adopting as an end some particular state of affairs or either maintaining or dropping or reordering some of your previously adopted goals. To speak in terms of acting or believing or desiring as the threefold classification of things that we have reasons for may miss the distinction, with respect to believing, between coming to believe something, on the basis of direct sensory perception, and retaining or 9 Most authors who take reason to be the central normative concept are non-naturalists: they take reason (e.g. [Scanlon, 1998]) or the favouring relation between a ground and what we can do reflected in a reason (see [Dancy, 2000a; 2005]) to be a primitive concept, unexplainable in terms of other non-normative, presumably naturalistic, concepts. A different position, however, which treats reason as the conceptual linchpin of our normative talk but holds that the concept of reason can be explained in non-normative naturalistic terms, is certainly conceivable. 10 This is how most authors see the relation between ‘a reason’ and ‘ought’. As Raz [1975] uses these terms, however, if one has a reason to do something one ought to do it; the case that others express by saying that one ought to do something is termed by Raz a case where one has a conclusive reason to do it. Gert [2004] defends a distinction between justifying reasons and requiring reasons and holds that no amount of justifying reasons ever add up to a requiring reason.

930

Maarten Franssen

adopting a belief, and it may similarly miss the distinction, with respect to desiring, between feeling a desire for something and maintaining or adopting something as an end. To bring this out, I will, in the remainder of this chapter, talk of all reasons as reasons for doing something. Acting is, on this usage, not identical to doing something; it is a form of doing. Since we should not lose the distinction between theoretical and practical rationality, however, we should distinguish theoretical and practical forms of doing. Finally, before going on to put this analysis to work for explicating the normative dimension of technical artefacts, it should be noted that the characterization of normativity adopted here is blind to the sorts of reasons that can be distinguished, for instance, whether they are moral reasons, or prudential reasons, or still other ones. Any conflicts one may think possible, or even inevitable, between various sorts of reasons, in particular between moral reasons and prudential reasons, are to be fought out on the scales where the balance of reasons is determined. 3 EVALUATIVE STATEMENTS AS NORMATIVE STATEMENTS With the above apparatus in place, it is not difficult to spell out how evaluative statements involving technical artefacts, such as ‘This camera is a good one’, can be interpreted as normative statements. Most generally taken, an evaluative statement points to a property of whatever is evaluated as in a particular way relevant to the question what to do. To be informative, the nature of the relevance has to be specified, but there is an obvious way to do so in the case of technical artefacts. Technical artefacts are used for a purpose, by which the user aims to realize a particular outcome, which he or she would not be able to realize, or much less satisfactorily, without using the artefact. The ‘goodness’ or ‘poorness’ of an artefact concerns features of the artefact that are relevant to their use. It has to be emphasized that evaluative statements like ‘This is a good knife’ presuppose the existence of kinds in which artefacts can be grouped, in this case the kind ‘knife’. It is as a knife that the artefact is evaluated as good. Artefacts are designed as kinds (even if only one copy is manufactured) to be used to achieve a particular purpose, which then is what they are for, what is their function. We may then suppose that artefacts kinds are in fact functional kinds. A drill is an artefact used for drilling holes, a saw is made for sawing through objects, a hammer for hammering pointed objects into a surface, and so forth. I will indeed refer to artefacts as representatives of functional kinds, say, the kind K, and refer to the activity constituted by their use in terms of this kind as K-ing. Nevertheless, to think that all artefact kinds are functional kinds is a mistake. For a start, the notion of a functional kind is ambiguous between objects being used for a particular purpose and objects being designed to be used for a particular purpose [Franssen, 2008]. The first category includes natural objects, the second includes objects that may never be used as designed or that may be incapable of being so used, due to, for instance, a broken or missing component. Unless stated otherwise, I will presuppose that the artefact kinds I refer to are the kinds they are by design, such

Artefacts and Normativity

931

that a particular artefact, designed to be, say, a drill, indeed classifies as a drill even when it is malfunctioning. The idea of malfunction presupposes that artefact kinds are subkinds of functional kinds: otherwise there would be nothing to prevent me from calling a sugar cube a malfunctioning telephone. By implication, the design history of an artefact is to be counted among its features. But even on the kind level, that is, disregarding malfunctioning tokens, the artefact kinds emerging from technical design do not match purely functional kinds. Take a knife. A knife is used for cutting, but the class of knives is not identical to the class of cutters, i.e., instruments designed for cutting. There are many ways to cut through some material, and using a knife is only one of them; one can also cut with a thread, with a cutting torch, with a laser beam, with a high-pressure water jet, and so forth. One may specify the nature of the use-activity, for instance by describing it as cutting-by-exerting-pressure-in-one-direction, or hand-steeredcutting, and this may remove some of the listed cutting instruments (e.g. the cutting torch), but it will not remove all but knives (e.g. not the thread). We would be hard put to circumscribe the activity of ‘kniving’ such that the concept of a knife can be defined in terms of it, rather than that it presupposes the concept of a knife. Similar considerations apply to all other artefact kinds, with the inclusion of drills, saws and hammers. Our definitions of artefact kinds include non-functional aspects, primarily morphological in relation to their operational principle, i.e., the way they ‘work’. But although the distinction between functional kinds and artefact kinds is of importance for a proper account of the notion of function as such, including the function of biological items, I do not think that it matters a great deal to the analysis of the normative dimension of artefacts, and therefore I will treat them as specimens of functional categories that are one-one related to practices of use. I will now start to make clear how evaluative artefact judgements can be interpreted as making normative claims. A first proposal for interpreting a statement like ‘x is a good K’ as expressing a normative fact, or making a normative claim, is the following: ‘x is a good K’ expresses the normative fact that x has certain features f that make x a K and that make it the case that if a person p wishes to K, then p has a reason to use x for K-ing. For example, if x is a knife, then the use I make of it is for cutting. Prima facie, on this analysis the statement ‘x is a good K’ is analyzed as stating two distinct facts at once: the fact that x is a K, and the fact that a person p who wishes to K has a reason to use x for K-ing. This is indeed so, but I will address this further in the next section. For now, let it just be noted that the addition of the condition ‘if a person p wishes to K’ is necessary and reflects that artefacts, if they are used, are used for a purpose. If someone does not have an appropriate purpose, he or she has no reason to use x, irrespective of x’s qualities. An important point contained in this analysis is that the particular features of x alone cannot give p a compelling or conclusive reason to use x for K-ing, such

932

Maarten Franssen

that p ought to use x for K-ing. However good a specimen of a K x may be, it is possible that another artefact x is at hand that is superior to x as far as K-ing is concerned. In this case, insofar as there is any artefact that p ought to use, it would be x rather than x. Alternatively, I may have strong, even conclusive reasons not to use x of a completely different sort: it may be illegal for me to use x, for example because I do not own x. Be that as it may, the above proposal cannot be correct. Suppose that x is a long and sharp knife and that Mrs p intends to have the throat of Mr p cut because of his loud snoring. On the present analysis it follows that calling the knife a good knife is a way of expressing the normative fact that the particular features of the knife, its sharpness for instance, give Mrs p a reason to use the knife to cut Mr p’s throat. But it cannot be correct that Mrs p is granted a reason to put the knife to her husband’s throat merely because she wishes to do so and a knife that would do the job is available.11 There are acts for which no reason for performing it can ever be given (barring perhaps extraordinary circumstances), and the cutting of someone’s throat to put an end to his snoring is one of them. In this case there are moral and perhaps also prudential reasons against doing so, since it may well be that at a later time Mrs. p comes to deplore her act. Whether on the force of a single compelling reason or on the balance of all reasons, Mrs p ought not to cut her husbands throat and she ought not to use the knife to do so. However, that she ought not to use the knife for cutting her husband’s snoring short does not diminish in any way the goodness of the knife. This difficulty could be repaired by letting the reason to use the artefact be conditioned by a justifiable wish or aim, such that the user can be said to have an antecedent reason, as the following second proposal has it: ‘x is a good K’ expresses the normative fact that x has certain features f that make x a K and that make it the case that if a person p has a reason for K-ing, then p has a reason to use x for K-ing. This move seems to miss the point, however. Although it is true that if x is a good K, then if p has a reason to K, p has a reason to use x for K-ing, it seems to me that this is no more than an implication of what ‘x is a good K’ expresses. The goodness of the knife is an instrumental value; the knife is good as a particular instrument, in this case, a knife. We may want to deny that Mrs p has a reason to use the knife to cut her husband’s throat, but that does not change in the least the fact that the knife is recommendable for cutting, abstracting from the purpose for which the cutting is done. To make this explicit, the analysis should not be in terms of simple reasons but of what Broome [1999] calls ‘normative requirements’ and Dancy [2000a] calls complex duties. The reasons that the features of the knife provide to a person are not direct reasons for an action but reasons for maintaining a connection between what the person aims to achieve, his or her end, and what this person uses to satisfy this aim, the means to the end. The case is similar 11 This is not to say that having a desire for doing something can never be a reason for doing it. For the analysis it suffices that desires are often incapable of giving one a reason.

Artefacts and Normativity

933

to, for example, the rationality of belief in the following way. Since p and p → q together entail q, it might seem that if you believe p and if p → q, then you ought to believe q. This does not follow, however, since if you ought not to believe p, because p is false, then it is not so that you ought to believe q. What you ought to do is a complex thing: you ought to uphold a certain connection between your beliefs: (if you believe p and if p → q, to believe q). There are two ways in which you can satisfy this requirement: one is by believing q, the other is by stopping to believe p, and if p is false, the latter is what you ought to do. For instrumental use, the features of an artefact x do not require of you that (if you wish to K, you use x for K-ing), they merely give you a reason to see to it that (if you wish to K, you use x for K-ing). So we arrive at a final proposal: ‘x is a good K’ expresses the normative fact that x has certain features f that make x a K and that make it the case that a person p has a reason to see to it that (if p wishes to K, then p uses x for K-ing). There are two ways in which p can satisfy what the features of x give her reason to: by hanging on to her wish to K and then use x for K-ing, or by dropping the wish to K. And if p has no good reasons for K-ing, then dropping her wish is what she has a conclusive reason to do. The term ‘normative requirement’ introduced by Broome refers only to a strict requirement or complex duty: x requires y for p means that p ought to see to it that (if x is the case then y be the case). The non-strict form that is at issue in the case of instrumental value is what Broome calls a normative recommendation: ‘x recommends y for p’ means that p has a reason to see to it that (if x is the case then y is the case). Using this notion of a normative recommendation, my final proposal can be given the following shorter form: ‘x is a good K’ expresses the normative fact that x has certain features f that make x a K and that make it the case that a person p’s wish to K recommends that p uses x for K-ing. The mirror image of the above proposal for interpreting positive evaluations as normative statements is the following interpretation of negative evaluations as normative statements: ‘x is a poor K’ expresses the normative fact that x has certain features f that make x a K and that make it the case that a person p’s wish to K recommends that p does not use x for K-ing. In this case, however, the reference to p’s wish to K is superfluous. The features of x make it the case that any person p has a pro tanto reason not to use x for K-ing. If p has no reason to K, then this gives p a conclusive reason not to use x for K-ing, but that does nothing to diminish the independent reason for not using x that is grounded in p’s features: ‘x is a poor K’ expresses the normative fact that x has certain features f that make x a K and that make it the case that a person p has a reason not to use x for K-ing.

934

Maarten Franssen

What remains is, again, that one cannot go further than saying that p has a reason not to use a poor x, not that p ought not to use x. However poor a K x may be, if no alternative is available p may still have a reason to use x, if only p’s need for K-ing is urgent enough. The basic evaluative statements express, therefore, that the features of the artefact itself are such that they provide a pro tanto reason for or against using the artefact for the purpose it was designed for. A final point to be noted is that the proposed analysis simplifies considerably in that it disregards a number of important issues related to the use of artefacts. Apart from p’s need for K-ing or wish for the result of K-ing, and x’s features f , usually other conditions must be satisfied as well. Take the following example: let x be a good car, in that it has features such that p, who needs to do the shopping for the coming week, has a reason to use it to drive to the local supermarket. Does this mean that p similarly has a reason to use the car to drive to her local school? Not if p is twelve years old. In this case she definitely has a reason not to use the car to drive to school. Houkes et al. [2002] argue that every technical artefact is imbedded in a use plan that specifies which operations of the artefact will lead to the end state that corresponds to the function of the artefact. A use plan tacitly or explicitly contains the circumstances that must obtain and the abilities the user must show for these operations to lead to the desired end state. A clinical thermometer cannot be used successfully to check the temperature of an oven, an electric drill will only work when connected to a life power socket, a torch needs a fresh set of batteries, and a chainsaw wielded by someone with insufficient muscle power and training will saw up something else rather than the wood that needs to be sawn. A more precise interpretation of positive evaluative statements as normative therefore reads: ‘x is a good K’ expresses the normative fact that x has certain features f that make x a K and that make it the case that (1) p’s wish to K and (2) the accordance with the use plan for x of (i) p’s abilities, (ii) p’s knowledge, and (iii) the circumstances in which p operates, jointly recommend that p uses x for K-ing. The addition of the conditions mentioned in clause (2) leaves the basic analysis intact, however, and having pointed them out I will forget about them in the remainder of this chapter. 4 FUNCTION AND MALFUNCTION STATEMENTS AS NORMATIVE STATEMENTS The results of the previous sections can be extended to statements about artefacts that are not explicitly evaluative judgements. This applies, for example, to the following statement: ‘x is a K in working order’ expresses the normative fact that x has certain features f that make x a K and that make it the case that a person p’s wish to K recommends that p uses x for K-ing.

Artefacts and Normativity

935

In other words, ‘x is a K in working order’ says that x is a K and that x is useful for K-ing. It must definitely be added that x is in working order; the mere fact that x is an artefact of a particular functional kind K by design is not enough to recommended its use to anyone with a wish for K-ing, since x could be a malfunctioning specimen of that kind. With malfunction statements, we have arrived at the second sort of normative statement distinguished in section 1. Malfunction statements are not on the face of it evaluative statements, and it is therefore unclear why one can be convinced so easily that they represent normative statements, without an analysis to ground this interpretation, as developed here. To say that some technical artefact malfunctions is to say that it is incapable of performing the function for which it was designed. Like ‘This is a drill in working order’, this contains no evaluative words at all. Nevertheless, a malfunction statement can be analyzed as a normative sentence quite like the evaluative statements from the preceding section, as follows: ‘x is a malfunctioning K’ expresses the normative fact that x has certain features f that make x a K and that make it the case that a person p has a reason not to use x for K-ing. As in the case of ‘x is a poor K’, the reference to p’s wish to or reason to K is now superfluous. Must we retain the restricted ‘use x for K-ing’ ? Surely, if x malfunctions, then one has a reason not to use x tout court? This, however, is not true, since the use that can be made of an artefact is not limited to what it is designed for. Suppose that p owns a burnt-out iron. It is not true that p now has a reason not to use the iron, since p may wish to use the iron as a paperweight, and it serves perfectly for that purpose. In a way it serves this purpose better than a functional iron would, since now no conflict of uses, wanting to iron a shirt while the papers on the desk are being organized, can occur. Similarly, it would seem that the difference between functioning poorly and malfunction should be brought out by saying that, whereas one may still have a reason to use a poor K for K-ing — when no better alternative is available and the value attached to the result of K-ing supports its being done poorly — one cannot have a reason to use a malfunctioning K for K-ing, since a malfunctioning K will not enable you to K in the slightest. So we would have to say that, in the case of a malfunctioning K, p has a compelling or conclusive reason not to use x for K-ing, in other words, that p ought not to use x for K-ing. Whether this is correct depends on the question how sharply artefact functions can be defined or how sharply they can be assumed to be defined, a question related to the issue of functional kinds referred to at the beginning of section 3. Suppose Jack intends to rob a bank but finds his gun is jammed. Does Jack now have a conclusive reason not to use the gun for the robbery? I would say he has not, since he may reasonably expect that merely waving the gun around will do the job. This, however, might be based on an overly congenial interpretation of the kind K to which x belongs. Guns are not for robbing banks, they are for shooting, and Jack surely ought not to use his jammed gun for shooting.

936

Maarten Franssen

It often requires some careful thinking, however, what the use of a particular artefact includes and whether one always has a conclusive reason against using a malfunctioning artefact in accord with a use plan matching the design requirements. Guns are made for shooting, but so are riot guns and air guns, and Jack would be ill-advised to rob the bank with an air gun. Guns are made for shootingto-kill, and it is on that basis that one can use a gun to threaten people with. So a gun can also be said to belong to the artefact kind of threatening weapons, and as a representative of that kind, a jammed gun need not necessarily classify as a malfunctioning weapon, whereas a knife with a broken blade probably would. Evaluative judgements concerning artefacts are, therefore, very sensitive to a precise match between design and use. A knife for slicing bread is not a good knife for slicing tomatoes, and vice versa. It may well be a pointless question to ask which of the two is the better knife, the bread-slicing knife or the tomato-slicing knife. I would say, however, that the fact that both are knives implies that, if something has to be cut, this pro tanto recommends the use of either of them for doing the cutting with. It can even be argued that what makes an artefact a good or a poor one not only depends on the relevant functional kind to be considered but also on the circumstances in which it is going to be used. If all is well, these circumstances have been taken into account in the functional requirements guiding the design and are attended to in the artefact’s use plan. Of two violins, one may be best for performing an indoor recital with (and this will probably be the violin we call the best one tout court) while the other one is considered the better violin for an outdoor concert on a rainy day. And if there was a market for violinsfor-performances-on-rainy-days, and a corresponding manufacturing industry, it should no longer be obvious which violin will be considered the best tout court. Or, extending this even to the evaluation of artefacts versus non-artefacts, if in an emergency situation in chaotic conditions, a wound needs to be cut open, then using a glass splinter may well be a better choice than picking up a knife that is lying there, because the knife, for being a knife, may have been used in ways that will put the victim at a higher risk of becoming infected than using the glass splinter will. In the previous section I pointed out that the notion of a functional kind is ambiguous between objects being used or usable for a particular purpose and objects being designed to be used for a particular purpose. I also suggested in the previous section that on the account adopted by me, the statement ‘x is a good K’ is analyzed as stating two distinct facts at once: the descriptive fact that x is a K and the normative fact that a person p who wishes to K has a reason to use x for K-ing. Indeed, these two issues are strongly related. Looking again at the judgement ‘x is a good K’, the following proposal for interpreting it as referring to normative facts brings this out: ‘x is a good K’ expresses the fact that x has certain features f  that make x a K and expresses the normative fact that x has certain

Artefacts and Normativity

937

physical features f  that make it the case that a person p’s wish to K recommends that p uses x for K-ing. The evaluative statement ‘x is a good K’, therefore, can be seen as a conjunction of two statements, one saying ‘x is a K’ and the other ‘x is useful for K-ing’ or ‘x is useful as a K’. I have left it open whether ‘x is a K’ is interpretable as stating a normative fact; this will be taken up in a moment. The second part carries the normative burden that has been central to my analysis up till now: ‘x is useful as a K’ expresses the normative fact that x has certain physical features f  that make it the case that a person p’s wish to K recommends that p uses x for K-ing. The features recommending the use of x for K-ing that the goodness of x points to are necessarily physical features of x. This is exactly how instrumental goodness is distinguished from other forms of goodness that give reasons for use. Instrumental goodness may refer to properties of the user, but these are again physical properties; they do not refer to individuating properties of the particular person who uses the artefact. The goodness of a knife, for example, lies in the physical properties by which it enables its user to make cuts of a particular smoothness. This cannot be defined independently of the pressure that is exerted on the knife, which must fit the average human. A knife that is able to cut smoothly but only when pressed with a force of 100 kgf is not a good knife, and arguably not even a knife that is useful for cutting. A knife that makes smooth cuts when handled with normal pressure is, then, a good knife, certainly a knife useful for cutting, and someone’s wish to cut recommends the use of this knife. However, a knife may also have non-physical properties such that a particular person’s wish to cut recommends the use of this knife: it may be a knife that this person received as a present; its use reminds him of the giver, and so his wish to cut recommends the use of this particular knife. Recommendations like these, however, have nothing to do with the knife’s quality as a knife. It may be thought that these cases are blocked because reasons that are grounded in non-physical properties can only be reasons for particular persons, and the analysis, formally speaking, quantifies over persons. This is not so, however. Suppose there is only one knife left on earth. Then anyone’s wish to cut recommends the use of this knife, but this does not make the knife a good knife. The properties that ground a recommendation for the use of an artefact, where the evaluation of the artefact reflects the strength of the recommendation, must therefore be specified to be physical properties of the artefact.12 Similarly, a malfunction statements like ‘x is a malfunctioning K’ can be seen as a conjunction of the two statements ‘x is a K’ and ‘x is useless as a K’. The latter conjunct, again, carries the normative load of use-(non)recommendation:

12 Plausibly,

the introduction of modality in my account would be a way of escaping this.

938

Maarten Franssen

‘x is useless as a K’ expresses the normative fact that x has certain physical features f  that make it the case that a person p has a reason not to use x for K-ing. These results can be presented as forming a hierarchy of normative facts about objects in relation to the use we can make of them: x can be useful for K-ing or it can be useless for K-ing. If x is useful for K-ing, it can either be the case that x is a working K or that x is not a K, being either a natural object or an artefact designed for some other function than K-ing, but that it can serve as a K, or make a K.13 Some xs, again, that are Ks are good Ks, just as some xs that can serve as Ks make particularly good Ks. If, in contrast, x is useless for K-ing, it can be the case that x is not a K and does not make a K either, or that x is a malfunctioning K. What remains now is to see whether the mere ascription of function can be construed as a normative fact. Since a specimen of an artefact kind can be a malfunctioning specimen, the fact that x is a K does not imply that it has features such that someone’s wish to K recommend that this person uses x for K-ing. One way to go around this difficulty is the following proposal: 14 ‘x is a K’ expresses the normative fact that x has certain features f  that make it the case that a person p’s wish to K recommends that p uses a specimen of the kind to which x belongs for K-ing. If we take ‘the kind to which x belongs’ to be the functional kind K itself, this proposal captures the fact that ‘typically’ Ks have features that recommend their use for K-ing — they must have in order to classify as Ks — but it does not solve the difficulty that for any particular x that is a K the fact that it is a K is not sufficient for recommending its use. If we take ‘the kind to which x belongs’ to be a narrower kind, an artefact kind of the sort distinguished from a functional kind in Section 3, the proposal may be downright false, due to the possibility of type malfunction. Hitherto malfunction was treated as a phenomenon restricted to the specimen level, so to speak: individual tokens may malfunction, but the majority of the tokens making up an artefact kind will perform as designed. We can speak of type malfunction when an error in the blueprint defining the kind causes all specimens manufactured according to this blueprint to fail, although the design is a sufficiently correct elaboration of a particular operational principle for the kind so defined to count as a K-kind. In complex high-tech artefacts there are numerous opportunities for such errors. Besides, causes of malfunction may also creep up in the details left unspecified in the blueprint. The Russian Vostok-2M rocket, for example, functioned properly when tin-based solder was used in a particular filter in the fuel system, but malfunctioned when lead-based solder was used. On the 13 If x is a K, its contribution to the user’s K-ing is called its proper function. If x is not a K but used for K-ing, its contribution is said to be its accidental function. 14 See [Dancy, 2006] for a suggestion in this direction.

Artefacts and Normativity

939

other end of the spectrum, there is a category of junk consumer products where type malfunction seems to result from sheer lack of effort.15 My alternative proposal to interpret mere artefact-kind attribution as expressing a normative fact is as follows: ‘x is a K’ expresses the normative fact that x has certain features f  that make it the case that a person p has a reason to believe that a wish of p to K recommends that p uses x for K-ing. To my mind this proposal is the correct expression of the idea that Ks typically have features that recommend their use for K-ing. We can say that on this proposal the mere function ascription to a particular artefact matches the justification of the use of that artefact. The (true) statement that an object is a particular artefact thus expresses a normative fact of a theoretical kind only. I have argued that the features f  that make an artefact useful for K-ing must be physical features of the artefact. The features f  , on the other hand, that determine whether x is a K must include historical features referring to the intentions of the designer(s) of x if x is to be considered a K by design.16 Additionally, however, they must include physical features, because the intentions of a (would-be) designer cannot suffice to determine artefact-of-kind-K-hood, as was argued in section 3. Functional kinds and artefact kinds are basically success terms; we can hardly conceive of the existence of an artefact kind K if there are no objects having the physical properties that allow them to be used for K-ing.17 Still, the features f  need not include all physical features included in f  and may include additional ones. A proper analysis requires a detailed investigation of the way artefact kinds are defined and the precise way in which individual artefacts are members of such kinds. I will not attempt this analysis here. Obviously we are almost always right in expecting a K-by-design to be better for K-ing than a object that just happens to be useful for K-ing. Although there is nothing that necessitates this, it is very much in the nature of engineering design to analyze how the use of a particular object, whether natural or already artefactual, achieves a purpose of ours, and to conceive, on the basis of that analysis, ways in which this purpose-achieving can be improved, by modifying the object. Much that goes on in engineering design is redesign. So typically, for every object x that is useful for K-ing, there will at a later time be an object y that is a K by design and that is more useful for K-ing than x is. It might be argued, as an objection to the analysis presented in this section, that if ‘x is a good K’ and ‘x is a working K’, or even just ‘x is a K’, can all be interpreted as expressing normative judgements, then the distinction between 15 For the Vostok-2M, see the Wikipedia article http://en.wikipedia.org/wiki/Vostok rocket. With the category of junk consumer products I am acquainted through a pair of nail scissors with which it proved virtually impossible to cut a nail. 16 If the functional kind K is defined in terms of use or usability for K-ing, the features f  need not include historical features, but if K is additionally understood to have only artefacts as members, they do. 17 See also [Thomasson, 2007] and [Houkes, 2006] for similar positions.

940

Maarten Franssen

evaluative and descriptive statements evaporates. There is some truth in this observation, but I do not see it as counting against the analysis. The normative content of ‘This is a knife’ carries little force in the conditions of our daily lives, but there are circumstances, for example when one has to improvise with the few objects that are at hand, where ‘This is a knife’ has the same normative recommendatory force as ‘This is a good knife’ has in circumstances where there are several knives available. On the other size of the scale, judgements like ‘This is a good knife’ or ‘This is a poor knife’ may be considered to have more normative content then my analysis gives them. It seems to me, however, that my analysis is as far as we can go as long as these judgements are taken in isolation. If you have never used a knife before, how would you decide that the particular one you are now handling is a good knife? Was Bell’s telephone a good telephone, or Edison’s phonograph a good phonograph? These considerations suggests that a judgement of the form ‘x is a good K’ has a meaning that goes beyond the rather anaemic ‘x is useful for K-ing’ only in a comparative setting. At many occasions when we use the expression, we probably do with an implicit background of other Ks in mind. The meaning of ‘x is a better K than y’ can be made much more precise, and its normative content can be specified as saying that anyone’s wish to K recommends that she uses x rather than y for K-ing.18 Still, however, we cannot easily strengthen this into saying that someone’s wish to K requires that he or she uses x rather than y. Minimally, both x and y must be available and there must be no institutional constraints against the use of x. And even then, a person with particular abilities or handicaps might be instrumentally better off using y rather than x.

5

UNDERSTANDING WHAT ARTEFACTS OUGHT TO DO

I turn, finally, to the third sort of statement distinguished in section 1, to which ‘This device ought to do x’ and ‘This device is supposed to do x’ belong. The way such sentences are used is obviously related to the two previous sorts: If x is a good (or poor) K, then it is said that x does, or does well (or poorly) what x, as a K, ought to do, or what Ks ought to do, and if x is a malfunctioning K, we say that x fails to do what x, as a K, ought to do, or what Ks ought to do. My analysis of the evaluative judgements in the previous two sections, however, does not imply much about how to think of the sentence ‘x, as a K, ought to do B’ as making a normative claim. This statements are deontic in form, but clearly it cannot be interpreted as literally deontic, and under that interpretation as normative ones, as sketched in section 2. People can be under an obligation to do something, for having conclusive reasons to do it, but not objects. These statements are, therefore, to be taken as a way of speaking. I propose that such statements, whether they are phrased in terms of ‘supposed’ or of ‘ought’ or ‘should’, are typically to be understood as stating that certain 18 Cf.

[Von Wright, 1963].

Artefacts and Normativity

941

expectations are in order. That is, by saying that an item ‘is supposed to show behaviour B’ or, equivalently, ‘ought to show behaviour B’, we mean that the speaker is justified in expecting behaviour B to occur. There are two quite different ways in which such expectations can be justified: the speaker’s expectations can be epistemically justified or can be morally justified. Epistemic justification involves the standard considerations of empirical evidence and logical implication. Being morally justified in expecting something to occur can alternatively be expressed by saying that one has a right to it that it occurs, for example because this occurrence was part of the content of a promise.19 As the following example shows, the epistemic and moral readings of ‘ought to’ statements directed at artefacts are often intermingled and difficult to separate. Suppose that I step into my car (and that I am in a philosophical mood). While turning the ignition key I could say: ‘It ought to start now.’ Partly I express an expectation that I consider justified. The basis for this justification, my knowledge, is in this case not of a technical nature – I know hardly anything about engines — but consists of my largely experiential knowledge as a member of a rich and highly technological society. Partly, however, my ‘ought to’ statement also expresses my belief that I am entitled to its starting. If it were so unreliable that I couldn’t expect it to start, it would not be worth the amount of money I paid for it; I certainly would not have paid that much had I known before that I could not count on its starting. So I feel I can hold the manufacturer, or, in the case of a second-hand car, the car dealer, to his or her part of a deal we made. Now suppose that my car does not start. I take it to the garage, where a mechanic is busy with it for some time. Then he wipes his hands, takes the driver’s seat, and while turning the ignition key he says: ‘It ought to start now.’ By that he means that as far as he understands the car, he has repaired what was obviously faulty and he sees no further reason why it shouldn’t start. This is a purely epistemic ‘ought to’. Next I come to collect my car. I pay and prepare to drive away. While I turn the ignition key I say: ‘It ought to start now.’ This is primarily a moral ‘ought to’; I am claiming a right to its starting. After all, when a garage returns a car to its owner as repaired there is the (implicit) promise that it is now in working order. By saying that the car ought to start now, I express my conviction that the garage should keep its promises, or should stick to a fair deal: money for repairs. However, an epistemic ‘ought to’, again expressing an expectation that one considers justified on the basis of probable though not fool-proof regularities of a psychological and sociological rather than a technical nature, would be just as well in order. You will wonder how a garage can do lousy repair jobs and manage to survive in the market place. 19 Cf. the view of Raz [1975, p. 31] that ‘This stone ought to break the window’ is logically equivalent to ‘There is reason to believe that this stone will break the window’. Whether being justified in believing something can be explicated exhaustively in terms of having reasons to believe is something I briefly discuss in the final section. Raz does not discuss a ‘moral’ reading of such statements.

942

Maarten Franssen

An important difference between the two types of ought to statements is that the moral ones apply only to artefacts, whereas the epistemic ones extend to any object concerning the behaviour of which we form expectations. When we say that object x ‘ought to’ show behaviour B, we generally mean that, although we cannot be certain that behaviour B will occur, we are justified in our expectation that it will occur. Take the following example: ‘When I let go of this stone (holding it under water), it ought to sink.’ But some unexpected stream could prevent it from sinking, or it could turn out to be pumice, which can float on water. Depending on the context, it may not even seem out of order to say: ‘When I let go of this stone (holding it up in the air), it ought to fall’. Perhaps I vaguely suspect that some trick is going to be played upon me. Or it may be that I myself intend to demonstrate to a class of students that one should never hold an empirical statement to be true with absolute confidence. Artefacts are, therefore, not special in that we have special expectations concerning their behaviour; only the grounds on which we form our expectations are different for artefacts as compared to natural objects. Undeniably both on the epistemic and the moral reading the ‘ought to’ statements at issue have a normative dimension. On the epistemic reading, the normativity lies in the notion of being justified to have a certain belief, and on the moral reading it lies in the notion of having a right to a particular behaviour in other people, and therefore of these other people having certain duties. Neither of these readings reveal something particular about technical artefacts. The epistemic ‘ought to’ statements reflect aspects of the rationality of belief formation in general, and the moral ‘ought to’ statements reflect aspects of the normative practice of promising and contracting among people. Indeed, when, on the moral reading of these ‘ought to’ statements, we see them as stating that the person who utters the statement, the user of the artefact, has a right to the artefact’s being able to perform its function, this implies that there rests a duty to see to it that this is the case on the shoulders of one or more other persons. Typically the designer or the manufacturer or retailer — or several of them — are responsible for any disappointed expectations, meaning that they ‘ought’, in the full-blooded normative sense of having conclusive reasons, to do something about it — either replace a non-functioning artefact by a working one or refund the user, and compensate those who were harmed as a direct consequence of the artefact’s malfunctioning. This ‘ought’ consists partly of moral reasons — a straightforward moral obligation to repair harm and losses caused by one’s reproachable behaviour — but also of prudential reasons, since disappointed clients will usually take legal action against non-delivering manufacturers or retailers. (See also the chapter by Pritchard, especially on standards of care.) Clearly, all this only applies to an, unsuccessful, attempt to use an artefact for the function it was designed for and sold for, and to use it according to the instructions for use and in the circumstances specified therein. If someone uses an artefact according to his own plan, based on his own inquiry after the artefacts

Artefacts and Normativity

943

capabilities, then if the plan misfires, the user can only blame himself, if anyone, for holding expectations that turn out to have been insufficiently justified. The notion of blame can help to understand the use of deontic language also for epistemic ‘ought to’ statements, although there are no obligations involved. We form our beliefs on the basis of our interaction with other people and with nature, and we expect the answers that we receive to be trustworthy in either case. We seem to hold nature to her part of a deal we supposedly made with her when we questioned her, to use Francis Bacon’s metaphor, just as much as we hold other people to the truth of what they are telling us. Because we did our best to check whether a belief about her is true, nature should see to it that it is indeed true when it seems we receive a positive answer. I suggest this is the reason why we use one expression to refer to two quite different situations. In forming our beliefs about the world and acting upon them, we always run the risk of being let down, either by our fellow men and women or by nature; that is apparently how we feel it. Although I think my interpretation of ‘is supposed to do’ or ‘should do’ statements about artefacts as expressing that certain expectations are in order fits the majority of cases, I am not claiming that such statements must invariably be interpreted in this way. Since justified moral expectations often, if not always, imply that there are duties resting upon the shoulders of certain people, ‘is supposed to do’ or ‘should do’ statements are sometimes simply a circumspect or polite way of saying to a particular person that he or she should do something, or at least has a strong reason to do something. Additionally, these statements sometimes merely express a necessary condition, being shortened forms of statements like ‘In order to fall in category X this artefact should do y’, or ‘In order to count as an X, this device should be like y’. If we tend to use ‘should’ here instead of ‘must’ (but seldom ‘ought to’), this may be because they occur in contexts where it is presupposed that there is someone who has the ability and/or the responsibility to make the thing in question do y or be like y.20 6

BEYOND TECHNICAL ARTEFACTS: THE NORMATIVITY OF FUNCTIONAL ITEMS IN GENERAL

In section 1 I said that any explanation of why artefacts support evaluative statements must be able to explain why specific biological items are thought to support them as well. This seems not too difficult for statements like ‘This is a good watchdog’ or ‘She is a poor milk-cow’. We could interpret these statements as analogous to a statement like ‘This rock makes a good hammer’: as statements that take the physical properties of particular natural objects to give us reasons to use them for a purpose. The dog has properties that make it fit to be used for watching over one’s house, and the cow has properties such that one has a reason not to use it for obtaining milk (although you still might have an overall reason 20 Cf. on these ‘should/ought to’ statements for artefacts Vaesen [2006] for a different account. Vaesen distinguishes between ‘should do’ and ‘should be’ statements for artefacts, but I think this distinction is largely illusory, since both are about properties an artefact may have or lack.

944

Maarten Franssen

to do so if no other source of milk is available and you are, for a good reason, in need of milk). It happens only very rarely, however, that someone uses a dog or a cow she just runs into in the fields for keeping a watch (and then release it) or for getting milk to put in the coffee. Watchdogs and milk-cows are better seen as genuine artefact kinds, obtained by meticulous breeding and training from the ‘natural material’. A statement like ‘This is a good watchdog’ is therefore quite similar to a statement like ‘This is a good water pump’. This will not work for statements like ‘She has a good heart’ or ‘His liver is bad’. It might perhaps be said that our heart or liver serves a purpose of ours, that of staying alive, but if it does, it is not by being used by us. We do not have voluntary control over our vital organs. We do have voluntary control over some parts of our body, such as our arms and legs, and we can rightfully say ‘Use your hands’, but exactly the parts that we have control over are rarely judged as ‘good’ or ‘bad’. We can say that someone has good eyes and we can tell this person to ‘use his eyes’, but these are not properly related: the goodness of eyes concerns sharpness of vision, most of all, while the using of your eyes concerns direction of gaze (and attention, but that is more a brain thing). What links a statement like ‘She has a good heart’ to ‘This is a good water pump’ is that both hearts and water pumps are items that are said to have a function. The function of the heart in a vertebrate body is to circulate the blood, the function of a water pump is to circulate water in whatever system it is a component of. However, it is a mistake, in my view, to take the concept of function as the basic one and as inherently supporting normative statements. In the case of artefacts, an artefact’s function is relegated to it by our using it as such. To have the water circulate in the system at hand is my purpose, and primary, and to place the water pump in the system in order to let me achieve my purpose is my using the pump, which comes next. If we ascribe to a water pump that is still on the shelf the function to circulate water in an as yet nonexistent system, we can do so because my purpose and my system are in a sense already present when the pump is designed, in the form of the functional requirements guiding the pump’s design. To be sure, this picture involves a considerable amount of simplification. In our society there are many artefactual items that are generally thought of as ‘being for’ a purpose and ‘for being used’, but where it is doubtful that all relevant forms of use were represented during the design phase. (Examples are large-scale, complex systems, such as infrastructures, which are currently much studied under the name of socio-technical systems; see the chapter by Bauer en Herder in this Volume). Additionally there are social items that can be seen as products of design, ranging from more or less concrete, such as money, to completely abstract, such institutions, sets of rules, laws, and so forth. Insofar as it makes sense to say these are used, which is particularly doubtful in the case of rules and laws, they involve many users whose many forms of use develop in the course of time, making it impossible to represent all of it adequately during their design even if it

Artefacts and Normativity

945

were true that these socio-technical and social ‘artefacts’ are designed just as cars and computers are. However, their design is at most piecemeal and stretches out over long periods of time, often in reaction to changes in the way they function induced by changed modes of use. The intricacies and difficulties, especially from the point of view of (engineering) design, that they lead to have only recently begun to receive serious scholarly attention and I can do no more here than point them out. The explication of what we mean by the function of an object in intentional terms is certainly unavailable in the case of biological organs and other biological items to which functions are ascribed, such as traits and forms of behaviour, since there are no purposes of and uses by intentional beings there. This extends to its explanatory use in so-called functionalist theories in sociology and social anthropology, since such explanations bypass completely the intentions of individual people. It depends on the precise way in which the functions of biological or social items are grounded in other terms than purposes and uses whether function attributions to natural or social items support normative statements of the various sorts distinguished here. There is, however, no consensus on the correct theory of the functions of biological items, and whether the attribution of functions to social items is ever in order is highly contested. Some theories of function deny that biological items can malfunction or that it makes sense to say, of a diseased liver, that although it does not metabolize as other livers do, it ‘ought to do so’. Theories of function that do allow this, on the other hand, would allow much more of such statements than biologists are prepared to utter, indicating that the grounding relation between the attribution of function and the assertability of a normative statement is not a ‘natural’ one. Indeed, as I have argued in the previous section, it is clear how we should interpret ‘ought to’ statements directed to artefacts, but on this interpretation they are normative in a way that does not reveal anything about artefacts in particular, and this extends without modification to similar statements directed to biological or social entities: they must be ‘read epistemically’ as expressing justified expectations of behaviour, and in this way they do not reveal anything about functional items in particular. Insofar as theories of function support malfunction statements about biological or social entities they are, in my view, unable to specify how these statements are normative statements. In [Franssen, 2009] I argue this point in more detail. How to explain, then, that statements like ‘She has a good heart’ or ‘His liver is bad’ seem so much in order to us? It is to be noted that we make such statements almost exclusively in the case of the organs of humans. One suggestion is to see them as referring to use, but not a straightforward use for the ‘purpose’ supported by their biological function, presumably ‘staying alive’ for most organs. Instead we could interpret ‘She has a good heart’ as being short for ‘She has a good heart for being or becoming a sportswoman’, similar to saying of a particular rock that it is ‘good for hammering’. It is difficult to see, however, what ‘He has a good liver’ would be short for; perhaps ‘He has a good liver for becoming an alcoholic’ ? I think this suggestion fails. Instead, my suggestion is that these

946

Maarten Franssen

normative judgements reflect a belief or hope that the quality of performance of these organs is to some extent under our control. A heart is bad because we can conceive of improving its performance, or of replacing it by a better one. And in a way everyone has the possibility to do so, through exercise or the abstinence from bad habits like smoking and excessive drinking. But thanks to technological developments, the scope of this possibility has dramatically increased. We can boost the heart’s function by implanting a pacemaker or by coronary surgery, or we can replace it altogether through a heart transplant. This situation makes our bodily organs increasingly similar to the components of a complex technical artefact, to be replaced when broken or when an ‘update’ becomes available. Thus, naturally, normative artefact talk seems to us to be in order for these organs as well, even though organs are not related to uses and purposes of people in the same way as artefacts are, if they are at all so related. 7

NON-INSTRUMENTAL VALUE IN ARTEFACTS

Up till now, evaluative statements referring to artefacts have been interpreted exclusively as addressing the instrumental value of these artefacts. The goodness or poorness of a particular artefact, say, a water pump, is its goodness or poorness as an instrument. Can artefacts also figure in evaluative statements that concern non-instrumental value? Two sorts of values for which this is prima facie plausible are aesthetic value and moral value. (Cf. the chapters by Van de Poel and by Schummer et al. in this Volume.) A technical artefact may be judged beautiful or ugly, and such judgements seem to play an important role in large areas of engineering design, in particular industrial design and architecture. A technical artefact may also be judged morally good or, perhaps more commonly, morally bad. Prime candidates are weapons, in particular the ones that aim to inflict particularly horrible wounds, or instruments of torture. How do these judgements fit into my account? In Section 2, I mentioned moral and aesthetic value as two sorts of value technical artefacts may have, but these forms relate to artefacts in a quite different way. Aesthetic value is traditionally attributed just as much to nature and to natural objects (animals, flowers, crystals, landscapes) as to artefacts. Many classical theories of art even hold that nature is the source of all beauty and that human art aims at imitating the beauty of nature. Moral value, on the other hand, is, certainly after Darwin’s theory of evolution, commonly denied to natural facts and objects (although this position has recently come under attack from environmental philosophers). Sticking to the account of normativity adopted here, for these evaluative judgements to be normative they must be interpretable as indicating a way in which properties they have are relevant to the question what to do, where the relevance is now not of an instrumental sort, the value that chosen means have for the realization of given ends. An account that fits in is the proposal by Thomas Scanlon [1998] that for an object to have value is to have features such that one has reason

Artefacts and Normativity

947

to adopt a certain attitude toward it, or to choose a certain course of action with respect to it. If an object is good, one has reason to promote it, to support it, to protect it, to admire it, and the like, and if an object is bad, one has reason to avoid it, to attack it, to ignore it, to deplore it, to loath it, and so forth. This analysis applies to aesthetic value and to moral value. It may also make clear how one can hold that “instrumental value is not a form of value at all”, as Dancy [2000b, p. 159], following Alf Ross, does. An object has instrumental value if its properties give us reasons, or recommendations, to use it in a particular way. This use it is, as a kind of action, entirely neutral; it is neither a form of promoting or supporting the artefact, nor a form of avoiding or attacking it, so the mere act of using an artefact brings to light neither a positive nor a negative value the artefact has.21 Technical artefacts, then, have aesthetic value in the same way works of art have, and their having this value means that we have reasons to admire them, protect them, and so forth. The distinction between the beauty or ugliness of technical artefacts as compared to works of art must lie in the features that give us these reasons. On this, the analysis is silent. Some features will be shared, such as a gracious form, harmony in dimensions, colour scheme. There are also features that are typical for technical artefacts, for instance the elegancy achieved in balancing several requirements, or in physically realizing a particular requirement. The moral goodness or badness of technical artefacts, on the other hand, is more difficult to establish. We may feel disgust when seeing the gun that a gang member used to kill a beloved one with, but how to establish that we have a reason to loath it or avoid it? And even if we can see this reason, can this reason be extended to other guns, such that we are saying something about guns as such, and not about one particular gun? The gang member may have been prevented from killing more people because he was shot with another gun before he could do so. Do we have reason to fight for the abolition of a particular weapon because of the damage it inflicts, if by using it our country was able to defeat a brutal regime and thereby bring to an end a period of slaughter and misery? This is the well-known problem of the supposed neutrality of technology as a means, and the idea that value resides exclusively in the ends that are achieved by using the technology. (For a discussion of the neutrality thesis, see also the chapter by Van de Poel in this Volume.) My own suggestion, which I will not elaborate here, is that technical artefacts can be called bad in a moral sense if its functional requirements, the characteristics that in a sense define it, explicitly refer to specific morally bad states of affairs as goal 21 In

Section 4, I said that someone may have a non-instrumental reason for using an artefact, for example, a sentimental reason for using an artefact that was given to the user as a present. If this reason for use reflects the value the artefact-present has, then use may not be the neutral action I claimed it is. However, one can have such a reason only if one also has an instrumental reason: one has no sentimental reason to use a malfunctioning present, however much one values the present and the person who gave it. So the sentimental reason is at most an additional reason for use, and is better interpreted as a reason to cherish-by-using the artefact, given that one has an instrumental reason for using it. This the right sort of positive attitude to satisfy Scanlon’s ‘buck-passing’ account of the artefact’s value, in the non-instrumental, ‘true’ sense of value.

948

Maarten Franssen

states to be realized by using the artefact, such that it will be optimized, through the accepted methods of engineering design, to realize precisely these outcomes. It is, I would guess, less controversial to establish the moral value of artefacts in a comparative sense only, that is, of two artefacts of the same functional kind, to state that one is morally better or morally worse than the other. If we have two instrumentally equivalent electric saws, one of which is more dangerous than the other by presenting some unexpected but easy-to-run-into way of getting electrocuted, we have good reason to call the more dangerous saw morally worse than the safer one. And similarly, in a lighter vein, if a public trash receptacle is designed such that it evidently invites more people more often to leave their trash in it than another trash receptacle does, we can call the former morally better than the latter. These examples may suggest that instrumental value is independent of true value, be it moral value or aesthetic value or still another form. It does not matter for the moral inferiority of the dangerous saw whether or not it is the instrumentally better or poorer saw of the two. To believe that instrumental value is unrelated to the value possessed by states of affairs or objects would be a serious mistake, however. One can, by using an instrumentally poor chisel or paint brush for a particular repair job, bungle the job and ruin a precious (in either a monetary or an emotional sense) piece of furniture. If, given that one is going to use an artefact and given that one is justified in doing so, one has the opportunity to pick one from a whole bunch, there will generally be strong reasons, if not a duty, to pick an instrumentally good one. 8

CONCLUSIONS AND FINAL CONSIDERATIONS

In this chapter I have shown that technical artefacts have in a precise sense a normative dimension, inherent to their status of objects designed to serve a purpose in being used. This normative dimension is not, however, typically expressed by the sorts of statements that, on the face of it, seem to be the most straightforwardly normative statements in which artefacts figure, such as ‘in circumstances C, artefact x ought to show behaviour B’. Instead, the normative side of artefacts is represented by statements like ‘artefact x is a good K’ or ‘artefact x is malfunctioning’. I have shown how all statements that fall into these categories allow for a normative reading on a general characterization of what normativity is about. The result is a very general analysis. Its force is, in my opinion, that it brings to the fore important general characteristics of the class of artefacts. It is a consequence of the fact that artefacts are designed as kinds, as blueprints (even if only a single copy is manufactured) and that they are designed to serve a purpose, by which they must have some characteristic features if they are to belong to a functional kind in the first place, that their use is recommendable to us. Exactly this is where artefacts differ from natural objects. It is not so that ‘This is a rock’, apart from the descriptive statement we usually take it for, can also be interpreted as a normative statement that says that x has certain features (the typical prop-

Artefacts and Normativity

949

erties of rocks) and that because of these features a wish to do a particular thing recommends the use of the rock to achieve it. The properties of rocks do not fix the sort of uses rocks can be put to. Some rocks make good paper weights, but most rocks do not. Some rocks are good for hammering, but most rocks not. Some rocks are good for smashing shop windows, but most are not. Although a particular rock may have properties such that a person with a need to cut a rope is recommended to use this rock for (rope-)cutting, no natural rock kind exists to which it belongs, such that its belonging to this kind can be taken to express these rope-cutting qualities. There are no physical laws that all and only ‘hand-sized rocks with sharp edges’ answer to. If only because the size of our hands need not be constant over the ages. The medal also has a reverse. The normative content of the statements I have analyzed is meagre, in the sense of implying very little on what people ought to do. This, however, is what is to be expected from an inquiry into the relations that artefact use and artefact performance have to normativity in general. I have charted how artefacts fit into the pattern of what people do in the world and how they, in that way, partake in the normative considerations that apply to human action in general. A more substantial shortcoming is that I have concentrated on the normative side of artefacts once in existence, which, therefore, mostly concern the users of these artefacts. Much less has been said on the coming into existence of artefacts, that is, on their design and manufacture, which would concern the designers and producers of artefacts. This part of the normative dimension of artefacts is dealt with in the chapters on rationality in engineering design by Kroes, Franssen and Bucciarelli and on values in engineering design by Van de Poel. The chapter by Pritchard focuses on the moral obligations and considerations that apply to the designers and producers of technical artefacts. I will close this analysis of normativity in relation to artefacts by pointing out some difficulties in the adopted philosophical account of normativity that have come to light in the course of applying it for a clarification of the normative dimension of technical artefacts. The philosophical account took the notion of a reason, or of the favouring relation that a reason expresses, as primitive. An important issue is whether it is uniquely primitive, meaning that all other normative concepts can be understood in terms of reasons. Raz, Scanlon, Broome and Dancy [2005, n. 14] all seem to think so. Now if one is prepared to accept primitive, i.e. irreducible, normative concepts (and I will ignore completely the controversy between naturalists and non-naturalists), it may seem an attractive idea to have only a single primitive concept grounding our normative talk. The consequences could well be less attractive, however. Broome’s analysis of the various ways in which reasons can affect us has led him to tear the normative domain in two halves, separating practical and theoretical rationality on the one hand from the deontic on the other hand. In rationality only requirements and recommendations are at work, or what Dancy calls complex duties, no straightforward reasons and oughts. I have analyzed instrumental value as belonging here, by using Broome’s notions of normative requirement and normative recommendation. Rationality puts restrictions

950

Maarten Franssen

on the totality of our beliefs and intentions, by pointing out the particular beliefs and intentions that other beliefs and intentions (and ends, I would add) can and cannot be combined with. Never can a reason for a particular action in isolation be derived from this, because rationality cannot look beyond the set of beliefs and intentions to see whether they are, one by one, reasonably grounded. Broome, then, feels forced to doubt whether rationality is normative, that is, reason-giving, at all (e.g. [Broome, 2005]), and he thinks that G.E. Moore may well be right that “you can never know for certain what you ought to do” [1999, p. 93]. The reasons for our actions lie in the facts about the world — which includes the facts about the sort of organisms we are — and they lie there mostly hidden. Dancy, whose treatment of the issues related to normativity is by far the most comprehensive one, is not prepared to accept Moore’s verdict that we may well be unaware of most of the reasons we have because we lack knowledge of the relevant facts. He has at least two strategies for a way out. One is to introduce the notion of an ‘epistemic filter’ [2000a, p. 138], which blocks facts that we cannot possibly have knowledge of from forming reasons for us. It is not determined by the facts alone, however, what we can have knowledge of, and Dancy concedes that his epistemic filter is partly normative in character. But in this way he invites normativity to a place where it will stay out of reach of an analysis in terms of reasons alone. Another strategy is meant for those cases that pass the epistemic filter. Suppose that one has a conclusive reason not to take flight UA175 from Boston to Los Angeles, since the aeroplane is going to be hijacked and deliberately crashed, and although one does not know this, it is assumed (just as an example) that one could have know, by paying careful attention, by eavesdropping on all conversations going on around one, and so forth. Few people, however, would deny that one was justified in taking the flight. Dancy accepts this intuition, but says that this notion of being justified belongs to the evaluative rather than the deontic part of the normative. Although our unfortunate air passenger, strictly speaking, took a wrong course of action, we will not blame him for doing so, for instance by holding him responsible for the grief his death caused to his relatives and friends. Relegation of being justified to the evaluative domain does not seem to make it a less normative notion, however, but Dancy does not indicate how the evaluative notion of being justified can be explicated in terms of reasons. What is more, notwithstanding his general characterization of normativity, in his [2005], as the way the facts give us reasons for actions (broadly conceived), Dancy has elsewhere [2000b] given arguments for being sceptical about Scanlon’s proposal for an overall reduction of the evaluative in terms of reasons. There is a proposal for analyzing ‘being justified to’ in terms of reasons, going back, again, to Raz, which says that what it means for someone to be justified in doing something is for this person to have a reason to believe that he or she has a reason to do this thing. In Section 4, in my discussion of the normative interpretation of mere function attribution, I also have analyzed ‘being justified to’ as meaning ‘having a reason to believe that one has a reason to’. This strategy may well be faulty, however. If the facts make it the case that either one has a reason

Artefacts and Normativity

951

to do something, say, X, or one does not have a reason to do X, then if one does not have a reason to do X, the belief that one has a reasn to do X is false. And if the idea that the facts make it the case that one has a reason to do or not to do something extends to reasons for the adoption of a belief, then, plausibly, one can only have reasons to believe things that are true, never to believe things that are false. Therefore, if one does not have a reason to do X, one cannot have a reason to believe that one has a reason to do X. Since we would be very hard-pressed to give up our conviction that one may, on the basis of the available evidence, perfectly be justified in believing that one has a reason to do X, to be justified in believing that one has a reason cannot be equated to having a reason to believe that one has a reason. My analysis, in Section 5, of statements saying that an artefact ‘ought to do such-and-such a thing’ also makes fundamental use of the notion of being justified, and although I have hardly tried, I do not see it could be all rephrased in terms of reasons. Additionally, the normativity of theoretical and practical rationality, which Broome, after recasting rationality in terms of complex requirements instead of reasons, now despairs of saving, may require a new analysis in terms of justification. Such complex requirements or duties also pose a problem for Dancy, since, in my opinion, he faces great difficulties in showing that they are grounded in the facts in the same way as simple reasons and duties. The present analysis of the normative judgements supported by artefact therefore strongly suggests that there is a place for another normative notion, ‘being justified to’, which is independent of and not reducible to the (primitive) normative notion of a reason.

BIBLIOGRAPHY [Broome, 2000] J. Broome. Normative requirements. Ratio 12, 398-419, 1999, repr. in Normativity, J. Dancy, ed., pp. 78-99. Basil Blackwell, 2000. [Broome, 2005] J. Broome. Does rationality give us reasons? Philosophical Issues 15, 321-337, 2005. [Dancy, 2000a] J. Dancy. Practical Reality. Oxford University Press, 2000. [Dancy, 2000b] J. Dancy. Should we pass the buck? In The Good, the True and the Beautiful, A. O’Hear, ed., pp. 159-173. Cambridge University Press, 2000. [Dancy, 2005] J. Dancy. ‘Non-naturalism’, in The Oxford Handbook of Ethical Theory, D. Copp, ed., pp. 122-145. Oxford University Press, 2005. [Dancy, 2006] J. Dancy. The thing to use. Studies in History and Philosophy of Science 37, 58-61, 2006. [Dipert, 1993] R. R. Dipert. Artifacts, Art Works, and Agency. Temple University Press, 1993. [Franssen, 2006] M. Franssen. The normativity of artefacts. Studies in History and Philosophy of Science 37, 42-57, 2006. [Franssen, 2008] M. Franssen. Design, use, and the physical and intentional aspects of technical artifacts. In Philosophy and Design: From Engineering to Architecture, P.E. Vermaas, P. Kroes, A. Light and S.A. Moore, eds., pp. 21-35. Springer, 2008. [Franssen, 2009] M. Franssen. The inherent normativity of functions in biology and technology. In Functions in Biological and Artificial Worlds: Comparative Philosophical Perspectives, U. Krohs and P. Kroes, eds., pp. 103-125. MIT Press, 2009. [Gert, 2004] J. Gert. Brute Rationality. Cambridge University Press, 2004. [Hansson, 2006] S. Hansson. Category-specified value statements. Synthese 148, 425-432, 2006.

952

Maarten Franssen

[Houkes, 2006] W. Houkes. Knowledge of artefact functions. Studies in History and Philosophy of Science 37, 102-113, 2006. [Houkes et al., 2002] W. Houkes, P. E. Vermaas, C. H. Dorst, and M. J. de Vries. Design and use as plans: an action-theoretic account. Design Studies 23, 303-320, 2002. [Raz, 1975] J. Raz. Practical Reason and Norms. Hutchinson, 1975. [Scanlon, 1998] T. M. Scanlon. What We Owe To Each Other. The Belknap Press of Harvard University Press, 1998. [Sorabji, 1964] R. Sorabji. Function. Philosophical Quarterly 14, 289-302, 1964. [Thomasson, 2007] A. L. Thomasson. Artifacts and human concepts. In Creations of the Mind: Theories of Artifacts and their Representation, E. Margolis & S. Laurence, eds., pp. 52-73. Oxford University Press, 2007. [Vaesen, 2006] K. Vaesen. How norms in technology ought to be interpreted. Techn´ e: Research in Philosophy and Technology 10, 117-133, 2006. [von Wright, 1963] G. H. von Wright. The Varieties of Goodness. Routledge & Kegan Paul / The Humanities Press, 1963.

PROFESSIONAL STANDARDS IN ENGINEERING PRACTICE Michael S. Pritchard

1

INTRODUCTION

As professionals, engineers are expected to commit themselves to high standards of conduct. The Preamble of Code of Ethics of the National Society for Professional Engineers (NSPE) puts it this way: Engineering is an important and learned profession. As members of this profession, engineers are expected to exhibit the highest standards of honesty and integrity. Engineering has a direct and vital impact on the quality of life for all people. Accordingly, the services provided by engineers require honesty, impartiality, fairness, and equity, and must be dedicated to the protection of the public health, safety, and welfare. Engineers must perform under a standard of professional behavior that requires adherence to the highest principles of ethical conduct. Although this Preamble insists that such conduct is expected of engineers, this is not a predictive statement about how engineers, in fact, conduct themselves. By and large, it is hoped, engineers do adhere to high principles of ethical conduct. However, the Preamble is a normative statement, a statement about how engineers ought to conduct themselves. This is based on the impact that engineering has on our quality of life. This impact is the result of the exercise of expertise that is the province of those with engineering training and experience. Such expertise carries with it professional responsibility. To talk about professional responsibility in this way is to enter the arena of ethics, or morality.1 Standards for engineers may be articulated in codes. These codes can be broad statements of principle, such as are found in engineering codes of ethics. Or they may be quite specific and prescriptive, such as building codes. Many engineering standards are also understood in terms of the “accepted practice” of engineers, whether formally stated or not. In each case we can ask what, if any, underlying moral basis engineering standards have and what these standards contribute to our understanding of the moral responsibilities of engineers. 1 In this paper ‘ethics’ and ‘morality’ will be used interchangeably. Neither textbooks nor ordinary language exhibit patterns of use that clearly distinguish them.

Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

954

Michael S. Pritchard

William F. May points out the seriousness of the responsibility that comes with professional expertise. Noting our growing reliance on the services of professionals whose knowledge and expertise is not widely shared or understood, May comments: [The professional] had better be virtuous. Few may be in a position to discredit him. The knowledge explosion is also an ignorance explosion; if knowledge is power, then ignorance is powerlessness. [May, 1998, p. 408] The knowledge that comes with expanding professional expertise is largely confined to specialists. Those outside these circles of expertise experience the ignorance explosion to which May refers. This includes the general public, as well as other professionals who do not share that expertise. May concludes: One test of character and virtue is what a person does when no one else is watching. A society that rests on expertise needs more people who can pass that test. [May, 1998, p. 408] May’s observations apply as much to engineers as accountants, lawyers, doctors, and other professionals. What this means is that, in its ignorance, the public must place its trust in the reliable performance of engineers, both as individuals and as members of teams of engineers who work together. It is not just the public that must place its trust in the reliable performance of engineers. Engineers’ employers, colleagues, and co-workers need to, as well. Thus, the need for mutual reliance and trust is pervasive. Fortunately, our common morality provides us with a resource that commends the establishment of such trust and enables us to understand, support, and critically evaluate standards for professionally responsible behavior. Thus, a good place to begin is with a discussion of common morality. This will help set the stage for a consideration of the function and limitations of codes of ethics regarding professional standards for engineers. Next will be a discussion of regulatory standards, commonly accepted standards of practice, and the broader notion of a “standard of care,” commonly invoked in judicial settings. Throughout it will be clear that engineers must rely on good judgment rather than merely algorithms. This will also be evident in the discussion of relationships between professional standards, on the one hand, and engineering imagination, innovation, and design, on the other. Finally, questions regarding the scope of professional standards will be considered, particularly in light of the rapidly growing international setting of much engineering practice. 2

COMMON MORALITY

Philosopher Bernard Gert observes that, regardless of our individual and cultural differences, there are some universal features of human nature that provide the basis for a system of common morality, such as our fallibility, rationality, and vulnerability [Gert, 2004]. He is careful to point out that common morality is

Professional Standards in Engineering Practice

955

not a system derived from his or any other philosopher’s moral theory. Common morality precedes theories that attempt to describe or evaluate it. Although it does reflect general acceptance by thoughtful people, common morality does not depend on the theorizing of moral philosophers. In fact, Gert says that common morality is accepted in all philosophical theories of morality [Gert, 2004, p. vii]. Gert characterizes common morality in terms of a set of moral rules and moral ideals. He does not claim that we explicitly endorse these rules and ideals as he formulates them. Rather, his account is best understood as a rational reconstruction of basic features of our moral lives — an account that attempts to represent faithfully something implicit in our moral lives. Gert’s list of rules and ideals is offered as a comprehensive representation of the central features of ordinary morality. To illustrate, Gert’s moral rules are: • Do not kill. • Do not cause pain. • Do not disable. • Do not deprive of freedom. • Do not deprive of pleasure. • Do not deceive. • Keep your promises. • Do not cheat. • Obey the law. • Do your duty. None of these rules is “absolute.” Each has legitimate exceptions. Sometimes they even conflict with one another. However, violations of these rules require a justification that can be publicly accepted by all reasonable persons. That there can be justified departures from a moral rule is a central feature of common morality. In controversial cases, reasonable persons might not come to the same conclusions about what to do. But, given basic agreement on the facts in particular situations, shared acceptance of the moral rules can be expected to result in widespread agreement on most matters. All moral agents, says Gert, agree that killing, causing others pain or disability, and depriving others of freedom or pleasure are morally wrong without some justification. This is in contrast to, for example, taking a walk or not taking a walk, neither of which normally requires any justification. Likewise, all moral agents agree that deceiving, breaking promises, cheating, breaking the law, and neglecting duties are in need of moral justification. Gert concludes:

956

Michael S. Pritchard

The claim that there are moral rules prohibiting such actions as killing and deceiving means only that these kinds of actions are immoral unless they can be justified. Given this understanding, all moral agents agree that there are moral rules prohibiting such actions as killing and deceiving. [Gert, 2004, p. 9] Given something like Gert’s account of common morality, the main ingredients of the codes of ethics of professional engineering societies should not be surprising. Nearly all of them identify the protection of public health, safety, and welfare as the paramount duty of engineers in the course of their engineering work.2 They also emphasize fidelity to employers and clients, honesty in their work, restricting one’s work to areas within which one has competence, the importance of confidentiality, and avoiding or minimizing conflicts of interest. In light of the Preamble to NSPE’s Code of Ethics, these are just the areas of concern one would expect common morality to address. Thus far, it has simply been assumed that engineering codes of ethics have an appropriate place in engineering practice. Given this assumption, common morality can be called upon to help formulate their provisions. However, it is important to examine this assumption itself, for serious questions have been raised about the function, limitations, and even the moral legitimacy of codes of ethics for professionals. 3 CODES OF ETHICS There is no universally accepted account of what professions are that distinguishes them from other occupations. However, engineering exemplifies the following features that, taken together, warrant regarding it as a profession:3 1) Engineering requires extensive preparation in the form of training, much of which is of an intellectual character; 2) mastery of this intellectual component typically requires formal education at an institution of higher education; 3) the knowledge and skills possessed by engineers make a vital contribution to the well-being of the larger society; 4) engineers exercise a considerable degree of autonomy, or professional judgment, in providing their services; and 5) engineering societies typically claim to be regulated by ethical standards, as evidenced by their codes of ethics. The first four features relate directly to May’s concern with the virtues of professionals such as engineers. However, when we turn to engineering codes of ethics, questions can be raised about both their moral status and scope. 2 This has not always been the case. Prior to the 1970’s, this was not explicitly acknowledged in the codes. A notable exception was the American Association of Engineers (AAE) code in the 1920’s. This code’s first principle stated, “The engineer should regard his duty to the public as paramount to all other obligations.” [Taeusch, 1926, p. 102]. However, AAE itself dissolved by 1930. 3 For a succinct discussion of features typically found in professions, see [Bayles, 1989]. For a nuanced discussion of somewhat contested questions about the status of engineering as a profession, see [Davis, 2002, Ch. 7, pp. 99-120].

Professional Standards in Engineering Practice

957

Engineering codes of ethics originate in particular professional societies: for example, the American Society of Civil Engineers (ASCE), the American Society of Mechanical Engineers (ASME), the Institute of Electronic and Electrical Engineers (IEEE), and the National Society for Professional Engineers (NSPE). As suggested by their titles, these societies typically are confined within the national boundaries within which they are adopted. Even within these boundaries membership is voluntary, and only a small percentage of practicing engineers actually join. Furthermore, there is some controversy about what role engineering codes of ethics should have. While conceding that a code of ethics may be necessary for an emerging profession to gain initial recognition, Heinz Luegenbiehl contends that engineering codes have now outlived their usefulness [Luegenbiehl, 1991, p. 137-138]. Supposedly, the codes constitute a “set of ethical rules that are to govern engineers in their professional lives.” However, he argues, practicing engineers seldom consult these codes, some of their provisions are in conflict with one another and provide no guidance for the resolution of these conflicts, and the codes are coercive in intent, which challenges the autonomy normally attributed to moral agents.4 In reply, it could be said that the usefulness of engineering codes of ethics does not depend on their being regularly consulted by practicing engineers. If a code does a good job of identifying the basic obligations of engineers, it can be called upon when needed — for example, as Michael Davis points out, when an employer expects an engineer to do something unethical [Davis, 1991, p. 150167]. Davis sees engineering codes of ethics as advising engineers how they should act as professionals and as conventions between professionals that enable them to cooperate in serving a shared ideal of public service better than they could if they stood alone. So, for Davis, a code of ethics is seen as an agreement among members of a profession to commit themselves to a common set of standards that serve the shared ends of their profession. The obligation to comply with a code is an obligation to one’s fellow professionals, and it is an obligation of fairness to one another — to do one’s part. This gives an engineer a reason for wanting to join a professional society with a code of ethics. It also gives those already in such a society reason to continue to support it and its code, and to work at recruiting new members. An advantage for individual engineers is that, by joining, conducting themselves ethically in their engineering work is no longer just a matter of personal conscience for them. Joined with others, an engineer can appeal to his or her society’s code and say, “As an engineer, I cannot do this.” In professional ethics, there can be strength in numbers. The NSPE Code of Ethics is the product of the collective reflection of its members. On Davis’s view, NSPE members are obligated to comply with the code’s provisions because of their agreement with each other that they will do so. However, the NSPE code is worded in such a way that it seems intended to address the ethical responsibilities of engineers as such, not solely members of NSPE. Given 4 This

is a point first argued by Ladd, [1991, pp. 130-136].

958

Michael S. Pritchard

this, the standards endorsed by the code should be supportable by reasons other than the fact that NSPE members publically endorse and commit themselves to those standards. That is, the standards should be supportable by reasons that are binding on even those engineers who are not members of NSPE. Are they? In answering this question it is important to note that the code’s Preamble makes no reference to its members creating or committing themselves to the NSPE code. Instead, it attempts to depict the role that engineering plays in society, along with the standards of conduct that are required in order for engineers to fulfill this role responsibly. Thus, this depiction is presumed to be apt regardless of whether engineers are members of NSPE. Engineers and non-engineers alike can readily agree that engineers do play the sort of vital societal role depicted by the Preamble. What about the normative implications of that role? Here, too, engineers and non-engineers can agree, at least broadly. This is because, as already noted, the basic ethical standards endorsed by the NSPE Code of Ethics are supported by common morality. However, even if common morality can be appealed to in support of the basic provisions of an engineering code of ethics, this does not ensure that all of its provisions will be free from controversy, or even inappropriate content. In part this is because those who deliberate about what should be included in a code can be expected to take into account not only ethical considerations, but also realities of the workplace of engineers. Most engineers are corporate employees, and corporate goals may be more or less receptive to, for example, engineering concerns about sustainable technological developments. Although some prominent engineering societies now include statements about environmental concerns, most still refrain from making any explicit statements on such matters. One way to minimize controversy is for a code’s provisions to be stated in such a way that there is broad room for interpretation. In fact, as is the case with Gert’s moral rules and ideals, this is to some extent a practical necessity. Actual situations cannot be anticipated in all their relevant nuances, and judgment is itself one of the hallmarks of professional practice. For example, although sometimes it is clear what would constitute a failure to protect public, health, and safety, often it is not. Not actively protecting public safety will fail to satisfy the public safety standard only if there is a responsibility to provide that level of safety. But, since no engineering product can be expected to be “absolutely” safe (at least, not if it is to be a useful product) and there are economic costs associated with safety improvements, there can be considerable controversy about what a reasonable standard of safety is. 4 TECHNICAL CODES AND STANDARDS Engineering codes of ethics typically state that the work of engineers is expected to conform with “applicable engineering standards,” such as technical codes and standards. These codes and standards have a life of their own, in the sense that they do not depend on engineering codes of ethics for either their origin or their

Professional Standards in Engineering Practice

959

bindingness.5 They have been developed and formulated in the course of time as engineering practice itself has developed. Technical codes are legal requirements that are enforced by a governmental body to protect safety, health and other relevant values [Hunter, 1977, p. 66-71]. Examples are building codes, sanitary and health codes, and fire codes. Technical standards are usually regarded as recommendations rather than legal requirements. Varying in length from a few paragraphs to hundreds of pages, they are usually written by engineering experts who sit on standardization committees. Technical codes are often based on standards, or they may refer to standards as either a required or possible way of meeting code requirements. Standards may become mandatory by inclusion in a business contract. Standards are often seen as specifying criteria for good design practice, and as such they may be relevant in liability claims against companies or designers [Hunter, 1977, p. 66-71]. While technical codes are formulated by government bodies, technical standards (on which codes are often based) are not. Standards may be internal to a company, to a consortium of companies, or industry-wide. Industry-wide standards are usually formulated by consensus. A main reason for standardization in industry is the desire for interchangeability and compatibility. Standardization ensures that replacement parts are interchangeable with the original ones. Standardization also ensures that different products can work together or can use the same technical infrastructure. Industry-wide standards are usually formulated through national standards institutes, like the American National Standards Institute (ANSI) and the International Organization for Standardization (ISO). ANSI is a privately funded federation of business and industry, standards developers, trade associations, labor unions, professional societies, consumers, academia, and government agencies. ANSI has accredited a number of organizations, like the American Society of Mechanical Engineers (ASME), as standards developing organizations. These standards developing organizations oversee the process of standard formulation, which involves the relevant stakeholders and which has to meet the requirements formulated by ANSI to guarantee openness, transparency, balance of interests and due process. Standards are achieved by consensus. European standards are formulated through the CEN, the European Committee for Standardization.6 The procedure is comparable to that of ANSI, but there are some significant differences. The process is organized primarily through the national standards bodies, which are members of the CEN. Stakeholder involvement is thus organized through these bodies. Moreover, a standard agreed upon by the technical committee for a specific standard is adopted as a harmonized European standard, or not, by a weighted vote of the national standards bodies. If a European standard is adopted, national standards bodies are obliged to withdraw 5 The next eight paragraphs on codes and standards are, with minor alterations, the contribution of Ibo van de Poel, to whom I am much indebted for allowing the inclusion of his work. 6 Information based on http://www.cenorm.be/cenorm/index.htm.

960

Michael S. Pritchard

conflicting standards and to adopt a standard that conforms to the harmonized standard. Apart from standards formulated through organizations like CEN, ANSI and ISO, also de facto standards can be distinguished. These standards are not approved by standardization organizations but are widely used and recognized by industry as being standard. Often such standards are effectuated through the market. Given this, de facto standards do not necessarily reflect the interests and values of the wider public. Industry consortia can also voluntarily agree on standards, not only to promote interchangeability and compatibility, but also in an attempt to create a de facto standard which may give an important economic advantage. So, it can be seen that technical codes and standards serve a number of values. They serve utility and prudential values like interchangeability, compatibility, and efficiency. They also serve moral values such as safety, health, environmental sustainability and privacy. These values are often translated in codes and standards in rather concrete terms. For example, in codes and standards for pressure vessels, the value of safety is translated into a certain wall thickness of the vessel — to avoid explosions. In a building code, sustainability may be translated in terms of certain maximum heat transfer through the windows of a building in certain circumstances. Such translations may, sometimes, be ethically questionable, as we shall see in the ASME/Hydrolevel case discussed below. Another problematic area is safety in car design. Most crash tests for cars stress the safety of people inside the car and not the safety of people outside the car [van Gorp, 2005]. However, for those inside a car the risks are more voluntary than for cyclists and pedestrians that are hit by the car in case of an accident. Moreover, those inside the car have the advantage of using the car while cyclists and pedestrians only face the risks. Both factors, the degree of voluntariness and the distribution of risks and benefits, mean that the moral acceptability of the risks to people outside the car is more problematic than the risks to those inside the car. This means that, from an ethical point of view, it is important to examine current codes and standards with a critical eye. Nevertheless, given the need for codes and standards in engineering practice, it seems reasonable to place a burden of proof on those engineers who would take exception to them. This would seem to be an implication of Gert’s last two moral rules of common morality: obey the law and do your duty (here, your job-related responsibilities). The other moral rules can be used to evaluate whether these two rules should carry the day in problematic cases. 5

ACCEPTED STANDARDS OF ENGINEERING PRACTICE AND THE STANDARD OF CARE

In requiring engineers to conform to accepted standards of engineering practice, engineering codes of ethics insist on compliance with technical codes and standards.

Professional Standards in Engineering Practice

961

These codes and standards are regulatory in intent. They may focus on desired results of engineering practice — for example, on whether the work satisfies certain standards of quality or safety. Technical codes and standards may also require that certain procedures be undertaken to ascertain that specific, measurable levels of quality or safety are met; or they may require that whatever procedures are used be documented, along with their results. Equally important, engineering codes of ethics typically insist that engineers conform to standards of competence, standards that have evolved through engineering practice and that presumably are commonly accepted, even if only implicitly, in ordinary engineering training and practice.7 Regulatory standards and standards of competence are intended to provide some assurance of quality, safety, and efficiency in engineering. It is important to realize, however, that they also leave considerable room for professional discretion in engineering design and its implementation. This calls for competence. There are few algorithms for engineers to follow here. Performance standards that do not specify particular procedures to be followed or materials to be used clearly leave room for professional discretion. But even more specific technical codes leave room for discretion (for example, as in the Citicorp Building illustration discussed below, whether to bolt or weld joints).8 So, the need for engineering judgment should not be overlooked.9 Regarding safety, for example, rather than leave the determination of what counts as safe solely in the hands of individual engineers, safety standards may be set by government agencies (such as the National Institute of Standards and Technology, the Occupational Safety and Health Administration, or the Environmental Protection Agency) or non-governmental organizations (such as professional engineering societies, ANSI, ISO, and CEN). Nevertheless, standards of safety, as well as standards of quality in general, leave room for considerable engineering discretion. Although some standards have a high degree of specificity (e.g., minimal requirements regarding the ability of a structure to withstand winds of a certain velocity striking that structure at a 90 degree angle), some simply require that unspecified standard processes be developed, followed, and documented [Shapiro, 1997, p. 290]. Underlying all of these more specific efforts to articulate particular codes and standards is a broader standard of care in engineering practice, a standard appealed to in law and about which experienced, respected engineers can be called upon to testify in the courts in particular cases. Although the standard of care is used as a standard in law, it can also be seen as a reasonable moral standard, reflected in common morality’s concern to avoid and prevent harm, suffering, and death, among other things. It also can be seen as instrumental in engineers’s efforts to protect public safety, health, and welfare in the course of their work, the 7 See, for example, the Association for Computing Machinery: ACM Code of Ethics and Professional Conduct, 2.2 Acquire and maintain professional competence. 8 For a good discussion of the importance of judgment, imagination, and responsibility in relation to standards and codes, see [Coeckelbergh, 2006, pp. 237-260]. 9 This is a major theme of Stuart Shapiro’s [1997].

962

Michael S. Pritchard

paramount duty of engineers according to virtually all of the codes of ethics of engineering societies in the USA, and in most other countries as well. Joshua B. Kardon characterizes this standard of care in this way [Kardon, 1999]. Although some errors in engineering judgment and practice can be expected to occur as a matter of course, not all errors are acceptable: An engineer is not liable, or responsible, for damages for every error. Society has decided, through case law, that when you hire an engineer, you buy the engineer’s normal errors. However, if the error is shown to have been worse than a certain level of error, the engineer is liable. That level, the line between non-negligent and negligent error is the “standard of care.” How is this line determined in particular cases? It is not up to engineers alone to determine this, but they do play a crucial role in assisting judges and juries in their deliberations: A trier of fact, a judge or jury, has to determine what the standard of care is and whether an engineer has failed to achieve that level of performance. They do so by hearing expert testimony. People who are qualified as experts express opinions as to the standard of care and as to the defendant engineer’s performance relative to that standard. For this legal process to be practicable and reasonably fair to engineers, it is necessary that there be an operative notion of “accepted practice” in engineering that is well understood by competent engineers in the areas of engineering under question. As Kardon puts it:10 A good working definition of the standard of care of a professional is: that level or quality of service ordinarily provided by other normally competent practitioners of good standing in that field, contemporaneously providing similar services in the same locality and under the same circumstances. [Kardon, 1999] Given this, we should not expect to find a formal statement of what specifically satisfies the standard. Rather, an appeal is being made to what is commonly and ordinarily done (or not done) by competent engineers. Engineers who have responsible charge for a project are expected to exercise careful oversight before putting their official stamp of approval on the project. However, what careful oversight requires will vary with the project in question in ways that resist an algorithmic articulation of the precise steps to be taken and the criteria to be used. Two well known cases are instructive. In the first instance, those in charge of the construction of the Kansas City Hyatt-Regency hotel were charged with professional negligence in regard to the catastrophic walkway collapse in 1981.11 Although those in charge did not authorize the fatal departure from the 10 Ibid. Kardon bases this characterization on Paxton v. County of Alameda (1953) 119c.C.A. 2d 393, 398, 259P 2d 934. 11 For further discussion of this case, see [Harris et al., 2009, p. 252]. See also [Shapiro, 1997, p. 287].

Professional Standards in Engineering Practice

963

original design of the walkway support, it was determined that responsible monitoring on their part would have made them aware of the proposed change. Had it come to their attention, a few simple calculations could have made it evident to them that the resulting structure would be unsafe. In this case it was determined that the engineers in charge fell seriously short of accepted engineering practice, resulting in a failure to meet the standard of care. Satisfying the standard of care cannot guarantee that failure will not occur. However, failure to satisfy the standard of care itself is not acceptable. In any particular case, there may be several acceptable ways of meeting the standard. Much depends on the kind of project in question, its specific context, and the particular variables that (sometimes unpredictably) come into play. The second case also involved a departure from the original design not noted by the chief structural engineer of Manhattan’s 59 story Citicorp Building [Morgenstern, 1995, p. 49-53].12 In contrast to the Hyatt Regency walkway, this was not regarded to be a matter of negligence. Chief structural engineer William LeMessurier was surprised to learn that Citicorp’s major structural joints were bolted rather than deep-welded together, as called for in the original design. However, he was confident that the building still more than adequately satisfied the New York City building code’s requirement that winds striking the structure from a 90 degree angle would pose no serious danger. Assuming he was correct, it is fair to conclude that either deep welds or bolts were regarded to be consistent with accepted engineering practice. The code did not specify which should be chosen, only that the result must satisfy the 90 degree wind test. Fortunately, LeMessurier did not rest content with the thought that the structure satisfied the city building code. Given the unusual features of the Citicorp structure, he wondered what would happen if winds struck the building diagonally at a 45 degree angle. This question seemed sensible, since the first floor of the building is actually several stories above ground, with the ground support of the building being four pillars placed in between the four corners of the structure rather than at the corners themselves. Further calculations by LeMessurier determined that bolted joints rendered the structure much more vulnerable to high winds than had been anticipated. Despite satisfying the city code, the building was unsafe. LeMessurier concluded that corrections must be made. The standard set by the city building code was flawed. The code could not be relied on to set reliable criteria for the standard of care in all cases. From this it should not be concluded that there is only one acceptable solution to the joint problem. LeMessurier’s plan for reinforcing the bolted joints worked. But the original plan for deep welds apparently would have, as well. Many other acceptable solutions may have been possible. So, a variety of designs for a particular structure could be consistent with professional engineering standards. The Hyatt-Regency case is a clear illustration of culpable failure. The original design failed to meet building code requirements. The design change made mat12 For further details, see [Harris et al., 2009, pp. 307-308]. See also [Morgenstern, 1995, pp. 49-53].

964

Michael S. Pritchard

ters worse. The Citicorp case is clear illustration of how the standard engineering practice of meeting code requirements may not be enough. It is to LeMessurier’s credit that he discovered the problem. Not doing so would not have been negligence, even though the structure was flawed. Once the flaw was discovered, however, the standard of care required LeMessurier to do something about it, as he clearly realized. Furthermore, it seems that foremost in mind for LeMessurier was his sense of moral responsibility for the safety of the Citicorp structure. To some extent, of course, the possibility of legal liability may have been a factor as well, but LeMessurier’s account of the course of events makes it clear that his primary focus was on his moral responsibility to do his best to correct the flaw in the building that he, and only he, had discovered through his own engineering conscientiousness and initiative.

6 INNOVATION AND PROFESSIONAL STANDARDS No doubt William LeMessurier was disappointed to discover a serious fault in the Citicorp Building. However, there was much about the structure in which he could take pride. A particularly innovative feature was a 400 ton concrete damper on ball bearings placed near the top of the building. LeMessurier introduced this feature, not to improve safety, but to reduce the sway of the building — a matter of comfort to residents, not safety. Of course, this does not mean that the damper has no affect on safety. Although designed for comfort, it is possible that it also enhances safety. Or, especially since it’s movement needs to be both facilitated and constrained, it is possible that, without other controls, it could have a negative effect on safety. In any case, the effect that a 400 ton damper near the top of a 59 story structure might have on the building’s ability to handle heavy winds is something that required careful attention. Supporting the structure on four pillars midway between the corners of the Citicorp building was another innovation — one that might explain why it occurred to LeMessurier that it was worthwhile to try to determine what effect 45 degree winds might have on the structure’s stability. Both innovations fall within the range of accepted engineering practice, provided that well conceived efforts are made to determine what effect they might have on the overall integrity and utility of the structure. The risk of relying exclusively on the particular directives of a building code is that its framers are unlikely to be able in advance to take into account all of the relevant effects of innovations in design. That is, it is quite possible for regulations to fail to keep pace with technological innovation. Although engineers and their employers might try to excuse failure to provide safety and quality by pointing out that they have met existing regulatory standards, it is evident that the courts will not necessarily agree. The standard of care in tort law (which is concerned with wrongful injury) is stated more broadly

Professional Standards in Engineering Practice

965

than regulatory standards are. The expectation is that engineers will meet the standard of care as expressed in Coombs v. Beede:13 The responsibility resting on an architect is essentially the same as that which rests upon the lawyer to his client, or upon the physician to his patient, or which rests upon anyone to another where such person pretends to possess some special skill and ability in some special employment, and offers his services to the public on account of his fitness to act in the line of business for which he may be employed. The undertaking of an architect implies that he possesses skill and ability, including taste, sufficient enough to enable him to perform the required services at least ordinarily and reasonably well; and that he will exercise and apply, in the given case, his skill and ability, his judgment and taste reasonably and without neglect. As Korden points out, this standard does not hold that all failure to provide satisfying services is wrongful injury. But it does insist that the services provided evidence reasonable care. What counts as reasonable care is a function of both what the public can reasonably expect and what experienced, competent engineers regard as acceptable practice. Given the desirability of innovative engineering design, it is unrealistic for the public to regard all failures and mishaps to be culpable; at the same time, it is incumbent on engineers to do their best to anticipate and avoid failures and mishaps.14 7

REGULATING THE REGULATORS

Ideally, regulatory standards are unbiased, as are the regulators. However, since the experts who help establish the standards typically are employed by the very companies whose products are being regulated, special efforts must be made to minimize the chances that they will unfairly favor their employers. Rather than naively assume that all conflicts of interest can be eliminated, Stephen Unger suggests the following: [One must] ensure that the membership of decision-making committees includes a variety of people with varying biases, and to carry out the entire process in an open fashion, that is, to make the decision-making process transparent to all interested parties. Inviting comments from all concerned groups and individuals and providing appeals processes are additional methods for ensuring fair play. [Unger, 1994, p. 210] Unfortunately, as Unger points out, his recommended process is not foolproof. A classic illustration is the case of the American Society for Mechanical Engineers 13 Coombs v. Beede, 89 Me. 187, 188, 36 A. 104 (1896). This is cited and discussed in [Strand and Golden, 1997]. 14 For good discussions of responsibility and innovation in engineering design, see [Grunwald, 2001] and [van de Poel and van Gorp, 2006].

966

Michael S. Pritchard

(ASME) vs. Hydrolevel, which was finally settled against ASME. The first half of the 19th century was marked by boiler explosions on steamboats that resulted in the deaths of thousands of Americans. ASME played the leading role in establishing uniform requirements for safe boilers. A special area of concern is boilers being heated when the water level in the boiler is insufficient. In the early 1970’s the ASME code specified: “Each automatically fired steam or vapor system boiler shall have an automatic low-water fuel cutoff, so located as to automatically cut off the fuel supply when the surface of the water falls to the lowest visible part of the water-gauge glass” [American Society of Mechanical Engineers, Boiler and Pressure Vessel Code, section IV, paragraph HG-605a]. Attempting to make competitive inroads in this area, Hydrolevel developed a mechanism that included a time delay in its fuel cutoff system, claiming that this would ensure a more reliable determination of water level in boilers whose water is in motion. ASME’s BPVC Boiler and Pressure Vessel Committee was headed by prominent representatives of two companies then dominating the market. Unknown to Hydrolevel, a letter was circulated that insisted that low-water fuel cutoff mechanisms should operate immediately. By the time Hydrolevel discovered the existence of this letter, it had suffered serious market losses. Hence, it undertook an anti-trust lawsuit against the two companies and ASME. The two companies settled out of court with Hydrolevel. However, ASME protested that it had done nothing wrong, despite the fact that some of its volunteer committee members had, on their own, acted unfairly in their companies’ behalf. ASME took its case all the way to the Supreme Court of the United States, but to no avail. In effect, the Supreme Court found ASME to be negligent in overseeing how its special committees operate in enforcing standards that can have a large impact on the economic success or failure of companies. The Court’s majority opinion said, in part:15 ASME wields great power in the nation’s economy. Its codes and standards influence the policies of numerous states and cities, and as has been said about “so-called voluntary standards” generally, its interpretation of guidelines “may result in economic prosperity or economic failure, for a number of businesses of all sizes throughout the country,” as well as entire segments of an industry. [Beardsley, 1984, p. 66] As a result of the Court’s ruling, ASME introduced a number of substantial changes in its procedures. Charles Beardsley sums up the changes: The most striking changes affect the Society’s handling of codes and standards interpretations. All such interpretations must now be reviewed by at least five persons before release; before, the review of two people was necessary. Interpretations are available to the public, with replies to nonstandard inquiries published each month in the Codes and Standards section of ME or other ASME publications. [Beardsley, 1984, p. 73] 15 [Harris

et al., 1009, p 254]. Cited from [Beardsley, 1984].

Professional Standards in Engineering Practice

967

In addition, ASME requires all staff and volunteer committee members to sign a conflict of interest disclaimer, and ASME provides them with copies of its code of ethics and a publication that discusses the legal ramifications of the standards. So, essentially, in an effort to reduce bias, ASME implemented the guidelines Stephen Unger suggests. ASME’s response to its adverse legal ruling is instructive. Rather than retreat from establishing and enforcing codes and standards, it introduced changes to improve its procedures. What remained intact was its rejection of the idea that everything should be left solely to the discretion of individual engineers or the firms for whom they work. There was no wavering from its commitment to uniform codes and standards regarding matters of safety and quality; and it continued to accept its responsibility to help frame and enforce them.

8 DESIGN The Hydrolevel case is instructive in another way. Hydrolevel came up with a departure from the more usual way of ensuring safety. This was challenged by its competitors. Initially, at least, Hydrolevel met with failure. However, in fact, its mechanism might well have satisfied reasonable standards of safety. The point here is that there is likely more than one way to satisfy safety standards, especially when stated broadly. Arguably, the ASME standard was interpreted too narrowly by Hydrolevel’s competitors. But if there is more than one way to satisfy safety standards, how are designers to proceed? If we are talking about the overall safety of a product, there may be much latitude, a latitude that, of course, provides space for considerations other than safety, as well (e.g., overall quality, usability, cost). For example, in the late 1960’s, operating under the constraints of coming up with an appealing automobile that weighed under 2000 lbs. that would cost consumers no more than $2000, Ford engineers decided to make more trunk space by putting the Pinto’s gas tank in an unusual place.16 This raised a safety question regarding rear end collisions. Ford claimed that the vehicle passed the current standard. However, some Ford engineers urged that a protective buffer should be inserted between the gas tank and protruding bolts. This, they contended, would enable the Pinto to pass a more demanding standard that it was known would soon be imposed on newer vehicles. They warned that, without the buffer, the Pinto would fail to satisfy the new standard, a standard that they believed would come much closer to meeting the standard of reasonable care enforced in tort law. Ford decided not to put in the buffer. It might have been thought that satisfying the current safety standard ensured that courts and their juries would agree that reasonable care was exercised. However, this turned out to be a mistaken view. As noted above, the courts can determine that existing technical standards are not 16 For further discussion of the Pinto case, see Case 27, “Pinto,” in [Harris et al., 2009, pp. 266-267].

968

Michael S. Pritchard

adequate, and engineers themselves are sometimes called upon to testify to that effect. Given the bad publicity Ford received regarding the Pinto and its history of subsequent litigation, Ford might regret not having heeded the advice of those engineers who argued for the protective buffer. This could have been included in the original design, or perhaps there were other feasible alternatives during the early design phases. However, even after the car was put on the market, a design change could have been made. This would have involved an expensive recall, but this would not have been an unprecedented move in the automotive industry. These possibilities illustrate a basic point about regulatory standards, accepted standards of engineering practice, and engineering design. Professional standards for engineers underdetermine design.17 In principle, if not in practice, there will also be more than one way to satisfy the standards. This does not mean that professional standards have no effect on practice. As Stuart Shapiro points out: Standards are one of the principal mechanisms for managing complexity of any sort, including technological complexity. Standardized terminology, physical properties, and procedures all play a role in constraining the size of the universe in which the practitioner must make decisions. [Shapiro, 1997, p. 290] For a profession, the establishment of standards of practice is typically regarded as contributing to professionalism, thereby enhancing the profession in the eyes of those who receive its services. At the same time, standards of practice can contribute both to the quality and safety of products in industry. Still, standards of practice have to be applied in particular contexts that are not themselves specified in the standards. Shapiro notes: There are many degrees of freedom available to the designer and builder of machines and processes. In this context, standards of practice provide a means of mapping the universal onto the local. All one has to do is think of the great variety of local circumstances for which bridges are designed and the equally great variety of designs that result.... Local contingencies must govern the design and construction of any particular bridge within the frame of relative universals embodied in the standards. [Shapiro, 1997, 293] Shapiro’s observation focuses on how standards of practice allow engineers freedom to adapt their designs to local, variable circumstances. This often brings surprises, not only in design but also in regard to the adequacy of formal standards of 17 Mark Coeckelbergh makes a distinction between prescriptive and goal-setting regulations, with the latter providing more room for autonomy in decision-making for engineers [Coeckelbergh, 2006]. However, even highly prescriptive regulations allow room for discretion, as there may be more than one mechanism, set of materials, or method that can satisfy even a fairly specific prescription. Coeckelbergh’s general point is that increasing responsibility goes with the increasing autonomy that comes with the absence of specific external determination of what, specifically, is required.

Professional Standards in Engineering Practice

969

practice. As Louis L. Bucciarelli points out, standards of practice are based on the previous experience and testing of engineers. Design operates on the edge of “the new and the untried, the unexperienced, the ahistorical” [Bucciarelli, 1994, p. 135]. Thus, as engineers come up with innovative designs (such as LeMessurier’s Citicorp structure), we should expect formal standards of practice themselves sometimes to be challenged and found to be in need of change. All the more reason why courts of law are unwilling simply to equate the standard of reasonable care with current formal standards of practice. Bucciarelli makes another important point about design. Design changes are often made during the process of implementation; that is, design itself can be seen as a work in process, rather than as a final plan that precedes and guides implementation. This is illustrated in the fictional case study An Incident in Morales, a video developed by the National Institute for Engineering Ethics.18 While implementing a design for a chemical plant in Mexico, the chief design engineer learns that his budget is being cut by 20%. To fall within the new budget, some design changes are necessary. Next the engineer learns that the effluent from the plant will likely cause health problems for local residents. The current design is consistent with local standards, but it would be in violation of standards across the border in Texas. A possible solution is to line the evaporation ponds, an additional expense. Implementing this solution provides greater protection to the public; but, as it turns out, this comes at the expense of putting some workers at the plant at greater risk because of a money-saving switch to cheaper controls within the plant — another design change. So, a basic question facing the engineer is, given the tight budgetary constraints, which standards of practice take priority? The moral of the story is that, from the very outset of this project, the engineer failed to take sufficiently into account signs of trouble ahead — including warnings from senior engineers at another facility that taking certain shortcuts would be unwise (if not unethical). 9

THE SCOPE OF STANDARDS OF PRACTICE

Some standards of practice are clearly only local in their scope. The New York City building code requirement that high rise structures be tested for wind resistance at 90 degree angles applied only within a limited geographic region. Such specific code requirements are local in their origin and applicability. Of course, one would expect somewhat similar requirements to be in place in comparable locales in the United States, as well as in other high rise locales around the world. This suggests that underlying local codes, particularly those that attempt to ensure quality and safety, are more general standards of safety and good engineering practice. 18 An Incident at Morales: An Engineering Ethics Story, developed and distributed by the National Institute for Engineering Ethics, the Murdough Center for Engineering Professionalism, and the Texas Tech University College of Engineering (2003). This video is available from the National Institute for Engineering Ethics, Box 41023, Lubbock, Texas 79409-1023. (Email: [email protected].)

970

Michael S. Pritchard

One test of whether we can meaningfully talk of more general standards is to ask whether the criteria for engineering competence are only local (e.g., New York City civil engineers, Chicago civil engineers, Kalamazoo, Michigan civil engineers). The answer seems clearly to be no within the boundaries of the United States, especially for graduates of accredited engineering programs at United States colleges and universities. However, as Vivian Weil has argued, there is good reason to believe that professional standards of engineering practice can cross national boundaries [Weil, 1998, p. 303-314]. She offers the example of early 20th century Russian engineer, Peter Palchinsky. Critical of major engineering projects in Russia, Palchinsky was nevertheless regarded to be a highly competent engineer in his homeland. He also was a highly regarded consultant in Germany, France, England, the Netherlands, and Italy. Although he was regarded as politically dangerous by Russian leaders at the time, no one doubted his engineering abilities — either in Russia or elsewhere. Weil also reminds readers of two fundamental principles of engineering that Palchinsky applied wherever he practiced: Recall that the first principle was: gather full and reliable information about the specific situation. The second was: view engineering plans and projects in context, taking into account impacts on workers, the needs of workers, systems of transportation and communication, resources needed, resource accessibility, economic feasibility, impacts on users and on other affected parties, such as people who live downward. [Weil, 1998, p. 306] Weil goes on to point out that underlying Palchinsky’s two principles are principles of common morality, particularly respect for the well being of workers — a principle that Palchinsky argued was repeatedly violated by Lenin’s favored engineering projects. At the outset of this chapter, it was noted that the codes of ethics of engineering societies typically endorse principles that seem intended to apply to engineers in general rather than only to members of those particular societies. Common morality was suggested as providing the ground for basic provisions of those codes (for example, concern for the safety, health, and welfare of the public). Whether engineers who are not members of professional engineering societies actually do, either explicitly or implicitly, accept the principles articulated in a particular society’s code of ethics is, of course, another matter. However, even if some do not, it could be argued that they should. Weil’s point, a point accepted in this paper as well, is that there is no reason, in principle, to believe that supportable international standards cannot be formulated and adopted. Furthermore, this need not be restricted to abstract statements of ethical principle. As technological developments and their resulting products show up across the globe, they can be expected to be accompanied by global concerns about quality, safety, efficiency, cost effectiveness, and sustainability. This, in turn, can result in uniform standards in many areas regarding acceptable and unacceptable engineering design, practice, and products.

Professional Standards in Engineering Practice

971

In any case, in the context of an emerging global economy, constructive discussions of these concerns should not be expected to be only local. BIBLIOGRAPHY [Bayles, 1989] M. D. Bayles. Professional Ethics. Belmont, CA: Wadsworth, 1989 (2nd edition). [Beardsley, 1984] C. W. Beardsley. The Hydrolevel Case — A Retrospective. Mechanical Engineering, June, 66, 1984. [Bucciarelli, 1994] L. L. Bucciarelli. Designing Engineers. MIT Press, 1994. [Coeckelbergh, 2006] M. Coeckelbergh. Regulation or Responsibility? Autonomy, Moral Imagination, and Engineering. Science, Technology & Human Values, 31 (3), pp. 237-260, 2006. [Coombs v. Beede, 1896] Coombs v. Beede, 89 Me. 187, 188, 36 A. 104, 1896. [Davis, 1991] M. Davis. Thinking Like an Engineer: The Place of a Code of Ethics in the Practice of a Profession. Philosophy and Public Affairs, 2, 150-167, 1991. [Davis, 2002] M. Davis. Profession, Code and Ethics. Burlington, VT: Ashgate, 2002. [Gert, 2004] B. Gert. Common Morality: Deciding What to Do. Oxford University Press, 2004. [Grunwald, 2001] A. Grunwald. The Application of Ethics to Engineering and the Engineer’s Moral Responsibility. Science and Engineering Ethics, 7, (3), pp. 415-428, 2001. [Harris et al., 2009] C. E. Harris, M. S. Pritchard, and M. Rabins, eds. Engineering Ethics: Concepts and Cases. Belmont, CA: Wadsworth, 2009 (4th ed.). [Hunter, 1977] T. A. Hunter. Designing to Codes and Standards. ASM Handbook. G.E. Dieter and S. Lampman eds., pp. 66-71, 1977. [Kardon, 1999] J. B. Kardon. The Structural Engineer’s Standard of Care. Presented at the OEC International Conference on Ethics in Engineering and Computer Science, 1999. This article is available at onlineethics.org. [Ladd, 1991] J. Ladd. The quest for a code of professional ethics: an intellectual and moral confusion. In, D. Johnson, ed., Ethical Issues in Engineering. New Jersey: Prentice Hall, 1991. [Luegenbiehl, 1991] H. Luegenbiehl. Codes of Ethics and the Moral Education of Engineers. In: Ethical Issues in Engineering. Deborah Johnson, ed., pp. 137-138. Prentice Hall, 1991. [May, 1988] W. F. May. Professional Virtue and Self-Regulation. In: Ethical Issues in Professional Life, Joan C. Callahan, ed. p. 408. Oxford University Press, 1988. [Morgenstern, 1995] J. Morgenstern. The Fifty-Nine Story Crisis. The New Yorker, May 29, pp. 49-53, 1995. [Shapiro, 1997] S. Shapiro. Degrees of Freedom: The Interaction of Standards of Practice and Engineering Judgment. Science, Technology, and Human Values, v22, 3, p. 290, 1997. [Strand and Golden, 1997] M. N. Strand and K. Golden. Consulting Scientist and Engineer Liability: A Survey of Relevant Law. Science and Engineering Ethics 3, 357-394, 1997. [Taeusch, 1926] C. Taeusch. Professional and Business Ethics. New York: Henry Holt & Co., 1926. [Unger, 1994] S. H. Unger. Controlling Technology: Ethics and the Responsible Engineer. WileyInterscience, 1994. [van de Poel and van Gorp, 2002] I. van de Poel and A. van Gorp. Degrees of Responsibility in Engineering Design: Type of Design and Design Hierarchy. Science, Technology & Human Values, 31 (3), pp. 333-360, 2002. [van Gorp, 2005] A. van Gorp. Ethical Issues in Engineering Design: Safety and Sustainability. Simon Stevin Series in the Philosophy of Technology, Delft University of Technology, 2005. [Weil, 1998] V. Weil. Professional Standards: Can They Shape Practice in an International Context? Science and Engineering Ethics, v.4, 3, pp. 303-314, 1998.

This page intentionally left blank

VALUES IN ENGINEERING DESIGN Ibo van de Poel

1

INTRODUCTION

Value is at the heart of engineering design. Design creates value for companies, users and, ultimately, for society. Few would disagree with such statements but what exactly do they imply? What is value? What types of values do technological artifacts have or contribute to? How are value considerations inherent to design choices? Is engineering design plagued by plural and conflicting values and, if so, how do, could and should engineers deal with such value conflicts? These are the types of questions that are explored in this contribution. The aim of this contribution is to philosophically explore the role of values in engineering design. Although, where relevant, I will draw on empirical evidence on the role of values in design, my aim will not be to merely empirically describe how values come to play a part in engineering design. Instead I shall aim to clarify, from a philosophical point of view, the role that values do, can and — according to some — should play in engineering design. I will not defend any specific approach to integrating values into design though I will discuss various approaches, especially in relation to conflicting design values together with the pros and cons of these approaches. The focus in this contribution will be on the values that are created through technical artifacts and, especially, on how the prospect of such value is integrated into the process of engineering design. I am not therefore focusing on the values, or virtues, of designing engineers (for this, see Pritchard’s chapter in this Volume, Part V) or on the value of engineering design as an activity (for engineers design may be an inherently valuable activity). This contribution starts with a brief overview of relevant notions from philosophical literature, like the notion of value. I shall then discuss the value of technological artifacts. In Section 4, I will discuss the role of values in the engineering design process. Sections 5 and 6 will deal with value conflicts in engineering design and approaches to dealing with them; Section 5 examines optimizing approaches and Section 6 looks at non-optimizing approaches. In the final section, I shall draw my conclusions.

Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

974

Ibo van de Poel

2 VALUES

2.1 Value statements Value statements are statements about whether certain things or state of affairs are good, i.e. valuable, or bad in a certain respect. If things or states of affairs are bad, they often not only lack value but also even have a negative value. I will use the term disvalue to refer to such negative value. Value statements are to be distinguished from statements of preference, i.e. statements about what individuals prefer. Establishing that something is a value or professing it to be valuable means not only claiming that it is valuable to me but also that it is or should be of value to others. Most things or states of affairs are not just plain good but good in a certain sense. A hammer may be good in the sense of being a useful instrument for driving nails into a piece of wood but it may simultaneously be aesthetically ugly. We might then say that it has a utility value with respect to driving nails into a piece of wood but that it has no, or merely a negative, aesthetic value. So conceived, values are varieties of goodness. Things or states of affairs may thus have — express, instantiate — different values at the same time. Statements about the value of things or state of affairs are evaluative statements: they evaluate something or a state of affairs in terms of a value. Value statements are therefore to be distinguished from descriptive and prescriptive statements. Descriptive statements are statements about how things are. We can, for example, describe a hammer in terms of its shape and the materials of which it is made; such descriptions do not attribute value to the object described. Still, the descriptive, non-evaluative, features of a hammer are relevant to answering the question of whether the hammer is a good hammer, in the sense of being useful for driving nails into a piece of wood. The value of a hammer as a useful instrument thus depends on the descriptive features of that same hammer. It can be expressed by saying that the value of a thing or state of affairs is “supervenient” on certain non-evaluative features of that thing or state of affairs [Zimmerman, 2004]. Value statements are also to be distinguished from prescriptive statements, i.e. statements about how to act (see, e.g. [Stocker, 1990; Dancy, 1993]). The statement “this is a good hammer for driving nails into a piece a wood” does not entail the statement “you should use that hammer to drive a nail into a piece of wood” (for a detailed discussion of the normative aspect of statements like these, see Franssen’s chapter in this Volume, Part V). This is not to deny that prescriptive statements may sometimes be derived from evaluative statements. In general, however, one should distinguish evaluative from prescriptive statements. In moral theory, there is a parallel distinction between “goodness” and “rightness.” States of affairs and in particular consequences of actions can be evaluated in terms of “goodness,” while actions themselves are evaluated in terms of “rightness.” Consequentialism is the doctrine which states that the goodness of consequences of actions determines the rightness of actions. However, consequentialism is cer-

Values in Engineering Design

975

tainly not the only theory in moral philosophy on the relation between goodness and rightness. If one takes a Kantian approach, for example, the rightness of an action in a certain sense determines or guarantees the goodness of the consequences of that action (cf. [Korsgaard, 1983, p. 183]).

2.2

Intrinsic versus instrumental value

Often a distinction is made between intrinsic and instrumental values. Intrinsic values are those that are good in themselves or for their own sake, while instrumental values are valuable because they help to achieve other values. It should be noted that in this respect an object can be instrumentally valuable and intrinsically valuable at the same time. A hammer may, for example, be instrumentally valuable as a useful tool for driving nails into wood while at the same time being intrinsically valuable as a beautiful object. Although the distinction between instrumental and intrinsic value may seem straightforward, it is not. Various philosophers have pointed out a number of terminological and substantive issues with respect to the distinction [for a discussion, see Zimmerman, 2004]. One issue is that the notion of intrinsic value is ambiguous. The notion is usually understood to refer to objects or states of affairs that are valuable in themselves. Intrinsic value is then value of a non-derivate kind. Intrinsic value may, however, also refer to things that are valuable due to their intrinsic natural, i.e. descriptive, properties. As Christine Korsgaard has pointed out things that are valuable due to their intrinsic properties are unconditionally good [Korsgaard, 1983]. Their goodness does not depend on the relation with other objects or with people; otherwise their value would not be intrinsic to the object. However, according to Korsgaard, some things may be good in a non-derivate sense, even if they are not unconditionally good. An example is human happiness in a Kantian respect. According to Kant, human happiness is non-derivate goodness. Happiness is good in itself, and not because it is a means to another end or contributes to another value. Nevertheless, according to Kant, happiness is only conditionally good; it is only good insofar as it corresponds to good will, i.e. respect for the moral law. To avoid the ambiguity to which Korsgaard refers, I will use the notion of ‘intrinsic value’ for non-derivate value and the notion of ‘value intrinsic to an object or state of affairs’ to describe value that only depends on the non-relational properties of an object or state of affairs. The notion of instrumental value is also more complex than it seems. It might refer to things that are useful for achieving some end, whether that end is valuable or not. Frankena [1973, p. 66] refers to such instrumental values as utility values. He puts forward the notion of extrinsic value to refer to “things that are good because they are a means to what is good” [Frankena, 1973, p. 66]. However, the term ‘extrinsic value’ is somewhat confusing because it is often used for all derivate values, i.e. all non-intrinsic values. Being a means to an end is, however, not the only way in which something can be valuable in a derivate way (see, e.g.

976

Ibo van de Poel

[Zimmerman, 2004]).1 Things can, for example, also be valuable because they enable a good life, just as privacy or health enable people to live a valuable life or because they contribute to a good life as do the virtues in an Aristotelian account of good life. I will therefore use the notion of instrumental value for the value of being a means to achieving a good end, i.e. another positive value.

2.3

Sources of value

What of the origins of value, i.e. what is it that brings value into the world? Here it is relevant to distinguish between intrinsic and extrinsic value. Extrinsically valuable objects or states of affairs derive their value from their contribution — for example as a means to an end — to intrinsic value. But where does intrinsic value come from? Three main doctrines can be distinguished here. Subjectivists believe that human desires, or more generally psychological states, constitute the source of value. Objectivists believe that values reside in the world outside us. Rationalists see human rationality as the ultimate source of value. Subjectivism does justice to the connection between values and human desires and interests. It runs, however, the risk of confusing value with preference. Not everything that is desired or preferred by people is valuable. Objectivism does justice to the fact that statements about value are not statements about preferences but rather about how the world is or ought to be from a normative point of view. Objectivists often hold that intrinsic value is value intrinsic to the valuable object (e.g. [Moore, 1912]). Since value that is intrinsic to an object cannot depend on the relational properties of that object human desires or interests can never be the, or even a, source of value. Rationalism can be seen as an in-between position. It restores the connection between human desires and values, which is lost in objectivism but strives to avoid confusing value with preference by claiming that things are valuable not just because people prefer them but because rational beings have sufficient practical reason to pursue them.2 With most of the issues discussed in this contribution it will not be necessary to take a stance in the debate about the source of value. The stance that I will take, where relevant, might be described as mildly rationalist. It is rationalist in the sense that both extreme subjectivist and extreme objectivist positions are avoided. I believe that values should be distinguished from preferences but not completely divorced from human desire, interest, interpretation and meaning-giving. I will call the position mildly rationalist because I will not assume one specific theory of practical rationality, nor will I make substantive assumptions about the ultimate sources of value in the world. 1 The point is not that the instrumental value may be insufficient to cause the end but that extrinsic values may contribute to intrinsic values in non-causal ways. They may, for example, be an indication of the achievement of an intrinsic value or they may be conceptually part of the intrinsic value (e.g. health and the good life). 2 Nevertheless, rationalism may make it difficult to express the fact that part of the reason for valuing an object may lie in the object itself.

Values in Engineering Design

2.4

977

Value conflicts and incommensurable values

A final general issue that is relevant to the discussion on the role of values in engineering is the notion of value conflict and value incommensurability. I speak of there being value conflict if two or more values conflict. Usually values will not generally conflict although that can happen but only in specific circumstances. Two or more values conflict in a specific situation if, when considered in isolation, they evaluate different options as best. For example, a governmental policy that is best from the viewpoint of individual freedom might not be best in terms of social justice. As we shall see, value conflict situations are common in engineering, especially in the engineering design process. Two or more values are incommensurable if they cannot be expressed or measured on a common scale or in terms of a common value measure. This can be an ordinal, an interval or a ratio scale. There is no agreement in the philosophical literature on the implications of the incommensurability of values for the comparability of objects or options (for a discussion, see [Chang, 1997; Hsieh, 2007]). It should be noted that the incommensurability of values does not as such entail the incomparability of options. For example, if all relevant values evaluate the same option as best in the absence of value conflict then incommensurability will not entail incomparability. In situations in which a lexicographical ordering of values exists, that is in situations where any small amount of one value is worth more than any large amount of another value, incommensurability will not lead to incomparability. We can then simply order the options with respect to the most important value and if two options score the same on this value we will examine the scores with respect to the second, less important, value. Even in cases of conflicting incommensurable values, between which no lexicographical ordering exists, there is no agreement on whether this is enough to entail the incomparability of options. In such cases the strongest argument for incomparability is probably the argument from small improvements [Raz, 1986; Chang, 1997] as illustrated in the following example. Suppose that one needs to choose between a career in engineering and a career in philosophy. Suppose too that there is good reason to judge that neither the philosophy nor the engineering career is better (i.e. more valuable overall) than the other. It is possible that we will still find that the philosophy career is not better than the engineering career, even if the former is slightly improved. Conversely, the engineering career may not be better than the philosophy career even if it is slightly improved. If this were the case then the two careers could not be said to be equally good because if that were so a slight improvement in one of the two would make it better than the other. Therefore the two options — a career in engineering and a career in philosophy — may in this case be said to be incomparable, because neither is better than the other and they are not equally good. The argument from small improvements can be interpreted in at least two, not necessarily mutually exclusive, ways. One solution is to interpret the incomparability as vagueness [Broome, 1997]. With this interpretation, on the scale of

978

Ibo van de Poel

overall goodness, options can be surrounded by a zone of mild indeterminacy; in this zone, the options are neither better nor worse than each other nor are they equally good. Another interpretation is to argue that the two careers are “roughly equal” or “on a par.” The latter has been proposed as a fourth comparative relation in addition to “better than,” “worse than,” and “equally good” [Chang, 1997]. With both interpretations, it is debatable whether they can erase incomparability between options. But even if they do not, both interpretations limit the extent of incomparability. Even if a specific career in engineering is not comparable with one in philosophy, a particularly good career in engineering may still be considered to be better than a particularly bad career in philosophy.3 In the argument given above, incommensurability has mainly been seen as a formal feature, as the impossibility to express or measure two values on a common scale. Some authors put forward a more substantive notion of incommensurability. Raz [1986], for example, has suggested that resistance to certain trade-offs is constitutive of certain values or goods. Consider, for example, the following trade-off: for how much money are you willing to betray your friend? It may well be argued that accepting a trade-off between friendship and financial gain undermines the value of friendship. On this basis it is constitutive of the value of friendship to reject the trade-off between friendship and financial gain. It has also been suggested that some values may resist trade-offs because they are ‘protected’ or ‘sacred’ [Baron and Spranca, 1997; Tetlock, 2003]. This seems especially true of moral values and values that regulate the relations between, and the identities of, people. One way to understand this phenomenon is by considering the notion of moral residue. Moral dilemmas — i.e. situations in which an agent is confronted with conflicting moral obligations and in which at least one moral obligation is not met whatever option one chooses — leave the agent with a moral residue or with a sense of guilt arising from the obligation not met [Williams, 1973; Marcus, 1980]. Similarly, trade-offs between protected values may create an irreducible loss because a gain in one value may not always compensate or cancel out a loss in the other. The loss of a good friend cannot be compensated by having a better career or more money. The point here is not that the value of friendship is lexicographically ordered over that of money or having a career (probably it is not). Even if a lexicographical ordering is absent, the nature of the values is so different that one cannot compensate for the other. Some philosophers have denied the existence of value incommensurability. They believe that all values can ultimately be expressed in terms of one overarching or super value. Utilitarianism often attributes such a role to the value of human happiness, but a similar role may be played by the value of the ‘good will’ in Kantianism or the value of ‘contemplation’ in Aristotelian ethics [Korsgaard, 1986].4 3 Such

comparisons are known as nominable/notable comparisons. way in which such super values operate may, however, be quite different in different approaches. In Kantianism, it is ‘good will’ that solves value conflicts by means of (practical) reasoning, while in utilitarianism value conflict is solved by expressing all values in terms of the super value ‘utility’. 4 The

Values in Engineering Design

979

The notion that there is ultimately only one value that is the source of all other values is known as value monism. Value monists do not necessarily deny the existence of more than one value but they do consider other values to be only conditionally good. A value monist believes that value conflicts can essentially be solved by having recourse to the super value. Nevertheless, value monists might recognize that there are practical or epistemological limitations to solving value conflicts. A utilitarian might, for example, believe that all values are only good in so far as they contribute to human happiness but she might recognize that in practice we cannot always exactly determine how much a specific value contributes to human happiness, so that some options might have an indeterminate ranking. 3 TECHNOLOGY AND VALUES Sometimes the thesis of technology being value-neutral is defended [Florman, 1987; Pitt, 2000]. The main argument usually given for this thesis is that technology is just a neutral means to an end which can be put to good or bad use. Value is thus created during use and is not located in technology. This also means that the objectionable effects of technology are to be blamed on the users and not on technological artifacts, or their designers. As the American National Rifle Association has expressed it: “Guns do not kill people, people kill people”. What does claiming that technology is value-neutral exactly entail? One interpretation would be to say that it means that the value of technological artifacts only depends on their extrinsic properties. In this interpretation, the thesis that technology is value-neutral is clearly false. It can be seen as follows. Technological artifacts have a physical or material component, in other words they are also physical objects, even if they are not mere physical objects. The value of physical objects as a means to an end depends — partly at least — on their intrinsic properties. A stone can be used to split a nut thanks to its intrinsic physical properties. A tree leaf would have a much smaller or no utility value when it comes to splitting nuts. Since it is implausible that the utility value of physical objects merely depends on their extrinsic properties, the same may be said of technologies.5 So the value of technological artifacts does not only depend on their extrinsic properties. The thesis that value is not intrinsic to technology may also be interpreted as implying that such value also partly depends on the extrinsic properties of a technology. To judge the plausibility of such a claim, it is crucial to define technology or technological artifacts because to a large extent that is what will determine what we consider to be the intrinsic and extrinsic properties of technological artifacts. Radder (this Volume, Part V), for example, presents a definition of technology in which the actual realizability of the function of a technological artifact is part of what technology actually is. This seems to make the actions of users internal to 5 It might be remarked that there is also such a thing as non-physical technologies, like software programs. It seems obvious that the utility value of such non-physical objects should also partly depend on the intrinsic properties of such objects, e.g. the intrinsic properties of the software program in question.

980

Ibo van de Poel

technology which, in turn, makes it more likely that at least some values will only depend on the intrinsic properties of technologies. In fact if we define technology sufficiently broadly, we can always make values internal to technology, it seems. But what happens if we start off with a minimal definition of technology? I think that any plausible minimal account of technology needs to refer to the notion of function, and/or comparable notions like ends, purposes and intentions. The fact technologies have a function implies that they have utility value, i.e. that they can be used for some end. On a minimal definition of technology, then, technology at least has utility value. This does not mean that such utility value is intrinsic to technological artifacts in the sense that it only depends on the intrinsic properties of technological artifacts. That, indeed, is not usually the case: the particular utility value of a particular hammer for driving nails into a piece of wood also depends, for example, on the physical abilities of users and such abilities are extrinsic to the hammer. So even if having utility value is part of what it means to be a technical artifact, that same utility value is not necessarily intrinsic to the technological artifact. We might express this by saying that having utility value is conceptually inherent to the notion of a technological artifact. The presence of such inherent values in technology seems to be a good reason for rejecting the thesis that technology is value-neutral. A further question is what types of value are precisely inherent to technology. I will answer this question below.

3.1 Instrumental value Whereas utility value is the value of being a means to any end, instrumental value is the value of being a means to a good end, of being a means to a positive value. A popular thesis is the one that holds that the instrumental value of technological artifacts, so defined, depends on the goals for which the artifact is used. Users may implement technology for good or bad. A macabre example of use for bad ends is the gas chambers designed by German engineers that were used to effect the planned extermination of the Jews held in the concentration camps during the Second World War. In this case the gas chambers clearly contributed to the morally objectionable goal of eradicating the Jews. So conceived, the gas chambers were a source of disvalue. They had a negative instrumental value but at the same time a positive utility value: they were useful for Nazi purposes. One might even argue that the larger the utility value of the gas chambers was, the greater their negative instrumental value would be.6 The distinction between utility value and instrumental value leads one to question how technological artifacts can achieve instrumental value or disvalue. One possibility is that they are thus invested during use: the gas chambers had a negative instrumental value due to the way in which they were used. After all, it is conceivable that this same technology could alternatively have been used to achieve morally good or neutral ends. In other words, technological artifacts only 6 This possibility is in fact a main reason why some philosophers deny that utility value is valuable at all.

Values in Engineering Design

981

initially have utility value. They acquire instrumental value or disvalue depending on how they are used. However, this thesis is problematic. In most cases technological artifacts are developed with certain uses in mind. It is hard to believe that the engineers who designed the gas chambers were unaware that they were intended for killing Jews. In view of these specific historical circumstances, it makes little sense to assert that the engineers designed the gas chambers and that subsequently, completely independent of the design efforts, it was decided to use these gas chambers to kill Jews. This point seems to have more general validity. Houkes and Vermaas argue that in the design process designers do not just design an artifact but also a use plan, that is, a plan in which the artifact functions as a means to achieve certain user ends [Houkes and Vermaas, 2004]. As a matter of fact, the gas chambers that were designed during the Second World War were part of an overall use plan aimed at eradicating the Jews. If it is true, as Vermaas and Houkes argue, that designing always involves designing a use plan, then user goals are not just added to the artifact later on but are intrinsic at least to the intentional history of a technological artifact. It is not so difficult to think of design processes in which an attempt is made to design an artifact that is particularly fit for achieving a certain positive value or good end. An example is the speed bump, which is intended to force drivers to slow down in built-up areas in order to increase safety. Speeds bumps do not literally force car drivers to drive more slowly but they certainly make it unattractive to drive fast. Speed bumps can thus be seen as an attempt to prescribe a certain moral maxim — do not drive too fast in a built-up area — for people, in order to uphold a positive moral value, in this case human safety. Speed bumps are thus not only designed to have utility value but also instrumental value. One reason why technological artifacts do not only have utility value but also instrumental value or disvalue is because the goals for which they are intended — and hence also the intended instrumental value or disvalue of an artifact — are not extrinsic to the artifact but are part of the intentional history of the artifact in question. Without such an intentional history, an object is not a technical artifact but rather a mere physical object [A similar position is defended by Franssen, this Volume, Part V, especially in Section 1]. Therefore intended instrumental value or disvalue is conceptually inherent to technology. This is not to deny that the actual instrumental value of a technological artifact may be different from the one intended in the design process due to the way the artifact is used. It remains true, after all, that artifacts may be used for ends not intended or foreseen by the designer. For this reason, instrumental value is conceptual inherent to technology even if it is not instrinsic to technology.

3.2

Economic value

The economic value of technological artifacts is often a main reason for developing them in a market economy. In economics, economic value is often viewed as the price (in monetary units) that people are willing to pay for tradable goods.

982

Ibo van de Poel

This value may be greater than the actual market price (so creating an economic surplus) but it is not usually less: otherwise, people would not buy the product. So conceived, the economic value of a product depends on the preferences of consumers. People might, of course, use it to create another positive value, but that contribution is not what gives the object its economic value. Therefore, economic value, in terms of the price people are willing to pay for goods, is a kind of utility value. Economic value is not entirely intrinsic to technological artifacts because it depends on the preferences of consumers which are external to the technological artifacts. Still, economic values may be said to be inherent to technological artifacts insofar as such artifacts are useful artifacts that serve a function. This usefulness seems to give technological artifacts an economic value, even if they are not actually traded. Technologies not only possess economic value as tradable goods but also as means of production. Many technologies facilitate the production of other economic goods and can therefore serve to generate income or profits. Technological innovation often makes it possible to create the same goods, or similar goods with the same function, for a lower price. If it is assumed that the economic value of the goods (i.e. the price that consumers are willing to pay ) produced by the new technology is the same as previously, this means that the producer can increase profits and/or the product becomes cheaper and, hence, more affordable for customers. For this reason, it might be argued that technology as a means of production does not only enable the preferences of producers and consumers to be fulfilled but it also has an overall positive value for human welfare. In as far as human welfare is considered a positive value, as it often is, the economic value of technology as a means of production is therefore not just a type of utility value but also something that is instrumentally valuable. Something similar applies to the economic value of infrastructural technologies. These are technologies such as roads, transportation, and information and communication technologies like the internet. Such technologies are not usually traded nor are they a direct means of production for other economic goods. They enable economic activities or make such activities easier or cheaper than they used to be. They may, for example, lower transaction costs or coordination costs. Again, this kind of economic value is a utility value from the point of view that it depends on the preferences of people, but it might be argued that it also contributes overall to other values such as human welfare, and so is instrumentally valuable.

3.3

Moral value

Technologies have utility or instrumental value because they have a consciously designed function. However, technological artifacts do not simply fulfill their function but in passing they also produce all kinds of valuable and harmful side-effects. Chemical plants produce useful substances but may also explode and kill people. Anti-conceptives are not only instrumental in preventing pregnancy, but they also

Values in Engineering Design

983

influence sexual morality and have an effect on emancipation. Automatic train ticket dispensing machines may be problematic for the elderly and may discourage them from traveling. Technologies thus have all kinds of effects, desirable and undesirable, beyond the goals for which they are designed or used. Through these side-effects, technologies create value or disvalue. New technologies can therefore be evaluated not only in terms of the value or disvalue they create through their (intended) use but also in terms of the value or disvalue created through their side-effects (see also the chapters by Hansson and by Grunwald in this Volume, Part V). Typical values that are relevant in this respect include safety, sustainability, human health, welfare, human freedom or autonomy, user-friendliness and privacy. All the referred to values are moral values since they are valuable for moral reasons. This does not mean that they are only valuable for moral reasons or that they are only strived for for moral reasons. In fact, it may be the case that an unsafe technological artifact is less useful than a safe one or is more difficult to sell. Most of the values in question will be enabling or contributory values, in other words values that enable people to live a good life or to contribute to making life good. Side-effects are often, but not always, unintentional. The designers and users of cars do not intend car accidents to happen nor do they — we may assume — intend to pollute the environment. Nevertheless, we know that the design and use of cars will result in car accidents and in environmental pollution. What is especially relevant here is the fact that different car designs and different modes of use will result in different numbers of car accidents and different degrees of pollution. This point can be generalized: given a desirable technological function or a certain user end, there are usually alternative ways to achieve that function or end. Usually these alternatives not only differ with respect to how effectively and efficiently they meet the formulated end or function, but also with respect to their side-effects, and hence with respect to the values upon which our evaluation of these side-effects are based. Sometimes side-effects are intentionally incorporated into artifacts. A famous but contested example is that of the low-hanging overpasses at Long Island in the USA [Winner, 1980]. According to Winner, Robert Moses — the architect who designed these overpasses — intentionally designed the road infrastructure in such a way as to prevent black people from reaching the beaches at Long Island. The idea was that black people could not afford private cars and would therefore have to use public transportation, i.e. buses that could not pass under the low-hanging overpasses (for a full discussion, see Radder’s chapter in this Volume, Part V, especially Section 4.1). It might be objected that the above analysis does not indicate that side-effects are either intrinsic or inherent to technology. Side-effects are not entirely intrinsic to technical artifacts because they partly depend on the ways and circumstances of use, even if they also partly depend on the intrinsic properties of technical artifacts. But are side-effects inherent to technological artifacts? If a technological artifact had no side-effects we would not stop calling it a technological artifact. Having

984

Ibo van de Poel

side-effects is, in other words, not a defining characteristic of technology and it is not therefore conceptually inherent to technology. I think, however, that it can be argued that side-effects are inherent to technology in a weaker sense. This weaker sense has to do with what I take to be a fact of the world in which we live: technological artifacts are never perfect in the sense that they always also produce effects other than their primary function or aim. Given this fact, side-effects — and the (dis)value they create — are inherent to technology not in a conceptual but in a practical or empirical sense.

3.4

Cultural and aesthetic values

A beautiful building or a carefully designed car is not only instrumentally valuable as a means for living or for transportation but it is also intrinsically valuable as a beautiful object. Technological artifacts may also acquire cultural value. They can be the bearers of meaning: a 2CV has a different cultural meaning than a Mercedes Benz. These two types of cars are not just different means to achieve more or less the same end (of going from A to B); they also embody different kinds of cultural values. As culturally valuable objects, cars are worthwhile in themselves, not just as means of transportation. Like economic and moral values, cultural and aesthetic values are not intrinsic to technological artifacts. The cultural value attached to an artifact is open to interpretation and may change in the course of time. Even ideals concerning what constitutes beauty may change in different historical periods and vary from culture to culture. If cultural and aesthetic values do indeed depend on interpretation, as is often argued, then such values will not be intrinsic to the valuable object but are rather relational properties of that object. The fact that the cultural and aesthetic value of technological artifacts depends on their relational properties should not be taken to mean that the attribution of such value to technological artifacts is arbitrary. Just as technological artifacts are often designed with a certain use plan in mind, so they are also often, but certainly not always, designed for a certain cultural or aesthetic value. Car design and styling is a good example of this. Given current user practices, cultural traditions, and social trends, one can to some extent predict what cultural or aesthetic value artifacts will acquire. This value does not only depend on the extrinsic, relational properties of the artifacts but also partly on their material, intrinsic properties. The intrinsic properties partly determine the aesthetic and cultural values of any artifact. Does this mean that cultural and aesthetic values are inherent to technology? It seems that technological artifacts can always be evaluated from an aesthetic or cultural point of view. However, this does not imply that they have cultural and aesthetic value as technological artifacts. A beautiful letter opener is instrumentally valuable as a technological artifact because it is intentionally designed and can be used to open letters. At the same time it might be a beautiful object and, as such, have intrinsic aesthetic value. Often, the aesthetic value of a technologi-

Values in Engineering Design

985

cal artifact will not be completely distinguishable from its utility or instrumental value. A beautiful letter opener might not just be a beautiful object, but it may also be beautiful as a letter opener, its beauty may be based on the specific way its function is translated into a certain form. It is not just beautiful but it is beautiful because it fulfils its function in a certain way. So, without understanding the function of the letter opener, its beauty may be unexplainable or even be absent. As Roger Scruton says about architecture: “ . . . our sense of the beauty in architectural forms cannot be divorced from our conception of buildings and of the functions they fulfil” [Scruton, 1979, p. 10]. The same is true of cultural values. A 2CV or Ferrari might amount to a certain expression of a way of living, and as such have cultural value, but this cultural value is often difficult, if not impossible, to understand or even absent altogether if one does not realize that it is a means of transport.

4

VALUES IN THE ENGINEERING DESIGN PROCESS

Design is aimed at the creation of useful things: it aims to create or achieve utility value. For the user, such utility value is achieved by embedding the artifact-to-bedesigned in a use plan. This use plan defines the functional requirements that the artifact should fulfill. These functional requirements, the function of the artifact, are achieved through a certain physical structure or make-up [Kroes, 2002]. During the design process, a translation is made from the desired function of the artifact to a physical structure. Put simply, the functional requirements are translated into technical specifications, which are embodied in a certain physical structure. Engineering design is thus the process by which certain functions are translated into a blueprint for an artifact, system, or service that can fulfill these said functions. Engineering design is usually a systematic process that makes use of technical and scientific knowledge. The design process is an iterative process that can be divided into different stages, like (see, e.g. [Pahl and Beitz, 1984]): 1. Problem analysis and formulation, including the formulation of design requirements and planning the design and development of the product, system, or service. 2. Conceptual design, including the creation of alternative conceptual solutions to the design problem and a possible reformulation of the problem. 3. Choosing one conceptual solution from a set of possible solutions. 4. Embodiment design. The chosen solution is worked out in structural and material terms. 5. Detail design. The design is further detailed, ending up with a design that can function as a blueprint for the production process.

986

Ibo van de Poel

Values can be relevant at all stages of the design process (cf. [van de Poel, 2005]). They are, however, most prominent and explicit during the first phase when the design requirements are formulated, and during the third phase when a specific design solution is chosen and trade-offs between different design requirements often have to be made. In Sections 4.1 and 4.2, these two phases will be analyzed in more detail.

4.1 Values and design requirements In the first phase of any engineering design process, requirements are formulated on the basis of the intended use of the artifact and the wishes of the client or user. In addition, economic constraints, legal requirements, technical codes and standards, as well as moral considerations will play their part. Below I will distinguish between functional design requirements and additional design requirements. A number of value issues will be discussed with respect to each of these requirements. 4.1.1

Functional design requirements

Functional design requirements are an indication of what the artifact-to-be-designed is supposed to do; they are expressed in functional language. For a pencil, for example, the possible functional requirements are ‘easy to hold,’ ‘does not smear,’ ‘point lasts’ and ‘does not roll’ (cf. [Wasserman, 1993]). During the design process, such functional design requirements are translated into technical specifications. In the case of the pencil, technical specifications may be expressed in terms of the length of the pencil, or the degree of hexagonality of the pencil. The functional design requirements are an expression of the intended utility value of the artifact-to-be-designed. They may be formulated by the client or the intended user; often, however, the designers will play an important role in translating the rather vague wishes and ideas of clients and prospective users into more concrete design requirements. The designers themselves may also formulate functional design requirements on the basis of a use plan. It should be noted that the formulation of functional design requirements is value-laden in itself. One issue is: for whom is one designing? As we have seen, the instrumental value of technological artifacts, i.e. their value as means to good ends, depends on how these artifacts are used and can be used. The instrumental value of a Kalashnikov is different from that of a cheap medicine to relieve AIDS in Africa. Of course, not all cases are so clear-cut, but choices concerning users and use plans are clearly value-laden. 4.1.2

Additional design requirements

Not all design requirements are based on the intended use or function of the artifact-to-be-designed. One source of additional design requirements is the possible different stakeholders who are affected by a technological artifact and who have

Values in Engineering Design

987

different desires, needs and interests. If one designs, for example, a production machine for paper clips, the client might be the company producing the paper clips. The direct users are the people who operate the production apparatus. In this case, there is also a range of indirect users and stakeholders including the paper clip production facility managers, the safety officer at the production facility, the paper clip users but also labor unions and environmental groups — the apparatus might cause environmental pollution. These stakeholders will place different requirements on the design, requirements that may conflict with each other or with those of clients or users. A second source of additional requirements is the larger socio-technical system within which the artifact that is designed will be embedded (see Bauer & Herder’s chapter in this Volume, Part III). For example, electrical apparatus has to be compatible with the electricity from the grid. The embedding of artifacts in larger socio-technical systems may also give rise to value issues. When designing a car, should one see the existing infrastructure for the distribution of gasoline as a fixed constraint and therefore design a car that runs on gasoline or should one — for example in conjunction with sustainability considerations — opt for a car that uses hydrogen or electricity and lobby for a infrastructure that fits the use of such cars? A third source of additional design requirements is moral considerations. Engineering codes of ethics, for example, suggest that engineers should hold “paramount the safety, health and welfare of the public” (for more on engineering codes of ethics, see Pritchard’s chapter in this Volume, Part V). This suggests that moral values should play an explicit role in the formulation of design requirements. A range of relevant moral values can be mentioned: safety, human health, human well-being, human welfare, privacy, autonomy, justice, sustainability, environmental care, animal health, animal well-being (cf. [Friedman et al., 2006]). Values are often too broad and vague to be used directly in the design process: they first have to be translated into more tangible design requirements. Often, different translations are possible. In the design of a chemical plant, one can look at the safety of employees and of people living close to the plant. Ethically, it would not be acceptable to limit safety to just the employees. Obviously, people in the direct vicinity of such a plant will experience the consequences of the design choices made without having the opportunity to agree or disagree or to benefit directly from the plant.

4.2

Trade-offs in design and value conflicts

While some design requirements are formulated as requirements that can be met or not — e.g. this electrical appliance should be compatible with 220V — others are formulated in terms of goals or values that can never be fully met. An example of the latter is safety. There is no such thing as an absolutely safe car: cars can only be safe to greater or lesser degrees.

988

Ibo van de Poel

If design requirements are formulated as strivings, they are more likely to conflict with each other. Cars that are made lighter in order to be more sustainable (less fuel consumption) are, for example, less safe [van de Poel and van Gorp, 2006]. The refrigerator coolant that replaced CFC 12 in most European household refrigerators — after the ban on CFCs — is more environmentally sustainable but flammable and therefore somewhat less safe than CFC 12 and other alternatives [van de Poel, 2001]. Most designs involve trade-offs between different design requirements. If the design requirements are motivated by different values these conflicts amount to value conflicts. Below I sketch two examples of value conflicts in engineering design. 4.2.1

Safety belts

A first example is the automatic seatbelt. A car with automatic seatbelts will not start if the automatic seatbelts are not put on. This forces the user to wear the automatic seatbelt. One could say that the value of driver safety is built into the technology of automatic seatbelts. This comes at a cost, however: the user has less freedom. Interestingly, there are various seatbelt designs which exist that would imply that hre are different trade-offs in terms of safety and user freedom. The traditional seatbelt, for example, does not enforce its use, but there are various systems that give a warning signal if the seatbelt is not being worn. This does not enforce seatbelt use, although it does encourage the driver to wear his seatbelt. 4.2.2

Refrigerants for household refrigerators

As a consequence of the ban on CFCs in the 1990s, an alternative had to be found to CFC 12 as a refrigerant in household refrigerators. Apart from utility value, three moral values played an explicit role in the formulation of design requirements for alternative coolants: safety, health and environmental sustainability. In the design process, safety was mainly understood as non-flammability, and health as non-toxicity. Environmental sustainability was equated with low ODP (Ozone Depletion Potential) and low GWP (Global Warming Potential). Both ODP and GDP mainly depend on the atmospheric lifetime of refrigerants. In the design process, a conflict arose between those three values. This value conflict can be illustrated with the help of Figure 1, which derives from a publication in the ASHRAE Journal of December 1987 by two engineers, McLinden and Didion, both employed by the National Bureau of Standards in the USA [McLinden and Didion, 1987]. For thermodynamic reasons, the most attractive coolants are hydrocarbons or CFC based on such hydrocarbons. Figure 1 is a graphic representation of CFCs based on a particular hydrocarbon. At the top, there is methane or ethane, or another hydrocarbon. If one moves to the bottom, the hydrogen atoms are replaced either by chlorine atoms (if one goes to the left) or fluorine atoms (if one goes to the right). In this way, all the CFCs based on a particular hydrocarbon are represented. The figure shows how the properties of flammability (safety), toxicity

Values in Engineering Design

989

Figure 1. Properties of refrigerants (health) and the environmental effects depend on the exact composition of a CFC. As can be seen, minimizing the atmospheric lifetime of refrigerants means maximizing the number of hydrogen atoms, which increases flammability. This means that there is a fundamental trade-off between flammability and environmental effects, or between the values of safety and of sustainability. 5 DEALING WITH VALUE CONFLICTS IN ENGINEERING DESIGN: OPTIMAL DESIGN When dealing with trade-offs, engineers are inclined to look for the best or optimal design solution. Below, I will investigate various approaches to optimal design: efficiency and effectiveness (Section 5.1), cost-benefit analysis (Section 5.2) and multiple criteria design analysis (Section 5.3).

5.1

Efficiency and effectiveness

A first-order approach to optimal design is to consider design to be optimal if it results in an artifact that optimally fulfills the desired function. But how do we know — or measure — whether a design optimally fulfills its function? Two measures come to the fore: effectiveness and efficiency. Effectiveness can be defined

990

Ibo van de Poel

as the degree to which an artifact fulfills its function. Efficiency could be defined as the ratio between the degree to which an artifact fulfills its function and the effort required to achieve that effect. As is pointed out in by Alexander (this Volume, Part V), efficiency in the modern sense is usually construed as an output/input ratio. The energetic efficiency of a coal plant may thus be defined as the ratio between the energy contained in the power produced and the thermal energy contained in the unburned coal. Historically, there seems to be a close connection between the rise of the modern notion of efficiency and optimal design. Frederick Taylor, for example, believed that production should be based on “the one best way” which, according to him, was simply the most efficient way [Taylor, 1911]. Two things need to be noted with respect to effectiveness and efficiency. Firstly, effectiveness and efficiency are different values that may well conflict. The design that most effectively fulfills its intended function may not necessarily be the most efficient one. A very effective vacuum cleaner that removes more dust than a less effective one may nevertheless be less energy-efficient, that is to say, it may use more energy per unit of dust removed than the less effective vacuum cleaner. So, we may be faced with a conflict between effectiveness and efficiency. A well-defined notion of optimal design requires a solution to this potential conflict. Secondly, effectiveness and efficiency are often very difficult to measure. Although this is partly a practical problem, this difficulty is often based on the more fundamental problem that often neither the function of an artifact (i.e. its output) nor the input can be uniformly formulated. This is witnessed, for example, by the fact that the desirable function of an artifact is often expressed in terms of a range of functional requirements, which may conflict. The following quote from Petroski about the design of paper clips illustrates this point: Among the imperfect things about the Gem [the classic paper clip, IvdP] that many a recent inventor has discovered or rediscovered when reflecting upon how the “perfected” paper clip is used to clip papers together are the following: 1. It goes only one way. Half the time, the user has to turn the clip around before applying. 2. It does not just slip on. The user first has to spread the loops apart. 3. It does not always stay on. The clip gets snagged on papers or other objects and gets pulled off. 4. It tears the papers. The sharp ends of the clip dig into the papers when it is removed. 5. It does not hold many papers well. The clip either twists badly out of shape or flies off the pile. 6. It bulks up stacks of papers. A lot of file space can be taken up by paper clips.

Values in Engineering Design

991

When a design removes one of the annoyances, it more likely than not fails to address some others or adds a new one. . . . All design involves conflicting objectives and hence compromise and the best designs will always be those that come up with the best compromise. Finding a way to bend a piece of wire into a form that satisfies each and every objective of a paper clip is no easy task, but that does not mean that people do not try. [Petroski, 1996, p. 29-30] This quote illustrates two points. One is that the ideal of optimal design or what Petroski calls “perfected” design is an important source of inspiration for designers. As long as the perfect paper clip does not exist, people will try to design it. The other is that in practice this ideal will probably never be achieved: the best is always the best compromise. The crucial question then is how to determine what the best compromise is. This requires trade-offs between the different requirements and it is unclear how we can make these trade-offs in a justified way (see also Kroes et al., this Volume, Part III). The actual situation is, however, even worse. Up until now, we have conceived of optimal design as design that optimally fulfills its intended function, or — put differently — as that which maximizes the (expected) utility value of the design. However, as argued in Section 3, the value of technological artifacts is not restricted to their utility value. The question that then arises is: what would it mean to try to maximize the overall value of technological artifacts during design, what would optimal design in such a broader sense amount to? Engineers have, in fact, dealt with this problem and have developed a number of approaches to the issue. The following two such approaches will be briefly discussed below: cost-benefit-analysis and multiple criteria design analysis.

5.2

Cost-benefit analysis

Cost-benefit analysis is a general method that is often used in engineering. What is typical of cost-benefit analysis is that all considerations that are relevant for the choice between different options are eventually expressed in one common unit, usually a monetary unit, like dollars or euros. Cost-benefit analysis may be an appropriate tool if one wants to optimize the expected economic value of a design. Still, even in such cases, some additional value-laden assumptions and choices need to be made. One issue is how to discount future benefits against current costs (or vice versa). The choice of discount rate may have a major impact on the outcome of the analysis. One might also employ different choice criteria once the cost-benefit analysis has been carried out. Sometimes all of the options in which the benefits are greater than the costs are considered to be acceptable. However, one can also choose the option in which the net benefits are highest, or the option in which the net benefits are highest as a percentage of the total costs. Cost-benefit analysis is more controversial if non-economic values are also relevant. Still, the use of monetary units does not mean that only economic values can

992

Ibo van de Poel

be taken into account in cost-benefit analysis. In fact, approaches like willingness to pay (WTP) have been developed to express values like safety or sustainability in monetary units (for some more details on WTP, see Hansson’s chapter in this Volume, Part V, Section 4.5 and Grunwald’s chapter in this Volume, Part V, Section 3.2.4). These approaches are often questioned but it would be premature to conclude that cost-benefit analysis necessarily neglects non-monetary or non-economic values. Moreover, when employing cost-benefit analysis, different ethical criteria might be used to choose between the options [Kneese et al., 1983; Shrader-Frechette, 1985]. One might, for example, choose an option with which nobody is worse off. By selecting a specific choice criterion, ethical considerations beyond considering which options bring the largest net benefits might be taken into account. In terms of values, cost-benefit analysis might be understood to be the maximization of one overarching or super value. Such a value could be an economic value like company profits, or the value of the product to users but it could also be a moral value like human happiness. If the latter is chosen, cost-benefit analysis is related to the ethical theory of utilitarianism. With Bentham’s classical variant of utilitarianism, for example, the assumption is that all relevant moral values can eventually be expressed in terms of the moral value of human happiness. One might question this assumption, however. One issue is that it is often difficult to indicate to what extent values like safety, health, sustainability, and aesthetics contribute to the value of human happiness, and to furthermore express this in monetary terms. A second, more fundamental issue, is that such an approach treats all these values as extrinsic values, whose worth should ultimately be measured on the basis of their contribution to the intrinsic value of human welfare. One might wonder whether values like human health, sustainability and aesthetics do indeed have only extrinsic value or are worthwhile in themselves. This potential objection to cost-benefit analysis amounts to an objection to the method being based on value monism. However, it might be argued that the above account of cost-benefit analysis is too substantive, from the point of view that it presupposes that the method is all about maximizing a specific value. Some proponents of cost-benefit analysis would probably maintain that it is merely a technical way of comparing alternatives in the light of heterogeneous considerations or values. The use of a common measure, they may admit, presupposes a common value but this value is merely formal, in other words, it is merely a means of comparison, rather than a substantial value. The claim, made above, that cost-benefit analysis presupposes value monism might thus be misconceived. I think two points need to be made in reply to such criticism. Firstly, even if cost-benefit analysis were merely a technical approach, interpretations of what this approach amounts to — even by most proponents of the approach – would often suppose a kind of value monism. Secondly, as a merely technical approach cost-benefit analysis might not indeed suppose value monism, but it does suppose value commensurability because it presumes that all values can be measured on a common scale. This may be a problem-

Values in Engineering Design

993

atic assumption (see Section 2.4 and Hansson’s chapter in this Volume, Part V, especially Section 4.5).

5.3

Multiple criteria design analysis

In multiple criteria analysis, different options are compared with each other in the light of several criteria. I focus here on one specific approach that is often used in engineering design: the method of weighted objectives. With this method, the relative importance of the criteria is first determined, because usually not all criteria are equally important. Next, each option is weighed for all the criteria and a numeric value is awarded, for example on a scale from 1 to 5. Finally, the value fi ∗ vij for each option is calculated according to the following formula: wj = over I, where wj is the value of the j th option, fi is the relative weight of the ith criterion, and vij is the score of the j th option on the ith criterion. The option with the highest value is then selected. The method can be illustrated using the case of coolants discussed in Section 4.2. I mentioned there three relevant values in the designing of coolants: safety, health and environmental sustainability. As we have seen, these values conflict in terms of the choice to be made between various coolants. How would the method of weighted objectives proceed in a case like this? The most simple and straightforward approach is to conceive of each of these values as a decision criterion. Table 1 gives a hypothetical interpretation of the choice between three of the alternatives that were considered: the traditionally used CFC 12; HFC 134a, the main alternative proposed by the chemical industry; and isobutene, an alternative proposed by environmental groups because it contributes less to greenhouse warming than HFC 134a, but is flammable. Safety (flammability)

Health (toxicity)

Environmental sustainability (atmospheric lifetime)

Total score

Weight of criterion

2

1

2

CFC 12

5

5

1

17

HFC 134a

4

4

3

18

Isobutane (HC 600a)

1

4

5

16

Table 1. Hypothetical application of the method of weighted objectives to the coolants case

994

Ibo van de Poel

Table 1 suggests that we should choose HFC 134a. But how sound is this advice? The first thing to note is the use of the scoring scale 1 — 5 for each decision criterion. The way this scale is used to calculate the overall worth of each alternative means that this scale is interpreted as a ratio scale for each of the design criteria.7 This means that it should make sense to say that HFC 134a scores three times as well on the criterion of environmental sustainability as CFC 12. If seems beyond debate to claim that HFC134a scores better on environmental sustainability than CFC 12 — so we can order the alternatives on an ordinal scale — but can we say that it does three times as well according to this criterion? To do this we would require an operationalisation of environmental sustainability that we can measure. Such operationalisation still seems to be lacking. Even if we can compare the atmospheric lifetime of two substances, it is not obvious that changing the atmospheric lifetime of a substance by a factor of 2 would correspond to a similar change in the environmental sustainability of that substance. The method of weighted objectives also suggests that the weights of the criteria can be measured on a ratio scale. This can, for example, be attained by asking the marginal rate of substitution question: “By how much should fi be increased to compensate for a loss of one unit in f1 ?,” in which fi refers to the weight of the ith criterion and f1 to the weight of the criterion that is selected as a metric case [Otto, 1995, p. 97]. Obviously, it is only possible to answer this question if the design criteria, and the underlying values, are commensurable. The method of weighted objectives is not, of course, the only multiple criteria method that can be applied in engineering. Any of these methods must, however, deal one way or another with value trade-offs and that seems to presuppose some form of value commensurability. Moreover, Franssen [2005] has shown that with all multiple criteria methods it is very hard to make justifiable trade-offs; as Kroes, Franssen and Bucciarelli conclude in their contribution to this volume: “. . . there is no general rational procedure for making trade-offs in engineering design” (see Kroes et al., this Volume, Part III).

5.4 The ideal of optimal design As we have seen, optimizing approaches to value conflicts in engineering are likely to come up against formal and substantive problems. Philosophically, these problems are mainly attributable to the fact that it is often impossible to identify one overarching or super value (value pluralism) and to value incommensurability. We should not, however, conclude from this that the ideal of optimal design has no importance whatsoever. Firstly, the ideal of optimal design often motivates and guides the design process. Since, at the start of the design process, engineers do not yet know what is technically feasible, the ideal of optimal design — and especially the more specific 7 If it is interpreted as an ordinal or an interval scale, it can easily be shown that the method does nor result in one but in a number of potentially conflicting overall orderings (see also [Franssen, 2005]).

Values in Engineering Design

995

design requirements that are relevant to optimal design — is important in looking for new technical possibilities. Secondly, the ideal of optimal design often spurs on investigations into technical parameters that are relevant to design improvement. This helps to provide a better understanding of the design problem while improving the resulting design. Thirdly, design problems can often be subdivided into smaller problems. Some of the smaller problems might be so well-defined that optimal solutions do exist. Fourthly, the costs of ignoring the philosophical problems related to value pluralism and value incommensurability and, for example, to carry out a cost-benefit analysis may be less than the costs of selecting another choice mechanism or just picking one design (cf. [Sunstein, 2005, p. 371]). This is especially the case in choice situations where one type of values, for example economic values, is obviously more important than others. Even in cases where there are a number of conflicting values, the crucial question is whether there are other, non-optimizing approaches that fare better than optimizing approaches. In the next section, I will suggest a number of such alternative, non-optimizing, approaches and discuss some of their advantages and weaknesses. 6

NON-OPTIMIZING APPROACHES TO VALUE TRADE-OFFS IN ENGINEERING DESIGN

Non-optimizing approaches are not alien to engineering. In fact, a number of authors have argued that it is not possible to optimize in engineering design (e.g. [Simon, 1973; Cross, 1989; Sch¨ on, 1992; Simon, 1996]). A major argument for the impossibility of optimizing in engineering given by Herbert Simon and Nigel Cross has to do with the ill-defined nature of engineering design problems. Simon lists a number of characteristics that problems should have if they are to be well-defined [Simon, 1973]. Three characteristics that are especially relevant in relation to engineering design are these: 1. A clear criterion needs to be available to judge possible solutions and this criterion can be applied uniformly; 2. A problem space can be defined in which the initial state, the desired state and all possible interim states — that can be considered or achieved during problem solving — can be represented; 3. Possible actions or solutions can be represented as transitions between different problem space states. Insofar as actions affect the real world — and are thus outside a formal language or play — the representation should match the natural laws of the external world. Most design problems do not meet these criteria. A clear and uniformly applicable choice criterion is not usually available, as is clear from the discussion presented in Section 5. Moreover, the problem space is not usually well-defined.

996

Ibo van de Poel

In engineering design we frequently do not have an overview of the complete set op possible solutions, let alone a representation of this set in a well-defined problem space. Design problems are thus usually ill-defined. The ill-defined nature of design problems makes optimizing difficult, if not impossible. There are, however, alternative, non-optimizing strategies in engineering design. One non-optimizing strategy that was first described by Simon and which engineers are reported to follow is satisficing involving the selection of an alternative that is not optimal but ‘good enough.’ Satisficing is also a possible strategy for dealing with conflicting values, as already noted by Simon [1955]. Other nonoptimizing strategies are possible with respect to conflicting values in engineering pertaining to reasoning about values, technological innovation and the choice of a diversity of products. Each of these strategies will be discussed below.

6.1 Satisficing In contrast to an optimizer, a satisficer does not look for the optimal option but first sets an aspiration level with respect to the options that are good enough and then goes on to select any option that exceeds that aspiration level [Simon, 1955, 1956, 1976]. Designers are reported to be satisficers in the sense that they set threshold values for the different design requirements and accept any design exceeding those thresholds [Ball et al., 1994]. So conceived, satisficing may also be seen as a way of dealing with conflicting values, i.e. by setting thresholds for each value and then selecting any option that exceeds those thresholds. Setting threshold values does not only occur in the design process but also in legislation and the formulation of technical codes and standards. An example of satisficing is to be found in the earlier-discussed case of the design of new refrigerants. On the basis of Figure 1, the engineers McLinden and Dion, drew more specific figure with respect to the properties of CFCs (Figure 2). According to McLinden and Didion the blank area in the triangle contains refrigerants that are acceptable in terms of health (toxicity), safety (flammability) and environmental effects (atmospheric lifetime). This value judgment is a type of satisficing because by drawing the blank area in the figure, McLinden and Didion — implicitly — establish threshold values for health (toxicity), safety (flammability) and the environment. These thresholds were partly based on technical codes and standards. For example, the threshold value for safety — non-flammability — was based on the then existing ASHRAE Code for Mechanical Refrigeration (ASHRAE Standard 15-1978), which prohibited the use of flammable coolants in equipment intended for household applications. Satisficing can also be combined with optimizing. For example, a designer who has to trade off safety and cost considerations when designing a chemical installation may well choose to make a design that meets the legal requirements with respect to safety and is as cheap as possible. This can be interpreted as satisficing behavior with respect to the value of safety, while optimizing with respect to cost within the safety constraints.

Values in Engineering Design

997

Figure 2. Properties of refrigerants. [Figure from McLinden and Didion, 1987] Philosophically, the important question is whether, and if so when, satisficing is a morally and rationally permissible strategy. If someone satisfies he does not aim for the best, but for an option that is good enough from a certain point of view. Some ethicists have argued that satisficing with respect to moral values might be allowable: in many situations we are not required to do what is morally best, but we should at least do what is morally good enough (see, e.g. [Stocker, 1990; Dancy, 1993]). Risking one’s life to save another person from a burning house might be morally praiseworthy, but that does not mean that it is morally required. So, not everything that is morally praiseworthy is also morally required because that might just demand too much of someone. This phenomenon is known as moral supererogation. One argument for why satisficing is not only allowed but may even be advisable in the case of moral values might go as follows. Moral values sometimes resist trade-offs as we have seen in Section 2.4. One possible explanation for this is that they may often be understood as moral obligations [Baron and Spranca, 1997], as the obligation to meet a certain value to a certain minimal extent. If interpreted thus moral obligations define thresholds for moral values. It seems plausible that below the threshold the moral value cannot be traded-off against other values because the moral obligation is more or less absolute; above the threshold, tradeoffs may be allowed. If this picture is right, it provides an additional argument for satisficing with respect to moral values: for each of the relevant moral values the threshold should be set in such a way that the corresponding moral obligation is at least met. If this is done, unacceptable trade-offs between moral values or between moral values and other values can be avoided.

998

Ibo van de Poel

Do the above arguments also apply to non-moral values? An important argument for the existence of moral supererogation is that we have other values (and reasons) besides moral values (and reasons), and that these may go against what is morally most praiseworthy. It is, for example, obvious that I have reasons not to risk my life to save somebody else from a burning house, even if I have good reason to believe that it would be best from a moral point of view to risk my life. It is the existence of such other reasons that can justify moral satisficing. The justification here is based on the — presumed — existence of a broader perspective that includes the moral perspective, for example the perspective of the entire life of a person. Can satisficing also be justified at the broadest perspective level, like the perspective of an entire life? Some philosophers, preeminently Michael Slote, believe that it can [Slote, 1984; 1989]. Slote argues that it can be rationally justifiable to forego the best choice even if we know what the best choice is and even if it is readily available; he calls this ‘rational supererogation.’ Slote’s argument is strongly contested, however (e.g. [Pettit, 1984; Schmidtz, 1995; Byron, 1998]). What makes it especially problematic is his claim that it is rationally allowed to choose some lesser option over an available better option even though we have no overriding reason for doing so. It seems that we either have an overriding reason to choose the lesser option, which makes it not the lesser option all things considered, or we are simply not rationally allowed to choose the lesser option. The argument against Slote’s position suggests that satisficing cannot be rational in the broadest perspective. It can only be rational with respect to a partial perspective; satisficing on such a sublevel can be rational because, seen from the wider perspective, it is the best way to achieve one’s overall values or aims. Some philosophers have therefore suggested that at the highest level we are always optimizing if we are rational, at least implicitly and tacitly [Byron, 1998]. The idea of implicit and tacit optimization, however, seems problematic. It suggests that we optimize even if we are not aware of it. However, this makes it impossible to empirically distinguish optimizing from non-optimizing, so that the thesis that we optimize cannot be empirically falsified. Maybe, however, the thesis is not intended to refer to an empirical fact but rather to a conceptual truth: the conceptual nature of rationality leaves no other option than to optimize at the highest level if we want to be rational. This however, is a very bold assertion. To make it plausible we should at least argue how people could, if they would wish, optimize at the highest level, and that they could do so always if they would wish so. One reason for doubting whether such an argument can be given has to do with the existence of plural and conflicting values and value incommensurability. This places doubt on the possibility of optimizing as we have seen: the notion of a best option may simply not be well-defined. Another relevant issue with respect to the acceptability of satisficing is that of whether we are considering a static or dynamic context [Simon, 1956; Schmidtz, 1995]. In static contexts, all the options are known, the consequences of the options are known with a certain probability and the options are readily available.

Values in Engineering Design

999

Slote sets out to argue that satisficing is rationally allowed in a static context. Such an argument is very hard — if not impossible — to substantiate, but what of the rational acceptability of satisficing in a dynamic context? In a dynamic context we are either not fully aware of all the options, or it requires effort to investigate the consequences of options or to make them available. In such a context, we are confronted with the following question: how much effort should be put into getting to know the solution space better? The effort may pay off because there is a chance that we will discover a better option than the options we already know. However, there are limits: in situations in which the solution space is not closed, we can go on endlessly searching for a better solution but at some point the results no longer justify the effort. In such a dynamic situation, satisficing may be a useful and rationally defensible stopping rule: look for a better option until you have found an option that meets all threshold values.8 What does this tell us about the acceptability of satisficing in engineering design? Firstly, it suggests that satisficing with respect to moral values – or more specifically morally motivated design requirements — can be permissible due to the phenomenon of moral supererogation. Secondly, it suggests that satisficing with respect to other values and design requirements can be rationally justified from a broader perspective. In the case of the design of a part or component, this broader perspective can be the design process of the entire artifact. In the case of a design process for an artifact the broader perspective can be the sociotechnical system in which the artifact is embedded. The broader perspective can also be the company that wants to make a profit with a certain design or it can be society at large that aims to sustain certain values through technology. Thirdly, satisficing can be rational in a dynamic context where the solution space is not closed. As we have seen this is often typically the case in engineering design. Satisficing can therefore provide a rationally defensible stopping rule for the search process that engineering design is.

8 It should be noted that it is also possible to formulate an optimizing stopping rule. A good candidate for the optimizer’s stopping rule is: stop looking for new options as soon as the expected value of finding a better option is lower than the expected costs of any further searching process. This stopping rule is different from that of the satisficer in that it requires calculations to decide whether to continue the search or not; such calculating takes time and might prove counterproductive. So the optimizer needs a stopping rule concerning the time spent on calculating whether it is worth continuing the search. If the optimizer chooses a satisficing stopping rule for the making of the calculation, he is not optimizing any longer; but, if he chooses an optimizing stopping rule he has to make even more calculations because the time to be spent on the other calculations needs to be accounted for (even though these calculations presumably take less time than if he had not applied the stopping rule). The point is this: at least in some circumstances, the time it takes the optimizer to make all these decisions is simply not worth the effort. The optimizer is usually, if not always, better off if, at some level, he chooses a satisficing stopping rule. He does not know, however, when it is the right time to choose the satisficing stopping rule and in that sense he is not optimizing even if he chooses the satisficing rule because he wants to optimize. Wanting to optimize is, after all, not the same as optimizing.

1000

6.2

Ibo van de Poel

Reasoning about values

The approaches to dealing with conflicting values that have already been discussed — efficiency, cost-benefit analysis, multiple criteria analysis and satisficing — are all calculative approaches. They strive to operationalize and measure value in one way or another. Of these approaches, satisficing does not aim at calculating the overall worth of an option, but it does presuppose that the worth of an option can be measured for each of the individual design requirements. I will now look at approaches that do not share this calculative approach, but which emphasize judgment and reasoning about values. In the philosophical literature on value incommensurability, certain authors maintain that the presence of value incommensurability does not impair our ability to compare options because we can exercise judgment (e.g. [Stocker, 1990]). Often precisely what such judgment implies and how it could lead to a justified choice in situations of value conflict remains unclear. Nevertheless, a number of things can be said about what such judgment could imply. The first thing to do when one wants to exercise judgment in cases of value conflict is to gain a better understanding of the values at stake. What do these values imply and why are these values important? Take the value of freedom in the case of safety belts. Freedom can be construed as the absence of any constraints on the driver; it then basically means that people should be able to do what they want. Freedom can, however, also be valued as a necessary precondition for making one’s own considered choices; so conceived freedom carries with it a certain responsibility. In this respect it may be argued that a safety belt that reminds the driver that he has forgotten to use it does not actually impede the freedom of the driver but rather helps him to make responsible choices. It might perhaps even be argued that automatic safety belts can be consistent with this notion of freedom, provided that the driver has freely chosen to use such a system or endorses the legal obligation for such a system, which is not unlikely if freedom is not just the liberty to do what one wants but rather a precondition for autonomous responsible behavior. One may thus think of different conceptualizations of the values at stake and these different conceptualizations may lead to different possible solutions to the value conflict. A second judgment step would be to argue for specific conceptualizations of the relevant values. Some conceptualizations might not be tenable because they cannot justify why the value at stake is worthwhile. For example, it may be difficult to argue why freedom, conceived of as the absence of any constraint, is worthwhile. Most of us do not strive for a life without any constraints or commitments because such a life would probably not be very worthwhile. This is not to deny the value of freedom; it suggests that a conceptualization of freedom only in terms of the absence of constraints misses the point of just what is valuable about freedom. Conceptualizations might not only be untenable for such substantial reasons, they may also be inconsistent, or incompatible with some of our other moral beliefs.

Values in Engineering Design

1001

A third step in judgment is to look for the common ground behind the various values that might help to solve the value conflict. Taebi and Kloosterman [2008], for example, have argued that various trade-offs in nuclear energy can be reduced to a trade-off between the present and the future, and can thus be best understood in terms of the notion of intergenerational justice. One might argue that looking for common ground between different values presupposes a form of value monism. I suppose this is true. It should be stressed, however, that this kind of value monism is different from that presupposed by cost-benefit analysis or utilitarianism. In the latter cases, the presupposition is that all values can be expressed in terms of an overarching value, like human welfare. Here the value monism is more a monism of reasons. It is the type of value monism that Kant seems committed to. According to Korsgaard, Kant recognizes only one unconditionally good value as the source of all other values and that is: ‘good will’ [Korsgaard, 1986]. It is likely that Kant would maintain that good will can solve all value conflicts, at least in principle. This seems too optimistic to me, but that does not reduce the need to look for common ground between values. Even if such common ground cannot always be found, it may be available in specific cases.

6.3

Innovation: value sensitive design

The previous approach treats the occurrence of value conflict merely as a philosophical problem to be solved by philosophical analysis and argument. However, in engineering design value conflicts may also be solved by technical means. That is to say, in engineering it might be possible to develop new, not yet existing, options that solve or at least ease the value conflict situation. In a sense, solving value conflicts by means of new technologies is what lies at the heart of engineering design and technological innovation. Engineering is able to play this part because most values do not conflict as such, but only in the light of certain technical possibilities and engineering design and R&D may be able to change these possibilities. An interesting example is the design of a storm surge barrier in the Eastern Scheldt estuary in the Netherlands. After a huge storm-flood hit the Netherlands in 1953, killing more than 1,800 people, the government decided to dam up the Eastern Scheldt as part of what came to be known as the Delta plan. The main value taken into account in the Delta plan was safety. The closure of the Eastern Scheldt was scheduled to start in the early seventies, as the final part of the Delta plan. However, by that time it had led to protests in conjunction with the ecological value of the Eastern Scheldt estuary, which would in that way be destroyed. Many felt that the ecology of the estuary should be considered. Eventually, a group of engineering students devised a creative solution that would meet both safety and ecological concerns. The idea was to construct a storm surge barrier that would be closed only in cases of storm floods. Eventually this solution was accepted as a creative, though more expensive, option that took into account both the values of safety and ecology.

1002

Ibo van de Poel

One approach that takes into account the possibility to solve, or at least ease, value conflicts through engineering design is Value Sensitive Design. Value Sensitive Design is an approach that aims at integrating values of ethical importance in a systematic way into engineering design [Friedman, 1996; Friedman and Kahn, 2003; Friedman et al., 2006]. The approach aims at integrating three kinds of investigations: conceptual, empirical and technical. Conceptual investigations aim, for instance, at clarifying the values at stake, and at making trade-offs between the various values. Conceptual investigations in Value Sensitive Design are similar to the kind of investigations that I described in Section 6.2. What Value Sensitive Design adds to this are empirical and technical investigations. Empirical investigations “involve social scientific research on the understanding, contexts, and experiences of the people affected by technological designs” [Friedman and Kahn, 2003, p. 1187]. It is not hard to see why this is relevant: people’s experiences, contexts and understanding are certainly important when it comes to appreciating precisely what values are at stake and how these values are affected by different designs. What remains somewhat unclear is just how the proponents of Value Sensitive Design see the relationship between conceptual and empirical investigations. For example, is it important how people perceive a value or how it should be understood on conceptual grounds? I would argue that people’s understanding of values is not irrelevant but that it should not be taken at face-value either, people might err after all. One could, for example, require people to justify their understanding of the values at stake in a broad reflective equilibrium [Daniels, 1979; 1996; Rawls, 1971/1999]]. This could also provide a model for integrating conceptual and empirical investigations; a model that still seems to be lacking in the literature on Value Sensitive Design. Technical investigations “involve analyzing current technical mechanisms and designs to assess how well they support particular values, and, conversely, identifying values, and then identifying and/or developing technical mechanisms and designs that can support those values” [Friedman and Kahn, 2003, p. 1187]. The second part is especially interesting and relevant because it provides the opportunity to develop new technical options that more adequately meet the values of ethical importance than do current options. As the example of the Eastern Scheldt barrier shows, technical investigations may also ease value conflicts. Usually, however, technical innovation will not entirely solve value conflicts, so that choices between conflicting values still have to be made. In this respect, Value Sensitive Design only presents a partial solution to value conflicts in engineering design.

6.4 Diversity, genre and value holism All approaches to value conflict discussed so far presuppose that only one option is to be chosen. This presupposition is indeed true of most specific product design processes. If we zoom out from this perspective, a somewhat different picture emerges though: engineering provides society with different products that have

Values in Engineering Design

1003

roughly the same function. A number of possible justifications can be given for this diversity. One is that different people have different needs and preferences. A can opener, which is useful for the average citizen, may not be the most apt device for the elderly who usually have less power in their hands. Due to divergences in, for example, physical make-up, different people have different needs, a fact of life which may justify the design of different products with roughly the same function for different groups of users. People may also have different preferences. Some people will prefer a fast but expensive car to a less expensive but slower car, whereas others will have opposite preferences. Even if not all preferences are justified or to be fulfilled, as such there is nothing wrong with differences in preferences and a diversity of products with roughly the same function may be instrumental in fulfilling such diverse preferences. Given the differences between people in terms of their needs and preferences, the existence of a diversity of technological products with roughly the same function may be one way of increasing the utility or economic value of technological products. The total utility or economic value for society is probably larger if a range of products is provided rather than just one. A second possible justification is the existence of cultural differences and different cultural and aesthetic traditions. As we have seen, technological artifacts have meaning and may express certain cultural or aesthetic ideals. This is clearly visible in architecture, where various traditions matched to different evaluation standards exist. In such cases evaluation tends to be genre-specific: we identify a work as an instance of a genre and judge it by the standards of the genre. As Joseph Raz expressed: We can admire a building, and judge it to be an excellent building for its flights of fancy, and for its inventiveness. We can admire another for its spare minimalism and rigorous adherence to a simple classical language. We judge both to be excellent. Do we contradict ourselves? Not necessarily, for each displays the virtues of a different architectural genre — let us say, romantic and classical. [Raz, 2003, p. 45] Of course, the relationship between a work and its genre is not always straightforward, since works can also ironically or ambiguously relate to a genre or to more than one genre at the same time. Nevertheless, also in such cases, we tend to judge the work in terms of how it is allied to existing genres and to the standards of excellence inherent in those genres [Raz, 2003, p. 41-42]. Typically, different genres — for example in architecture — do not often differ merely in terms of the general relevant values, but also in terms of the preferred ‘mix’ or ‘ideal’ combination of values [Raz, 2003, p. 39]. In such cases, an optimizing approach that seeks to optimize one overarching value or each of the values in isolation may actually destroy this ‘ideal’ mix and create less value instead of more. A satisficing approach might not be appropriate either because what is valued is not a certain degree of each value but a specific combination of values. Such cases of value holism, where we cannot reasonably appreciate the values in isolation of

1004

Ibo van de Poel

each other, usually require educated judgment, which is genre-specific. Therefore, with respect to values that are genre-specific — especially some cultural and aesthetic values — it may be worthwhile having a number of technological artifacts with roughly the same function that each excel in their own genre. One might wonder whether such diversity is also worthwhile in the case of moral values. The idea of genre-based evaluation seems much harder to defend with respect to moral values than with respect to cultural or aesthetic values. It might be defensible with respect to perfectionist moral values, in other words those that go beyond what is morally required. The value of a good life, conceived as a moral value, may also be genre-specific. However, genre-specific evaluation seems hard to defend with respect to minimum moral standards. Still, it might be argued that minimum moral standards are not universal but situation-specific (cf. e.g. [Dancy, 1993]). For example, the minimum moral standard for environmental sustainability or animal welfare may be higher in a society of abundance than in a society of scarcity, especially if meeting high standards in, for example, animal welfare would involve further deterioration in living conditions for humans. This suggests that the thresholds for moral values — if one employs a satisficing approach — cannot been established completely independently of each other. This entails some degree of value holism; we cannot appreciate the values in complete isolation of each other. However, it does not extend to the type of genre-specific evaluations Raz is thinking of. 7

CONCLUSION

I began this contribution with the observation that it is the creation of value that lies at the heart of the engineering design process. We may now conclude that value conflict is in fact at the heart of the design process. In many cases, value conflict is the engine that fuels innovation and design, as is underlined by an approach such as the Value Sensitive Design approach. I have discussed various approaches to value conflict in engineering design. None of them is obviously superior to any of the others. The discussion suggests though that the approach that perhaps turns out to be most fruitful will depend, partly at least, on the kinds of values that are at stake. Cultural and aesthetic values are often genre-specific and will often consist of ideal combinations. Optimizing or satisficing approaches are probably of little help in such cases. Rather, one might adopt the approach that I have described under the heading of diversity. Most moral values, on the other hand, are not genre-specific. Here we might initially try to solve, or at least ease, a moral value conflict by employing Value Sensitive Design or by reasoning. A possible resultant value conflict might be dealt with by satisficing which can, as we have seen, amount to a justified approach to moral values. Optimizing might be a less desirable approach, especially if we are dealing with heterogeneous moral values that resist trade-offs. With economic values and other utility values, optimizing approaches might be fruitfully applied, even if such approaches still come up against a number of methodological problems. As we have seen, under certain conditions satisficing

Values in Engineering Design

1005

approaches and diversity might also be useful approaches for utility and economic value. BIBLIOGRAPHY [Ball et al., 1994] L. J. Ball, J. S. B. T. Evans and I. Dennis. Cognitive processes in engineering design: A longitudinal study. Ergonomics 37, 1753-86, 1994. [Baron and Spranca, 1997] J. Baron and M. Spranca. Protected values. Organizational Behavior and Human Decision Processes, 70, 1-16, 1997. [Broome, 1997] J. Broome. Is incommensurability vagueness? In Incommensurability, Incomparability and Practical Reason, R. Chang ed. pp. 67-89, Harvard University Press, 1997. [Byron, 1998] M. Byron. Satisficing and optimality. Ethics 109, 67-93, 1998. [Chang, 1997] R. Chang, ed. Incommensurability, Incomparability, and Practical Reasoning. Harvard University Press, 1997. [Cross, 1989] N. Cross. Engineering Design Methods. John Wiley & Sons, 1989. [Dancy, 1993] J. Dancy. Moral Reasons. Blackwell Publishers, 1993. [Daniels, 1979] N. Daniels. Wide reflective equilibrium and theory acceptance in ethics. Journal of Philosophy 76, 256-82, 1979. [Daniels, 1996] N. Daniels. Justice and Justification. Reflective Equilibrium in Theory and Practice. Cambridge University Press, 1996. [Florman, 1987] S. C. Florman. The Civilized Engineer. St. Martin’s Press, 1987. [Frankena, 1973] W. K. Frankena. Ethics. Prentice Hall, 1973. [Franssen, 2005] M. Franssen. Arrow’s theorem, multi-criteria decision problems and multiattribute preferences in engineering design. Research in Engineering Design 16, 42-56, 2005. [Friedman, 1996] B. Friedman. Value-sensitive design. Interactions 17-23, 1996. [Friedman and Kahn, 2003] B. Friedman and P. H. J. Kahn. Human values, ethics and design. In Handbook of Human-Computer Interaction, J. Jacko and A. Sears eds., pp. 1177–201, Lawrence Erlbaum Associates, 2003. [Friedman et al., 2006] B. Friedman, P. H. J. Kahn and A. Borning. Value sensitive design and information systems. In Human-Computer Interaction in Management Information Systems: Foundations, P. Zhang and D. Galletta eds., pp. 348-72, M.E, Sharpe, 2006. [Houkes and Vermaas, 2004] W. Houkes and P. E. Vermaas. Actions versus functions. A plea for an alternative metaphysics of artefacts. Monist 87, 52-71, 2004. [Hsieh, 2007] N.-h. Hsieh. Incommensurable values. In The Stanford Encyclopedia of Philosophy (fall 2007 edition), E. N. Zalta ed. http://plato.stanford.edu/archives/fall2007/entries/valueincommensurable/, 2007. [Kneese et al., 1983] A. V. Kneese, S. Ben-David and W. D. Schulze. The ethical foundations of benefit-cost analysis. In Energy and the Future, D. MacLean and P. G. Brown eds., pp. 59-74, Rowman and Littefield, 1983. [Korsgaard, 1983] C. M. Korsgaard. Two distinctions in goodness. Philosophical Review 92, 169-95, 1983. [Korsgaard, 1986] C. M. Korsgaard. Aristotle and Kant on the source of value. Ethics 486-505, 1986. [Kroes, 2002] P. Kroes. Design methodology and the nature of technical artefacts. Design Studies 23, 287-302, 2002. [Marcus, 1980] R. B. Marcus. Moral dilemmas and consistency. Journal of Philosophy 77, 12136, 1980. [McLinden and Didion, 1987] M. O. McLinden and D. A. Didion. Quest for alternatives. ASHRAE Journal 32-42, 1987. [Moore, 1912] G. E. Moore. Ethics. Oxford University Press, 1912. [Otto, 1995] K. N. Otto. Measurement methods for product evaluation. Research in Engineering Design 7, 86-101, 1995. [Pahl and Beitz, 1984] G. Pahl and W. Beitz. Engineering Design: A Systematic Approach. Springer-Verlag, 1984. [Petroski, 1996] H. Petroski. Invention by Design. Harvard University Press, 1996. [Pettit, 1984] P. Pettit. Satisficing consequentialism. Proceedings of the Aristotelian society suppl. 58, 165-76, 1984.

1006

Ibo van de Poel

[Pitt, 2000] J. C. Pitt. Thinking About Technology: Foundations of the Philosophy of Technology. Seven Bridges Press, 2000. [Rawls, 1999] J. Rawls. A Theory of Justice. The Belknap Press of Harvard University Press, 1999 [1971]. [Raz, 1986] J. Raz. The Morality of Freedom. Oxford University Press, 1986. [Raz, 2003] J. Raz. The Practice of Value (with commentaries by Christine Korsgaard, Robert Pippin, & Bernard Williams; edited and introduced by R.Jay Wallace). Oxford University Press, 2003. [Schmidtz, 1995] D. Schmidtz. Rational Choice and Moral Agency. Princeton University Press, 1995. [Sch¨ on, 1992] D. A. Sch¨ on. Designing as reflective conversation with the materials of a design situation. Research in Engineering Design 131-47, 1992. [Scruton, 1979] R. Scruton. The Aesthics of Architecture. Methuen, 1979. [Shrader-Frechette, 1985] K. S. Shrader-Frechette. Science Policy, Ethics and Economic Methodology. Some Problems of Technology Assessment and Environmental-Impact Assessment. Reidel, 1985. [Simon, 1955] H. A. Simon. A behavioral model of rational choice. Quarterly Journal of Economics 69, 99-118, 1955. [Simon, 1956] H. A. Simon. Rational choice and the structure of the environment. Psychological Review 63, 129-38, 1956. [Simon, 1973] H. A. Simon. The structure of ill-structured problems. Artificial Intelligence 4, 181-201, 1973. [Simon, 1976] H. A. Simon. Administrative Behaviour. A Study of Decision-Making in Administrative Organization. The Free Press, 1976. [Simon, 1996] H. A. Simon. The Sciences of the Artificial. MIT, 1996. [Slote, 1984] M. Slote. Satisficing consequentialism. Proceedings of the Aristotelian Society suppl. 58, 139-64, 1984. [Slote, 1989] M. Slote. Beyond Optimizing. A Study of Rational Choice. Harvard University Press, 1989. [Stocker, 1990] M. Stocker. Plural and Conflicting Values. Clarendon Press, 1990. [Sunstein, 2005] C. R. Sunstein. Cost-benefit analysis and the environment. Ethics 115, 351-85, 2005. [Taebi and Kloosterman, 2008] B. Taebi and J. L. Kloosterman. To recycle or not to recycle? An intergenerational approach to nuclear fuel cycles. Science and Engineering Ethics 14, 177–200, 2008. [Taylor, 1911] F. W. Taylor. Principles of Scientific Management. 1911. [Tetlock, 2003] P. E. Tetlock. Thinking the unthinkable: Sacred values and taboo cognitions. Trends in Cognitive Sciences 7, 320-24, 2003. [van de Poel, 2001] I. van de Poel. Investigating ethical issues in engineering design. Science and Engineering Ethics 7, 429-46, 2001. [van de Poel, 2005] I. van de Poel. Engineering design ethics. In Encyclopedia of Science, Technology, and Ethics C. Mitcham ed. pp. 622-25, Macmillan Reference USA, 2005. [van de Poel and van Gorp, 2006] I. van de Poel and A. van Gorp. The need for ethical reflection in engineering design the relevance of type of design and design hierarchy. Science, Technology & Human Values 31, 333-60, 2006. [Wasserman, 1993] G. S. Wasserman. On how to prioritize design requirements during the QFD planning process. IIE Transactions 25, 59-65, 1993. [Williams, 1973] B. Williams. Problems of the Self. Philosophical Papers 1956-1972. Cambridge University Press, 1973. [Winner, 1980] L. Winner. Do artifacts have politics? Daedalus 121-36, 1980. [Zimmerman, 2004] M. J. Zimmerman. Intrinsic vs. extrinsic value. In The Stanford Encyclopedia of philosophy (fall 2004 edition), E. N. Zalta ed. http://plato.stanford.edu/ archives/fall2004/entries/value-intrinsic-extrinsic/. 2004.

THE CONCEPT OF EFFICIENCY: AN HISTORICAL ANALYSIS Jennifer K. Alexander

1

INTRODUCTION

The concept of efficiency expresses a specific form of rationality, used in attempts to control a changing situation by bringing it into conformity with a vision of how the world works. Efficiency became an important technological value during the nineteenth and twentieth centuries, as part of the construction of modern industrial society. It was integral in achieving the purposeful and measurable effects in an industrial modernity that championed rationality, foresight, and planning in the control and manipulation of the social and material worlds, and it remains an important post-industrial value, particularly in continuing concern about waste and wise resource management. This article examines efficiency both as a concept in contemporary engineering use and as a historical artifact. The article begins with a discussion of how efficiency might be described currently, in colloquial use and as defined both generally and technically. Especially important in characterizing efficiency is its identity as a form of control. The article then examines efficiency’s historical background, as it moved from a philosophical concept describing the workings of the Christian God in the medieval and early modern periods to its linking with human powers and abilities during industrialization. Its history offers important illustrations of the breadth of efficiency, and of nuances and valences not apparent in an analysis of its current use. In particular, efficiency’s history reveals a deep and long association with power and authority; how integral was its use of tools of surveillance, accounting, and control; and its support of visions for reforming or remaking the world. The section that follows suggests several terms that are useful currently in distinguishing between varieties of efficiency, both in engineering and common use. The distinctions turn on the metaphors used, how efficiency is measured, and whether the context is one of abundance or scarcity. The final two sections survey efficiency as a design value in contemporary engineering, and discuss important critiques of the concept and its use. Two observations lay the foundation for what follows. First, efficiency is a central value in engineering. It is, and has been, most salient where project or system specifications are governed by clear and unambiguous parameters setting limits that must be observed. The most fundamental of these is energy and its Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

1008

Jennifer K. Alexander

availability. Efficiency is thus central to projects concerned with thermodynamics, heat transfer, or power generation, and it has historically been more important in mechanical engineering than in the other engineering professions. Efficiency is also central to projects whose specifications pose limits other than energy availability, limits for example on materials, costs, or physical size. In these instances, efficiency may be valued not for its own sake but because it is part of specifications that must be met [Newberry, 2005]. Second, in its technical sense efficiency is an intellectual construction designed to bring machines, systems, or processes under material control. Efficiency is not an exercise in pure intellect. Its goal is not knowledge but an intellectual understanding that can be made practically effective. Efficiency is a way of bringing human will to bear in the world. It is a measurement with an apparently objective form, but it carries with it a history as a tool designed to make the natural world conform to the way in which it is intellectually understood. Even as a technological concept, efficiency carries inherent social and political implications.

2 THE SCOPE OF EFFICIENCY

2.1 Definitions: fundamental features Efficiency may be used in two different ways, as a general term, usually of approval, indicating a job well and economically done; and as a specific technical assessment, growing out of the experience of industrialization and tied to measurements of performance in machines and the thermodynamics of energy. Efficiency in general use may be quantified; in engineering traditions it is quantified, almost without exception. The interplay of these technical and common uses characterizes contemporary forms of efficiency. One may speak of thermal efficiency, for example, or of mechanical efficiency, different yet precise concepts and with an identical form of measurement: a mathematical ratio of yield achieved to resources used. In contrast, one may use efficiency colloquially, referring not to a precise measurement but to the ease, speed, and good sense with which a task is performed. People speak of an efficient administrator, or of the efficient use of time. This may be an informal reference to an output/input relationship, but it may also be a remnant of its pre-industrial use, in which efficiency was not measured but was a qualitative reference to competence and power. However understood, efficiency in this common sense generally denotes approval: better efficient than not. The Oxford English Dictionary (OED) offers a suite of useful definitions of efficiency.1 Its first meaning, “The fact of being an operative agent or efficient cause”, 1 These are definitions in English, of course. Efficiency had no exact counterpart in French. Rendement, or performance, is perhaps the closest equivalent. Efficacit´ e denotes planned effectiveness more generally and does not refer primarily to comparative measurements. Historically, the term perfectionnement encompassed much of the same territory as efficiency, denoting not only the mathematical ratio of output to input but also broader issues of the social and economic

The Concept of Efficiency: An Historical Analysis

1009

is now only in philosophical use, and in fact such uses are increasingly antique, as mentioned below in the discussion of historical background. But this definition does emphasize the role of efficiency as the tool of an agent seeking to have active effect in the world. The OED’s second definition comes closer to what people now mean when they use the term: “Fitness or power to accomplish, or success in accomplishing, the purpose intended; adequate power, effectiveness, efficacy.” This establishes the connection of efficiency to material power and the achieving of goals, but it is not yet precise enough to describe its engineering use. The OED’s next definition is: “The ratio of useful work performed to the total energy expended or heat taken in.” The OED definitions, general and technical, buttress the important point that the concept of efficiency is not merely an intellectual conception but is intended to have material effect in the world. This becomes especially clear in engineering uses. Nayler’s Dictionary of Mechanical Engineering gives three meanings of efficiency: “(a) The performance of a machine as a percentage of its theoretical performance. (b) The ratio of the energy output to the energy input of a machine. . . . (c) The ratio of the mechanical advantage to the velocity ratio. . . .” [Nayler, 1985]. These definitions come from efficiency’s mechanical core, from the engineering discipline with which efficiency is most closely allied and in which it was most rigorously defined. Efficiency had to do with machines, as Nayler’s definition makes clear. Not so apparent is that engineering efficiency is allied not only with machines but with engines; this is the key to the second of Nayler’s definitions, the energy ratio. Significant about an engine, as opposed to a mere machine, is that it generates motion itself, and does not merely transmit or transform motion introduced from outside. A pulley is a machine, but it cannot move things itself; it requires an outside source of motion. Someone or something must pull the cord. In contrast, an internal combustion engine itself generates the motion it then transfers to its working parts. An engine is not a closed system, because it relies on fuel supplied from outside, but that fuel, although it can be converted to motion or work, is not motion itself. This conversion, of fuel to motion or work, is at the heart of engineering efficiency. Such a conversion operates through the intervening medium of heat, and the association with heat links efficiency to the science of thermodynamics, especially to the laws of energy codified by physicists and engineers in the midnineteenth century. The first law of thermodynamics recognized heat as a form of energy and specified that energy is conserved, and that although energy may be transformed it is neither created nor destroyed. The first law does bear on efficiency, as may be seen in the term “first law efficiency” or “plant efficiency”, which denotes the ratio of useful energy out to energy available in the fuel used performance of machines. In German, the term Wirkungsrad expresses the engineering measurement of efficiency, especially in terms of energy use. Wirtschaftlichkeit and Leistungsf¨ ahigkeit are also sometimes translated as efficiency. Wirtschaftlichkeit is a general term, denoting economic efficiency. Leistungsf¨ ahigkeit is similar to the French rendement, describing performance, or more properly the capacity to perform.

1010

Jennifer K. Alexander

by a power plant, for example. It is the second law of thermodynamics that has been more closely allied with conceptions of efficiency. According to the second law it is inevitable that some energy will be lost, by being rendered irrecoverable, in the process of converting energy from its native forms, as in coal or sunlight, into useful forms such as electricity, mechanical work, or heat. This arises because there is a preferred direction to the process of energy conversion and transfer, as Rudolf Clausius made clear in a classic formulation: “Heat can never pass from a colder to a warmer body without some other change, connected therewith, occurring at the same time” [Clausius, 1867, p. 117]. William Thomson (Lord Kelvin), himself the author of a classic formulation of the law, put it more plainly in an 1851 draft of a paper on heat: “Mechanical effect escapes not only from agencies immediately controlled by man, but from all parts of the material world, in the shape of heat, and escapes irrecoverably. . . .” [Thomson, 1851/1989]. The complete conversion of energy from one form into another was impossible, for there would always be losses through processes like friction or radiant heat. The second law thus made the issue of lost or wasted energy important for engineers and physicists interested in measuring the performance of machines, because it provided them with a theoretical upper limit: an engine or machine could at best, and only theoretically, give out as much energy as it had taken in. Efficiency became a matter of accounting for the transformation of energy through an engine or machine. Definitions cannot fully capture the variety of ways in which efficiency is used, but they do underscore several of its fundamental features. First, efficiency is more than an intellectual conception, for it is connected to material power and material effect in the world. Second, it is linked not only with machines but with engines, in other words not only with the transmission of motion but with its generation. Efficiency is thus not only managerial but also creative. Third, important varieties of efficiency are fundamentally linked with the notion of scarcity, through their reliance on the laws of the conservation of energy and the recognition that energy is lost irrecoverably when it is converted into a useful form.

2.2 Efficiency in use: control, effects, and means-ends rationality In use, efficiency evinces a commitment to bringing things under managerial control. The issue of control bears significantly on how efficiency is characterized, whether as an intrinsic or instrumental value. Contemporary engineering efficiency uses techniques of observation and control to guarantee that action will be effective, and detailed accounting principles, tracking the use and transformation of materials, to measure that effect and balance it against costs, whether in energy, materials, or funds. Efficiency operates through the control and management of resources. Efficiency is linked with control because it is a means ‘to produce predetermined effects’ through systematic and rational processes [Levin, 2000, p. 16]. This is apparent in general uses of efficiency, in which it describes a job well and economically

The Concept of Efficiency: An Historical Analysis

1011

done. Efficiency does not necessarily provide full control, however; in any given instance the parameters important to efficiency are many and complex. An internal combustion engine provides an example. To measure its efficiency a variety of different factors, such as heat value of the fuel and the torque on the driveshaft, must be converted into equivalent units, usually of energy, and their interconversions must then be tracked through the system. Attempts to increase the engine’s efficiency require observing it closely enough to find avenues of greater control. Efficiency operates as a technique of control both by providing a model of how a machine or process should function, and a technique for measuring how closely that functioning matches the model. The model thus functions as a benchmark or yardstick. Its reliance on techniques of control helps to distinguish efficiency from the closely allied concept of effectiveness. Efficiency, through control, describes actions that are more than simply effective; it describes effectiveness achieved through the precise apportioning of resources to task, so that enough resources are used but no more. The power of efficiency lies not in producing a great effect, but in producing a desired effect using precisely the desired amount of resources. Effectiveness, in contrast, may often be achieved by dumping resources into a problem, by using more resources than a desired outcome might in fact require if the problem could be specified with precision [Mitcham, 1994, pp. 226-227]. Efficiency is more elegant than effectiveness; it is narrow and targeted. The precise apportioning of means to ends is not integral to effectiveness as it is to efficiency, so it is easier to see effectiveness as an instrumental value. The characterization of efficiency is complicated by its necessary intertwining of means and ends. Efficiency is thus more than a form of means-ends rationality, and more than a mere instrumental value. Its close association with control, and particularly its function as a way to measure control, indicates that efficiency is a value attached to processes rather than to goals. Although efficiency does entail the apportioning of means to ends, it does not follow that means may be considered mere instruments for achieving such ends. With efficiency, the end to be achieved is the mastery of the process itself, or, more precisely, the relationship between means and ends. “[N]ew horizon[s] of efficiency calculations” will indeed emerge in different social and cultural contexts, and they will emphasize different goals and use different means [Feenberg, 1999, p. 97]. Such calculations nevertheless, as far as they continue to concern efficiency, will also continue to measure mastery or control. 3 HISTORICAL BACKGROUND: ENDURING NUANCES Efficiency has a wide variety of meanings in contemporary use, but its history suggests a deep and shared resonance beneath them. Consistent throughout efficiency’s history has been its equation with direct and effective action, from the Aristotelian system of causes through medieval conceptions of the nature of the Christian God. Uniting the variety of efficiency’s contemporary meanings is not

1012

Jennifer K. Alexander

only the pursuit of control but also overarching concerns for productivity and economy, and an acceptance of Enlightenment values of calculating rationality. Efficiency is often accompanied by a seemingly moral imperative, occasionally made explicit but more often understood: that efficiency is a good thing, on its face, in the way that good use is opposed to waste, and that planning and foresight are opposed to accident and happenstance.2 The resonance of efficiency is with both intellectual and material power, in its association with the ability to take precisely effective action in the world. Although these connections may be extrapolated from the definitions offered above, in Section 2, only a survey of historical developments can reveal the depth of such resonances and their accompanying moral valences. Two historical junctures were particularly important: the transition from preindustrial to industrial society, and the perceived crises of western industrial society at the turn of the twentieth century. In the transition to industrial society a medieval and early modern conception of efficiency as an attribute of the creator fell away and efficiency came to be described as a value rather than an attribute, and as something achievable by humans themselves. In the crises of industrial civilization at the turn of the twentieth century efficiency became widely popular and offered policy makers the possibility of remaking societies destroyed by war. Although historical examples appear through this article, as illustrations supporting arguments about efficiency’s contemporary character, it is important to note that in this section not only the examples but the argument itself is historical. The point is that efficiency has a character of its own, developed over a long period and in tandem with the development of industrial culture. Its historical character means that efficiency is not merely a neutral instrument without inherent qualities of its own – in other words, it is not an empty shell waiting to take on the values and goals of whomever invokes it. Its history reveals not only that efficiency has character, but that it has depth.

3.1 Pre-industrial conceptions Pre-industrial conceptions of efficiency differed importantly from the definitions of the term offered above. Examining these early conceptions makes clear both a deep continuity in the association of efficiency with power and goodness, and discontinuities in the use of the concept, which was neither quantified nor associated with human powers. Two pre-modern sources for efficiency are plausible: scattered examples of comparisons of output and input in machines that scholars have traced back to antiquity; and a tradition growing out of philosophy and theology associating efficiency with action and with the power and goodness of God. The first plausible source, of output/input comparisons, helps to account for efficiency’s contemporary math2 Economist and Nobel Laureate Robert William Fogel notes the persistence of “the widely held assumption that technological efficiency is inherently good” in a recent lecture on the history of debates over slavery’s efficiency [Fogel, 2002, p. 46].

The Concept of Efficiency: An Historical Analysis

1013

ematical form but also illustrates important discontinuities, for such comparisons were incidental and not general, and not quantified. The second, the philosophical and theological tradition, does help to account for efficiency’s later moral and value-laden character, although it came to be associated with human rather than divine action. 3.1.1

Output/input comparisons

Although contemporary efficiency may be accurately described as an output/input relationship, an interest in such relationships can actually be traced back quite far. Archimedes asserted, as legend has it, that given a lever long enough he could “raise the world,” implicitly comparing the accomplishment of a task with the resources he would need to achieve it. Pseudo-Aristotle’s analysis of simple machines, including the lever, suggests a similar comparison [Mitcham, 1994]. Historians have also seen an interest in efficiency in the seventeenth-century, in Galileo’s theory of machines, and in the early British Royal Society’s interest in improving mining, transportation, and military technologies. Although early investigations of output/input relationships do provide antecedents for the technical and comparative varieties of efficiency, they do not account for its later emphasis on quantification, nor for its later value-laden character. Neither Galileo nor the Royal Society used the term “efficiency”, nor did they quantify their comparisons [Cardwell, 1995, pp. 83-91; Merton 1938, pp. xx-xxi, 521]. Ancient output/input comparisons also remained unquantified, having been expressed in geometrical ratios. Such comparisons suggest an antecedent for efficiency’s contemporary mathematical form while simultaneously underscoring the lack of a general concept to express such relationships. 3.1.2

Wisdom and actions of power

Antecedents for the moral and value-laden qualities of efficiency appear in premodern conceptions of the simplicity, economy, and power of the deity, for whom the term efficiency was largely reserved. Pre-modern conceptions of God’s goodness, power, and simplicity informed the developing idea of efficiency through doctrines of divine economy. In the fourteenth-century Gabriel Biel described God’s omnipotence in terms of efficiency [Lindberg, 1992, pp. 241-244, 390; Funkenstein, 1986, pp. 117-201]. Early ideas of economy associated with William of Ockham and Ockham’s razor, or the principle of economy, were also influential; according to Ockham’s razor, the simple or economical explanation of an event or phenomenon was always to be preferred [Grant, 1977; Adams, 1987; Copleston, 1950-; Gilson, 1961]. The term economy had been used to describe household management since at least 1530, but it also referred to God’s managing of the cosmos, and through the eighteenth century economy denoted the “grand organization and government of life on earth” with God as “Supreme Economist” who had designed the household and kept it functioning [Worster, 1994, p. 37; Screpanti and Zamagni, 1993, pp. 20-23]. Efficiency of this sort was neither a measurement nor a comparison. It

1014

Jennifer K. Alexander

denoted adequacy or sufficiency for accomplishing something, rather than a precise match between abilities or resources and task. The term efficiency itself suggests a particular antecedent in the medieval concept of the efficient cause, based upon the ancient Aristotelian system of four causes, in which the efficient was the active and immediate principle that produced change.3 Thomas Aquinas was prominent among the medieval scholars who adopted Aristotle’s system as part of a thirteenth-century program reconciling the pagan philosopher’s work with the principles of Catholicism; the intellectual tradition of Scholasticism was founded upon this Aristotelian-Catholic synthesis. One outgrowth of this synthesis was the description of God as the efficient cause or prime mover [Kaiser, 1997], a practice that endured into the seventeenth century and can be seen in Spencer’s analysis of God as “the Efficient Cause of man,” for he had given human form to base matter [Spencer, 1628, p. 31]. The philosophical concept of efficient cause accentuates three characteristics of what efficiency came to mean: it was active, its actions were immediate, and they were effective. Pre-industrial concepts of efficiency included interest in output/input relationships in mechanics (although the term was not used), theories of divine economy and simplicity, and the Aristotelian idea of the efficient cause. Such sources help account for the moral character efficiency took on as people came to see in it a positive social and economic good. They also help explain efficiency’s association with authority and power, and especially with managerial power. The modern concept of efficiency resulted from the intersection of output/input measures with theories of divine simplicity, economy and power, and with a theory of immediate causal agency. What would prove discontinuous with later conceptions were the lack of a quantified, general sense of output/input relationships, and efficiency’s association with the creative power of God.

3.2 Conceptions during and after industrialization The meanings and uses of efficiency changed greatly during industrialization. Although non-quantified uses of the term did endure, quantified measures of performance became increasingly important, especially for people dealing with machines. As important as quantification were the developing recognition of the natural limits or boundaries that surrounded and contained human efforts, the most conspicuous of these being the law of the conservation of energy, and the variety of mechanical devices and observational techniques that made it possible to recognize, analyze, and record ever smaller mechanical effects. Dropping out in the transition between pre-industrial and industrial efficiency was the idea of efficiency as mere effectiveness, or as mere adequacy or sufficiency. This was increasingly replaced by a similar but more exacting idea, in which efficiency referred to adequate or sufficient powers and nothing more. Anything more generated waste. This idea turned on the closeness of the match between 3 The other three Aristotelian causes dealt with the material and formal aspects of change, and with its ultimate purpose, or final cause.

The Concept of Efficiency: An Historical Analysis

1015

resources expended and effect achieved, and although it is especially evident in the developing mathematics of efficiency, it is also reflected in the more general sense in which efficiency came to be opposed to waste. This more exacting use of efficiency corresponded to interest in the natural limits increasingly seen as governing what one could expect to get out of a machine. It also helped to distinguish efficiency from the related concept of effectiveness. In this transition, efficiency became a human rather than divine attribute, in keeping with a developing belief in the possibilities of rational, systematic, and effective human action. Two attitudes deriving from the Renaissance and the Enlightenment were crucial: a humanistic view of people and their achievements as valuable in and of themselves; and a reliance on measurement and quantification in understanding and manipulating the world. Three features of efficiency’s industrial development have particular significance in discussions of its contemporary uses and meanings: industrial efficiency’s roots in technical practices of motion control in machines, which underscore its alliance with control more generally; differences between the limiting parameters that efficiency could not overcome (in thermodynamics) and those it could (in economics, for example), which inform a fundamental distinction between efficiency when allied with balance and when allied with growth; and visions of efficiency as a way to remake society in response to social crisis, which illustrate the historical connection between efficiency and visions of how the world should or does work. 3.2.1

Motion control in machines

Industrial efficiency has its roots in technical practices of motion control in machines. It is closely linked to physical and mechanical measurements, developed from the eighteenth through the mid-nineteenth centuries to help quantify the performance of machines, and stemming from a tradition of analyzing machines and their effects in terms of motion. This tradition gave rise to a variety of devices both to contain and direct motion in machines and to assess and measure that motion, although the term “efficiency” was not in common use until well into the nineteenth century. Mechanics and engineers used instead a variety of terms, such as “mechanical effect” and “mechanical power”. Mechanical traditions have long linked efficiency to how things move. Efficiency in machine performance came to emphasize a mechanical discipline that used physical structures to eliminate extraneous and wasteful motions, and to control and direct productive motion along predetermined paths. British engineer John Smeaton, in a series of celebrated experiments on waterwheel efficiency in the 1750s, designed his model to minimize splashing and turbulence, and to eliminate disturbances that might keep the water from moving smoothly and directly through the system [Smeaton, 1759; Skempton, 1981, pp. 35-57; Alexander, 2008a].4 Gerard Joseph Christian, French machine theorist and director of the Conservatoire des arts et m´etiers during the Restoration, described the most 4 Smeaton

did not use the term “efficiency”; he wrote instead of “mechanical effect.”

1016

Jennifer K. Alexander

perfect machine in terms of efficiency, as the one that produced “the greatest mechanical effect, while using the least amount of fuel,” only possible if all but a machine’s working parts were immobilized [Christian, 1825, II p. 374, III pp. 18, 37; Alexander, 1999]. In the mid-nineteenth century, W. J. M. Rankine, at the University of Glasgow, found in efficiency a way to link the precise mathematical formulations of the energy concept with measurements of machine performance: the best or most efficient machines lost the least energy in useless and extraneous motion [Marsden, 1992; Wise and Smith, 1989-1990]. The influential machine theorist Franz Reuleaux defined a machine in terms of motion control: a well-designed and effective machine allowed only predictable and controlled motions [Reuleaux, 1876]. Motion control offers a particularly potent illustration of the types of control affiliated with efficiency. It requires that disturbances be eliminated, that the machine or system be kept under detailed surveillance, and that only predicted motions be allowed. The most efficient machine is the most thoroughly controlled.

3.2.2

Thermodynamics, balance, and growth

Two developments of the mid-nineteenth century were important to the maturing concept of efficiency: the working out of the laws of thermodynamics, including the conservation of energy and the tendency to entropy; and a shift in the sciences away from ideas of balance to ideas of growth. In establishing relationships between motion, heat, and energy, and by postulating that although these quantities could be transformed they could not be created, conservation laws provided a theoretical upper limit to the efficiency of a machine. Glasgow engineer and physicist W.J.M. Rankine relied on the conservation laws in giving efficiency its mechanical definition, which expressed the effectiveness of a machine in terms of its use of energy, in the ratio of effect produced to energy used. Also important were new concepts of dynamism and change that characterized work in the physical sciences during the period in which Rankine gave efficiency its mechanical definition. Dynamic understandings of natural phenomena increasingly replaced characterizations of nature or natural systems as in equilibrium, in which changes had been seen as offsetting each other and resulting in a balanced state. The shift from ideas of balance to ideas of growth or dynamic change was expressed most clearly in Darwin’s theory of descent through natural selection [Wise and Smith, 1989-1990]. Efficiency also began to be applied in economics, where it was allied with rational management in pursuit of growth [Alexander, 2008a]. Thus, by the turn of the twentieth century, efficiency had both matured in applied mechanics and become part of a larger intellectual shift in how natural systems were conceptualized. Efficiency had been named, and given a formal mathematical definition in the classic form of the output/input ratio, as the ratio of work performed by a machine or system to energy used in producing that work. In this formal sense, efficiency was allied with the notion of the balance. The

The Concept of Efficiency: An Historical Analysis

1017

term also came to be used in ways corresponding with growth, most commonly in economics. These developments underlie the metaphors of the balance (statics) and the engine (dynamics) that inform efficiency’s contemporary meanings, as discussed below in Section 5.1. 3.2.3

Response to industrial crisis

Efficiency was a central concern in the reform efforts that characterized American and European history from the turn of the twentieth century until the Great Depression, reforms spurred by worries over the effects of industrialization and urbanization in a changing international order, and expressed in efficiency movements tied to national health, governmental reform, military prowess, and protection of empire, nation, or race. Efficiency became ubiquitous in the United States during the progressive era, a time of intellectual, social, and political turbulence. Efficiency described not only technical matters, like the thermal economy of an engine, but personal ones as well: careful spending habits, fastidious bodily hygiene, and good childhood education. Technical features like quantification and calculation jumbled together with social, governmental, and personal concerns to produce a word resonating of technical expertise, personal integrity, and good government. Efficiency expressed both sober qualities of hard and patient work, and enormous hopes for remaking society and the world. Frederick Winslow Taylor’s system of scientific management is the most recognizable American efficiency marker of this era.5 Efficiency as a response to crisis is best illustrated by the efficiency and rationalization movements that characterized European society and politics early in the twentieth century. In Britain, at the turn of the century, embarrassment over how long it had taken the Empire to subdue the rebels in the Boer War led to worries about national efficiency and calls for efficiency reform [Searle, 1990]. The ravages of the First World War and threatened economic collapse led to widespread enthusiasm for efficiency in Italy, France, and especially Germany; “rationalization” was the key term describing the prominent international movement of the interwar years that sought in efficiency a solution to problems of economic scarcity and social turbulence [Brady, 1933; Maier, 1970; Weiss, 1987; Nolan, 1994]. These movements sought in efficiency a way to achieve visions of how society 5 Although Taylor was a mechanical engineer, and one-time president of the American Society of Mechanical Engineers, his system of scientific management lay outside the mechanical tradition; many did not consider his management work to be engineering and the Society refused to publish his seminal management papers [Kanigel, 1997]. The discipline of industrial engineering traces its roots to Taylor, and it is from Taylor’s system that confusion has arisen over the nature of efficiency in engineering. Scholars have seldom noted the differences between Taylor’s system of management and the techniques used with the engines and machines at efficiency’s mechanical core. Connecting the two were Taylor’s early experiments with the cutting of metals, which established his international reputation and which turned on control of the precise angle and conditions of the cutting blade: this work was placed firmly within engineering traditions. Taylor’s system of management was concerned with the control of labor as much as with increased productivity; this concern linked him with management traditions rather than engineering.

1018

Jennifer K. Alexander

should be organized and how it should function. They were responses to social crises, in some cases extreme, and they illustrate a significant feature of how efficiency has often been used in the contemporary world: in attempts to control changing or threatening situations by bringing them into conformity with visions of how the world ought to behave [Alexander, 2008a].

4

A VOCABULARY OF EFFICIENCY: IMPORTANT CONTEMPORARY DISTINCTIONS

Efficiency takes on a variety of forms in modern and contemporary use. This section develops three distinctions that are useful in analyzing how the concept functions in any given case, and that suggest ways to narrow and focus discussions of an enormously broad issue. Discriminating between these varieties of efficiency reveals subtle but significant shadings in connotation and use: social distinctions in how efficiency is applied; varying emphases on human agency depending on how efficiency is measured; and the paradoxical nature of the concept itself, which addresses scarcity but is most effectively applied in contexts of abundance. The sections that follow use historical examples, but in an argument about efficiency’s contemporary use.

4.1 Root metaphors: static and dynamic forms of efficiency Efficiency appears in two ways that are conceptually distinct, where the difference lies not in how they are measured but in the root metaphor they employ. The two root metaphors are the balance, illustrating static efficiency, and the engine, illustrating dynamic efficiency.6 Static efficiency emphasizes conservation and predictable performance, and dynamic efficiency emphasizes effective management rewarded by growth [Alexander, 2008b]. Although efficiency in its formal sense emphasizes balance (through the conservation of work, mechanical effect, or power), people also use the term in ways in which it corresponds with growth, or more specifically with visions of change and progress. In this sense efficiency connotes rational management in pursuit of the greatest effect. What is efficient is effective, and available inputs, whether resources, or existing systems or institutions, do not limit possible output in the narrow mathematical sense, but instead challenge the management skill of whomever is in charge. In pursuing a dynamic efficiency of growth, a manager is not bound by an equation of conservation or balance, but is free, in the pursuit of profits, to broaden what he or she considers the raw materials, or inputs, in his or her control. The distinction between efficiency’s static and dynamic forms lies not in how efficiency is measured nor does it lie in whether something remains at rest or moves. The distinction lies instead in the connotations that accompany efficiency’s use. 6 The steam engine provided the most potent illustration of the new possibilities offered by engines in the nineteenth century, but the metaphor of the engine need not be limited to steam.

The Concept of Efficiency: An Historical Analysis

1019

When efficiency measures are emphasized as tools to generate stability, predictability, and manageability, the root metaphor is of the balance. The connotation is one of things evening out, moving smoothly and without turbulence. When efficiency measures are emphasized as tools of transformation the root metaphor is of an engine, which creates motive force from materials themselves at rest. The connotation is of change and progress, progress here meaning procession toward a goal, which may or may not be itself a positive good. Through industrialization steam power was transforming society, bringing people from the countryside into the factories of the cities and stimulating deeper scouring of the earth in search of coal for its fuel. But the steam engine, because it not only produced motive force but also generated irrecoverable losses of energy through waste heat, also carried the world further on the path of energy dissipation described in the second law of thermodynamics. Opinions are, and were, divided on the benefits of industrialization and the new engines, but there is concensus on their transformative effects [Smith, 1998; Landes, 1969]. Although efficiency’s static and dynamic connotations exist side-by-side, an important social distinction has come to govern how they appear. In its static and balanced sense, efficiency describes machines, processes, and people who are subject to management; in its dynamic sense, it describes the efficiency of those who do the managing. Those efficient in the static or conservative sense provide the stable and balanced elements manipulated by those in positions of relative privilege in pursuit of the greater rewards of dynamic efficiency. A detailed example will help to make clear not only the distinction itself, but also the significant issues that can be at stake. This is illustrated in the painful development of Robert William Fogel’s thinking on efficiency and slavery in antebellum America. Fogel, Nobel Laureate (1993), economist, and cliometrician, devoted his early career to demonstrating that plantation agriculture in the American south had been efficient because of the good work ethic of slaves, and because slaves were well treated by their masters. He offered the analysis in an attempt to recover a respectable past for slaves’ descendants, and to find something in which they could take pride. His work, published in 1974 with co-author Stanley Engerman, was met with widespread and acrimonious criticism, and occasioned one of the most public academic disputes in post-war America. Devastating criticism led Fogel to reevaluate his position on slavery and efficiency, and to confront his own earlier assumption that efficiency was itself a positive good. Fifteen years later Fogel admitted that his assumption had been false, and that the efficiencies he had found in plantation agriculture were the result not of a good work ethic or personal initiative on the part of slaves, but of exploitation through gang labor, brutally enforced. The efficiencies he had measured had accrued to the owners and masters of slaves; they were dynamic efficiencies, measured in profits and growth. But the profits and growth were built on static efficiencies, which required keeping slaves in order and hard at task, and, in particular, in preventing them from creating any disturbance or turbulence, or change, in the plantation system [Fogel and Engerman, 1974; Fogel, 1989, 2002; Alexander, 2008a].

1020

Jennifer K. Alexander

4.2 Measurements: bounded and arbitrary forms of efficiency Another important distinction in how efficiency is used concerns the terms by which it is quantified. Efficiencies that are quantified take one of two forms: a bounded form or an arbitrary form. The distinction was first noted during the American progressive era by Walter Polakov, engineer and contemporary of the well-known efficiency theorist Frederick Winslow Taylor. Polakov used the term “arbitrary” to emphasize the element of human choice in constructing efficiency ratios that are not measured in terms of energy. Bounded forms of efficiency are dependent upon quantities limited by natural law such as energy; efficiency is bounded by the laws of the conservation of energy in that they pose an upper limit beyond which it cannot reach. Arbitrary forms of efficiency take the form of ratios between otherwise independent quantities, in which a particular value of the ratio is chosen as the standard of one hundred percent efficiency, and against which other values of the ratio are measured. Using efficiency measurements predicated on the laws of energy may also be described a matter of human choice, specifically the choice to use highly authoritative scientific law rather than to create arbitrarily a practical measurement suited to the quantities at hand. Arbitrary measures allow a very broad use of efficiency, making it possible to apply the term, precisely and quantitatively, to almost anything. Historical examples provide the clearest illustrations of this breadth; Polakov himself used the ratio of cost per BTU of coal as an example. Measurements of efficiency based on the ratio of the number of widgets produced per hour to the rate of ventilation in a factory would also take the arbitrary form. One arbitrarily sets a standard of 100 percent efficiency (a certain cost per BTU of coal, or so many widgets per hour at a certain ventilation rate), and measures other examples of the same ratio against it. Setting an arbitrary standard of 100 percent efficiency for the yield of potatoes at 100 bushels per acre, Polakov calculated that the potato yield efficiency champion in 1907 was the “desert state of Wyoming,” with an average yield of 200 bushels per acre, for an efficiency of 200 percent [Polakov, 1909]. Arbitrary measurements of efficiency remain in widespread use. The distinction between efficiency’s bounded and arbitrary forms is sometimes used to illustrate the difference between engineers and economists. Engineers, governed by natural limits such as the second law of thermodynamics, cannot reach a one hundred percent return on energy invested in a machine. Economists, in contrast, can exceed the one hundred percent limit by arbitrarily setting efficiency standards according to terms other than energy, which allow them to achieve profits, (i.e., returns greater than one hundred percent).

4.3 Root contexts: efficiency under scarcity or abundance Efficiency may be worked out elegantly and carefully over time, or it may be adopted in desperation as a response to scarcity. The distinction lies not in the type of efficiency measures adopted or in why they are chosen. It depends instead on the context within which efficiency is pursued. Engineering uses of efficiency

The Concept of Efficiency: An Historical Analysis

1021

generally depend on resources allowing time for the analysis of a problem, the formulation of various approaches, and the careful and managed implementation of whatever option is chosen. The context is one of abundance. Frederick Winslow Taylor’s well-known system of scientific management depended on an abundance of resources allowing time for management and consultants to analyze the flow of work and to consider various efficiency approaches. Even if the point of efficiency measures is to maximize the use of scarce resources, a context that allows careful consideration and a variety of options remains a context of abundance. Efficiency measures are also adopted in desperation as a response to a scarcity of resources. The context of scarcity refers not only to the scarcity of resources generally but of efficiency resources, when the time, expertise, and materials needed to analyze a problem thoroughly and choose a well-suited plan of action are unavailable. This is efficiency in the sense in which most people use it, in the everyday decisions they make about how to apportion their limited resources — people may make a few calculations but do not generally mount a full-scale efficiency analysis. But efficiency in a context of scarcity is especially marked by desperation, by the attempt, in a true crisis of resources, to make a few variables under one’s control stand in for other matters that cannot be helped. An example may be found in attempts by the Kaiser Wilhelm Institute for Labor Physiology to stimulate the productivity of coal miners in Germany during the last years of the Second World War. Institute researchers were unable to ameliorate the inhumane treatment of the miners, many of whom were Soviet prisoners of war and subjected to conditions described by the Red Cross as appalling. Improving the efficiency of the miners came down to improving their diets; the racial commitments of the Reich did not allow it to ameliorate other factors recognized as crucially important to efficiency, such as working and living conditions. Dietary measures were adopted in desperation, when a scarcity of political and material options allowed little other intervention [Alexander, 2006b]. 5 EFFICIENCY AS A DESIGN VALUE IN ENGINEERING The concept of efficiency plays an important role in engineering design in several ways. It is important in the design of engineering systems and artifacts, but perhaps even more fundamental has been the importance of engineering artifacts themselves, and their design, in the formulation of efficiency theory. The theoretical apparatus of efficiency was built on close observations of machines at work. In practice efficiency analysis also remains closely tied to the particular configuration of an artifact or system, and especially to the arrangement of its mechanical parts. General principles of efficiency may be derived from practice, that certain types of motor are more mechanically efficient for example, but even motors of the same design will evidence differences in efficiency when put to the test. Also important is the prestige that its association with physics brings to efficiency, and its scientific status may be offered as proof of technocratic expertise on the part of members of a design team. Efficiency is not so apparent an indicator of optimal

1022

Jennifer K. Alexander

engineering design, principally because of its malleability; what is to be optimized must first be decided, and then there will remain a variety of options for efficient achievement of the goal. That said, efficiency does serve as a design criterion when it is included among project specifications, and may be an important consideration in both the original conceptualization, and any necessary re-conceptualization, of the design problem itself. What efficiency can offer most potently is its recognized value as a goal, to which members of a design team may share a commitment despite contests over how it should be reached.

5.1 Reflects the role of design in building theory Formal definitions of efficiency in thermodynamics were historically linked to attempts to understand and quantify the performance of machines. They give evidence not only of the interrelationship between engineering theory and scientific theory, but of the importance of the design of the technological object in the process of theory building. Peter Kroes gives as an example Pambour’s nineteenthcentury theory of the conservation of steam in steam engines; another example is the role of the configuration of the test waterwheel in the famed studies of its mechanical effect made by eighteenth-century British engineer John Smeaton [Kroes, 1992; Reynolds, 1983; Alexander, 2008a]. In both cases, the physical configuration of the machine bore directly on how it was theoretically construed. The most famous example is Sadi Carnot’s work on the ideal engine cycle, which was fundamental to the formulation of engineering definitions of efficiency. Carnot’s conceptual design of an ideal engine played a crucial role in the development of thermodynamics theory [Cardwell, 1971].

5.2 Brings the prestige of science to bear on the design process The concept of efficiency brings the prestige of science into the design process, especially when energy is at issue, because it is founded on one of the most authoritative conceptual structures in modern physics. Appeals to efficiency may trump other claims of value in the design process. Efficiency’s value here depends on a ladder of assumptions: most generally that science is a good foundation for engineering design, but also that the laws of thermodynamics do reflect how things work (that they are true) and that they are an adequate guide to building real systems in the world; and that science offers a politically neutral ground for consensus. This can be seen in the appeal to science as an authoritative and consensual basis for the redesign of society according to principles of efficiency and rationalization in Weimar Germany. Germany’s Reichskuratorium f¨ ur Wirtschaftlichkeit offered scientific measurements of efficiency as a non-political and culturally neutral solution to social and political strife, when it was charged by the Reichstag in the 1920s with ameliorating the tensions of a nation experiencing extreme economic and social dislocation following the First World War [Maier, 1970; Nolan, 1994]. Efficiency is perceived as especially valuable in design considerations that bear

The Concept of Efficiency: An Historical Analysis

1023

on environmental use and sustainability, because of its association with waste reduction in addition to its scientific foundations. In such cases efficiency offers not only scientific prestige, but also a way to set goals and gauge progress. In discussions of the design of environmental systems, claims to efficiency often carry with them claims of technocratic expertise [Hays, 1959; Worster, 1994].

5.3

Offers a measure of optimal design

The concept of efficiency plays an important role in considerations of the optimal, or most perfect, design of engineering systems. Efficiency offers a measure of perfected design in cases that bear on the use and transformation of energy; the closer efficiency approaches 100 percent, or perfect unity, the more perfect the design. But it is important to note that even with energy efficiency there remain a wide variety of ways to measure it, and how it is to be measured may become an issue of significant disagreement among persons involved in the design process. With regard to engines, questions that will remain may include whether thermal or mechanical efficiency should be maximized, and what instruments should be used to measure it, for example an indicator within the engine or prime mover itself or a brake or other device at the output end of the system, or both. Even the heat value of the fuel may be calculated by various methods. In general, efficiency offers only a very indistinct guide for optimal design, and even when it is agreed that a design will be chosen on efficiency grounds, there will remain a wide variety of efficiency options. That said, there does remain a sense in which efficiency is indeed an important guide for optimal design, and that is when control of a moving or changing system is the primary desired function. Control values include stability of operation, predictability of results, and minimal interference or turbulence. When control of a system is the desired output, rather than some other measure of productivity, the concept of efficiency offers a model for optimal design: a system under complete surveillance, where motion is restricted to predetermined paths and resources are primarily directed toward preventing turbulence. This is efficiency in a totalizing sense and it is not restricted to engineering use. Below, in Section 6, I discuss several important critiques of efficiency’s totalizing function within society more generally.

5.4

Serves as a design criterion

Efficiency serves as a design criterion when it is among the design specifications that a project must fulfill. In such cases efficiency is not regarded as having a value of its own, but is instead valued as one among many limiting parameters specified by the owner or manager of a project. It functions to guide the design process insofar as it is congruent with the other limiting parameters, and may be accorded higher or lower priority if it comes into conflict with them. If specifications do come into significant conflict with each other, a design team generally turns to

1024

Jennifer K. Alexander

the project owner or manager to have them resolved, and does not itself choose which values to give priority. The issue then becomes one of re-conceptualizing the design problem itself. Efficiency can offer a common language to those involved in the design of a project, and because it is so widely presumed to be valuable it can serve to legitimate a wide variety of engineering options. The very breadth of its presumed value means that even competing groups may make claims to efficiency during the design process. Rather than imposing a rigid set of requirements, efficiency may instead provide a shared general goal. Those involved in the design process might see efficiency in an economic sense, or as reflective of social or cultural values, but may, despite such differences, nonetheless be able to design a project within a shared commitment to the general goal of efficiency [van de Poel, 1998].

6

CRITIQUES OF EFFICIENCY

The most famous critics of efficiency have been critics of industrial society more generally, seeing in efficiency not a harbinger of progress but a technique of control and exploitation. The American architectural critic and author Lewis Mumford rejected capitalist uses of efficiency, which he believed were tied to profit-seeking and individual enrichment at the expense of true social and cultural advance [Mumford, 1963]. A much more influential critique was mounted by the early Frankfurt school critical theorists Max Horkheimer and Theodor Adorno. Members of the Institute for Social Research, Frankfurt am Main, sought to emancipate society from ideological captivity by “bringing to awareness the conditions of our own knowledge of the world” [Anderson, 2000]. In their famous and provocative work The Dialectic of Enlightement (1947), Horkheimer and Adorno described efficiency as encapsulating a form of instrumental rationality that had been dominant since the Enlightenment. Efficiency bespoke the human attempt both to know and to manipulate the world; the test of true knowledge was how completely it allowed its object to be controlled. The greater the knowledge the greater the control; the greater the control the greater the efficiency. As Horkheimer and Adorno put it, writing in response to the nightmare of Nazism, “The totalitarian order has granted unlimited rights to calculating thought . . . . Its canon is its own brutal efficiency” [Horkheimer and Adorno, 2002, pp. 67-68]. Important support for Horkheimer and Adorno’s argument was found in the influential concept of rationalization advanced earlier by sociologist Max Weber; rationalization described the efforts of Calvinist Protestants to reassure themselves of their salvation through the good and measurable effect of their works in the world. Such effectiveness seemed proof that God was disposed graciously and positively toward them [Weber, 1904-5, 2001].7 An influential and sustained critique of Horkheimer and Adorno came from a member of the second generation of the Frankfurt school, 7 It should be noted that the term ‘rationalization’ encompassed a variety of efficiency measures as part of the rationalization movement in Europe between the world wars.

The Concept of Efficiency: An Historical Analysis

1025

J¨ urgen Habermas, who found Horkheimer and Adorno’s condemnation of rationalization so far-reaching as to undercut the possibility of critical theory itself. Such a dark description of rationality tainted even the critical and constructive use of reason, Habermas maintained, and in a series of works he painstakingly developed important distinctions between rationalization’s various forms [Habermas, 1984]. Habermas’s concern was to identify the conditions that would allow human interrelationships to flourish, free from controlling or dominating interests [Anderson, 2000]. In particular he offered an analysis of how the public sphere, under the right circumstances, could function as a critical check on political and economic authority, in contrast to the fragmented and weak check it had posed to Nazism. To buttress this argument, it was important that Habermas develop a respectable ground for rationality and thus for the critical theory that could be used to identify and help to create conditions more open to humane progress [Habermas, 1989]. The most sustained and far-reaching critique of efficiency has come from the French jurist and theologian Jacques Ellul. Like Horkheimer and Adorno he saw in efficiency a totalizing function, arguing that the quest for efficiency is antithetical to human freedom because it ultimately requires all things human to be analyzed and integrated into orderly and manageable systems [Ellul, 1964]. But Ellul’s method was notably opaque and his work cannot be considered to be within the framework of critical theory. His rejection of critical theory was in fact radical, for he argued that planning itself is implicated in the dominating effects of efficiency; the requirement that critical theory be not only theoretical but practical was itself an invitation to strategy and planning, moves that themselves give rise to the desire for efficiency. Ellul’s work is not as widely known as that of the Frankfurt school, in part because no theorist has yet interpreted his deeply disturbing work in as constructive a manner as did Habermas in response to the despair evidenced in the work of Horkheimer and most particularly Adorno. Another reason for its neglect is that Ellul’s work demands to be interpreted within a theological framework. His analysis of efficiency does not rest on theological argument nor does theological argument play a role in his major work on the subject (The Technological Society, 1964), but it is apparent in the important role he gives to human hope in much of his other work, in which hope offers an open system in contrast to the closed system that requires efficiency. More limited critiques of efficiency have been mounted in recent years, most notably of efficiency in law and economics. In American law criticism of efficiency has been direct, whereas criticism of efficiency in economic practice, particularly of globalization, has taken the form of analyses of foundational practices that invoke efficiency. Efficiency is an important and contested concept in American law, and there, as elsewhere, it has several different meanings. Efficiency describes measures to streamline the settling of lawsuits, where it is promoted as a tool to help manage increasing burdens on the legal system. It has other and more potent meanings, more particularly legal, in constitutional law, where it is recognized as an important interest to be balanced against claims such as free speech, and among jurists

1026

Jennifer K. Alexander

who see efficiency as itself the embodiment of justice. Both efficiency’s practical and administrative use, and its normative role in determining judicial or legal action, are disputed. Chief among critics of efficiency in the law has been Thane Rosenbaum, who objects to the bureaucratic streamlining of the legal process and its emphasis on settling criminal cases through negotiated pleas. Truth, he argues, has become a “hostage” to efficiency [Rosenbaum, 2004]. The critique of efficiency in economics takes two primary forms. The first is embedded within a critique of economic globalization, and characterizes market efficiency as incompatible with equality. It sees efficiency in markets as a strategy of wealthy industrial or post-industrial societies for maximizing their own benefits; protests at the meeting of the Ministerial of the World Trade Organization in Seattle in 1999 can be construed as part of this critique. The second form of critique lies in an analysis of the foundational suppositions of economic theory, tying them directly to the “gospel of efficiency” of the American progressive era, and describing efficiency in economics as in essence a religious value [Okin, 1975; Porter et al., 2001; Nelson, 2001]. It should be noted that the Chicago school of law and economics combines law and economics by seeing in economic efficiency a demonstration of natural principles the law should follow. It has been criticized for attaching an especially high value to efficiency, and justifying that high valuation by an analysis of the common law that finds it to have achieved an especially efficient distribution of resources. A good illustration of Chicago school theory may be found in Richard A. Posner, Economic Analysis of Law [Posner, 1983]. Critiques of efficiency underscore the broad reach of the concept. That they are embedded in more general social criticism highlights the deep resonance between efficiency and industrial and post-industrial society, and that important strands of critique have emphasized efficiency’s association with domination further substantiates the concept’s identity as a form of control. Theorists have not mounted a full-scale defense of efficiency; its continued use in engineering contexts constitutes a rebuttal in practice. 7 CONCLUSION The concept of efficiency is central to contemporary engineering, and it remains fundamental in both industrial and post-industrial contexts. It is broad and complex, and there is such great flexibility in how it is measured that it can be applied, precisely and quantitatively, to nearly any situation. But beneath its complexity and breadth lie three fundamental features. First, efficiency is a practical tool, an intellectual construct designed to make intellectual understanding of the world practically effective. It is not primarily about understanding the world but about acting in it. Second, efficiency is comparative and thus requires a vision of how the world ought to work, against which it may be measured or with which it may be compared. Without a vision of how a system or process should perform, there is no standard against which efficiency can be assessed. Third, efficiency functions

The Concept of Efficiency: An Historical Analysis

1027

by eliminating resistance to achieving the vision, by controlling change or motion in the process or system to which it is applied. Controlling resistance increases the likelihood that the vision will be achieved, and lessens the costs of achieving it by reducing the waste that is associated with resistance. It is important to note that resistance may take many forms, such as turbulence in the tailrace of a waterwheel or shirking by slaves in a cotton field. These last examples, of possible sites of resistance to efficiency, illustrate how its engineering and social implications overlap. Engineering efficiency has a mechanical core, tied to mechanical engineering traditions and to the physics of energy and thermodynamics through historical attempts to understand and to control the performance of machines. Efficiency’s reliance on control is key, for it distinguishes efficiency from the closely allied concept of effectiveness. Efficiency is associated not only with being effective, but with having such control over a process that effect is achieved with the least waste. Efficiency thus turns on the controlled apportioning of resources, whereas effectiveness denotes an effect but not the costs of achieving it. Its alliance with control suggests that efficiency is not only a tool but an ideology. This suggestion is supported by many of the historical instances in which it has appeared, such as debates over the productivity of slavery, and it is against efficiency as an ideology that its most potent critics have raised their voices. ACKNOWLEDGEMENTS Please note that Sections 4 and 5 contain excerpts from [Alexander, 2008a]. BIBLIOGRAPHY [Adams, 1987] M. M. Adams. William Ockham. University of Notre Dame Press, 1987. [Alexander, 1999] J. K. Alexander. The Line Between Potential and Working Machines: C´esar Nicolas Leblanc and Patent Engravings, 1811-1835. History and Technology, 15, 175-212, 1999. [Alexander, 2006a] J. K. Alexander. Efficiency and Pathology: Mechanical Discipline and Efficient Worker Seating in Germany, 1929-1932. Technology and Culture, 47, 286-310, 2006. [Alexander, 2006b] J. K. Alexander. An Efficiency of Scarcity: Using Food to Increase the Productivity of Soviet Prisoners of War in the Mines of the Third Reich. History and Technology, 22, 391-406, 2006. [Alexander, 2008a] J. K. Alexander. The Mantra of Efficiency: From Waterwheel to Social Control. Johns Hopkins University Press, 2008. [Alexander, 2008b] J. K. Alexander. Efficiencies of Balance: Technical Efficiency, Popular Efficiency, and Arbitrary Standards in the Late Progressive Era U.S. Social Studies of Science, 38, 323–34, 2008. [Anderson, 2000] J. Anderson. The “Third Generation” of the Frankfurt School. Intellectual History Newsletter 22 (2000). http://www.phil.uu.nl/ joel/research/publications/3rdGeneration.htm. [Brady, 1933] R. A. Brady. The Rationalization Movement in German Industry: A Study in the Evolution of Economic Planning. University of California Press, 1933. [Cardwell, 1995] D. S. L. Cardwell. Norton History of Technology. W.W. Norton & Co., 1995. [Cardwell, 1971] D. S. L. Cardwell. From Watt to Clausius: The Rise of Thermodynamics in the Early Industrial Age. Cornell University Press, 1971.

1028

Jennifer K. Alexander

[Cardwell, 1993-94] D. S. L. Cardwell. Steam Engine Theory in the 19th Century: From Duty to Thermal Efficiency, from Parkes to Sankey. Transactions of the Newcomen Society, 65, 117-28, 1993-94. [Carnot, 1960] S. Carnot. Reflections on the Motive Force of Fire and Other Papers. Dover, 1960. [Channell, 1989] D. F. Channell. The History of Engineering: An Annotated Bibliography. Garland, 1989. [Channell, 1982] D. F. Channell. The Harmony of Theory and Practice: The Engineering Science of W. J. M. Rankine. Technology and Culture 23, 39-52, 1982. [Christian, 1825] G.-J. Christian. Trait´ e de m´ ecanique industrielle, Bacherlier, 1825. [Clausius, 1867] R. Clausius. “On a Modified Form of the Second Fundamental Theorem in the Mechanical Theory of Heat,” trans. John Tyndall, in Clausius, T. Archer Hirst, ed., The Mechanical Theory of Heat, with its Applications to the Steam-Engine and to the Physical Properties of Bodies. John van Voorst, 1867. [Copleston, 1950-] F. C. Copleston. A History of Philosophy. Newman Press, 1950-. [Ellul, 1964] J. Ellul. The Technological Society. Knopf, 1964. [Feenberg, 1999] A. Feenberg. Questioning Technology. Routledge, 1999. [Feenberg, 2002] A. Feenberg. Transforming Technology: A Critical Theory Revisited. Oxford University Press, 2002. [Fogel, 1989] R. W. Fogel. Without Consent or Contract: The Rise and Fall of American Slavery. W. W. Norton & Company, 1989. [Fogel, 2002] R. W. Fogel. The Slavery Debates, a Retrospective 1952-1990. Louisiana State University Press, 2002. [Fogel and Engerman, 1974] R. W. Fogel and S. L. Engerman. Time on the Cross: The Economics of American Negro Slavery. W. W. Norton & Company, 1974. [Funkenstein, 1986] A. Funkenstein. Theology and the Scientific Imagination from the Middle Ages to the Seventeenth Century. Princeton University Press, 1986. [Grant, 1977] E. Grant. Physical Science in the Middle Ages. Cambridge University Press, 1977. [Gilson, 1961] E. Gilson. Saint Thomas Aquinas and Philosophy. Saint Joseph College, 1961. [Haber, 1964] S. Haber. Efficiency and Uplift: Scientific Management in the Progressive Era, 1890-1920. University of Chicago Press, 1964. [Habermas, 1984] J. Habermas. The Theory of Communicative Action, Vol I. Reason and the Rationalization of Society. Harvard University Press, 1984. [Habermas, 1989] J. Habermas. The Structural Transformation of the Public Sphere: An Inquiry into a Category of Bourgeois Society. MIT Press, 1989. [Hays, 1959] S. P. Hays. Conservation and the Gospel of Efficiency: The Progressive Conservation Movement, 1890-1920. Harvard University Press, 1959. [Horkheimer and Adorno, 2002] M. Horkheimer and T. W. Adorno. Dialectic of Enlightement: Philosophical Fragments. Stanford University Press, 2002. [Kaiser, 1997] C. B. Kaiser. Creational Theology and the History of Physical Science: The Creationist Tradition from Basil to Bohr. Brill, 1997. [Kanigel, 1997] R. Kanigel. The One Best Way: Frederick Winslow Taylor and the Enigma of Efficiency. Viking, 1997. [Kortabinski, 1965] T. Kotarbinski. Praxiology: An Introduction to the Sciences of Efficient Action. Pergamon, 1965. [Kroes, 1992] P. Kroes. On the Role of Design in Engineering Theories: Pambour’s Theory on the Steam Engine. In Technological Development and Science in the Industrial Age, P. Kroes and M. Bakker, eds., pp. 69-98, 1992. [Landes, 1969] D. S. Landes. The Unbound Prometheus: Technological Change and Industrial Development in Western Europe from 1750 to the Present. Cambridge University Press, 1969. [Levin, 2000] M. R. Levin. Contexts of Control. In Cultures of Control, M. R. Levin, ed., pp. 13-69, 2000. [Lindberg, 1992] D. C. Lindberg. The Beginnings of Western Science. University of Chicago Press, 1992. [Maier, 1970] C. Maier. Between Taylorism and Technocracy: European Ideologies and the Vision of Industrial Productivity in the 1920s. Journal of Contemporary History 5, 27-61, 1970. [Marsden, 1992] B. Marsden. Engineering science in Glasgow: economy, efficiency and measurement as prime movers in the differentiation of an academic discipline. British Journal for the History of Science, 25, 319-46, 1992.

The Concept of Efficiency: An Historical Analysis

1029

[Merton, 1938] R. K. Merton. Science, Technology, and Society in Seventeenth Century England. Osiris, 1938. [Mitcham, 1994] C. Mitcham. Thinking Through Technology: The Path Between Engineering and Philosophy. University of Chicago Press, 1994. [Mumford, 1963] L. Mumford. Technics and Civilization. New York, 1963. [Nayler, 1985] G. H. F. Nayler. Dictionary of Mechanical Engineering. Butterworth’s, 1985. [Nelson, 1992] D. Nelson. A Mental Revolution: Scientific Management Since Taylor. Ohio Sate University Press, 1992. [Nelson, 1980] D. Nelson. Frederick W. Taylor and the Rise of Scientific Management. University of Wisconsin Press, 1980. [Nelson, 2001] R. H. Nelson. Economics as Religion: From Samuelson to Chicago and Beyond. Pennsylvania State University Press, 2001. [Newberry, 2005] B. Newberry. Efficiency as a design value in engineering. Unpublished paper, 2005. [Nolan, 1994] M. Nolan. Visions of Modernity: American business and the modernization of Germany. Oxford University Press, 1994. [Okin, 1975] A. M. Okin. Equality and Efficiency: The Big Tradeoff. The Brookings Institution, 1975). [Polakov, 1909] W. Polakov. Efficiency in the Purchase of Fuel for Power Generation. Engineering Magazine 38, 215-225, 1909. [Porter et al., 2001] R. B. Porter et al., eds. Efficiency, Equity, Legitimacy: The Multilateral Trading System at the Millennium. Center for Business and Government of Harvard University, and The Brookings Institution, 2001. [Posner, 1983] R. Posner. Economic Analysis of Law, 3rd ed. Little, Brown. 1983. [Rabinbach, 1990] A. Rabinbach. The Human Motor: Energy, Fatigue, and the Origins of Modernity. Basic Books, 1990. [Reuleaux, 1876] F. Reuleaux. The Kinematics of Machinery: Outlines of a Theory of Machines. Macmillan, 1876. [Reynolds, 1983] T. S. Reynolds. Stronger than a Hundred Men: A History of the Vertical Water Wheel. Johns Hopkins University Press, 1983. [Rosenbaum, 2004] T. Rosenbaum. The Myth of Moral Justice: Why Our Legal System Fails to Do What’s Right. HarperCollins, 2004. [Screpanti and Zamagni, 1993] E. Screpanti and S. Zamagni. An Outline of the History of Economic Thought. Clarendon Press, 1993. [Searle, 1990] G. R. Searle. The Quest for National Efficiency: A Study in British Politics and Political Thought, 1988-1914. Ashfield Press, 1971, 1990. [Skempton, 1981] A. W. Skempton, ed. John Smeaton, FRS. Thomas Telford Limited, 1981. [Smeaton, 1759] J. Smeaton. “An Experimental Enquiry concerning the Natural Powers of Water and Wind to turn Mills, and other machines, depending on a circular motion,” Philosophical Transactions of the Royal Society, 51, pt. 2: 100-174, 1759. [Smith and Wise, 1989] C. Smith and M. N. Wise. Energy & Empire: A Biographical Study of Lord Kelvin. Cambridge University Press, 1989. [Smith, 1998] C. Smith. The Science of Energy: A Cultural History of Energy Physics in Great Britain. University of Chicago Press, 1998. [Spencer, 1628] T. Spencer. Logick, 1628. [Thomson, 1851/1989] W. Thomson. Draft for ‘Dynamical theory of heat’ (1851). In Energy and Empire: A Biographical Study of Lord Kelvin, C. Smith and M. N. Wise, p. 317. Cambridge University Press, 1989. [van de Poel, 1998] I. van de Poel. “Why are Chickens Housed in Battery Cages?” in Getting New Technologies Together: Studies in Making Sociotechnical Order, C. Disco and B. v.d. Meulen, eds., 143-178, Walter de Gruyter, 1998. [Weber, 2001] M. Weber. The Protestant Ethic and the Spirit of Capitalism. University of Chicago Press. 1904-5, rpt 2001. [Weiss, 1987] S. F. Weiss. Race hygiene and national efficiency: the eugenics of Wilhelm Stallmayer. University of California Press, 1987. [Wise and Smith, 1989-1990] M. N. Wise and C. Smith. Work and Waste: Political Economy and Natural Philosophy in Nineteenth Century Britain. History of Science, pt. I, v. 27, 263301, 1989. Pt. II, v. 27, 391-449, 1989. Pt. III, v. 28, 221-261, 1990.

1030

Jennifer K. Alexander

[Worster, 1994] D. Worster. Nature’s Economy: A History of Ecological Ideas, 2nd ed. Cambridge University Press, 1994.

AESTHETIC VALUES IN TECHNOLOGY AND ENGINEERING DESIGN Joachim Schummer, Bruce MacLennan, and Nigel Taylor∗

1

INTRODUCTION

With few exceptions, most notably in architecture and product design, engineers have been used to pay little explicit attention to aesthetics. Most philosophers of technology have followed that model and excluded anything related to aesthetics from the philosophy of technology. The neglect seems to be justified on basic conceptual grounds. Indeed, many definitions of technology start with discussing the ambiguity of the term “art”, in order to distinguish the useful arts from the fine arts: While both kinds of arts are productive or poi¨etical in the Aristotelian sense, it is said that they fundamentally differ from each other by their values. The fine arts are governed by aesthetic values and thus constitute the proper realm of aesthetics, whereas the useful arts, i.e. technology, are governed by functional values, such as product performance, durability, cost, safety, and so on, as well as by epistemic values in so far as they produce technological knowledge. From such an approach it follows that aesthetic values play only a marginal and at most additional role in certain engineering fields such as product design, in order to please consumers and increase sales. There is a long philosophical tradition of defining the fields of science, technology, fine arts, and ethics in terms of their dominating goals and values, with prominent examples by Aristotle and Kant.1 However, even if certain values are dominating in or characteristic of a certain field, it would be naive to exclude them by definition from other fields in order to maintain a pure systematics. The distinctions between the useful and fine arts and between science and technology have always been debated and indeed redefined many times in the course of history, frequently reflecting the changing social status of the corresponding professions. Moreover, pure science that ignores any functional and ethical values is as hard to find as fine arts that completely exclude these values. Such as ethical values have always played a role in engineering, for instance by inspiring technological ideas of human progress or by prohibiting harmful technologies in codes of conduct, such have aesthetic values been influential by informing design processes, whether ∗ Sections 1, 3 and 5 were written by Joachim Schummer. Section 2 was written by Nigel Taylor and Section 4 by Bruce MacLennan. 1 For instance, Aristotle: Metaphysics, 980a ff.; Kant: Critique of Judgement, §§ 43ff.

Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

1032

Joachim Schummer, Bruce MacLennan, and Nigel Taylor

knowingly or not. Thus, the question is not if aesthetic values do or should play a role in technology. Instead, the question to be dealt with in this article is how aesthetic values inform technology and how they compete or harmonize with other values. Aesthetic values are difficult to define and to identify in engineering activities for several reasons. One reason is that the professional aesthetics discourse has been too narrowly focused on the fine arts including literature, such that, particularly for many Anglo-Saxon aestheticists, aesthetics has become equivalent to the study of the fine arts or art criticism (e.g. [Cooper, 1992]). Unfortunately that makes their conceptual apparatus largely inappropriate for other fields of aesthetics, including engineering aesthetics. Another reason is that scientists and engineers frequently use terms such as “beautiful”, which would otherwise be typical indicators of aesthetic appreciation, to express epistemic or functional approval or to popularize their activity to a broader public. It is useful therefore to start with a broad concept of aesthetic values by considering any values that are not of epistemic, functional, or ethical nature. The remaining values typically include familiar aesthetic values such as beauty, elegance, harmony, (non-epistemic) simplicity and clarity, and familiarity, as well their opposites on which aesthetic disapproval is based. In addition, something can aesthetically please or displease by resemblance to something else that pleases or displeases for aesthetic reasons only, which is typically expressed by analogies or metaphors and which sometimes leads to the formation of aesthetic styles. Whenever such aesthetic values contribute to preferences in engineering decisions, there is evidence that they inform the engineering activity. The focus of this article is on how aesthetic values inform the process of functional engineering design. Rather than looking at how the engineering products are aesthetically received by consumers, we look at how they are designed by engineers and what role aesthetic values play in the research and design process. Of course the distinction is not always clear-cut because, depending on the engineering field, the design is frequently influenced by the anticipated reception by consumers. For instance, in industrial design, the anticipated aesthetic reception by consumers has become an important part of ergonomics that essentially informs design decisions. Moreover, our focus is less on the product itself than on the process of designing the product. That is, we are interested in how aesthetic values have an impact on the various activities and steps that contribute to the design process. That particularly includes the initial choice of the engineering problem to be solved, different steps of the cognitive process of functional design, and various representational tools and media that engineers use in their design process for visualizing and structuring the engineering problem, the strategies to solve it, and the final product. Unlike the frequently assumed uniformity of technology, the various engineering fields have quite different historical traditions and methodologies, so that it is perhaps not surprising that the impact of aesthetic values, as well as the kind of aesthetic values that matter, differ accordingly. Since this article cannot cover all

Aesthetic Values in Technology and Engineering Design

1033

the engineering fields, we have made a selection of three fields that might represent to some extent the diversity of aesthetic values and traditions. The rationale behind the selection is that the size and visibility of the engineering product makes a crucial difference in how the product is designed, both regarding the cognitive processes involved and the representational tools used in that process, and that the role and kind of aesthetics might differ accordingly. Thus the first section, written by Nigel Taylor, deals with aesthetic values in the design of large-scale objects, exemplified by urban landscape planning and architecture. The second section, by Joachim Schummer, investigates the aesthetics of chemistry, which is usually not considered an engineering field, but fits well in our systematics because it has a major focus on the design of small-scale molecular objects. Finally, the subject of the third section, written by Bruce MacLennan, is the role of aesthetic values in the design of virtual objects as performed in software engineering. Both our selection of the engineering fields and our focus on the design process are clear departures from the few classical treatments of aesthetics and technology, which are largely confined to architecture and industrial design. Of course, in architecture, and more recently in industrial design, aesthetics is frequently part of the standard curriculum, either in the form of historical accounts of styles or in the normative-didactic form of teaching students the principles of aesthetically preferred products. That tradition indeed goes back to Vitruvius who, in the oldest extant textbook of architecture from the first century BC, devoted a whole chapter to that topic (De Architectura, III.1). The other classical topic is the “aesthetic assimilation” of machinery by modern artists (e.g., Mumford [1934, ch. VII.3]) and the complementary view of how industrial production has enabled a kind of mass art and influenced a mass aesthetics [Benjamin 1936], which became a standard critique of modern civilization. That has inspired many later sociological, frequently Marxist, studies on industrial design products and the social, political, and economical contexts that have determined their aesthetics (e.g., Haug [1971], Gartman [1994], Brummett [1999]). Furthermore, many critics of modern culture have argued that technology, because it would focus on functional values alone and ignore aesthetic values, drives our culture into an aesthetic vacuum. In so far as the critique was directed at the functionalist movement in 20th-century architecture and industrial design, it turned out, however, that functionalism is an aesthetic style in its own right, which tries to express functionality in its products by aesthetic elements, sometimes to the degree that its products become dysfunctional. 2

2.1

DESIGNING LARGE-SCALE OBJECTS: URBAN LANDSCAPE PLANNING AND ARCHITECTURE

Introduction

The concern of this section is to identify and discuss the main aesthetic values that inform the process of contemporary urban landscape planning and architec-

1034

Joachim Schummer, Bruce MacLennan, and Nigel Taylor

ture. Although the distinction is a fine one, it is worth emphasising that our primary focus will be on the aesthetic values that have informed the process of urban landscape planning and architecture rather than planning and architectural outcomes, and it follows from this that our primary focus will be on the main aesthetic ideas and values that have informed contemporary planning and architecture. Below, I begin by clarifying what, for the purposes of this section, we are refering to when we speak of contemporary urban and landscape planning, and architecture, and what, for the purposes of this review, I take aesthetic values to be. The aesthetic values that underpin contemporary architecture and planning have themselves been largely shaped by the ideas and values that came to dominate the Modern movement in architecture and planning during the late 19th and early 20th centuries.2 Accordingly, following the clarification of basic terms, I shall provide a brief historical overview of the attitudes to aesthetic considerations that prevailed in the Modern movement in architecture and planning before I come to describe the main aesthetic ideas and values that inform contemporary planning and architecture. I shall conclude with some more general reflections on the status and political significance of aesthetics in contemporary planning and architecture. First, our terms of reference. “Urban and rural/landscape planning”3 is often run together with “architecture”, as if these were two aspects of a single — or at least an integrated — discipline and practice. And in fact, this used to be the case, in that “town” planning (as it used to be called) was once widely viewed as an extension of architecture, in the sense that it was concerned with the physical planning and design of whole towns, cities, and even regions (see e.g. Keeble [1952] and Taylor [1998, Ch. 1]). However, since at least the 1960s this “physical design” conception of town and country (or urban and rural/landscape) planning has undergone radical questioning and transformation, so that today urban and landscape planning is viewed as a much more complex process of managing the social and economic functioning of urban settlements and regions, as well as just their physical design. Further, the process of making decisions about urban and rural development (and hence the process of urban and rural planning) is now viewed as a highly political process, since different interest groups and “stakeholders” will often hold differing and sometimes conflicting views and values about the nature and location of new developments such as new roads, airports, major shopping or other commercial developments, and so on. Consequently, the contemporary urban planner is no longer someone who plans towns and cities in the same way that architects design individual buildings, but rather someone who seeks to manage a complex process of arriving at plans and decisions about future urban develop2 Throughout this paper I shall employ the upper case “M” to distinguish the Modern movement in architecture and planning from what is modern in the everyday sense of being recent. 3 For shortness sake, I shall often just write “planning”, instead of the full formulation “urban and rural/landscape” planning. It also needs to be noted that what is termed “urban and rural” planning is sometimes also referred to as “town and country” planning, “city and regional” planning, “environmental” planning, “spatial” planning, etc. In this review I treat all of these as synonymous.

Aesthetic Values in Technology and Engineering Design

1035

ment in ways that reflect a range of different interests and values.4 In short, urban planners are typically not themselves the authors and designers of urban plans in the same way that architects are of their buildings, and this has a bearing on our subject since, in architecture, individual architects may come to express their aesthetic values and visions in the buildings they design whereas, in contemporary urban and rural planning, individual planners have far less creative autonomy in the making of urban plans and so are less able to stamp their particular aesthetic visions on large-scale plans for urban and rural development.

2.2

Aesthetic values in modern architecture and urban planning

The aesthetic values of contemporary urban planning and architecture have been greatly influenced by attitudes to the aesthetic aspects of urban form and buildings that developed during rise of the Modern movement that came to dominate architecture and planning for most of the 20th century, and so a brief overview of the main tenets of Modernism is needed as a foundation to this account. Two points, in particular, emerge from Modernist planning and architecture, each of which stands in some tension to each other. 2.2.1

The Modernists’ concern with functional design independent of aesthetic style

During the latter years of the 19th century and the early part of the 20th century, the pioneers of Modern architecture and planning found inspiration in the great, and seemingly purely utilitarian or functional forms of 19th century civil engineering, such as the iron bridges of Telford and Brunel, Paxton’s Crystal Palace, Eiffel’s famous tower for the Paris exhibition of 1889 (see e.g. [Giedion, 1941]). This fascination with the great works of 19th century engineering combined with an equally powerful disdain for the “revivalist” architecture that had dominated the 19th century, and the matching architectural theory that claimed that the quality which distinguished architecture from merely utilitarian building was “style”. For most architects (and architectural theorists) of the 19th century, the distinguishing mark of architecture was a self-conscious concern with the aesthetics of built form independent of utilitarian or functional considerations. Such a view was articulated by the influential 19th -century critic John Ruskin who, in his book The Seven Lamps of Architecture, defined architecture in terms of the aesthetic aspect of building. And since the aesthetic content of a building was not strictly necessary to its utilitarian purpose, it followed for Ruskin that architecture concerned that aspect of building design that was, literally, “unnecessary”. As he put it: “Let us [...] confine the name (of architecture) to that art which, taking up and admitting [...] the necessities and common uses of the building, impresses on its form certain characters venerable and beautiful, but otherwise unnecessary. 4 For an account of the evolution of changing conceptions of urban planning over the last century see [Taylor, 1998; 1999].

1036

Joachim Schummer, Bruce MacLennan, and Nigel Taylor

[...] that is Architecture” [Ruskin, 1849, Ch. 1, pp. 14-15]. For Ruskin, then, considerations of style and aesthetics in their own right were what distinguished architecture from building. And this also distinguished architecture from engineering, for whereas the prime concern of architecture was with the aesthetics of built form (and hence style), the prime concern of engineering and construction was with structures that were functional to their purpose; indeed, this is what, for Ruskin, made architecture, but not engineering, an “art”.5 Now, it was just this distinction between “aesthetic architecture” and “functional building and engineering” that the early Modernists contested. Thus they challenged, on the one hand, the view that purely functional buildings and forms could not also be works of architectural art, and, on the other hand, the view that architecture required ornamentation and “style” to be architecture. Indeed, some of the pioneers of Modernism went so far as to assert that a truly modern architecture — that is, an architecture consonant with the contemporary age of industry, science and engineering — should be an architecture whose forms were primarily, if not solely, governed by functional considerations, and hence an architecture that would be free of independent stylistic or aesthetic considerations. Thus the early Modern American architect Louis Sullivan suggested that the “form” of buildings should “follow” their function, and, as regards the use of “unnecessary” ornament on buildings, he wrote: “I should say that it would be greatly for our aesthetic good if we should refrain entirely from the use of ornament for a period of years, in order that our thought might concentrate upon the production of buildings well formed and comely in the nude” [Sullivan, 1892]. It was in this way that early Modernist architects downplayed aesthetic considerations, even to the extent regarding them as irrelevant to the design of (a genuinely modern) architecture.6 2.2.2

The aesthetics of geometrical “purism” in modern architecture and planning

In spite of the rhetoric of functionalism that played such a central role in early Modernist architecture and planning, we find that most of the leading figures of the Modern movement of architecture and planning were actually very particular about the aesthetics of the new Modern architecture, so much so that, by the 1930s, a distinguishable style — the so-called “International Style” — had come to dominate Modern architecture, so that henceforth Modern architecture became instantly recognisable because of its style. In particular, the kind of architectural forms preferred by the pioneers of Modern architecture were “pure” geometrical 5 Ruskin

also defined architecture as “the art which [...] disposes and adorns the edifices raised by man, for whatsoever uses, that the sight of them may contribute to his mental health, power, and pleasure” [Ruskin, 1849, Ch. 1, p. 13]. What Ruskin failed to acknowledge was that the great 19th -century engineers also sought to create aesthetically beautiful forms whilst solving functional problems, as the writings of David Billington have made clear (see e.g. [Billington, 1979; 1983]. 6 As we shall see, not all Modernists adhered to this purely functionalist view of architectural design, but one who did was Hannes Meyer, the second director of the Bauhaus School of Art and Design (see e.g. [St John Wilson, 1995]).

Aesthetic Values in Technology and Engineering Design

1037

forms — forms that were “pure” in two senses. First, the actual forms out of which the new architecture was composed were themselves pure geometrical forms, such as cubes, rectangles, cylinders, and cones, and second, these forms were pure in the sense that they were left plain and undecorated. Thus Le Corbusier, in one of the seminal texts of architectural Modernism, advocated an architecture of “primary” forms: Primary forms are beautiful forms because they can be clearly appreciated. [...] Architecture is the masterly, correct and magnificent play of masses brought together in light. [...] cubes, cones, cylinders or pyramids are the great primary forms which light reveals to advantage. [...] it is for this reason that these are beautiful forms, the most beautiful forms. [Le Corbusier, 1927, pp. 26, 31] In fact, Le Corbusier was once held to be responsible for characterising Modern architecture as “functional” architecture because, in a letter he wrote to the Italian Alberto Sartoris about a book the latter was writing about the new architecture titled Architettura Razionale (Rational Architecture), Le Corbusier commented that: “The title of your book is limited: it is a real fault to be constrained to put the word Rational on one side of the barricade, and leave only the word Academic to be put on the other. Instead of Rational say Functional [...]” (quoted in Banham [1960, Ch. 22, p. 320]).7 In spite of this, Le Corbusier was one of the leading advocates of the purist aesthetics of Modern architecture (see St John Wilson [1995]), and in his Towards a New Architecture, he drew essentially the same distinction between architecture and engineering as Ruskin had earlier drawn: it will be a delight to talk of ARCHITECTURE after so many grain stores, workshops, machines and skyscrapers. ARCHITECTURE is a thing of art, a phenomenon of the emotions, lying outside questions of construction and beyond them. [...] You employ stone, wood and concrete, and with these materials you build houses and palaces; that is construction. Ingenuity is at work. But suddenly you touch my heart, you do me good, I am happy and I say: ‘This is beautiful’. That is Architecture. Art enters in. My house is practical. I thank you, as I might thank railway engineers or the telephone service. You have not touched my heart. But suppose that walls rise towards heaven in such a way that I am moved. [...] The relationships between them have not necessarily any reference to what is practical or descriptive. They are a mathematical creation of your mind. They are the language of Architecture. By the use of inert materials and starting from conditions more or less utilitarian, you have established certain relationships 7 Sartoris quoted Le Corbusier’s letter approvingly in the preface to his book, which was published in 1932 with the revised title Gli Elementi dell’Architecttura Funzionale (The Elements of Functional Architecture).

1038

Joachim Schummer, Bruce MacLennan, and Nigel Taylor

which have aroused my emotions. This is Architecture. [Le Corbusier, 1927, pp. 23, 187] In short, in spite of the rhetoric about functional design in the Modern movement, there emerged a distinct “purist” aesthetic style that governed Modern architecture, and indeed, this often prevailed at the expense of genuine functional design (see again [St John Wilson, 1995]). And it was the very ubiquity of the aesthetic values of pure geometrical form, plain undecorated surfaces, and harmonious geometric proportions that, by the 1930s, led Modern architecture to be instantly recognisable as a distinctive style. And it was because this style was adopted by Modernists right across Europe and North America that the style was later dubbed the “International Style”. Three points are worth adding to this account of the aesthetic values that underpinned Modern architecture. First, whilst the early Modern architects praised, and in many ways sought to imitate, the purist aesthetics of the functional forms designed by the civil engineers of the 19th and 20th centuries, it is misleading to presume that, for their part, the great engineers were only preoccupied with functional considerations in their designs. To be sure, bridges and grain stores had, of necessity, to be fit for their purposes of spanning rivers and storing grain. But as David Billington has pointed out in a succession of writings (e.g. [Billington, 1979; 1983]), the great engineers of the 19th and 20th centuries sought also to create beautiful forms and, in so doing, most of them also sought to create forms of geometrical purity and harmonious proportions. In this respect, Billington [1979] is correct to describe the memorable works of 19th and 20th century civil engineering — such as Robert Maillart’s bridges — as works of “structural art”. Second, the aesthetic values — of both the great engineers and the early Modernist architects — which emphasised purity of plain geometrical forms and harmonious proportion were values that had been emphasised before in Western architectural and engineering design, namely, in the great tradition of classicism. For this reason, in spite of its seemingly radical new forms, Modern architecture has sometimes been viewed as a revival of classical aesthetic values (see e.g. [Gelernter, 1995, Ch. 8, pp. 225-229; Ch. 9, pp. 252-254]). Third, the purist aesthetics of Modern architecture also found expression in Modernist urban planning which, in the era of the rise of Modernism, was dominated by Modern architects’ visions of the “city of the future” (see e.g. [Taylor, 1998, Ch. 2]). Le Corbusier was again an important figure here, arguing for the modern city to be more clearly organised than its historical counterpart, with the major urban land uses distinguished from each other and allocated to distinct “zones”, and also advocating a radically new urban morphology in which urban activities such as housing and commerce would be accommodated in large tower and slab blocks standing in open parkland and connected by urban expressways (see [Le Corbusier, 1924; 1933]). Such a vision represented a new conception of urban space in which buildings would stand out as individual forms punctuating a sea of free-flowing space, rather than being viewed in combination with other buildings that together bounded and defined distinct urban spaces such as streets

Aesthetic Values in Technology and Engineering Design

1039

and squares. This open-ended conception of space was also a central aesthetic value in the new architecture, where great sheets of glass curtain walls broke down the visual barrier between the outside and inside of buildings, thus allowing space seemingly to flow freely through buildings.

2.3

The critique of Modernist architecture and urban planning, and aesthetic values in architecture and planning in the contemporary “post-modern” era

According to some accounts, the Modernist architectural ideas and values described in the previous section had all been worked through by the end of the 1920s, so that this excursion into early Modernist aesthetic values might not seem so relevant to an examination of aesthetic values in architecture and urban planning today. But such a response would be superficial. For the Modernist ideas and values that developed at the beginning of the 20th century cast a spell over the remainder of that century and, in many respects, continue to hold sway today. Thus the design and form of many contemporary large-scale works of architecture and engineering continue to be informed and shaped by the purist geometrical aesthetics that were the mark of Modernism, so much so that some architects and critics continue to refer to this work as stylistically (if not ideologically) “Modernist” or, perhaps, as “Late Modernist” (see e.g. [Jencks, 1980]). What has complicated the picture over the last forty years since the 1960s has been the emergence of a vigorous critique of Modernism in architecture and planning, and the consequent assertion of alternative aesthetic values. The reaction against Modernism came first in urban planning in the early 1960s. The heyday of Modernism in architecture and planning was, in fact, in the quarter century following the Second World War, when the Modernist ideas that had emerged in the first half of the 20th century were enthusiastically adopted for large-scale schemes of “post-war” urban planning, most famously and notoriously in the great schemes of “slum clearance”, “comprehensive redevelopment”, and social housing that took place in most of the old industrial cities of Europe and North America, but also in the planning and building of completely new cities, such as the new capital cities of Brasilia and Chandigarh. And very quickly most of these schemes of Modernist urban surgery came to be regarded as having failed, abysmally, to create either the functionally efficient or the aesthetically beautiful utopias that the pioneers of Modernism had dreamed of. Indeed, so quick was the revulsion against what Alison Ravetz described as the Modernist “clean sweep” approach to planning that, in 1972, the award-winning Pruitt-Igoe social housing scheme in St Louis, Missouri, was itself largely swept away by being dynamited. And paradoxically (given the functionalist rhetoric that had attended the rise of Modern architecture), it was precisely the lack of an adequate analysis and understanding of how successful cities (and the human beings within them) functioned that was at the heart of the trenchant criticisms of Modernist urban planning of the 1960s. Thus Jane Jacobs [1961] and Christopher Alexander [1965] pointed out

1040

Joachim Schummer, Bruce MacLennan, and Nigel Taylor

that the Modernist urban utopias of architects such as Le Corbusier showed no real understanding of the subtle and complex inter-relationships between people and activities in real cities. Here, again, lay that conflict between the requirements of genuine functional planning and design, and the “purist” aesthetic values that, under the influence of such figures as Le Corbusier, had come to dominate Modern urban planning as well as Modern architecture. But the new post-war architecture and planning was also criticised simply in aesthetic terms. For the repetition of the same (or similar) bare blocks across large tracts of towns was seen as visually monotonous and alienating. As Charles Jencks put it, in reflecting with relief on the passing of the aesthetics of mass-produced Modern architecture, “ the era of stupid and inarticulate slabs is over, the age of the repetitive clich´e is finished” [Jencks, 2002, p. 2]. It was the lack of aesthetic richness and variety in Modern architecture that was the focus of Robert Venturi’s criticism of “International Style” Modernism in his seminal publication Complexity and Contradiction in Architecture of 1966. At the outset, Venturi parodied the accepted aesthetic norms of “International Style” Modernism in opening his book by countering these with their opposites and implying that the choice between the two was not, as some Modernists had claimed, a rational one, but rather one of personal aesthetic taste and preference: I like complexity and contradiction in architecture. [...] Everywhere, except in architecture, complexity and contradiction have been acknowledged [...]. Architects can no longer afford to be intimidated by the puritanically moral language of orthodox Modern architecture. I like elements that are hybrid rather than ‘pure’, compromising rather than ‘clean’, distorted rather than ‘straightforward’, ambiguous rather than ‘articulated’ [...]. I am for messy vitality over obvious unity [...]. I am for richness of meaning rather than clarity of meaning. [Venturi, 1966/1977, Ch. 1, p. 16] In place of the aesthetic purism and minimalism of Modern architecture, Venturi advocated an architecture that is aesthetically diverse and complex, even ambiguous and contradictory, and which is also unafraid to re-introduce ornament and decoration into its forms and surfaces. It was thus that Venturi’s Complexity and Contradiction became the inspiration for “Post-Modern” architecture that overtly challenged the aesthetic values of Modernism, even to the extent of reviving historical architectural styles and motifs (see e.g. [Jencks, 1977]). Indeed, just as some past periods of architecture have witnessed a “battle of styles” (e.g. between neo-classical and neo-gothic styles in the 19th century), so the last forty years have witnessed a similar battle between those architects and designers who have wished to perpetuate whilst further developing the project of Modernism, and those Post-Modernists and others who have sought to challenge and go beyond Modernism. To be sure, the on-going protagonists of Modernism have not simply persisted with reproductions of the “International Style”. As Modernists, they have insisted

Aesthetic Values in Technology and Engineering Design

1041

that a genuinely Modern architecture must be modern in the everyday sense of designing buildings that use the materials and possibilities generated by the latest technology; hence the coining of the terms “Late Modern” and “Hi-Tech” to describe the varieties of architecture that have been spawned by this late efflorescence of Modernism. Nonetheless — and at the risk of some crudeness of generalisation — what makes this Late Modern architecture still aesthetically Modern is the persistence of a preference for pure form and plain, undecorated surfaces. We see this, for example, in contemporary glass architecture, which, in its pristine elegance and light transparency, is the inheritor of that formally minimalist tradition pioneered by Mies van der Rohe. In contrast to this contemporary Modernist architecture, “Post-Modern” architecture is characterised by an eclectic mix of building styles — some “progressive” in the sense that its authors are unafraid to adopt recognisably Modern forms and yet equally unafraid to mix these with more traditional echoes and decorative motifs, some conservative to the extent of reviving — lock, stock and barrel — the architectural styles of a past age. The architecture of (for example) Michael Graves illustrates the former tendency, while the replica Georgian terraces and mansions of Quinlan Terry are an example of the latter. This account has come to focus on the contemporary scene in architecture, and a brief word needs to be added about the aesthetic values of contemporary engineering and urban planning to provide a more balanced picture. In the case of those large-scale objects in townscapes and landscapes that are chiefly the products of contemporary civil engineering, such as new bridges and power stations, the aesthetic form of these has continued to be — perhaps for necessary functional as well as aesthetic reasons — to reflect those aesthetic values of pure geometrical form and harmonious proportions that became the hall-mark of Modern design (see again [Billington, 1983]). By contrast, the state of affairs in urban planning following the disasters of large-scale modernist development of the 1960s is altogether more complex. But two generalisations can be made. First, the general failing of modernist architect-planners to understand and plan sensitively for the greater complexity of cities and human settlements led to a radical change of view about the nature of urban planning as a profession and the most appropriate qualifications for its practice. In particular, the view emerged that — at least at the larger scales of the planning of the land uses and development of whole cities and regions — planners with a prior training in geography and/or the social sciences were better equipped properly to undertake the necessary analysis for this kind of urban planning. This refocusing of the profession of urban planning away from architecture and design led in turn to a greater emphasis being placed on “social and economic” (and also political) considerations in urban planning, and a corresponding downplaying of aesthetic considerations (see e.g. [Taylor, 1998, pt. II]). Secondly, in so far as aesthetic considerations did continue to play a part in urban planning and decision-making about large-scale development, there emerged a renewed emphasis on the “rehabilitation” of old buildings, rather than their comprehensive redevelopment, and more generally a concern for urban conservation and environmental protection. This more conservative aesthetic has arisen partly

1042

Joachim Schummer, Bruce MacLennan, and Nigel Taylor

in reaction against the large-scale comprehensive planning of the 1960s that had so insensitively swept away large tracts of cities to which many citizens were emotionally attached, and partly as a result of the emergence of ecological thinking and the “green” movement after the 1960s, (see e.g. [Bishop and Phillips, 2004; Larkham, 1996]). It can also be seen to be a result of the greater politicisation of urban planning as a result of the protests of communities against the insensitive planning of the 1950s and 60s; as noted in the introduction, to a much greater extent than architecture, what now happens in the name of urban planning is often the result of many voices and interests, rather than a single creative designer.

2.4 Conclusion: the importance of aesthetic values in contemporary urban development and planning I shall conclude this brief resume of the aesthetic values that have informed contemporary architecture and urban planning by commenting more explicitly on the importance attributed to aesthetic considerations in the process of design and planning. I offer three general reflections. First, although the Modern movement in architecture and planning was heavily infused with the rhetoric of “functionalism” and, connectedly, with the notion of “rational design”, as we have seen, its chief protagonists advanced some clear aesthetic values, namely, those that emphasised the “classical” virtues of pure geometry and proportion, combined with a utilitarian cleansing away of “unnecessary” ornamentation and decoration. Nonetheless, the rhetoric of functionalism and rationality did have an important impact in undermining the importance attributed to aesthetic considerations in the training of architects and planners, and — by extension — in the process of architectural design and decision-making in urban planning. Thus still today many university programmes in architecture and planning do not contain any separate courses on, or provide any systematic education in, aesthetics. In this respect, the process of design in contemporary architecture and urban planning remains what Hearn [2003, Ch. 4, p. 81] has termed an “inside-out” approach in which the design, and hence the exterior form of built structures is determined primarily by the prior demands of how best to accommodate the internal functions, rather than by some preconceived idea of what the exterior form should look like. In this respect, Le Corbusier’s observation that “The plan is the generator” has indeed come to pass [Le Corbusier, 1927, pt. III, pp 44-45]. Second, this relegation of aesthetic values to (at best) a subsidiary role in the process of contemporary architectural design and planning mirrors the relatively minor importance attributed to environmental aesthetic considerations in political decision-making, and in contemporary developed societies more generally. Thus in considering proposals to build new power stations or shopping centres, roads or port facilities, the primary focus of debate in contemporary societies tends to be on current or projected demands for energy, consumer products, travel and transhipment, rather than on the aesthetic impact that the large-scale structures

Aesthetic Values in Technology and Engineering Design

1043

associated with these activities will necessarily have on the landscape. To be sure, since (roughly) the 1980s, the wider “environmental” impact of planned largescale developments has become an ever increasingly important consideration. But these wider environmental considerations focus mostly on the degree to which new development is “environmentally sustainable”, and here the prime considerations are ecological, not aesthetic. But the marginalisation of aesthetic considerations in contemporary developed societies is also associated with more deeply ingrained attitudes, values, and philosophical assumptions. Thus, even when it is acknowledged that the aesthetic quality of our surroundings is important, the objection is frequently made that qualitative aesthetic judgements are a matter of “subjective” personal taste, from which it is typically inferred that there can be no generally accepted norms or principles to govern the aesthetic form or “style” of major new developments. Then again, and perhaps because of its association with “the arts”, the aesthetics of architecture and planning are widely assumed to be just a “luxury” in comparison with allegedly more fundamental “social and economic” matters, and even, because of this, only of importance to privileged elites or the “middle-class”.8 And yet, in spite of these prevailing attitudes and values, almost every day numerous cases surface where it is apparent that ordinary people do care enormously about the aesthetic quality of their surroundings and the large-scale objects that threaten to alter the character of places. Recent controversies over the siting of wind turbines in open landscapes (in spite of their otherwise beneficial environmental effects) are a vivid illustration of this concern, as are the numerous other campaigns that have been fought throughout the developed world, often with great passion, against new development proposals for roads, airports, etc, that bring about the destruction of cherished landscapes and townscapes. We touch here on a paradox concerning this subject. This is that, on the one hand, it often appears that aesthetic considerations in relation to the built environment are not highly valued. And yet, on the other hand, there is also plenty of evidence — for example, from people’s choices about where to live and take their leisure — that suggests that the aesthetic quality of places is of central importance to the quality of people’s lives. Certainly, the latter has been revealed in some studies. Thus, in a very thorough investigation into people’s attitudes to green urban open spaces (urban parks) in London, Jacqueline Burgess and her fellow researchers showed that ordinary working class people, including ethnic minorities, valued very highly the sensory, aesthetic pleasures of having access to green urban spaces, so that this was very far from being an unimportant, or simply a “middle class”, issue (see [Burgess et al., 1988]). Whilst, then, aesthetic values rarely figure prominently in everyday public and political debate, and so might seem to be an unimportant political issue, in people’s everyday lives, and in numerous local campaigns and forms 8 One of the characters in Milan Kundera’s novel The Unbearable Lightness of Being — Sabina, a painter — considers that, for all our material wealth, our age is an age of ugliness, or at least an age where a concern with the beauty of our surroundings is not widely thought to be a central factor in the quality of life [Kundera 1984, p 93].

1044

Joachim Schummer, Bruce MacLennan, and Nigel Taylor

of “community action”, the aesthetic quality of the built environment emerges as a most important, and often highly charged political matter. 3 DESIGNING SMALL-SCALE OBJECTS: CHEMISTRY

3.1 Introduction Chemistry is the study of material substances, their properties and, particularly, their chemical transformations into each other. The focus on chemical transformations explains chemistry’s rather undefined position according to standard distinctions between science and technology [Schummer, 1997a]. On the one hand, chemical transformations constitute chemical properties, for instance the capacity to react under certain conditions with another substance to form a third substance, which are characteristic properties for each chemical substance. On the other, such chemical knowledge about substances enables one to actually perform these transformations in order to create new substances. Since the mid-19th century, theoretical developments, most notably molecular structure theory, have allowed chemists to design and synthesize new substances on a regular basis at a tremendous speed, such that there are in 2006 about 90 million known substances [CAS, 2007]. The synthetic activity, which dominates chemistry overall, establishes a clear similarity to the activity of engineers. However, only a small, but increasing, fraction of chemical synthesis is performed with the goal of providing useful applications outside of chemistry. Indeed most new substances are produced in the course of research to further improve the synthetic capacity of chemistry both on the experimental and theoretical level [Schummer, 1997b; 1997c]. The unclear status of synthetic chemistry, between being a science and being a technology in the received sense, which has recently been called technoscience, has opened a space for values other than scientific truth and technological performance. The task of this section is to investigate the role of aesthetic values in that space and if and how they have influenced research decisions and directions both in a positive and negative way from epistemic and utilitarian perspectives. Unlike many previous studies of aesthetics in science, I do not start with the a priori assumption that the impact of aesthetics is always positive with regard to other values. If there is an impact at all, it seems more likely that this results in a conflict of values. However, because of the unclear status, it is sometimes difficult to determine the exact impact of aesthetic values on classical epistemic and engineering values in chemistry. Moreover, because synthetic chemistry is embedded in the wider context of general chemistry, it is useful to consider also the role of aesthetic values in non-synthetic fields of chemistry. Aesthetic studies of chemistry are still in a rudimentary state (for a first collection of essays, see Spector and Schummer [2003]). That is in contrast to the frequent references to aesthetics by many chemists. Indeed, since the mid-19th century chemists have frequently compared their synthetic work to that of artists, because like artists and unlike other natural scientists, they create their own ob-

Aesthetic Values in Technology and Engineering Design

1045

jects of study on a regular basis or are creative inventors in a similar sense as artists are [Jacobs, 2006, Ch. III.4-5]. However, such references sometimes make use of the ambiguity of the term ‘art’, which comprises the fine arts as well as engineering and crafts. Thus, not every reference to ‘art’ entails an unambiguous reference to aesthetic values. Moreover, since the 1960s, when the public justification of high energy physics for weapon research lost considerable ground, physicists have tried to rehabilitate their public image by pointing to the alleged beauty of their theories [Stevens, 2003]. The reference to beauty along with the comparison to the work of fine artists thus became standard rhetoric in the popularization of physics and other disciplines including chemistry. For an analysis of the role of aesthetic values in the actual research process, it is therefore necessary to exclude such popularizations efforts via beautification as long as they are, despite their popularity, just a fa¸cade. The following analysis focuses on four aspects of chemical research in which aesthetic values have played a discernible role: microscopic structures with a particular focus on symmetry (3.2), molecular representations (3.3), chemical experimentation (3.4), and mathematical modeling in chemical engineering and physical chemistry (3.5). Because knowledge of chemistry is frequently not common among philosophically minded readers, I will draw only on very prominent research examples that were mostly awarded by Nobel Prizes. The underlying concept of aesthetic values, and of aesthetics more generally, is intentionally broad. As a guiding principle, I identify aesthetic values through appreciations by chemists that are clearly not based on epistemic, instrumental, or consequentialist-ethical values. The identification by exclusion has the advantage of clearly distinguishing aesthetic values from other value fields, which enables determining their mutual impact.9

3.2

The quest for symmetry as guiding and misguiding research principle

From ancient Greece to the early 19th century, symmetry was a purely aesthetic concept to describe the balanced proportions, which were taken from the model of the perfect human figure, both between the parts of an artwork and between each part and the whole. In contrast the modern concept of symmetry, which was developed only in mid-19th century crystallography [Schummer, 2006a], is a mathematical description of forms according to the invariance with regard to certain transformations, such as reflection on a mirror plane, rotation around an axis at a certain angle, or lateral translation by a certain length. In this approach, the 9 The disadvantage of this approach is that some sophisticated relationships between aesthetic and non-aesthetic values might be overlooked, such when aesthetic values and epistemic values tend to coincide or mutually contribute to each other or when aesthetic values are articulated in epistemic terms. However, as in any applied field, there is a limit to conceptual sophistication in applied aesthetics, because the concepts need to be useful to distinguish clearly between real cases based on the available evidence, whereas overly conceptual fine-tuning might only result in confusion.

1046

Joachim Schummer, Bruce MacLennan, and Nigel Taylor

higher the symmetry, the simpler is the form, which makes symmetry a measure of mathematical simplicity. Because of the influence of Plato, who considered mathematical simplicity a measure of natural beauty, and because of the double meaning of the term “symmetry”, mathematical symmetry has become an aesthetic criterion in science, unlike in art and aesthetics. Following Kant (Critique of Judgement, §22) one could argue that scientists value symmetry/simplicity because it pleases their epistemic rather than their aesthetic sense. However, in as much as symmetry/simplicity is not an accepted epistemic criterion in the experimental sciences, it describes an extra-epistemic value and an important heuristic research principle, and only as such it may be called aesthetic. Mathematical symmetry plays a fundamental role in chemistry to describe crystal structures and molecules, to identify forms of molecular isomerism, to develop quantum-chemical models, to analyze spectroscopic results, and so on. There are even quantum-chemical rules, the Woodward-Hoffman rules for which Roald Hoffmann received the 1981 Nobel Prize in Chemistry, that predict the products of certain reactions from the symmetry of molecular orbitals. Apart from such routine uses, however, symmetry is also a guiding principle of research by suggesting certain explanations about the natural order of substances or certain synthetic strategies for the design of new products. In these extra-epistemic contexts, symmetry functions as an aesthetic principle that can guide or misguide research from an epistemic point of view. Three examples may illustrate that. One of the most flourishing fields of chemistry since the late 19th century has been the synthesis and study of transition metal complexes. These compounds, which were long neglected because they belonged neither to organic nor to inorganic chemistry, have received particularly attention because of their potential use as catalysts in petrochemical processes and polymer production. In liquid solution their structure is rather instable, so that they are complexes rather than molecules, which made their structural analysis very difficult. Synthesizing and studying hundreds of such compounds in the 1890s, Alfred Werner (1866-1919) brought order to the matter and thus established the entire field, for which he eventually received the Chemistry Nobel Prize in 1913. Since he found that other atoms combine with transition metals only at the numbers of 3, 4, 6, and 8, he suggested that these atoms are coordinated around the transition metals in a regular way. And because Werner, like Plato, believed that “nature” prefers simple and symmetrical structures, he suggested that complexes form regular polyhedra, for instance, that coordination number 6 corresponds to a regular octahedron. Werner’s aesthetic intuition proved largely successful in later x-ray diffraction studies, but exceptions began to grow. In a theoretical study of 1937, Hermann A. Jahn and Edward Teller showed that in certain cases regular polyhedra are instable, such that the actual structures are distorted polyhedra. The result was a blow to all Platonist, because it suggested that “nature” sometimes prefers distortion to regularity. However, Werner’s aesthetically driven choice has survived as a first-order approach to structural classification that distinguishes between regular structures as the norm and the distorted ones as exceptions.

Aesthetic Values in Technology and Engineering Design

1047

While these distortions cannot be corrected by chemical means, there are many other examples where chemists have worked hard to produce the ideal, aesthetically preferred form. The most prominent one is the ideal crystal, which requires tremendous efforts at purification and recrystallization, without being ever achieved in practice because of remaining impurities and entropy effects. The ideal crystal has perfect translational symmetry such that a small unit represents the whole crystal, which allows for theoretical representation. In addition to these theoretical advantages, approximately ideal crystals sometimes have distinguished properties of practical importance. For instance, the perfect metal crystal has maximum electric conductivity and the perfect diamond has maximum transparency and stability. However, there is no general rule or law according to which only ideal crystals have properties optimized to material needs. In contrast, artisans such as smiths and steel-makers have long benefited from impurities and crystal defects in their products. Chemists, on the other hand, when synthesizing new materials for technical applications, have virtually always worked towards pure and ideal crystals and then checked for their suitable properties. The engineering approach by chemists thus follows the aesthetic preference of the pure and ideal form. While that has proved successful in some cases, it completely ignores, and despises, the entire field of impure, disordered, and defect crystals for aesthetic reasons. However, since the 1970s, that field has been explored by the newly emerging discipline of materials science and engineering. In particular, nanostructured materials, with crystal defects and disorders in the nanometer range, are the most flourishing and promising field, because tailoring the defects has become a means of tailoring unprecedented properties. The example illustrates that aesthetic values can be deeply misleading to the extent that they make you blind for rich opportunities, which, in this case, were harvested by others who either ignored or embraced the opposite of the aesthetic values. Another chemical field in which the aesthetics of symmetrical forms has played a dominant role is the synthesis of molecules in which carbon atoms bind to form regular polyhedra or Platonic bodies. Since carbon atoms usually bind with bond angles of 109˚, such molecules require increasingly distorted bonds if one goes from octahedron to cube to tetrahedron. Therefore, such molecules are extremely unstable and difficult to make, which requires sophisticated synthetic strategies. Indeed many research groups worked for years, if not for decades, on the synthesis of regular carbon polyhedra since the 1970s. It was rather like a sports competition, in which the goal was aesthetically attractive but extremely difficult to achieve [Grahn, 1981; de Meijere, 1982; Hoffmann, 1990; Hargittai, 2000, pp. 419f.]. Apart from the aesthetic attraction, it is questionable if there were at the beginning any aims involved other than that achieving the goal would require major improvements in the synthetic toolbox for the benefit of synthetic chemistry. Only later they discussed possible spin-offs, such as the use of these extremely unstable compounds as explosives or as cages for the inclusion of ions. The aesthetic fascination with regular carbon polyhedra even involved a broader public in 1985 when Harold Kroto, Robert Curl, and Richard Smalley incidentally

1048

Joachim Schummer, Bruce MacLennan, and Nigel Taylor

made and discovered a soccer-ball-like stable carbon structure, which they called Buckminster fullerene and for which they received the 1996 Nobel Prize in Chemistry. Although that opened up the field of fullerenes as a new class of carbon compounds, for which technological applications were soon desperately sought for, the original fascination was a purely aesthetic one. Taken together, the three examples discussed above prove that the classical aesthetic preference of symmetry and pure forms can play mixed roles with regard to expistemic and functional research values. It can provide a useful (first-order) guide, as with Werner’s structural classification of transition metal complexes; it can be deeply misleading, as with the chemists’ neglect of impure and defect crystals; and it can provide arbitrary orientation for research whose usefulness needs to be established only afterwards.

3.3 Aesthetics of molecular representations Like in other fields of science and engineering, colorful images are nowadays omnipresent in chemistry both in research publications and in public presentations. Enabled by recently improved print and display technologies, these images help make a field more attractive to colleagues, students, and a general public and as such are tools of popularization. However, visual representations of molecules have also been very important in chemical research at least since the mid-19th century. Indeed chemists have developed their own sign languages which they use not only for presentations but also for their own research planning and contemplation. They have built their own molecular model sets or used stereo images for threedimensional representations and eagerly embraced the latest innovations, including interactive Internet images and virtual reality sets for the visual understanding of molecular structures. These visualizations are necessary tools in the research process, as they help formulate questions and find solutions. Thus, it is more than likely that the graphic styles and aesthetic elements have an important impact on chemical research directions, that research is frequently stimulated by aesthetic experiences. While case studies are rare is this area, chemists have frequently expressed ideas in that regard. There is at least one example illustrating that such aesthetic experiences can stimulate the development of an entirely new research field, here the fields of supramolecular chemistry and molecular nanotechnology [Schummer, 2006b]. In addition to their fascination with symmetrical molecules (see above), chemists have been particularly enthralled since the early 1980s by molecules that “look” like ordinary objects. Because molecules are invisible, indeed the result of a model approach that reasonably applies only to certain substance classes, it is rather a set of molecular images that have raised their fascination. These images are captivating because of their ambiguity. On the one hand they refer to entities in the molecular world, on the other hand they refer to objects of the ordinary world, like a basket with a handle, a wheel on an axis, or a two interlocked links of a chain. From a classical chemical point of view, these two worlds are quite

Aesthetic Values in Technology and Engineering Design

1049

disparate and disconnected from each other, because all the molecular properties that chemists are interested in are just missing in ordinary objects and vice versa. However, owing to their ambiguity, the images connected these two worlds in a productive manner that stimulated the imagination of combining both worlds into one. One way to combine both worlds appeared in cartoons of little humans walking through and playing with molecules like ordinary objects. Another way was to reproduce by chemical means the ordinary world in miniature. Indeed, since the 1980s, chemists have imitated all kinds of ordinary world objects on the molecular level, from funny things like dogs and pigs to technological artifacts like gears, turnstiles, and elevators. They have developed a whole battery of molecular systems and devices with various mechanical and electrical functions, like molecular machines and circuits. The field thus inspired by the aesthetic phenomenon of ambiguous images came to be known as supramolecular chemistry and, more recently, as molecular nanotechnology. Umberto Eco’s semiotic theory of aesthetics [Eco, 1962/1989] is a useful approach to understand the aesthetic inspiration that has triggered the historical development [Schummer, 2006b]. Faced with ambiguous signs, the interpreter is prompted to lower the tension of ambiguity by developing new, potentially reconciling interpretations and by contemplating and revising the form of the signs. Indeed supramolecular chemist have not only tried to solve the ambiguity by reproducing the ordinary world on the molecular level, they have also developed a new chemical language of technomorph signs which they frequently use in combination with classical structural formulas. In accordance with Eco’s aesthetic theory, this creates a new productive tension that calls for reinterpretation and semiotic revision as a reiterative process, which chemists perform by exploring further parts of the ordinary world on the molecular level and adjusting their sign language. In Eco’s theory, the process eventually reveals more about the interpreters and their imagination than about the original signs. Estimated from the specific areas of the ordinary world that chemists have selected to imitate on the molecular level, chemists revealed a deep fascination with mechanical and electrical engineering. The aesthetic experiences that stimulated the emergence of supramolecular chemistry and molecular nanotechnology are difficult to grasp by the classical aesthetics of beauty. Moreover, it is hard to identify the aesthetic values underlying the chemists’ aesthetic fascination with certain molecular representations. The example thus illustrates that the field of aesthetics in science is much richer than a simple product-oriented aesthetics of beauty would suggest, that intermediate representations and their symbolic references play an important role, and that more sophisticated aesthetic theories, like Eco’s, are able to explain important research dynamics, which would otherwise remain miraculous.

3.4

Aesthetic virtues of chemical experimentation

Scientists frequently use aesthetic categories like beauty to denote the importance, historical significance, or model character of certain experiments, as in top ten lists

1050

Joachim Schummer, Bruce MacLennan, and Nigel Taylor

of “the most beautiful experiments” [Freemantle, 2003]. In so doing, they make some kind of value judgments without expressing the specific kind of value they mean. In order to identify the aesthetic kernel of such statements it is useful to exclude first the non-aesthetic values that are frequently confused with beauty. If an experiment is valued only because it produced new knowledge or confirmed or refuted a theory, the underlying value is not of aesthetic but of epistemic nature. Likewise, historical significance or importance is clearly not an aesthetic but an instrumental value, because it values something only because it enabled something else, for instance the subsequent development or the present state of the art, which here alone is considered valuable in its own right. More generally, if an experiment is valued only because of its result, for instance the synthesis of an important substance or some economic improvement, it is not the experiment but the result that matters. This also includes all cases in which the experiment is ethically valued in a consequentialist sense, for instance if it helps avoid harm by providing useful insights or by replacing harmful procedures like animal experiments. If we thus exclude all epistemic, instrumental, and ethical values and focus on the experiment itself, any further evaluation is likely to be guided by aesthetic values. It might be recalled that in the experimental sciences like chemistry, an experiment is not just a test for hypotheses as in mathematical physics, but also an explorative approach under controlled conditions that might be related to improving theoretical knowledge but is more frequently aimed at discovering new effects or phenomena, including new substances as in synthetic chemistry. In a recent book, Philip Ball has scrutinized historical experiments in chemistry for their aesthetic appreciation by the chemical community [Ball, 2005] (see also [Schummer, 2006c]). He found ten aesthetic traits that apply both to particular experiments and to the particular attitude of the experimenters in performing these experiments. By analogy with virtue ethics, one can speak of experimental virtues that are valued for aesthetic rather than epistemological reasons. Ball’s ten virtues and the experimenters who exemplified them are: exact quantification (Johan Baptista van Helmont); attention to details (Henry Cavendish); patience in the conduct of the experiment (Marie Curie); elegance in the design of the experiment (Ernest Rutherford); miniaturization and acceleration of the experiment (various nuclear chemistry groups); conceptual simplicity (Louis Pasteur); imagination that transcends common views (Stanley Miller); simple-minded and straightforward reasoning (Neil Bartlett); economy and avoidance of deviations (Robert B. Woodward, see also [Woodward, 1989]); and conceptually straightforward design (Leo Paquette). One might object that these experimental virtues are also valued for epistemological and instrumental reasons because they would enable experimental success. However, even if they enabled experimental success in the particular historical cases, on which later chemists might place their hopes, these virtues do not guarantee success. There is no logical or proven statistical relation between the virtues and experimental success. Even worse, some virtues seem to contradict each other, for instance, imagination that transcends common views and simple-minded and

Aesthetic Values in Technology and Engineering Design

1051

straightforward reasoning. Ball’s analysis rather provides categories to describe different styles of experimentation that have been valued at different times by different communities or research groups. Such styles include, beyond the standard methodology of the discipline, particular ways to approach a problem, particular foci and care on certain aspects of experimentation, and particular ways of reasoning or designing. Beyond epistemic and instrumental values, experimental styles meet aesthetic preferences that might resonate with general aesthetic preferences of the corresponding socio-historical context. Aesthetic values thus perform an intermediate function in chemical experimentation. On the one hand, they are believed to enable experimental success, which qualifies them for provisional instrumental or epistemic rather than for aesthetic values proper. On the other hand, because these beliefs have no methodological basis but rather refer to general aesthetic preferences, they provide aesthetic guidance of research. If such guidance is successful in the long run, the aesthetic values can be incorporated into the standard methodology of the discipline and thus become epistemic or instrumental values.

3.5

Aesthetic values in mathematical modeling of chemical engineering and physical chemistry

There is a long Platonic tradition in mathematics that considers mathematical simplicity an aesthetic value in its own right. Based on the metaphysical belief that nature has a simple mathematical structure, mathematical physicists have tried to combine aesthetics with epistemology in order to derive mathematical simplicity as an epistemological criterion in science. For instance, the Cambridge professor of mathematics Paul Dirac [1963] famously claimed that for a physical theory the mathematical beauty of its equations, here its algebraic symmetry, is more important than its accordance with experiments. Dirac’s controversial claim reflects the particular epistemological tension between experimental and theoretical physics. His allusion to beauty helped him downplay the epistemological standards of the experimental sciences in favor of the epistemological standards of his own field. However, apart from such epistemological struggles, there is also an aesthetic appreciation of certain mathematical structures in fields that use mathematical models in a more instrumentalist way, particularly in chemical engineering and physical chemistry. A major issue in chemical engineering is to develop mathematical models of industrial processes where standard physical approaches of analysis do not work for complexity reasons, for instance the fluid flow or heat transfer through a complicated system that cannot fully be described in simple geometrical and physical terms or that require too many parameters with too many functional dependencies. A standard modeling approach for such systems is dimensional analysis. The art of dimensional analysis consists in combining all possible parameters into a few terms such that all units cancel. In addition, these terms, which are called dimensionless numbers, must have a physical meaning and be accessible by the measurement of

1052

Joachim Schummer, Bruce MacLennan, and Nigel Taylor

the system elements — for many standard engineering problems the data is even catalogued. If the analysis is successful, the modeling problem wondrously reduces from sheer overcomplexity to a simple equation with few retrievable parameters. This sudden mathematical simplicity frequently arises an aesthetic appreciation among engineers (see, for instance, Aris [1997]), which is above the suspicion of Platonist epistemology because the model must be feasible in industrial processes. However, as with all appreciations of mathematical simplicity, it would be wrong to say that the solution of the modeling problem is guided by aesthetic values, because reducing the mathematical complexity is actually the proper engineering goal. Instead, the aesthetic feeling arises only in addition to the satisfaction from solving the problem. Apart from simplicity, there are other mathematical features that are aesthetically valued by chemists. In particular, formal analogies are prominent candidates. If the mathematical structure of one equation is analogous to the mathematical structure of another, this suggests that the two systems described by these equations are somehow related to each other. For instance, studying the phenomenon of osmosis of liquid solutions Jacobus Henricus van ‘t Hoff (1852-1911) derived in 1887 an equation that was formally analogous to the ideal gas law and for which he eventually received the first Chemistry Nobel Prize in 1901. The formal analogy made a deep aesthetic impression on many chemists and does so still today (see, for instance, [Root-Bernstein, 2003, p. 36]), because it connected two formerly disparate fields. It suggested that solutions and gases behave in similar ways and thereby eventually opened up the entire field of thermodynamics of solutions. Besides being scientifically productive, such analogies seem to be aesthetically satisfying because they suggest an underlying holistic structure of nature in which, despite the analytical approach of science, everything is related to each other. One of the most impressive examples in this regard are the reciprocal relations by Lars Onsager (1903-1976), for which he received the Chemistry Nobel Prize in 1966. It was long known that a pressure difference causes matter flow, that a temperature difference causes heat flow, and so on for each pair of thermodynamic forces and flows. Yet, studying such forces and flows in more detail, Onsager found that a pressure difference can also cause heat flow and that a temperature difference can cause matter flow, and so on for each combination of thermodynamic forces and flows. Moreover, for each combination the flows are equal, which is mathematically expressed by the numerical equality of the reciprocal coefficients or by the symmetry of the coefficient matrix. Although Onsager’s relations meet the need for mathematical simplicity, they clearly oppose the idea that nature is simple, because any flow is now related to any force, albeit in a regular way. Thus, the aesthetic satisfaction rather arises from the fact that, contrary to previous analytical approaches, the reciprocal relations reveal a deeply holistic structure of nature. In general, there seem to be two different sources of aesthetic appreciation in mathematical modeling. One arises from unexpected or surprising mathematical simplicity, which equally applies to the modeling of natural and engineering

Aesthetic Values in Technology and Engineering Design

1053

systems. Other than an inclination to over-simplification, aesthetic values here cannot provide any extra-guidance of research, because what is aesthetically valued is at the same time the sought-after solution of the research problem. The other source of aesthetic appreciation seems to be rooted in metaphysical views of nature. Whether mathematical simplicity or the holistic constitution of nature, such metaphysical preconceptions are likely to have an impact on the personal choice of those research fields which promise aesthetic satisfaction to the individual researcher. In particular, the appreciation of analogous and holistic structures seems to be epistemologically productive because the exploration of analogies frequently opens up new insights and research directions.

3.6

Conclusion

In science as well as in everyday life, “beauty” is frequently used as a proxy for values that cannot be clearly defined. In this section I have tried to identify aesthetic appreciations of chemists by exclusion of appreciations that are based on epistemic, instrumental, or ethical values. Although the distinction is not always clear-cut, the results proof that there is ample space for aesthetic values in various areas of chemical research. Indeed aesthetic values have played important roles in selecting and designing synthetic targets, in designing and interpreting molecular representations, in designing and performing chemical experiments, and in developing mathematical models. The impact of aesthetic values has not always been productive with regard to epistemic and functional goals. Particularly the extreme fascination of chemists for symmetry and purity has led to a strong and persistent neglect of “dirty” and disordered materials, which the new discipline of materials science and engineering has systematically explored instead with many surprising results of economical importance. In other fields, however, particularly in supramolecular chemistry, the aesthetic fascination with molecules that “look” like ordinary objects has opened up an entire promising research field that is nowadays called nanotechnology. In chemical experimentation, where aesthetic values shape the particular styles of experimentation in the form of experimental virtues, aesthetics allows for an intermediate space for provisional and tentative methodological values. In all cases, whether productive or not, the aesthetic values of the individual researchers have been an important research motivation. Because of its focus on epistemology and the justification of physical theories, philosophy of science has long neglected aesthetic values in science, unless they are treatable as quasi-epistemological criteria of mathematical equations in the Platonic tradition. However, scientific research is about the production of new knowledge rather than about the justification of old knowledge, and much scientific and all engineering knowledge is ultimately aimed at developing useful products. This makes science an arena for a multitude of different values, including aesthetic, ethical, economic, and epistemological, which may harmonize or be in conflict with each other. Understanding the role of aesthetic values in scientific research is therefore essential to the philosophical understanding of science. And because of

1054

Joachim Schummer, Bruce MacLennan, and Nigel Taylor

its unclear position between science and technology in the received sense, chemistry is an excellent candidate to start with. 4 DESIGNING VIRTUAL OBJECTS: SOFTWARE ENGINEERING

4.1 Introduction Software engineering is a historically new activity, so it does not have a long aesthetic tradition as do the other arts and engineering disciplines. Therefore it is helpful to begin our aesthetic inquiry with analogies to longer-established disciplines, always keeping in mind the distinctive characteristics of software. In this article I will draw analogies principally from two sources. First, because software systems are large and complex, often constructed by teams, intended to serve a useful function, and capable of causing injury and economic loss if they fail, I will draw analogies from the structural engineering of towers and bridges (e.g., [Billington, 1983]), which shares these characteristics. This will lead to an exploration of the practical importance of elegance for both the designers and users of software systems. Second, because of the abstract and formal character of software I will draw analogies with aesthetics in the exact sciences, including mathematics (e.g., [Heisenberg, 1975]). Here we find that beauty depends on a harmonious interrelationship among the parts of an organic whole. Next I will discuss means for making abstract aesthetic qualities perceptible, including visual programming languages and models grounded in human embodiment. Finally, I will consider how we may advance the aesthetic dimension of software engineering.

4.2 Importance of aesthetics in software engineering 4.2.1

Designer’s perspective

Following Billington [1983] we may identify three dimensions along which designs may be evaluated: efficiency, economy, and elegance — “the Three E’s.” These correspond to three aspects of any artifact, the scientific, social, and symbolic (“the Three S’s”). Efficiency deals with the physical resources used by the system, which in the case of software artifacts is primarily computer time and memory. Typically there are tradeoffs involved, with efficiency weighed against factors such as functionality, reliability, and maintainability. These are scientific issues because they concern the physical resource utilization of the system’s design. Economy refers to all aspects of the cost of the system, including hardware and human costs, in all phases, including development, use, and maintenance. These are social issues because costs depend on market forces, social processes, governmental policies, etc. Due to the uncertainties in these factors, the economy of a design is more difficult to evaluate than its efficiency, and it is subject to change and local context. Furthermore, it cannot be assumed that all costs can be reduced to a common denominator, such as money, as is often the case with human suffering.

Aesthetic Values in Technology and Engineering Design

1055

This brings us to the explicitly aesthetic dimension of a design, its elegance, which depends on the aspects that Billington calls symbolic. Although we may take for granted that aesthetically appealing designs should be preferred, other things being equal, there are other compelling reasons for preferring elegant designs, but to understand them we need to review some of the characteristics of software systems. (See also [MacLennan, 1999, pp. 156–60].) Modern software systems can be enormously complex, often comprising millions of lines of instructions. Even a text editor, generally considered a basic software tool, can be hundreds of thousands of lines in length (e.g., the open source “vim 7.0” editor has approximately 300 000 lines of source code). The steady increase in software complexity has resulted from a number of factors (both scientific and social), including the increasing capacity and speed of computer systems, users’ demands for new features and richer interfaces, and competing systems with more features. Software systems with such large numbers of instructions are among the most complex systems ever constructed, and analytic tools for understanding them (such as program verifiers and test generators) are still quite limited. The complexity results in part from the fact that these millions of components interact with each other (and with other software and hardware systems) in real time, and that the number of interactions to be considered increases with at least the square of the number of components. Furthermore, the components (e.g., computer instructions) are far removed from physical objects and interactions for which we have an intuitive basis for understanding (e.g., the physical components and interactions of a mechanical system). Therefore, our intuition is set adrift, and our analytical tools do little to anchor it. Every analysis makes idealizing simplifications, and generally, the more complex the system, the greater will be the simplifications in its analysis. In the case of physical systems, for example, we may assume that the dynamics is linear, because that simplifies the mathematical analysis (or makes it feasible), even though we know that it is nonlinear. In the case of a software system, we may assume that any two numbers can be added and that the result will be correct, although we know that computer arithmetic is limited in range and precision. Similarly we may assume that input-output processes and other system services will operate correctly and be completed within real-time constraints. It is important to realize that simplification is an inherent limitation in the analysis of complex systems, since an analysis is supposed to separate out the relevant features of the system, so that we can understand them better (with our limited cognitive capacities), from the irrelevant features (which we intend to ignore). Therefore the validity and usefulness of an analysis depends on decisions (sometimes tacit) as to what the analysis should include or omit, which derive from assumptions (often unconscious) as to what is relevant or irrelevant. Furthermore, since human cognitive capacities are limited, the more complex a system is, the more must be omitted from its analysis so that the analysis itself will not exceed

1056

Joachim Schummer, Bruce MacLennan, and Nigel Taylor

our understanding. Thus there are inherent limitations to the analysis of very complex systems, such as modern software systems. Similar problems arise in structural engineering, and Billington observes that the best structural engineers are guided by aesthetics as well as by mathematical analysis. In elegant designs the dispositions of masses and forces are manifest in the design, and therefore the designs that look good (look balanced, secure, stable, etc.) correspond to the designs that are safe, efficient, and economical. For example, although extensive mathematical analysis was used in the design of the Tacoma Narrows Bridge, it collapsed four months after it was completed because aerodynamic stability had not been included in the analysis (it was not considered relevant). In contrast earlier bridges, designed without the benefit of complex mathematical models but in accord with aesthetic principles, were aerodynamically stable. How is it possible that good aesthetics can lead to good engineering? Billington observes that in structural engineering, designs are under-determined, that is, there are many designs that will solve a particular structural engineering problem, such as bridging a certain river (see also [Ferguson, 1992, p. 23]). Therefore, in contrast to Louis Sullivan’s architectural maxim, form follows function, which suggests that the design is strongly determined by its function, Billington argues that the more appropriate structural engineering maxim is function follows form, because there are many structures that will accomplish a particular function. The same arguments are even more applicable in software engineering, in which typically many different software designs will satisfy the system requirements. Therefore in software engineering we have a great deal of freedom in the choice of solutions to a software problem. In particular, software engineers (like structural engineers) can choose to work in a region of the design space in which experience has shown that designs that look good in fact are good (e.g., safe, efficient, and economical). In the case of towers and bridges, such designs make the interaction of forces manifest, so that designers (and, as we will see, users) can perceive them clearly. Since aesthetic judgment is a highly integrative cognitive process, combining perception of subtle relationships with conscious and unconscious intellectual and emotional interpretation, it can be used to guide the design process by forming an overall assessment of the myriad interactions in a complex software system. The discussion thus far has focused on the cognitive aspects of aesthetics, for an elegant software system is easier to understand and can be designed more reliably than an inelegant one. Thus there are practical, engineering reasons for striving for elegance. However, aesthetics also plays a less tangible role, which may be called ethical, for a design also symbolizes a set of moral values. Specifically, if a designer is seeking an elegant design, then they are being guided by a set of aesthetic values (which imply engineering values in the chosen subset of the design space). A design may be robust or delicate, spare or rich in features, straight forward or subtle, ad hoc or general, and so forth, and the values exemplified in the design will call forth extensions and modifications consistent with those

Aesthetic Values in Technology and Engineering Design

1057

values. By keeping certain values, embodied in the design aesthetic, before the designers’ eyes, these values will be kept in their attention and persist as conscious goals. Conversely, values incompatible with the aesthetic, or not exemplified by it, will tend to recede into the background of the designers’ minds, and will be underrepresented in the design. Thus a software system may embody a coherent ethical-aesthetic character, which is difficult to state in words but can guide the aesthetically sensitive engineer. Aesthetic appreciation can unite a software development organization through a common set of values embodied in a shared sense of elegance. We can see a similar role for aesthetics among mathematicians and theoretical scientists, who strive for proofs and theories that are elegant. For example, Heisenberg [1975, p. 176] says that science “also has an important social and ethical aspect; for many men can take an active part in it.” Scientists, he says, are like the master masons who constructed the medieval cathedrals, for “[t]hey were imbued with the idea of beauty posited by the original forms, and were compelled by their task to carry out exact and meticulous work in accordance with these forms” (ibid.). Like cathedrals and scientific theories, large software projects are the result of the efforts of many people, and aesthetic standards provide criteria by which individual contributions can be objectively evaluated (ibid.). Thus, in software engineering, as in mathematics and theoretical science, correctness is required, but among the correct solutions, the more elegant are preferred. (The education of mathematicians and theoretical scientists also provides models for how a shared sense of software elegance might be learned.) Therefore a shared aesthetic sense can unite a software engineering team in a common purpose. 4.2.2

User’s perspective

Hitherto I have stressed the importance of aesthetics for the designers of software artifacts, but it is also important for the users. In the modern information economy many people spend much of their working lives interacting with one or a few software systems (e.g., a word processor, database system, or reservation system); further, in their recreational time, people may be engaged with the same or other software artifacts (e.g., a web browser or computer game). Therefore the external aesthetics of software systems can have a significant effect on the quality of many people’s lives. Other things (such as functionality) being equal, most people would prefer to work with a beautiful tool than with an ugly one. Furthermore, for many people the computer is not simply one tool in an otherwise uncomputerized occupation; rather, the computer and its software constitute, to a large degree, the entire occupation. In these cases the software system defines the work environment as fundamentally as the physical workspace does. Therefore, the aesthetics of the software systems deserves at least as much attention as that due the architecture, decor, etc. (From this perspective, many contemporary programs are the software equivalent of sweatshops: cluttered, dangerous, ugly, alienating, and dehumanizing.) As architecture deals with the functionality and

1058

Joachim Schummer, Bruce MacLennan, and Nigel Taylor

aesthetics of physical space, organizing it for practicality and beauty, so software engineers organize cognitive (or virtual) space toward the same ends. Thus software aesthetics can have a major effect on quality of work and quality of life. An elegant software design can also promote confident use of the system, for eventually users will acquire an aesthetic sense of the design space and will come to recognize that the designs that look good also are functionally good. As is well known, many people approach software fearfully, and part of this fear arises from the fact that software is unpredictable (for them, but often also for the designers!). In elegantly designed software, however, the dynamical interaction of the parts is manifest in the external form, and so aesthetic comprehension of the form can guide the user’s understanding of the system’s operation. Therefore, as in mathematics and theoretical science, the goal is that beauty coincide with intelligibility, for then users (as well as programmers) will experience pleasure through understanding. This is possible for both beauty and intelligibility are grounded in the interrelation of the parts (Section 4.3; cf. [Heisenberg, 1975, pp. 169–70]). It is well-known that people’s ability to use technological devices with pleasure, confidence, and fluency depends on their ability to build a cognitive or conceptual model of the device’s behavior [Norman, 1988, Ch. 7; 1998, Ch. 8; 2005, Ch. 3]. An effective cognitive model of a system is not required to reflect its actual internal structure or operation, but it must be accurate enough not to mislead the user (thus resulting in a loss of confidence and in frustration). By implying an intelligible dynamical structure, an elegant design can help the user to form an effective cognitive model. Therefore an elegant design aids users’ understanding of a system in much the same way it aids that of the system’s designers. Similarly, just as for the designers the aesthetics of a design has an ethical dimension and exemplifies certain values to the exclusion of others, so also the design aesthetics has ethical implications for users. At very least, by making some practices easy and others awkward, and by bringing some concerns into the foreground while leaving others in the background, the external aspect of the system will influence users in its use. Indeed, such non-neutrality is an unavoidable characteristic of the phenomenology of all tools [Ihde, 1986, Chs. 5, 6; 1993; MacLennan, 1999, pp. 33–35]. In addition to this, however, is the symbolic dimension, for by exemplifying particular aesthetic norms, the system keeps these before the eyes of the users, and increases the likelihood that they will be guided by these norms in their own work. Finally, there is a social aspect for the users of an elegant design just as there is for the designers. As users come to appreciate the beauty of an elegant design, they will develop an appreciation for its aesthetic principles and come to expect similar elegance in other software systems. Thus the users (and consumers) of software systems are included in a feedback loop that encourages the development of elegant software and discourages the inelegant. This will accelerate the development of software that is efficient, economical, reliable, and a pleasure to use. (Billington notes the role of an aesthetically educated public in improving bridge design in Europe.)

Aesthetic Values in Technology and Engineering Design

4.3

1059

Intelligibility

All arts have their formal and material characteristics, but software engineering is exceptional in the degree to which formal considerations dominate material ones. All the issues that are most fundamental in software engineering (e.g., correctness, efficiency, understandability, maintainability) depend primarily on the formal characteristics of the program and only secondarily on its material embodiment (i.e., the effect of the hardware on the software). Clearly, the hardware cannot be ignored (especially in cases in which the engineering is pushed to its limits), but in general hardware considerations are secondary and often an afterthought. Software engineering is a new discipline and so it does not have a well-established aesthetic tradition. We may look to other arts for suggestions and analogies, but software’s lack of essential material embodiment implies that perceptual qualities will not have so great a role as they do in the other arts. Rather, aesthetic considerations in software engineering will be comparable to those in mathematics and theoretical science. Indeed, discussions of the aesthetics of mathematics and theoretical science often focus on such qualities as correctness (either consistency or empirical adequacy), generality, simplicity, and (abstract) beauty, and the same qualities are central to the aesthetic evaluation of software. (See, for example, [Curtin, 1982; Farmelo, 2002, Pref.; King, 2006; Wechsler, 1988].) Science attempts to comprehend a multiplicity of phenomena under a single principle, expressed as a simple, elegant mathematical relationship among abstract ideas. Most commonly the phenomena are dynamical relationships and processes evolving in time, and so, as Heisenberg explains (in the case of Newtonian mechanics), “The parts are individual mechanical processes . . . And the whole is the unitary principle of form which all these processes comply with [and which is expressed] in a simple system of axioms” [Heisenberg, 1975, p. 174]. In science, then, as in art, “Beauty is the proper conformity of the parts to one another and to the whole” (ibid.). The goals of the software engineer are similar to those of the scientist in that both are attempting to give a static abstract description of material processes and interactions taking place in time. One difference, of course, is that the scientist is trying to describe naturally occurring phenomena, whereas the engineer is attempting to design a static structure (program) that will generate the desired temporal interactions. As mechanical processes are described by the axioms of Newtonian mechanics, so a program, contingent on external events, describes a set of possible execution sequences. Individual execution sequences are the parts with respect to the infinite set of all sequences, for which the program provides an intensive (finite) definition. Beauty, then, resides in the conformity of the execution sequences to each other and to the program. They should form a harmonious ensemble (extension) and have a simple relation to the program (intension). For elegant programs the dynamic possibilities (extension) will be easy to visualize from the generative

1060

Joachim Schummer, Bruce MacLennan, and Nigel Taylor

form (intension). The engineers will have a reliable intuitive understanding of the consequences of their design. Conversely, in designing a program, software engineers have certain desired execution sequences in mind, and they have to expand these in their minds into a coherent infinite set of possible sequences (conformity of the parts to one another). From this multiplicity of possible dynamics they need to derive a finite and unified static generative form (conformity of parts to the whole). Beauty resides in the simplicity, harmoniousness, orderliness, and symmetry of these relations, which elicit simultaneous intellectual and aesthetic appreciation.

4.4

Visual beauty

Sensuously perceivable beauty is a means toward apprehension of the intellectual beauty of abstract forms. Therefore I will consider briefly the role of visual beauty in software design. Vision is our richest sensory modality, and thus it is not surprising that visual representations have played an essential role in the development of engineering, in engineering design, and in engineering education [Ferguson, 1992]. In particular, aesthetic evaluation of designs is aided by such “tools of visual analysis” as characteristic curves, which provide memorable and comprehensible visual representations of the relations of relevant variables, and graphic statics, which afford intuitive assessment of the relative forces in a structure [Ferguson, 1992, ch. 5]. Are there comparable means for visualizing the relevant abstract structures in software? Visual programming languages (VPLs), in which programs are represented as two-dimensional figures rather than as text, have been investigated since the earliest days of electronic computing (e.g., AMBIT/G, SKETCHPAD), and VPLs continue to be developed (e.g., Alice, StarLogo TNG), especially for introductory programming instruction (e.g., [Eades and Zhang, 1996; Stasko, et al., 1998]). In these languages formal relations between program parts are represented as spatial relations between visual forms. Early VPLs represented programs as flowcharts, in which connecting edges represented possible paths of control flow, but after the introduction of structured programming around 1970 it became more popular to represent visually the hierarchical structure of the program, which reflects both the logical and dynamical organization of a structured program. Often visual representations of hierarchical program structure take the form of some kind of tree diagram. Sometimes these are graphs, in which leaves represent atomic program components (individual programming language statements), interior nodes represent composite program components, and edges connect composite components to their immediate constituents. A more recent style, facilitated by improved computer graphics capabilities, represents program components by twodimensional shapes reminiscent of jigsaw puzzle pieces, which can be interlocked only in conformity with the programming language’s syntax (e.g., Alice, StarLogo TNG).

Aesthetic Values in Technology and Engineering Design

1061

Visual representations of hierarchical program structure would seem to be ideal as a medium for elegant program design, and they are certainly superior in this regard to flowcharts. By representing abstract relations spatially, they create a correspondence between the domains of abstract forms and of spatial forms, and facilitate the visual perception (and aesthetic appreciation) of well-organized, symmetric, and balanced structures; that is, beauty coincides with intelligibility. Unfortunately, in practice these visual representations have limitations, for even small program modules can be quite deeply nested (and it can be argued that for very small modules visual representation is not important). As a consequence, the visual representations can be quite large in whatever dimension represents nesting depth. Due to the limitations of human visual perception and practical computer screen size, we are faced with the undesirable alternatives of displaying the entire structure, but with many tiny components, which are difficult to discern, or of displaying only a portion of the structure at one time and having to use devices such as panning and zooming to explore the structure sequentially. Neither is conducive to Gestalt recognition of the program’s structure, or to an intuitive intellectual comprehension and aesthetic appreciation of it. Perhaps the problem is that VPLs result in a too literal representation of program structure in perceptible form, and that an aesthetically satisfying expression of the design will require a less literal representation.

4.5

Embodiment

Fishwick and his colleagues have explored a more metaphorical approach to programming aesthetics (e.g., [Fishwick, 2002]). Noting that graphs are “largely devoid of texture, sound, and aesthetic content,” he seeks to make software “more useful, interesting, and comprehensive” by an approach that begins with a model; this is the “craft-worthy, artistic step.” The model is intended to be the usual representation of the software design, the textual program being relegated to a secondary, marginalized status comparable to assembly language. However, since most software concepts are abstract and do not have real-world correspondents, they are represented metaphorically. Therefore, once the model is determined, an aesthetic must be chosen as a foundation for the metaphors. For example, if architecture were chosen, then abstract control-flow relations in the program could be represented by corridors in a building through which avatars move. Notice that such a metaphorical representation recruits our embodied understanding of physical space and motion to improve our understanding of the program (see below). Similarly, our aesthetic understanding of architectural space guides the design of the program and our aesthetic and intellectual appreciation of it. The metaphorical model is the principal representation of the software, which becomes an object of aesthetic expression and appreciation, thereby enriching the experience of software. Fishwick notes that even three-dimensional visual programming languages tend to use simple iconography rather than sensuously rich objects: “One is aesthetically-challenged and Platonic whereas the other promotes famil-

1062

Joachim Schummer, Bruce MacLennan, and Nigel Taylor

iar sensory appeal”. [Fishwick, 2006] contains recent contributions to aesthetic computing (“the impact and effects of aesthetics on the field of computing,” p. 3). Recent developments in psychology have illuminated the essential role played in cognition by embodiment, thus confirming insights from phenomenological philosophy and psychology (e.g., [Gibbs, 2006]). Much of our understanding of the world is rooted in our sensorimotor capacities, both those that are part of our genetic inheritance, and those that are acquired, especially in early childhood. Indeed, Lakoff and N´ un ˜ez [2000] have argued that our understanding even in such abstract domains as mathematics is built on a network of interrelated metaphors grounded in sensorimotor skills. For example, at an intuitive level abstract sets are understood as physical containers, abstract trajectories as paths through physical space, and so forth. All human beings have an enormous repertoire of sensorimotor skills, and it is normal to feel pleasure when acting skillfully, competently, and fluently, and to be dissatisfied otherwise; this is part of the feedback mechanism that increases the range and depth of our skills. Therefore to the extent that users’ interactions with a system, such as a program, are accomplished through an existing repertoire of sensorimotor skills, they will feel competent and satisfied when they use it. In this way, aesthetic appreciation arises from the correspondence between people’s embodied skills and the sensorimotor interface and abstract structure of the system, which is a different sort of resonance or congruence between the system and human cognitive structures. Therefore aesthetic appreciation and satisfaction will be improved if a system and its parts, including the interface, behave similarly to the physical world, including the objects and processes that are familiar to most people. For example, if when we pull on or drag an object on a computer screen it behaves similarly to a physical object (e.g., in terms of stretching or inertia), then our sensorimotor skills will be engaged, and our skillful manipulation will be pleasurable [Karlsson and Djabri, 2001].

4.6 Applying aesthetic principles in software engineering The foregoing remarks have merely sketched an approach to an aesthetic theory appropriate to software engineering, and so it will be worthwhile to say a little about how such a theory might be further developed. We can progress by four simultaneous activities, which we may call experiment, criticism, theory, and practice. Experiment refers to learning by means of the self-conscious practice of the art of program design and the empirical evaluation of the results. For this to be effective, software engineers must be aware of aesthetic issues during the design, and they must evaluate the aesthetics of the resulting designs as experienced by themselves and others (evaluated phenomenologically and statistically). This entire activity presupposes greater aesthetic awareness in programmers.

Aesthetic Values in Technology and Engineering Design

1063

Criticism plays an important role in all of the arts, most obviously to provide the general public with aesthetic evaluations, but more importantly to make various aesthetic issues salient, which influences the aesthetic sensibilities of both the producers and consumers of art. Even when artists disagree with criticism, they are encouraged to defend their aesthetic choices in word or deed. Thus criticism provides an important feedback loop that can improve artistic quality. To accomplish this we need more published aesthetic criticism of software, both of its external appearance and behavior, and of its internal structure and design, focusing on aesthetics in both cases. Theory refers to the use of research results from cognitive neuropsychology and allied fields, which will continue to provide insights into the qualities that make something simultaneously intellectually comprehensible and aesthetically pleasing (that is to say, elegant). Theoretical understanding contributes by explaining the results of previous aesthetic experiments, and by suggesting new ones. In spite of all the foregoing, the art of program design is neither a body of theory nor a set of design rules; rather, it is a practice. Both the long history of aesthetic debate and the analogy of aesthetic considerations in mathematics and the exact sciences suggest that beauty is an illusive concept. Therefore, in programming as in the other arts, while many aesthetic principles can be stated explicitly, others must remain implicit and essentially embodied in the practices of skilled artisans. 5 CONCLUSION As was mentioned in the Introduction, the three engineering fields have been selected to represent different modes regarding the size and visibility of their respective engineering products, from large-scale to small-scale to virtual objects. In conclusion we may asked how that affects the role of aesthetics in the design process, if aesthetic differences correspond to this order, and if there are common aesthetic features in all fields. Because aesthetics of engineering is still in its infancy and far from being a canonized field, the authors of the previous sections have each discussed their engineering domain from a personal angle, such that one should be careful with premature generalizations. However, if we move from the engineering of large-scale to small-scale to virtual objects, there are four trends in the aesthetic emphasis, some of which are obvious and less surprising, with important exceptions though. The first trend is the decreasing importance that the anticipated aesthetic experience by consumers plays in the design process. Of course that is a trivial observation, because the less visible and comprehensible the product structure by consumers is, the less need engineers in their design process consider the aesthetic experience by consumers. As a consequence, aesthetic considerations are less connected to general aesthetic discourses, which allows engineers to develop their specific aesthetic preferences in either a reflected or unconscious manner. In chemistry that has occasionally led to unlucky popularization efforts in which chemists publicly praised the alleged beauty of their molecules, which, however, nobody else was

1064

Joachim Schummer, Bruce MacLennan, and Nigel Taylor

able to comprehend. Instead, for much of the general public, chemical products, like plastics, have become a symbol of the synthetic, artificial, and anti-natural, if not of excessive Modernism, which for aesthetic reasons alone have been rejected, regardless of their molecular structure. Software engineering seems to be the exception to the rule, since, as Bruce MacLennan argues in Section 4.2.2, the software structure that is aesthetically preferred by engineers is at the same time the software that users aesthetically enjoy most because of their transparency and intelligibility. The second, equally less surprising, trend is that, if one moves from the engineering of large visible objects to that of virtual objects, the aesthetic role of primary sensual experience decreases. The second trend is compensated for, however, by the third trend of the increasing importance and increasingly deliberate use of representational tools and media, which become the primary objects of sensual experience for engineers in the design process and at the same time move the impact of aesthetic values to the early research state. The creation and selection of representational tools and media imply aesthetic choices and preferences that guide and shape the research and design process and its final products. In architecture the effect might be observable in slight design changes because of the recent shift from drawing boards to computer-aided design programs. In chemistry the creation and use of molecular models is so influential that it can inspire entirely new research fields (Section 3.3). Software engineers have even moved one step further by deliberately employing the latest psychological understanding of our sensorimotor capacities to build various metaphorical models, like physical spaces, for the representation of software in the design process (Section 4.6). What appears aesthetically preferable in the metaphorical model is thus translated into decisions about the preferred abstract structure of the software. The forth trend concerns the relationship between aesthetic and epistemological values. In architecture and urban planning, the connection is less present, partly because architecture has been more removed from the epistemological mainstream discourse. However, from ancient architecture, as exemplified by Vitruvius, to Renaissance and Modernist architecture, particularly in the works of Le Corbusier, there were much stronger ties, since mathematical proportions figured prominently not only as aesthetic ideals but also as epistemic guidelines for adjusting constructions to human nature. In chemistry aesthetic values frequently assume the role of proto-epistemological criteria, i.e. they guide epistemological decisions in case of epistemological indetermination and, if they turn out to be successful in the long run, might be incorporated in the methodological standard canon (Section 3.4). Finally, in software engineering, at least in the aesthetics suggested by Bruce MacLennan (Section 4), aesthetic and epistemological values merge to form a common basis for assessing the quality of software. A common feature in all three areas of engineering discussed in this article is the prominent, albeit slightly different, role that classical aesthetics, with its emphasis on mathematical purity and conceptual clarity, still plays today. This is perhaps less obvious in architecture and urban landscape planning; but, as Nigel

Aesthetic Values in Technology and Engineering Design

1065

Taylor argues in Section 2, contemporary aesthetic debates in architecture are still deeply influenced by early 20th century Modernism and its aesthetic preference of pure geometrical form and the clear expression of function, which post-modernist approaches have tried to overcome. As guidelines for identifying and designing the “ideal” human environment, these classical aesthetic values have certainly failed in the excessive Modernists projects of post-WWII urban landscape planning. While classical aesthetics has been debated, and periodically embraced and rejected, in architecture for more than two thousand years, chemists discovered these aesthetic values only recently. As with the excesses of architectural Modernism, the chemists’ obsession with geometrical symmetry and purity has led to many misconceptions and the almost complete neglect of “impure” materials, which others have very successfully harvested instead (Section 3.2). In the design of virtual objects, software engineering has inherited much from mathematics to the extent that the classical ideal of “beauty coinciding with intelligibility” becomes meaningful in as much as the criteria for beauty are related to mathematical features of abstract structures. Another common feature in all three engineering field is the neglect of explicit treatments and serious investigations of aesthetics, although for different reasons. In architecture and urban landscape planning, which one would expect to make use of their long aesthetic tradition, the neglect is largely a heritage of the “antiaesthetic” attitude of early 20th century functionalism. In addition, as Nigel Taylor points out (Section 2.4), the more recent move of urban planning into the political sphere has led to the paradox that aesthetic aspects, although highly valued by citizens, are difficult to articulate in the political decision process. In chemistry the lack of serious aesthetic investigations is in accordance with a general neglect of chemistry, if not chemophobia, by most humanists, which chemists, on the other hand, are likely to increase rather than to overcome by popularization efforts that refer to beauty. In software engineering the neglect seems to be largely because of the youth of the discipline, because, as Bruce MacLennan emphasizes (Section 4.2.1), the use of aesthetic criteria is increasingly required because of the increasing complexity and functional underdetermination of software products. The neglect of explicit considerations of aesthetics in engineering thus coincides with the richness of aesthetic values and their strong impact on the engineering design process at various stages, whether consciously or not. Since the aesthetic impact can be both productive and counter-productive with regard to purely functional values, as many examples in this chapter have illustrated, even the most functionalist-minded engineer or philosopher might become easily convinced of the need of further serious investigations of aesthetics in engineering.

1066

Joachim Schummer, Bruce MacLennan, and Nigel Taylor

BIBLIOGRAPHY References for Section 1. [Benjamin, 1936] W. Benjamin. Das Kunstwerk im Zeitalter seiner technischen Reproduzierbarkeit (The Work of Art in the Age of Mechanical Reproduction). Zeitschrift f¨ ur Sozialforschung, 1936. [Brummett, 1999] B. Brummett. Rhetoric of Machine Aesthetics. Praeger, 1999. [Cooper, 1992] D. Cooper, ed. A Companion to Aesthetics. Blackwell, 1992. [Gartman, 1994] D. Gartman. Auto-Opium: A Social History of American Automobile Design. Routledge, 1994. [Haug, 1986] W. F. Haug. Kritik der Waren¨ asthetik. Suhrkamp, 1971 (Engl. trans. as Critique of commodity aesthetics. Polity Press, 1986). [Mumford, 1934] L. Mumford. Technics and Civilization. Harcourt Brace, 1934. References for Section 2. [Alexander, 1996] C. Alexander. A City is not a Tree. Architectural Forum, 122 (1), 1965 (reprinted in The City Reader, R.T. LeGates and F. Stout eds., pt. 2, pp. 118–131. Routledge, 1996. [Banham, 1960] R. Banham. Theory and Design in the First Machine Age. The Architectural Press, 1960. [Billington, 1979] D. P. Billington. Robert Maillart’s Bridges: The Art of Engineering. Princeton University Press, 1979. [Billington, 1983] D. P. Billington. The Tower and the Bridge: The New Art of Structural Engineering. Basic Books, 1983. [Bishop and Phillips, 2004] K. Bishop and A. Phillips, eds. Countryside Planning: New Approaches to Management and Conservation. Earthscan, 2004. [Burgess et al., 1988] J. Burgess, C.M. Harrison, and M. Limb. People, Parks and the Urban Green: A Study of Popular Meanings and Values for Open Spaces in the City. Urban Studies, 25, 455–473, 1988. [Frampton, 1985] K. Frampton. Modern Architecture: A Critical History, 2nd edition. Thames and Hudson, 1985. [Gelernter, 1995] M. Gelernter. Sources of Architectural Form: A Critical History of Western Design Theory. Manchester University Press, 1995. [Giedion, 1941] S. Giedion. Space, Time and Architecture: The Growth of a New Tradition. Harvard University Press, 1941. [Goodman, 1985] N. Goodman. How Buildings Mean. Critical Inquiry, 11, 642–653, June 1985. [Hearn, 2003] F. Hearn. Ideas That Shaped Buildings. MIT Press, 2003. [Jacobs, 1964] J. Jacobs. The Death and Life of Great American Cities. Penguin, 1964. [Jencks, 1977] C. Jencks. The Language of Post-Modern Architecture. Academy Editions, 1977. [Jencks, 1980] C. Jencks. Late-Modern Architecture and Other Essays. Academy Editions, 1980. [Jencks, 2002] C. Jencks. The New Paradigm in Architecture: The Language of PostModernism. Yale University Press, 2002. [Keeble, 1952] L. Keeble. Principles and Practice of Town and Country Planning, The Estates Gazette, 1952 (4th edition published 1969). [Kundera, 1984] M. Kundera. The Unbearable Lightness of Being. Faber, 1984 (trans. M. Henry). [Larkham, 1996] P. J. Larkham. Conservation and the City. Routledge, 1996. [Le Corbusier, 1924] Le Corbusier. Urbanisme. Editions Cr` es, 1924 (English trans. by F. Etchells published as The City of Tomorrow. The Architectural Press, 1971). [Le Corbusier, 1927] Le Corbusier. Towards a New Architecture, trans. by F. Etchells. The Architectural Press, 1927 (originally published as Vers Une Architecture. Editions Cr` es, 1923). [Le Corbusier, 1933] Le Corbusier. La Ville Radieuse. Vincent, Freal et Cie, 1933. (English trans. by P. Knight, E. Levieux and D. Coltman published as The Radiant City. Faber and Faber, 1967). [Ruskin, 1849] J. Ruskin. The Seven Lamps of Architecture. George Allen, 1849 (1904 edition).

Aesthetic Values in Technology and Engineering Design

1067

[St John Wilson, 1995] C. St John Wilson. The Other Tradition of Modern Architecture: The Uncompleted Project. Academy editions, 1995. [Sullivan, 1985] L. Sullivan. Ornament in Architecture. Extract quoted in Frampton [1985, pt. 2, ch. 2, p. 51]. [Taylor, 1998] N. Taylor. Urban Planning Theory since 1945. Sage, 1998. [Taylor, 1999] N. Taylor. Anglo-American town planning theory since 1945: three significant developments but no paradigm shifts. Planning Perspectives, 14 (4), 327–345, October 1999. [Venturi, 1966] R. Venturi. Complexity and Contradiction in Architecture. Museum of Modern Art, 1966, (2nd edition, 1977). References for Section 3. [Aris, 1999] R. Aris. Reflections on Keats’ equation. Chemical Engineering Science, 52 (15), 2447–2455, 1997 (reprinted in Aris, R. Mathematical Modeling: A Chemical Engineer’s Perspective, pp. 400–414. Academic Press, 1999). [Ball, 2005] Ph. Ball. Elegant Solutions: Ten Beautiful Experiments in Chemistry. Royal Society of Chemistry, 2005. [Chemical Abstract Service, 2007] Chemical Abstract Service (CAS). Statistical Summary, 1907-2006. Chemical Abstract Service, 2007. [de Meijere, 1982] A. de Meijere. Sport, Spiel, Spannung — die Chemie kleiner Ringe. Chemie in unserer Zeit, 16, 13–22, 1982. [Dirac, 1963] P. Dirac. The evolution of the Physicist’s Picture of Nature. Scientific American, 208 (May), 45–53, 1963. [Eco, 1962] U. Eco. Opera aperta, Bompiani 1962 (English trans. as The Open Work. Hutchinson, 1989). [Freemantle, 2003] M. Freemantle. Chemistry at its most beautiful. Chemical and Engineering News, 81(34), 27–30, 2003. [Grahn, 1981] W. Grahn. Platonische Kohlenwasserstoffe. Chemie in unserer Zeit, 15, 52–61, 1981. [Hargittai, 2000] I. Hargittai. Candid Science: Conversations with Famous Chemists. Imperial College Press, 2000. [Hoffmann, 1990] R. Hoffmann. Molecular Beauty. The Journal of Aesthetics an Art Criticism, 48, 191–204, 1990. [Jakobs, 2006] S. Jakobs. “Selbst wenn ich Schiller sein k¨ onnte, w¨ are ich lieber Einstein”: Naturwissenschaftler und ihre Wahrnehmung der “zwei Kulturen”. Campus, 2006. [Laszlo, 2003] P. Laszlo. Foundations of Chemical Aesthetics. Hyle: International Journal for Philosophy of Chemistry, 9, 11–32, 2003. [Root-Bernstein, 2003] R. Root-Bernstein. Sensual Chemistry: Aesthetics as a Motivation for Research. Hyle: International Journal for Philosophy of Chemistry, 9, 33–50, 2003. [Schummer, 1995] J. Schummer. Ist die Chemie eine sch¨ one Kunst? Ein Beitrag zum Verh¨ altnis ¨ von Kunst und Wissenschaft. Zeitschrift f¨ ur Asthetik und Allgemeine Kunstwissenschaft, 40, 145–178, 1995. [Schummer, 1997a] J. Schummer. Challenging Standard Distinctions between Science and Technology: The Case of Preparative Chemistry. Hyle: International Journal for Philosophy of Chemistry, 3, 81–94, 1997a. [Schummer, 1997b] J. Schummer. Scientometric Studies on Chemistry I: The Exponential Growth of Chemical Substances, 1800-1995. Scientometrics, 39, 107–123, 1997b. [Schummer, 1997c] J. Schummer. Scientometric Studies on Chemistry II: Aims and Methods of Producing New Chemical Substances. Scientometrics, 39, 125–140, 1997c. [Schummer, 2003] J. Schummer. Aesthetics of Chemical Products: Materials, Molecules, and Molecular Models. Hyle: International Journal for Philosophy of Chemistry, 9, 73–104, 2003. [Schummer, 2006a] J. Schummer. Symmetrie und Sch¨ onheit in Kunst und Wissenschaft. In ¨ Asthetik in der Wissenschaft, W. Krohn ed., pp. 59–78. Meiner, 2006a. [Schummer, 2006b] J. Schummer. Gestalt Switch in Molecular Image Perception: The Aesthetic Origin of Molecular Nanotechnology in Supramolecular Chemistry. Foundations of Chemistry, 8, 53–72, 2006b.

1068

Joachim Schummer, Bruce MacLennan, and Nigel Taylor

[Schummer, 2006c] J. Schummer. The End of Silent Rites. Book Review of Philip Ball: Elegant Solutions: Ten Beautiful Experiments in Chemistry. Hyle: International Journal for Philosophy of Chemistry, 12, 157–159, 2006c. [Spector and Schummer, ] T. I. Spector and J. Schummer, eds. Aesthetics and Visualization in Chemistry, special issue of Hyle: International Journal for Philosophy of Chemistry, 9 (1 & 2), 2003. [Stevens, 2003] H. Stevens. Fundamental physics and its justifications, 1945-1993. Historical Studies in the Physical and Biological Sciences, 34, 151–197, 2003. [Woodward, 1989] C. E. Woodward. Art and elegance in the synthesis of organic compounds: Robert Burns Woodward. In Creative people at work: twelve cognitive case studies, D.B. Wallace and H.E. Gruber, eds., pp. 227–253. Oxford University Press, 1989. References for Section 4. [Billington, 1983] D. P. Billington. The Tower and the Bridge. Princeton University Press, 1983. [Curtin, 1982] D. Curtin, ed. The Aesthetic Dimension of Science. Philosophical Library, 1982. [Eades and Zhang, 1996] P. Eades and K. Zhang, eds. Software Visualization. World Scientific, 1996. [Farmelo, 2002] G. Farmelo, ed. It Must Be Beautiful. Granta Books, 2002. [Ferguson, 1992] E. S. Ferguson. Engineering and the Mind’s Eye. MIT Press, 1992. [Fishwick, 2002] P. Fishwick. Aesthetic programming: Crafting personalized software. Leonardo, 35 (4), 383–390, 2002. [Fishwick, 2003] P. Fishwick. Aesthetic computing manifesto. Leonardo, 36 (4), 255–256, 2003. [Fishwick, 2006] P. Fishwick, ed. Aesthetic Computing. MIT Press, 2006. [Gelernter, 1998a] D. Gelernter. Machine Beauty: Elegance and the Heart of Technology. Basic Books, 1998. [Gelernter, 1998b] D. Gelernter. Aesthetics of Computing. Phoenix House, 1998. [Gibbs, 2006] R. W. Gibbs Jr. Embodiment and Cognitive Science. Cambridge University Press, 2006. [Heisenberg, 1975] W. Heisenberg. The Meaning of Beauty in the Exact Sciences. In Across the Frontiers, W. Heisenberg, trans. P. Heath, pp. 166–83. Harper & Row, 1975. [Ihde, 1986] D. Ihde. Consequences of Phenomenology. State University of New York Press, 1986. [Ihde, 1993] D. Ihde. The Philosophy of Technology: An Introduction. Paragon House, 1993. [Karlsson and Djabri, 2001] P. Karlsson and F. Djabri. Analogue styled user interfaces: An exemplified set of principles intended to improve aesthetic qualities in use. Proceedings of Mobile HCI 2001: Third International Workshop on Human-Computer Interaction with Mobile Devices, 2001. [King, 2006] J. King. The Art of Mathematics. Dover Publications, 2006. [Lakoff and N´ un ˜ez, 2000] G. Lakoff and R. E. N´ un ˜ez. Where Mathematics Comes From: How the Embodied Mind Brings Mathematics Into Being. Basic Books, 2000. [MacLennan, 1999] B. J. MacLennan. Principles of Programming Languages: Design, Evaluation, and Implementation, 3rd ed. Oxford University Press, 1999. [Norman, 1988] D. Norman. The Psychology of Everyday Things. Basic Books, 1988. [Norman, 1998] D. Norman. The Invisible Computer. MIT Press, 1998. [Norman, 2005] D. Norman. Emotional Design. Basic Books, 2005. [Stasko et al., 1998] J. Stasko, J. Dominue, M. H. Brown, and B. A. Price, eds. Software Visualization. MIT Press, 1998. [Wechsler, 1998] J. Wechsler. On Aesthetics in Science. Birkh¨ auser Verlag, 1988.

RISK AND SAFETY IN TECHNOLOGY Sven Ove Hansson

1

INTRODUCTION

Risk is ubiquitous in technology, and safety has been a central concern of engineering as long as there have been engineers. In codes of engineering ethics, the engineer’s responsibility for the safety of workers and the public is strongly emphasized. This chapter begins with a section on the definition of key terms in technological risk and safety. After that follow two sections that describe the two major, complementary approaches to technological risk and safety: safety engineering (Section 3) and risk analysis (Section 4). The final Section 5 is devoted to ethical analysis of risk. 2

DEFINING THE KEY TERMS

Technological risk and safety is an area in which the terminology is far from wellestablished. The definition of key terms often differs between different branches and traditions of engineering. These differences depend largely on lack of communication between different expert communities, but there is also a normative or ideological element in the terminological confusion. Different uses of “risk” and “safety” can express different priorities concerning what hazards should be subject to preventive or mitigating measures. In this section, six key terms will be discussed: risk, uncertainty, hazard, safety, security, and the precautionary principle.

2.1

Risk

The word “risk” has several clearly distinguishable meanings, both in technology and other social areas. This can be illustrated with statements about the risks associated with nuclear energy. First: “A reactor melt-down is the most serious risk that affects nuclear energy.” Here, we use “risk” in the following sense: 1. risk = an unwanted event that may or may not occur. Next, consider the following statement: “Hidden cracks in the tubing is one of the major risks in a nuclear power station.” Here we use “risk” in another sense: Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

1070

Sven Ove Hansson

2. risk = the cause of an unwanted event that may or may not occur. We can quantify the risk of a major nuclear accident for instance in the following way: “The risk of a melt-down during this reactor’s life-time is less than one in 10,000.” This is yet another concept of risk: 3. risk = the probability of an unwanted event that may or may not occur.1 In risk analysis, nuclear energy is often compared to other energy sources in terms of the statistically expected number of victims. We may say, for instance: “The total risk from nuclear energy is smaller than that from coal energy.” Then the word is used in the following sense: 4. risk = the statistical expectation value of unwanted events that may or may not occur. Expectation value means probability-weighted value. Hence, if 200 deep-sea divers perform an operation in which the individual risk of death is 0.1 % for each individual, then the expected number of fatalities from this operation is 200 × 0.1 % = 0.2. The risk of fatalities in this operation can then be said to be 0.2. Expectation values have the important property of being additive. Suppose that a certain operation is associated with a 1 % probability of an accident that will kill five persons, and also with a 2 % probability of another type of accident that will kill one person. Then the total expectation value is 0.01 × 5 + 0.02 × 1 = 0.07 deaths. In similar fashion, the expected number of deaths from a nuclear power plant is equal to the sum of the expectation values for each of the various types of accidents that can occur in the plant. In decision theory, an essential distinction is drawn between decisions “under risk” and “under uncertainty”. The difference is that in the former case, but not the latter, probabilities are assumed to be known. Hence we may say: “The probabilities of failure of different types of electric switches for the control room are so well known that a decision which of them to install can be classified as a decision under risk”. This corresponds to the following definition: 5. risk = the fact that a decision is made under conditions of known probabilities (“decision under risk”) With this we have identified five common meanings of “risk”. Many attempts have been made to put an end to the ambiguity of “risk” through a stipulative definition, but unfortunately these attempts have gone in different directions. In a joint book from 1981, several of the leading researchers in the field wrote that 1 The probabilities referred to in risk analysis are usually taken to be objective probabilities (frequencies or tendencies in the real world). When the term “subjective probability” is used, it usually refers to subjective estimates of objective probabilities. (In decision theory, “subjective probabilities” can also refer to degrees of belief that are not necessarily the outcome of an attempt to estimate an objective probability.)

Risk and Safety in Technology

1071

“[w]hat distinguishes an acceptable-risk problem from other decision problems is that at least one alternative option includes a threat to life or health among its consequences. We shall define risk as the existence of such threats” [Fischhoff et al., 1981, p. 2]. This is close to our definition (1). In 1983, a Royal Society working group defined risk as “the probability that a particular adverse event occurs during a stated period of time, or results from a particular challenge”, i.e. our definition (3) [Royal Society, 1983, p. 22]. In the same vein, the US National Research Council [1983] defined risk assessment as an assessment of the “probability that an adverse effect may occur as a result of some human activity.” At present, by far the most common technical definition of risk is our (4), namely risk as statistical expectation value. This is in fact the newest of the five meanings of risk referred to above. Although expectation values have been calculated since the 17th century, the use of the term “risk” in this sense has much more recent origin. It was introduced into technological risk analysis in the influential Reactor Safety Study (WASH-1400, the Rasmussen report) from 1975 [Rechard, 1999, p. 776]. Today it is the most common technical meaning of the term “risk”. The International Organization for Standardization [2002] defines risk somewhat vaguely as “the combination of the probability of an event and its consequences”, where “combination” is often (but need not be) interpreted as multiplication. All the major variants of technological risk analysis are based on the identification of risk with expectation value. In addition, this concept of risk is used in related fields such as chemical risk assessment. In cost-benefit analysis, the risks that enter the analysis are expectation values. In studies of risk perception, the “subjective risk” reported by the subjects is compared to the “objective risk”, which is identified with the expectation value [Hansson, 1993]. The identification of risk with expectation value requires that the severity of outcomes can be measured in numerical terms. Ideally such a measure should refer to over-all utility, in which case risk analysis becomes a branch of expected utility theory. In risk-benefit analysis, monetary values are used as proxies for utilities. In many engineering applications, the number of deaths is used, somewhat simplistically, as a measure of the severity of outcomes. Since “risk” has been widely used in various senses for more than 300 years, it should be no surprise that attempts to reserve it for a technical concept that was introduced 30 years ago give rise to communication problems. It is advisable to be respectful to common usage. When there is a risk of misunderstanding, it is preferable to employ a technical term such as “expectation value” for the technical concept, rather than trying to eliminate the established colloquial uses of “risk”.2

2.2

Uncertainty

Not all dangers come with probabilities assigned to them. In decision theory, the terms “risk” and “uncertainty” are used to distinguish between those that do and 2 In the 1983 Royal Society report, the term “detriment” was proposed to denote the product of risk and harm, but this proposal never caught on.

1072

Sven Ove Hansson

those that do not. In one of the most influential textbooks in decision theory, the terms are defined as follows: We shall say that we are in the realm of decision making under: (a) Certainty if each action is known to lead invariably to a specific outcome (the words prospect, stimulus, alternative, etc., are also used). (b) Risk if each action leads to one of a set of possible specific outcomes, each outcome occurring with a known probability. The probabilities are assumed to be known to the decision maker. For example, an action might lead to this risky outcome: a reward of $10 if a ‘fair’ coin comes up heads, and a loss of $5 if it comes up tails. Of course, certainty is a degenerate case of risk where the probabilities are 0 and 1. (c) Uncertainty if either action or both has as its consequence a set of possible specific outcomes, but where the probabilities of these outcomes are completely unknown or are not even meaningful. [Luce and Raiffa, 1957, p. 13] This usage of the key terms “risk” and “uncertainty” differs distinctly from everyday usage. In everyday conversations, we would not hesitate to call a danger a risk although there are no means to determine its probability. By uncertainty we would mean a state of mind rather than the absence of information. It is not uncommon for this difference between technical and non-technical usage to give rise to confusion. The gambler’s decisions at the roulette table are clear examples of decisions under risk, i.e. decisions with known probabilities. Given that the wheel is fair, the probabilities of various outcomes — gains and losses — are easily calculable, and thus knowable, although the gambler may not take them into account. For a clear example of a decision under uncertainty, consider instead the decision of an explorer to enter a distant part of the jungle, previously untrod by human foot. There are tigers and poisonous snakes in the jungle, but no estimates better than guesses can be given of the probability of being attacked by them. Such attacks are known dangers with unknown probabilities. In addition, the jungle may contain a large number of other species — from microorganisms to mammals — some of which may be dangerous although they are completely unknown. Not only their probabilities but also their very existence is unknown.3 In real life we are seldom in a situation like that at the roulette table, when all probabilities are known with certainty (or at least beyond reasonable doubt). Most of the time, we have to deal with technological dangers without knowing their probabilities, and often we cannot even foresee what dangers we will have to deal with. This is true not least in the development of new technologies. The social and environmental effects of a new technology can seldom be fully grasped 3 Unknown dangers are not included in Luce and Raiffa’s definition of uncertainty, as quoted above, but in subsequent discussions the concept of uncertainty has been extended to include them.

Risk and Safety in Technology

1073

beforehand, and there is often considerable uncertainty with respect to the dangers that it may give rise to [Porter, 1991; Rosenberg, 1995]. There is, however, a tendency in risk studies to proceed as if decisions on technologies were made under conditions analogous to gambling at the roulette table. This has been called the tuxedo syndrome, and it has been argued that the metaphor of entering an unexplored jungle better describes the plight of engineers and others responsible for new technology [Hansson, 2008]. It is important to observe that even in cases when the plausibility of dangers can be meaningfully summarized in terms of probabilities, there may yet remain significant uncertainties about the accuracy of these estimates. There is often a drift in the sense of the word “risk”, in the following sense: A discussion or an analysis begins with a general phrase such as “risks in the building industry” or “risks in modern energy production”. This includes both dangers for which probability estimates are already available and dangers for which it is doubtful whether meaningful probability estimates can at all be obtained. As the discussion or analysis goes more into technical detail, the term “risk” is narrowed down to refer only to that which can be quantified in probabilistic terms. In the course of this narrowing-down, the analyst often loses sight of uncertainties that should have been taken into account in an accurate analysis of the dangers at hand. To avoid the neglect of such dangers, it is important to treat uncertainties explicitly, in particular if a narrow concept of risk is used in the technical analysis.

2.3 Hazard In many engineering contexts, a distinction is made between a risk and a hazard. A hazard can be defined as a potential risk. Hence, consider two persons who set off the same type of firework. One of them is an expert pyrotechnician, the other a drunken teenager who never lit fireworks before. The hazard is the same in both cases, namely that the device may explode close to the person who handles it. The risk (in sense 3 or 4 as defined in Section 2.1) is much smaller in the former case, provided that the pyrotechnician takes the standard precautions. Hazard is mostly treated as a non-quantitative concept. Attempts have been made to quantify it. In such quantifications, the focus is on consequences, not on probabilities [Rahman et al., 2005]. The limit between a hazard and a non-hazard is vague. If an airplane is just about to crash, and this airplane is in the air very close to you, we would say that you are exposed to a hazard. However, if the plane is flying normally at high altitude, we would not think of it as a hazard as it is passing by. More generally speaking, when the risk associated with a hazard is considered negligible, we tend not to call it a hazard at all.

1074

Sven Ove Hansson

2.4 Safety The concept of safety is sometimes used in an absolute, sometimes in a relative sense. In order to illustrate the meaning of absolute safety, suppose that you buy a jacket that is promised to be of fire-proof fabric. Later, it actually catches fire. Then you might argue, if you apply the absolute notion of safety, that you were not safe from fire in the first place. If the producer of the jacket tries to argue that you in fact were safe since the fabric was highly unlikely to catch fire, you would say that he was simply wrong. In some contexts, therefore, “I am safe against the unwanted event X” is taken to mean that there is no risk at all that X will happen. Technical safety has often been defined as absolute safety. For example, in research on aviation safety it has been claimed that “Safety is by definition the absence of accidents” [Tench, 1985]. However, in practice absolute safety is seldom achievable. For most purposes it is therefore not a very useful concept. Indeed, the US Supreme court has supported a non-absolute interpretation, stating that “safe is not the equivalent of ‘risk free’ ” [Miller, 1988, p. 54]. With this interpretation, a statement such as “this building is fire-safe” can be read as a short form of the more precise statement: “The safety of this building with regard to fire is as high as can be expected in terms of reasonable costs of preventive actions.” In this vein, the American department of defence has stated that safety is “the conservation of human life and its effectiveness, and the prevention of damage to items, consistent with mission requirements” [Miller, 1988, p. 54]. Usage of the term “safe” (and derivatives such as “safety”) in technical applications, e.g. in aviation safety, highway safety etc, vacillates between the absolute concept (“safety means no harm”), and a relative concept that only requires the risk reduction that is considered feasible and reasonable. It is not possible to eliminate either of these usages, but it is possible to keep track of them and avoid confusing them with each other. Safety is usually taken to be the inverse of risk: when the risk is high, then safety is low, and conversely. This may seem self-evident, but the relationship between the two concepts is complicated by the fact that, as we saw in Subsection 2.1, the concept of risk is in itself far from clear. It has been argued that if risk is taken in the technical sense as statistical expectation value (expected harm), then safety cannot be the antinomy of risk, since other factors such as uncertainty have to be taken into account when assessing safety [M¨ oller et al., 2006]. With a broader definition of risk, an antonymic relationship between the two concepts may be more plausible.

2.5 Security Safety and security are closely related concepts. (Some languages have the same word for these two English terms, such as the German “Sicherheit”.) With safety we usually mean protection against unintentional threats, and with security protection against intentional threats. Typical security issues are protection of a coun-

Risk and Safety in Technology

1075

try against war, protection of individuals against violence, protection of property against theft, malicious or wanton destruction and economic crime, and protection of computers against unauthorized intrusion. Many of the measures that are taken to improve safety also increase security, and vice versa. Hence, the same sprinkler system can extinguish a fire caused by arson or by an accident. Security checks that prevent unauthorized persons from entering a building with dangerous machinery contribute to preventing both accidents and sabotage, etc. However, safety and security have traditionally been treated as separate issues, and they have been delegated to different professions. In many organizations that have to deal with both types of issues, surprisingly little has been done to coordinate them.

2.6

The precautionary principle

The precautionary principle is frequently invoked in debates on environmental issues. Strictly speaking, it is a principle for decision-making under scientific uncertainty that has been codified in a number of international treaties.4 Formulations of the precautionary principle can be divided into two major groups: argumentative and prescriptive versions of the principle. An argumentative version of the precautionary principle is found in Principle 15 of the Rio Declaration [UNCED 1993]. It requires that “lack of full scientific certainty shall not be used as a reason for postponing cost-effective measures to prevent environmental degradation”. Prescriptive versions of the precautionary principle prescribe actions. Perhaps the most famous of these is the so-called Wingspread Statement: “When an activity raises threats to the environment or human health, precautionary measures should be taken, even if some cause-and-effect relationships are not fully established scientifically” [Raffensperger and Tickner, 1999, pp. 354-355]. Most prescriptive versions of the precautionary principle share four common components [Sandin, 1999]. Consider the following possible formulation of the precautionary principle: It is mandatory to limit, regulate, or prevent potentially dangerous technologies even before scientific proof is established. We find four different components in this formulation, namely 1. the threat component, expressed in the phrase “potentially dangerous technologies”; 2. the uncertainty component, expressed in the phrase “even before scientific proof is established”; 3. the action component, expressed in the phrase “to limit, regulate, or prevent”; 4 The phrase “precautionary principle” is often used less specifically to denote cautious decision-making in general. It is better to use other terms for this, such as “cautious decisionmaking”.

1076

Sven Ove Hansson

4. the prescription component, expressed in the phrase “is mandatory”. The first two of these can be summarized as the trigger of the precautionary principle, whereas the last two constitute the precautionary response [Ahteensuu, 2008]. The uncertainty dimension ensures that action is triggered even in the absence of full scientific evidence. It is the most characteristic part of the principle. It is what distinguishes the precautionary principle from other principles or argumentation forms for the protection of health and the environment. In summary, the precautionary principle proclaims that policy decisions in environmental decisions can legitimately be based on scientific evidence of a danger that is not strong enough to constitute full scientific proof that the danger exists [Sandin et al., 2004]. However, it is not unproblematic to describe this as a special principle for environmental policies, as some sort of extra cautiousness that is presumed not to apply in other decisions. From a decision-theoretical point of view, allowing decisions to be influenced by uncertain information is not a special principle that needs to be specially defended. To the contrary, doing so is nothing else than ordinary practical rationality, as it is applied in most other contexts [Hansson, 2006a]. If there are strong scientific indications that a volcano may erupt in the next few days, decision-makers will expectedly evacuate its surroundings as soon as possible, rather than waiting for full scientific evidence that the eruption will take place. Furthermore, as we will see in the next section, traditional safety engineering is largely based on cautious thought patterns that are similar to the precautionary principle but of much older origin. 3 SAFETY ENGINEERING Engineering is primarily a practice. For as long as humanity has used technological artefacts, we have taken measures to protect humans against various risks associated with these artefacts. Since the 19th century, many engineers have specialized in worker’s safety and other safety-related tasks. With the development of technological science, the ideas behind safety engineering have been subject to academic treatments. There are now many ways to systematize the practices of safety engineering, but none of them has gained general acceptance. A major reason for this is that the discussion on safety engineering is fragmented between different areas of technology. In this section, three major principles of safety engineering will be discussed, and it will be argued that there is an underlying principle that unites them all.

3.1 Three principles of safety engineering The following principles are in general use in many fields of engineering. 1. Inherently safe design. A recommended first step in safety engineering is to minimize the inherent dangers in the process as far as possible. This means

Risk and Safety in Technology

1077

that potential hazards are excluded rather than just enclosed or otherwise coped with. Hence, dangerous substances or reactions are replaced by less dangerous ones, and this is preferred to using the dangerous substances in an encapsulated process. Fireproof materials are used instead of inflammable ones, and this is considered superior to using flammable materials but keeping temperatures low. For similar reasons, performing a reaction at low temperature and pressure is considered superior to performing it at high temperature and pressure in a vessel constructed for these conditions. 2. Safety factors. Constructions should be strong enough to resist loads and disturbances exceeding those that are intended. A common way to obtain such safety reserves is to employ explicitly chosen, numerical safety factors. Hence, if a safety factor of 2 is employed when building a bridge, then the bridge is calculated to resist twice the maximal load to which it will in practice be exposed. 3. Multiple independent safety barriers. Safety barriers are arranged in chains. The aim is to make each barrier independent of its predecessors so that if the first fails, then the second is still intact, etc. Typically the first barriers are measures to prevent an accident, after which follow barriers that limit the consequences of an accident, and finally rescue services as the last resort. One of the major lessons from the Titanic disaster is that an improvement of the early barriers (in this case: a hull divided into watertight compartments) is no excuse for reducing the later barriers (in this case: lifeboats). Several caveats have to be appended to this list of safety engineering practices: The terminology is not universally accepted, and some of these principles are also known under other names. The three principles are partly over-lapping. Furthermore, safety engineering includes many more principles and practices than the three mentioned above: negative feedback mechanisms, education of operators, maintenance of equipment and installations, incidence reporting etc. In the following subsections, each of the three principles will be treated in somewhat more detail.

3.2

Inherent safety

Inherent safety, also called primary prevention, consists in the elimination of a hazard. It is contrasted with secondary prevention that consists in reducing the risk associated with a hazard. This can be done by reducing either the probability or the consequences of an adverse event such as an accident in which the hazard is realized. For a simple example, consider a process in which inflammable materials are used. Inherent safety would consist in replacing them by non-inflammable materials. Secondary prevention would consist in removing or isolating sources of ignition and/or installing fire-extinguishing equipment. As this example shows, secondary prevention usually involves added-on safety equipment.

1078

Sven Ove Hansson

Traditionally, four types of safety measures are recommended in inherently safer design of plants: • minimize (intensify): use smaller quantities of hazardous materials • substitute: replace a hazardous material by a less hazardous one • attenuate (moderate): use the hazardous material in a less hazardous form • simplify: avoid unnecessary complexity in facilities and processes, in order to make operating errors less likely [Khan and Abbasi, 1998; Bollinger et al., 1996]. It is important not to overstate the case of inherent safety. Many safety problems cannot be solved with inherent safety. Often, socially desirable means cannot be achieved without hazardous materials or processes. In such cases, reliance must be put on well-developed secondary prevention. However, there are also many cases with a choice between eliminating and managing the hazard. Proponents of inherent safety have shown that other things being equal, elimination is the better option. The major reason for this is that, as long as the hazard still exists, it can be realized by some unanticipated triggering event. Even with the best of control measures, some unforeseen chain of events can give rise to an accident. Even the best added-on safety technology can fail, or be destroyed in the course of an accident. An additional argument for inherent safety is its usefulness in dealing with security threats. Add-on safety measures can often easily be deactivated by those who want to do so. When terrorists enter a chemical plant with the intent to blow it up, it does not matter much that all ignition sources have been removed from the vicinity of explosive materials (although this may perhaps have solved the safety problem). The perpetrators will bring their own ignition source. In contrast, most measures that make a plant inherently safer will also contribute to diverting terrorist threats. If the explosive substance has been replaced by a non-explosive one, or the inventories of explosive and inflammable substances have been drastically reduced, then the plant will be much less attractive to terrorists and will therefore also be a less likely target of attack. Inherent safety has a particularly important role in the chemical industry. Most of the development of techniques for inherent safety has taken place within or in cooperation with chemical companies. The other major industry where inherent safety is often discussed is the nuclear industry. Inherent safety is often referred to in efforts to construct new, safer types of reactors. A reactor will be inherently safer than those currently in use if, even in the case of failure of all active cooling systems and complete loss of coolant, the fuel element temperatures will not be exceed the limits at which most radioactive fission products remain confined within the fuel elements [Brinkmann et al., 2006].

Risk and Safety in Technology

3.3

1079

Safety factors

Probably, humans have made use of safety reserves since the origin of our species. We have added extra strength to our houses, tools, and other constructions in order to be on the safe side. However, the use of numerical factors for dimensioning safety reserves seems to be of relatively recent origin, probably the latter half of the 19th century. The earliest usage of the term recorded in the Oxford English Dictionary is from W.J.M. Rankine’s book A Manual of Applied Mechanics from 1858. In the 1860s, the German railroad engineer A. Wohler recommended a factor of 2 for tension [Randall, 1976]. The use of safety factors is now since long well established in structural mechanics and in its many applications in different engineering disciplines. Elaborate systems of safety factors have been developed, and specified in norms and standards [Clausen et al., 2006]. A safety factor is typically intended to protect against a particular integritythreatening mechanism, and different safety factors can be used against different such mechanisms. Hence one safety factor may be required for resistance to plastic deformation and another for fatigue resistance. A safety factor is most commonly expressed as the ratio between a measure of the maximal load not leading to the specified type of failure and a corresponding measure of the maximal load that is expected to be applied. In some cases it may instead be expressed as the ratio between the estimated design life and the actual service life. In some applications safety margins are used instead of safety factors. A safety margin differs from a safety factor in being additive rather than multiplicative. In order to keep airplanes sufficiently apart in the air a safety margin in the form of a minimal distance is used. Safety margins are also used in structural engineering, for instance in geotechnical calculations of embankment reliability [Duncan, 2000]. According to standard accounts of structural mechanics, safety factors are intended to compensate for five major categories of sources of failure [Knoll, 1976; Moses, 1997]: 1. higher loads than those foreseen, 2. worse properties of the material than foreseen, 3. imperfect theory of the failure mechanism in question, 4. possibly unknown failure mechanisms, and 5. human error (e.g. in design). The first two of these refer to the variability of loads and material properties. Such variabilities can often be expressed in terms of probability distributions. However, when it comes to the extreme ends of the distributions, lack of statistical information can make precise probabilistic analysis impossible. Let us consider the variability of the properties of materials. Experimental data on material properties are often insufficient for making a distinction between e.g. gamma and lognormal distributions, a problem called distribution arbitrariness [Ditlevsen, 1994]. This

1080

Sven Ove Hansson

has little effect on the central part of these distributions, but in the distribution tails the differences can become very large. This is a major reason why safety factors are often used as design guidance instead of probabilities, although the purpose is to protect against failure types that one would, theoretically, prefer to analyze in probabilistic terms. Theoretically, design by using structural system reliability is much more reasonable than that based on the safety factor. However, because of the lack of statistical data from the strength of materials used and the applied loads, design concepts based on the safety factor will still dominate for a period. [Zhu, 1993] The last three of the five items on the list of what safety factors should protect against all refer essentially to errors in our theory and in our application of it. They are therefore clear examples of uncertainties that are not easily amenable to probabilistic treatment. In other words: The eventuality of errors in our calculations or their underpinnings is an important reason to apply safety factors. This is an uncertainty that is not reducible to probabilities that we can determine and introduce into our calculations. It is for instance difficult to see how a calculation could be accurately adjusted to compensate self-referentially for the possibility that it may itself be wrong. However, these difficulties do not make these sources of failures less important. Safety factors are used to deal both with those failures that can be accounted for in probabilistic terms and those that cannot.

3.4 Independent safety barriers The use of multiple safety barriers is based on the simple principle that even if one measure that we take to avert a danger should fail, there should be some other measure in place that averts it. The archetype of multiple safety barriers is an ancient fortress. If the enemy manages to pass the first wall, there are additional layers that protect the defending forces. Some engineering safety barriers follow the same principle of concentric physical barriers. Interesting examples of this can be found in nuclear waste management. The waste will be put in a copper canister that is constructed to resist the foreseeable challenges. The canister is surrounded by a layer of bentonite clay that protects the canister against small movements in the rock and “acts as a filter in the unlikely event that any radionuclides should escape from a canister”.5 This whole construction is placed in deep rock, in a geological formation that has been selected to minimize transportation to the surface of any possible leakage of radionuclides. The whole system of barriers is constructed to have a high degree of redundancy, so that if one the barriers fails the remaining ones will suffice. With the usual standards of probabilistic risk analysis, the whole series of barriers around the waste would not be necessary. Nevertheless, sensible reasons can be given for this approach, namely reasons that refer to uncertainty. Perhaps the 5 http://www.skb.se/templates/SKBPage

8762.aspx.

Risk and Safety in Technology

1081

copper canister will fail for some unknown reason not included in the calculations. Then, hopefully, the radionuclides will stay in the bentonite, etc. In this particular case, redundancy can also be seen as a means to meet public scepticism and opposition (although it is not self-evident that redundant safety barriers will make the public feel safer). The notion of multiple safety barriers can also refer to safety barriers that are not placed concentrically like the defence walls of a fortress, but are arranged consecutively in a functional sense. The essential feature is that the second barrier is put to work when the first one fails, etc. Consider for instance the protection of workers against a dangerous gas such as hydrogen sulphide that can leak from a chemical process. An adequate protection against this danger can be constructed as a series of barriers. The first barrier consists in constructing the whole plant in a way that excludes uncontrolled leakage as far as possible. The second barrier is careful maintenance, including regular checking of vulnerable details such as valves. The third barrier is a warning system combined with routines for evacuation of the premises in the case of a leakage. The fourth barrier is efficient and well-trained rescue services. The basic idea behind multiple barriers is that even if the first barrier is wellconstructed, it may fail, perhaps for some unforeseen reason, and that the second barrier should then provide protection. For another illustration of this principle, we can consider what is possibly the most well-known example of technological failure in modern history, the Titanic that sank with 1500 persons in April 1912. It was built with a double-bottomed hull that was divided into sixteen compartments, constructed to be watertight. Four of these could be filled with water without danger. Therefore, the ship was believed to be unsinkable, and consequently it was equipped with lifeboats only for about half of the persons onboard. We now know that the Titanic was far from unsinkable. But let us consider a hypothetical scenario. Suppose that tomorrow a ship-builder comes up with a convincing plan for an unsinkable boat. Calculations show that the probability of the ship sinking is incredibly low and that the expected cost per life saved by the life-boats is above 1000 million dollars, a sum that can evidently be more efficiently used to save lives elsewhere. How should the naval engineer respond to this proposal? Should she accept the verdict of the probability calculations and the economic analysis, and exclude lifeboats from the design? There are good reasons why a responsible engineer should not act in this way: The calculations may possibly be wrong, and if they are, then the outcome may be disastrous. Therefore, the additional safety barrier in the form of lifeboats (and evacuation routines and all the rest) should not be excluded. Although the calculations indicate that such measures are inefficient, these calculations are not certain enough to justify such a decision. The major problem in the construction of safety barriers is how to make them as independent of each other as possible. If two or more barriers are sensitive to the same type of impact, then one and the same destructive force can get rid of all of them in one swoop. Hence, any number of concentric walls around a fortified

1082

Sven Ove Hansson

city could not protect the inhabitants against starvation when they run out of provisions. Similarly, three consecutive safety valves on the same tube may all be destroyed in a fire, or they may all be incapacitated due to the same mistake by the maintenance department. A major geologic event or an attacking enemy can destroy all the barriers of a subterranean nuclear waste repository at the same time, etc. Therefore, it is essential, when constructing a system of safety barriers, to make the barriers as independent as possible. Often, more safety is obtained with fewer but independent barriers than with many that are sensitive to the same sources of destruction.

3.5 The common trait The principles of safety engineering discussed in this section have one important trait in common: they all aim at protecting us not only against risks (in the technical sense) but also against hazards that cannot be assigned meaningful probability estimates, such as the possibility that some unforeseen event triggers a hazard that is seemingly under control. Although explicit discussions of risk and uncertainty are new to engineering, the insights encoded in safety engineering are much older than that – indeed older than probability theory. 4

RISK ANALYSIS

In the late 1960s, rapidly growing public opposition to new technologies gave rise to a new market for applied science: a market for expertise on risks and on the public’s attitudes to risks. The demand came mostly from companies and institutions associated with the technologies that had been subject to public opposition. The supply was met by professionals and academics with training in the natural, behavioural, and social sciences. Most of their undertakings focused on chemicals and on nuclear technology, the same risk factors that public opposition had targeted on. The new field was institutionalized as the discipline of risk analysis, with professional societies, research institutes, and journals of its own.

4.1

The subdisciplines of risk analysis

The short history of risk analysis can be summarized in terms of five major approaches to risk that made their appearance consecutively, and now all coexist [Otway, 1987]. The first of the five approaches was acceptable risk. Many of the earliest studies in the field aimed at determining a level of “risk”, i.e., of the statistically expected number of fatalities, that is accepted or that should be accepted. A common procedure was to compare new technological risks to risks that are accepted in everyday life. The next step was risk-benefit analysis, in which both the risks and the benefits of a technology were quantified so that they could be compared. The standard method is to assign a monetary value to all relevant outcomes (including the loss of

Risk and Safety in Technology

1083

human lives) in order to make risks and benefits computationally comparable. This approach was largely the result of economists entering the stage of risk science. The third step resulted in part from the growing involvement of psychologists in risk analysis. This step consisted in studies of risk perception that were much in vogue in the early 1980’s. The ordering of risks obtained from questionnaires was said to measure “subjective risk”, and was compared to the expected number of deaths that was called “objective risk”. The difference was commonly conceived as a sign of irrationality or misperception. The next approach was risk communication that aims at providing lay people with information that helps them to see risks in a certain way. Typically, risk communication is considered to be successful if it has made people adjust their “subjective risk” to fit in with the “objective risk”. The fifth approach was studies of trust that emanated from the difficulties that both public authorities and companies have encountered when trying to change the public’s opinion on risk through various measures of risk communication. The problem, seen in their perspective, seems to be that the public does not have trust in the sources of information. How such trust can be achieved is currently a major theme at conferences on risk. All the major variants of risk analysis are associated with the same formal model of risk, namely the expectation value definition of risk that was introduced above in Subsection 2.1. In other words, the common procedure is to multiply “the probability of a risk with its severity, to call that the expectation value, and to use this expectation value to compare risks” [Bondi, 1985, p. 9]. The following is a typical example of the jargon: The worst reactor-meltdown accident normally considered, which causes 50 000 deaths and has a probability of 10−8 /reactor-year, contributes only about two per cent of the average health effects of reactor accidents. [Cohen, 1985, p. 1] In what follows, we will consider three central methodological issues in risk analysis (Subsections 4.2–4.4) and three additional such issues in risk-benefit analysis (Subsections 4.5–4.7). The final Subsection 4.8 is devoted to a comparative discussion of risk analysis and safety engineering.

4.2

Fault tree methodology

When there is statistically sufficient experience of an event-type, such as a machine failure, then we can determine its probability by collecting and analysing that experience. Hence, if we want to know the probability that the airbag in a certain make of car fails to release in a collision, we should collect statistics from the accidents in which such cars were involved. For new and untested technologies this method is not available. Accident statistics is not at hand to determine the probability of airbag failure in a new car model that is just to be introduced. If the construction is essentially unchanged since

1084

Sven Ove Hansson

previous models, then we may rely on statistics from these earlier models, but this is not advisable if there have been significant changes. Even after many years’ experience of a technology there may be insufficient data to determine the frequencies of unusual types of accidents or failures. As one example of this, there have (fortunately) been too few severe accidents in nuclear reactors to make it possible to estimate their probabilities. In particular, most of the reactor types in use have never been involved in any serious accident. It is therefore not possible to determine the risk (probability) of a severe accident in a specified type of reactor. One common way to evade these difficulties is to calculate the probability of major failures by means of a careful investigation of the various chains of events that may lead to such failures. By combining the probabilities of the subevents in such a chain, a total probability of a serious accident can be calculated. Such calculations were in vogue in the 1970s and 1980s. Today there is a growing scepticism against them, due to several difficult problems with this methodology. One such problem is that accidents can happen in more ways than we can think of beforehand. There is no method by which we can identify all chains of events that may lead to a major accident in a nuclear reactor, or any other complex technological system. Another problem with this methodology is that the probability of a chain of events can be very difficult to determine even if we know the probabilities of each individual event. Suppose for instance that an accident will happen if two safety valves both fail. Furthermore suppose that we have experience showing that the probability is 1 in 500 that a valve of this construction will fail during a period of one year. It does not follow from this that the probability that both will fail in that period is 1/500 × 1/500, i.e. 1/250,000. The reason for this is that failures in the two valves are not independent events. In spite of these difficulties, the construction and analysis of such event chains (often called fault-trees) is not a useless exercise. To the contrary, it can be an efficient way to identify weaknesses in a complex technological system. It is important, though, to keep in mind that an exhaustive list of negative events cannot be obtained, and that therefore total risk levels cannot be determined in this way.

4.3 Indetectable effects Much of modern science is devoted to the study of composite systems: ecosystems, the human body, the world economy, etc. Each of these contains so many components and potential interactions that it is in practice unpredictable. Some of these systems are unpredictable not only in practice but also in principle, due to chaotic phenomena. In addition, science is always subject to another type of uncertainty, namely that of unknown factors. Only seldom do we have good reasons to believe that our scientific models are complete in the sense that no important components or interactions have been left out.

Risk and Safety in Technology

1085

In some cases, knowledge about complex systems can be gained through systematized experience. This applies for instance to the effects of therapeutic agents on the human body. Due to the complexity of the body, it is in practice impossible to theoretically predict the effects of a new drug. Instead, after preliminary trials have been completed, medicines are tried out experimentally on groups of patients. Based on statistics from such studies (clinical trials) the effects of medical drugs can be ascertained with reasonable accuracy. However, there are many cases in which this type of “statistical bypass” to knowledge about complex systems is not available. We do not have access to one hundred earths on which we can experiment to determine a tolerable level of greenhouse gas emissions. Furthermore, even in the cases when statistical information is available, it does not always reduce uncertainties as efficiently as one might hope. To see this, let us consider the example of health effects of chemical substances. To what extent is it possible to determine the presence or absence of such effects through direct studies of exposed humans? Unfortunately, the answer to that question is rather disconcerting. To simplify the discussion, let us focus on lifetime risks of lethal effects of toxic substances. To begin with, suppose that 1000 persons are all subject to a chemical exposure that gives rise to hepatic angiosarcoma (a rare cancer of the liver) among 0.5 % of the exposed. Among unexposed individuals, the frequency of this disease is very close to zero. If a proper investigation is made, chances are very high that the overrepresentation of this disease among the exposed population will be discovered. Next, suppose that another group of 1000 persons are subject to an industrial exposure that increases the incidence of lung cancer from 10.0 to 10.5 %. The expected number of additional cancer cases is the same as in the previous case. However, as can be shown with probability calculus, the difference between 10.0 and 10.5 % is in this case indistinguishable from random variations. Hence, the effects of this substance cannot be detected in a study of the exposed population [Hansson, 1999]. As a rough rule of thumb, epidemiological studies cannot reliably detect excess relative risks if they are about 10 % or smaller. For the more common types of lethal diseases, such as coronary disease and lung cancer, lifetime risks are of the order of magnitude of about 10 %. Therefore, even in the most sensitive studies, an increase in lifetime risk of the size 10−2 (10 % of 10 %) or smaller may be indetectable (i.e. indistinguishable from random variations). In animal experiments we have similar experimental problems, and in addition problems of extrapolation from one species to another. There is no objective answer to the question how small health effects should be of concern to us. However, many attempts have been made to set a limit of concern, expressed either as “acceptable risk” or “de minimis risk”. Most people seem to agree that if a human population is exposed to a risk factor that will, statistically, kill one person out of 109 , then that risk will not be an issue of high priority. Arguably, it is no big problem that our risk assessment methods are

1086

Sven Ove Hansson

insufficient to discover risks of that order of magnitude. On the other hand, most of us would consider it a serious problem if a risk factor that kills one person out of 100 or 1000 cannot be detected. The most common proposals for limits of concern for lethal risks are 1 in 100,000 and 1 in 1,000,000. It is difficult to find proposals above 1 in 10,000. These values are of course not objective or scientific limits; they belong to the ethical realm. However, it is important to note the presence of what has been called an ethical gap, a gap between those risk levels that are scientifically detectable and those that are commonly regarded to be ethically acceptable or at least of minor concern [Hansson, 2002]. This gap has the breadth of 2–4 orders of magnitude. Hence, even if no adverse effects have been found in exposed populations, there may still be effects at risk levels that are at least 100 to 1000 times higher than commonly proposed levels of concern or acceptability.

4.4 The unreliability of probability estimates When probabilities cannot be estimated from empirically known frequencies, the standard method is to instead use experts’ estimates of probabilities. The reliability of risk analysis will then depend on the assumption that there are no or only small systematic differences between objective probabilities and experts’ estimates of these probabilities. However, this assumption is not correct. Significant differences between such objective frequencies and expert’s estimates of them are well known from experimental psychology, where they are described as lack of calibration. Probability estimates are (well) calibrated if “over the long run, for all propositions assigned a given probability, the proportion that is true equals the probability assigned” [Lichtenstein et al., 1982, pp. 306-307]. Thus, half of the statements that a wellcalibrated subject assigns probability 0.5 are true, as are 90 per cent of those that she assigns probability 0.9, etc. Experimental studies indicate that there are only a few types of predictions that experts perform in a well-calibrated manner. Professional weather forecasters and horse-race bookmakers make well-calibrated probability estimates in their respective fields of expertise [Murphy et al., 1984; Hoerl and Fallin, 1974]. In contrast, most other types of prediction that have been studied are subject to substantial overconfidence. Physicians assign too high probability values to the correctness of their diagnoses [Christensen-Szalanski and Bushyhead, 1981]. Geotechnical engineers were overconfident in their estimates of the strength of a clay foundation, etc. [Hynes and Vanmarcke, 1976]. We do not know how well calibrated the experts’ estimates of probabilities are that are used in risk analysis. To the extent that they are badly calibrated, the outcome of risk analysis is correspondingly inaccurate.

Risk and Safety in Technology

4.5

1087

Problems of incommensurability

Most of the philosophical discussion about risk-benefit analysis has been concerned with the difficulties involved in assigning an economic value to that which we conceive as invaluable, such as a human life or an animal species. Clearly, human lives do not have a monetary price in the common sense of the word. A risk-benefit analyst who assigns a monetary value to the loss of a human life does not thereby imply that someone can buy another person, or the right to kill her, for that price. In any sensible interpretation of a risk-benefit analysis, the values of lives are for calculation purposes only. The underlying motivation for values of life is that they can be used as a means to reduce multi-dimensional decision problems to uni-dimensional ones. The common way to do this, technically, is to convert all dimensions of a decision to monetary values, even those that are incommensurable with money. The essential problem — or perhaps even dilemma — is that in order to achieve this we need to comparatively evaluate entities that we conceive as incomparable. Defenders of risk-benefit analysis tend to emphasize that comparisons between lives and money are not unique to risk-benefit analysis. They are in fact unavoidable components of many of the decisions that we have to make in different social sectors. We could, for instance, always spend more money than we do on traffic safety, taking the resources from activities that do not save lives. Our decision not to spend more than we do contains an implicit value of life; we do not pay to save more lives than we do since it would be too costly. The problem does not come with risk-benefit analysis, it is only more clearly exhibited when a risk-benefit analysis is performed in order to guide the decision. On the other hand, money has connotations not shared by non-monetary units that can sometimes be used for the same or similar purposes, such as QALYs (quality-adjusted life years). The use of money instead of some other unit may therefore send a message that can be conceived as desecrating the value of life. One of the most common methods used to derive calculation values for nonmarket goods is contingent valuation (willingness to pay studies, WTP). This means that the values are based on people’s answers to questions of the type “How much would you be prepared to pay for saving the giant panda from extinction?” The presumption is that the sum of everyone’s answers to that question determines the value that the non-extinction of the giant panda should be assigned in an economic analysis. It turns out, however, that our answers to such questions do not give good indications of our priorities. Hence, Beattie and coworkers [1998] found that many respondents tend to report an amount that would not seriously disturb their normal expenditure and savings patterns, typically a sum in the range £50–200 per annum. Respondents were also insensitive to the magnitude of the risk reduction. No other, more reliable method seems to be available for eliciting calculation values from questionnaire respondents. Some methods used in risk-benefit analysis, including contingent valuation, tend to give more influence to affluent people since they can pay more than others to

1088

Sven Ove Hansson

have it their way [Copp, 1987]. Although this is a common feature in risk-benefit analysis, it is methodologically easy to avoid for instance by relating (actual or hypothetical) payments by an individual to that individual’s income.

4.6 Transferability across contexts In risk-benefit analyses, cost estimates are regularly transferred across contexts. This applies, in particular, to estimates of life values. Two examples can be given to illustrate this practice. The first example is a risk-benefit analysis of mammography that was performed in 1992 by the American FDA. The analysis made use of values of life to determine, in monetary terms, the economic benefits from saving a life with mammography. The life values were derived from estimates of how much more male workers are paid when working in occupations with a high risk of fatal accidents [Heinzerling, 2000, pp. 205–206]. The second example is a risk-benefit analysis performed in 2000 by the American EPA in order to determine a new standard for arsenic in drinking water. Here again, values of life were taken from studies of how much compensation male workers receive for risks of fatal accidents [Heinzerling, 2002, p. 2312]. In both these cases, it would have been possible to use life values derived from the very context of the risk-benefit analysis in question. Women could have been asked how much they are prepared to pay for mammography, given realistic assumptions about the risk reduction it gives rise to. Their willingness to pay for reduced risks could then be used in a risk-benefit analysis for mammography. Although the use of such values would not have been unproblematic, it would at least have been much closer to the relevant context than the life value that was actually used. Similarly, people could have been asked how much they were prepared to pay for reduced levels of arsenic in drinking water, given realistic assumptions about the health effects of such a reduction. Alternatively, willingness to pay for healthy water (in the form of mineral water) could have been obtained from actual markets. Instead, life values were transferred from another context, namely that of wage compensation for occupational health risks. The assumed transferability across contexts that is illustrated in these examples is in fact an essential condition without which current methods of cost-benefit analyses cannot be justified. If all values used in a risk-benefit analysis had to be derived from the precise context of the particular analysis, then the practice of risk-benefit analysis would come close to that of performing opinion polls on the topic to be analyzed. Once transferability across contexts is given up, we seem to enter a slippery slope in which the characteristic features of risk-benefit analysis as we know them today would be lost. In order to avoid this, and defend transferability across contexts, one would have to claim that it is better to use values from a certain context (such as wages that compensate for workplace risks) in a risk-benefit analysis concerning another context (such as mammography), than to use values derived in the context of the analysis in question. No such argument seems to be available. In particular, it has not been shown that our

Risk and Safety in Technology

1089

decisions on what employment offers to seek and accept are better informed than most other decisions that we make (such as decisions on mammography and on contaminants in drinking-water).

4.7

Interpersonal aggregation

In a risk-benefit analysis, all risks and all benefits are combined in one and the same balance. This means that a disadvantage affecting one person can be fully compensated for by an advantage of the same size that affects some other person. In other words, interpersonal compensability of advantages and disadvantages is assumed [Hansson, 2004].6 This is an assumption that risk-benefit analysis shares with utilitarianism. We can express it as a weighing principle, as follows: The collectivist weighing principle: An option is acceptable to the extent that the sum of all individual risks that it gives rise to is outweighed by the sum of all individual benefits that it gives rise to. This is not the only way in which risks can be weighed against benefits. Another possibility is to perform the weighing individually for each affected person, and require a positive balance for each person: The individualist weighing principle: An option is acceptable to the extent that the risks affecting each individual are outweighed by benefits for that same individual. Individualist weighing has a strong tradition in social practices that have their origin in the physician–patient relationship. For an example, consider a physician who selects patients for a clinical trial with a new, experimental treatment. Such a treatment involves risks and benefits that have to be weighed against each other. If the physician based this decision on a conventional risk-benefit analysis, then she would include a patient in the study if the risk to this patient is outweighed by the total social benefit. The total social benefit includes the expected gains from the study for future patients. With such a criterion, a patient can be included in the trial even if the risks by far exceed the expected gains to her personally. This, of course, is not how such decisions are made. Instead, they are made in accordance with the individualist weighing principle. A patient is not offered to participate in a clinical trial unless it is believed that the risks to which she will be exposed are outweighed by the expected advantages for her of the experimental treatment [Hansson, 2006b]. 6 Interpersonal compensability should not be conflated with the related but distinct issue of interpersonal comparability. Even if a benefit is greater than a harm, it need not cancel out the harm. Interpersonal comparability does not imply interpersonal compensability, but they are nevertheless closely related since the former is a necessary prerequisite for making the latter operative.

1090

Sven Ove Hansson

For another example, consider recommendations by health authorities on fish consumption. Although fish is generally speaking healthy food, contaminants in fish caught in certain waters give reason to recommend limits in fish consumption. Such recommendations are based on the positive and negative health effects on the individual (and in the case of pregnant or breast-feeding women, on corresponding effects on the child) [Knuth et al., 2003]. It would be regarded as inappropriate to base such recommendations on a full risk-benefit analysis that included other factors, such as the effects of diminished fish consumption on employment in the fishing industry or on regional economics. Hence we perform risk-benefit analyses, with collectivist risk-weighing, when deciding on road projects and other engineering projects, but use other types of calculations, based on individual risk-weighing, when deciding on clinical trials and dietary advice. In some policy areas we have a tradition of sacrificing individual interests for the sake of collective goals, whereas individual interests have a much stronger protection in other areas. It is a problem for risk-benefit analysis to motivate why we should employ total (collective) aggregation instead of alternative methods that protect individuals against sacrifice of their interests for collective goals.

4.8 Risk analysis versus safety engineering Probabilistic risk analysis and cost-benefit analysis have sometimes been seen as competitors of traditional forms of safety engineering. This is a too narrow view of the matter. Instead, it should be recognized that neither of these methods can in practice tell the full truth about risk and safety. It is more constructive to see them as complementary. Probabilistic risk analysis is often an indispensable tool for priority-setting and for the effect evaluation of safety measures. On the other hand, some of the uncertainties that safety engineering deals with successfully tend to be neglected in probabilistic calculations. Methodological pluralism, rather than monopoly for one single methodology, is to be recommended. Currently there is a trend in several fields of engineering towards increased use of probabilistic risk analysis. This trend will strengthen safety engineering, provided that it leads to a broadening of the knowledge base and not to exclusion of the wide range of dangers — from one’s own miscalculations to terrorist attacks — for which no meaningful probability estimates can be obtained. 5

THE ETHICS OF RISK

Throughout the history of moral philosophy, moral theorizing has for the most part referred to a deterministic world in which the morally relevant properties of human actions are both well-determined and knowable. In recent years, moral philosophers have in most cases left it to decision theorists to analyse the complexities that the indeterminism of real life gives rise to. Mainstream ethical (and metaethical) theories still focus on deterministic problems; in fact they lack the

Risk and Safety in Technology

1091

means to deal with problems involving risk and uncertainty [Hansson, 2003]. In this section we will investigate how moral theory can be extended so that it can support decisions on technological risks.

5.1

The causal dilution problem

How can we generalize ethical theories so that they can be effectively applied to problems involving risk and uncertainty? The problem of how to perform this generalization can be specified in terms of the causal dilution problem. The causal dilution problem (general version): Given the moral appraisals that a moral theory T makes of valuecarriers with well-determined properties, what moral appraisals does (a generalized version of) T make of value-carriers whose properties are not well-determined beforehand? The term “moral appraisal” covers a wide range of assignments of moral status, such as declarations that something is forbidden, permitted, morally required, good, bad, better than something else to which it is compared, etc. The term “value-carriers” refers to all entities that can be assigned (moral) value, including in particular human actions and the outcomes of human actions. Under conditions of risk, we can restate the causal dilution problem as follows: The causal dilution problem (probabilistic version): Given the moral appraisals that a moral theory T makes of valuecarriers with well-determined properties, what moral appraisals does (a generalized version of) T make of probabilistic mixtures of such value-carriers?

5.2

Actualism

We will begin with utilitarianism, the moral theory that has most often been applied to problems of risk. One fairly obvious approach to the causal dilution problem for utilitarianism is the following [Carlson, 1995] Actualism The utility of a (probabilistic) mixture of potential outcomes is equal to the utility of the outcome that actually materializes. To exemplify the actualist approach, consider an engineer’s decision whether or not to reinforce a bridge before it is being used for a single, very heavy transport. There is a 50 % risk that the bridge will collapse if it is not reinforced. Suppose that she decides not to reinforce the bridge and that everything goes well; the bridge is not damaged. According to the actualist approach, what she did was right. This is, of course, contrary to common moral intuitions.

1092

Sven Ove Hansson

The actualist solution requires that we use moral terms such as “right” and “wrong” in a way that differs radically from ordinary usage. If we accept the actualist usage, then it will in most cases be impossible to know what is right or wrong (or permitted, morally required, good, best, etc.) to do. In this way, actionguidance is expelled from moral discourse. However, action-guidance is largely what we need ethics for. Therefore, this is an unusually unhelpful approach. If we follow it, then action-guidance will have to be reintroduced in some other way.

5.3 Expected utility The standard decision-theoretical solution to the causal dilution problem for utilitarianism is the maximization of expected utility. To maximize expected utility means to choose among a set of alternatives one of those that have the highest expected, i.e. probability-weighted utility: Expected utility: The utility of a probabilistic mixture of potential outcomes is equal to the probability-weighted average of the utilities of these outcomes. The strongest argument in favour of maximizing objectivist expected utility is that this is a fairly safe method to maximize the outcome in the long run. Suppose, for instance, that the expected number of deaths in traffic accidents in a region will be 300 per year if safety belts are compulsory and 400 per year if they are optional. Then, if these calculations are correct, about 100 more persons per year will actually be killed in the latter case than in the former. We know, when choosing one of these options, whether it will lead to fewer or more deaths than the other option. If we aim at reducing the number of traffic casualties, then this can, due to the law of large numbers, safely be achieved by maximizing the expected utility (i.e., minimizing the expected number of deaths). The validity of this argument depends on the large number of road accidents that levels out random effects in the long run. Therefore, the argument is not valid for case-by-case decisions on unique or very rare events. Suppose, for instance, that we have a choice between a probability of 0.001 of an event that will kill 50 persons and a 0.1 probability of an event that will kill one person. Here, random effects will not be levelled out as in the traffic belt case. In other words, we do not know, when choosing one of the options, whether or not it will lead to fewer deaths than the other option. In such a case, taken in isolation, there is no compelling reason to maximize expected utility. Nevertheless, a decision in this case to prefer the first of the two options (with the lower number of expected deaths) may very well be based on a reasonable application of expected utility theory, namely if the decision is included in a sufficiently large group of decisions for which a metadecision has been made to maximize expected utility. As an example, a case can be made that a criterion for the regulation of safety equipment in motorcars should be one of maximizing expected utility (minimizing expected damage). The consistent application of this criterion

Risk and Safety in Technology

1093

in all the different specific regulatory decisions should minimize the damage caused by technical failures of motor vehicles. The larger the group of decisions that are covered by such a rule, the more efficient is the levelling-out effect. In other words, the larger the group of decisions, the larger catastrophic consequences can be levelled out. However, there is both a practical and an absolute limit to this effect. The practical limit is that decisions have to be made in manageable pieces. If too many issues are lumped together, then the problems of information processing may lead to losses that outweigh any gains that might have been hoped for. Obviously, decisions can be partitioned into manageable bundles in many different ways, and how this is done may have a strong influence on decision outcomes. As an example, the protection of workers against radiation may not be given the same priority if it is grouped together with other issues of radiation as if it is included among other issues of work environment. The absolute limit to the levelling-out effect is that some extreme outcomes, such as a nuclear war or a major ecological threat to human life, cannot be levelled out even in the hypothetical limiting case in which all human decision-making aims at maximizing expected utility. Perhaps the best example of this is the Pentagon’s use of secret utility assignments to accidental nuclear strike and to failure to respond to a nuclear attack, as a basis for the construction of command and control devices [Pat´e-Cornell and Neu, 1985]. Even when the levelling-out argument for expected utility maximization is valid, compliance with this principle is not required by rationality. In particular, it is quite possible for a rational agent to refrain from minimizing total damage in order to avoid imposing high-probability risks on individuals. To see this, let us suppose that we have to choose, in an acute situation, between two ways to repair a serious gas leakage in the machine-room of a chemical factory. One of the options is to send in the repairman immediately. (There is only one person at hand who is competent to do the job.) He will then run a risk of 0.9 to die due to an explosion of the gas immediately after he has performed the necessary technical operations. The other option is to immediately let out gas into the environment. In that case, the repairman will run no particular risk, but each of 10,000 persons in the immediate vicinity of the plant runs a risk of 0.001 to be killed by the toxic effects of the gas. The maxim of maximizing expected utility requires that we send in the repairman to die. This is also a fairly safe way to minimize the number of actual deaths. However, it is not clear that it is the only possible response that is rational. A rational decision-maker may refrain from maximizing expected utility (minimizing expected damage) in order to avoid what would be unfair to a single individual and infringe her rights.

5.4

Deontological and rights-based theories

The causal dilution problem for rights-based theories was formulated (in its probabilistic version) by Robert Nozick: “Imposing how slight a probability of a harm that violates someone’s rights also violates his rights?” [Nozick, 1974, p. 7]. In

1094

Sven Ove Hansson

somewhat more general language we can restate it, and its deontological counterpart, as follows: The causal dilution problem for deontological/rights-based moral theories (general version): Given the duties/rights that a moral theory T assigns with respect to actions with well-determined properties, what duties/rights does (a generalized version of) T assign with respect to actions whose properties are not well-determined beforehand? The causal dilution problem for deontological/rights-based moral theories (probabilistic version): Given the duties/rights that a moral theory T assigns with respect to actions with well-determined properties, what duties/rights does (a generalized version of) T assign with respect to probabilistic mixtures of such actions? An extension of a deontological theory to indeterministic cases can be obtained by just prescribing that a prohibition to bring about a certain outcome implies a prohibition to cause an increase in the risk of that outcome (even if the increase is very small). Similarly, for a rights-based theory, it could be claimed that if I have a right that you do not bring about a certain outcome, then I also have a right that you do not perform any action that has a non-zero risk of bringing about that outcome. Unfortunately, such a strict extension of rights and prohibitions is socially untenable. Your right not to be killed by me certainly implies a prohibition for me to perform certain acts that involve a risk of killing you, but it cannot prohibit all such acts. Such a strict interpretation would make human society impossible. I am allowed to drive a car in the town where you live, although this increases the risk of being killed by me. Hence, rights and prohibitions have to be defeasible so that they can be cancelled when probabilities are small. The most obvious way to achieve this is to assign to each right (prohibition) a probability limit. Below that limit, the right (prohibition) is cancelled. However, as Nozick observed, such a solution is not credible since probability limits “cannot be utilized by a tradition which holds that stealing a penny or a pin or anything from someone violates his rights. That tradition does not select a threshold measure of harm as a lower limit, in the case of harms certain to occur” [Nozick, 1974, p. 75]. Clearly, a moral theory need not treat a slight probability of a sizable harm in the same way that it treats a slight harm. The analogy is nevertheless relevant. The same basic property of traditional rights theories, namely the uncompromising way in which they protect against disadvantages for one person inflicted by another, prevents them from drawing a principled line either between harms or between probabilities in terms of their acceptability or negligibility. In particular, since no rights-based method for the determination of such probability limits seems to be

Risk and Safety in Technology

1095

available, they would have to be external to the rights-based theory. Exactly the same problem obtains for deontological theories. Probability limits do not solve the causal dilution problem for these types of theories. No other solution of the causal dilution problem for these theories seems to be available.

5.5

Contract theories

Contract theories may perhaps appear somewhat more promising. The criterion that they offer for the deterministic case, namely consent among all those involved, can also be applied to risky options. Can we then solve the causal dilution problem for contract theories by saying that risk impositions should be accepted to the degree that they are supported by a consensus? Unfortunately, this solution is far from unproblematic. Consent, as conceived in contract theories, is either actual or hypothetical. Actual consent does not seem to be a realistic criterion in a complex society in which everyone performs actions with marginal but additive effects on many people’s lives. According to the criterion of actual consent, you have a veto against me or anyone else who wants to drive a car in the town where you live. Similarly, I have a veto against your use of coal to heat your house, since the emissions contribute to health risks that affect me. In this way we can all block each other, creating a society of stalemates. When all options in a decision are associated with risk, and all parties claim their rights to keep clear of the risks that others want to impose on them, the criterion of actual consent does not seem to be of much help. We are left then with hypothetical consent. However, as the debate following Rawls’s Theory of Justice has shown, there is no single decision-rule for risk and uncertainty that all participants in a hypothetical initial situation can be supposed to adhere to [Hare, 1973; Harsanyi, 1975]. It remains to show that a viable consensus on risk-impositions can be reached among participants who apply different decision-rules in situations of risk and uncertainty.7 Apparently, this has not been done, and hence, contract theory does not either have a solution to the causal dilution problem.

5.6

Restating the problem

The difficulties that we encounter when trying to solve the causal dilution problem within the frameworks of common types of moral theories are indications of a deeper problem. The attempted solutions reviewed above are all based on an implicit derivation principle: It is assumed that given moral appraisals of actions with deterministic outcomes, we can derive moral appraisals of actions whose outcomes are probabilistic mixtures of such deterministic outcomes. In other words, 7 If a unanimous decision is reached due to the fact that everybody applies the same decisionrule, then the problem has not been solved primarily by contract theory but by the underlying theory for individual decision-making.

1096

Sven Ove Hansson

Values of (deterministic) outcomes

Probabilities

Values of actions with uncertain outcome

Figure 1. The standard view of how values of indeterministic options can be determined.

Values of (deterministic) outcomes

Probabilities Rights

Intentions Consent Agency

Values of actions with uncertain

Equity

outcome

Figure 2. A less incomplete picture of the influences on the values of indeterministic options.

Risk and Safety in Technology

1097

it is assumed that probabilities and (deterministic) utilities are all the information that we need (Figure 1). However, this picture is much too simplified. The morally relevant aspects of situations of risk and uncertainty go far beyond the impersonal, free-floating sets of consequences that decision theory operates on. Risks are inextricably connected with interpersonal relationships. They do not just “exist”; they are taken, run, or imposed [Thomson, 1985]. To take just one example, it makes a moral difference if it is one’s own life or that of somebody else that one risks in order to earn a fortune for oneself. Therefore, person-related aspects such as agency, intentionality, consent etc. will have to be taken seriously in any reasonably accurate account of real-life indeterminism (Figure 2). A moral analysis of risk that includes considerations of agency and responsibility will be an analysis more in terms of the verb (to) “risk” than of the noun (a) “risk”. Major policy debates on risks have in part been clashes between the “noun” and the “verb” approach to risk. Proponents of nuclear energy emphasize how small the risks are, whereas opponents question the very act of risking improbable but potentially calamitous accidents. Based on this analysis, the causal dilution problem can be replaced by an defeasance problem that better reflects the moral issues of risk impositions: The defeasance problem: It is a prima facie moral right not to be exposed to risk of negative impact, such as damage to one’s health or one’s property, through the actions of others. What are the conditions under which this right is defeated, so that someone is allowed to expose other persons to risk?

5.7

Solving the defeasance problem

In social practice, the prima facie moral right not to be exposed to risk has to be defeated quite often. Social life would be impossible if we were not allowed to expose each other to certain risks. It is important to observe that a right can be meaningful and socially important even if it can often be defeated. That it is a right means that it prevails whenever there are no overweighing reasons not to realize it. To make this more concrete: as car-drivers we put each other’s lives at risk. However, if we are all allowed to drive a car, exposing each other to certain risks, then we can all lead more mobile lives, and this will on balance be to the benefit of all of us (or so we may assume). The same principle can be applied to exchanges of different types of risks and benefits, as long as these exchanges are mutually beneficial. However, there is (or should be) a limit: No single person should be exposed to risks to an extent or in ways that are to the benefit only of others, not herself. We cannot require that every single risk-exposure be to the risk-exposed person’s benefit, but we can demand that the totality of risk-exposures be arranged so that everyone gains, and no one is exploited. This will lead us to the following solution to the defeasance problem:

1098

(E)

Sven Ove Hansson

Nobody should be exposed to a risk unless it is part of an equitable social system for risk-taking that works to her advantage.

This rule needs, of course, to be specified in several respects, both for theoretical purposes and to make it useful in concrete applications. It should be compared to the dominating approach in risk analysis that can be summarized as follows: (RA)

A risk imposition is acceptable if the total benefits that it gives rise to outweigh the total risks, measured as the probability-weighted disutility of outcomes.

By choosing a rule such as (E), rather than (RA), we change the agenda for discussions on risk. We choose to treat each risk-exposed person as a sovereign individual who has a right to a fair treatment, rather than as a carrier of utilities and disutilities that would have the same worth if they were carried by someone else. In order to argue according to (RA) that it is acceptable to impose a risk on a particular person, one has to give sufficient reasons for accepting the risk as such, as an impersonal entity. According to (E), one instead has to give sufficient reasons for accepting that this particular person is exposed to the risk.

5.8 Hypothetical retrospection With further developments, the approach introduced in the previous subsection can help us to deal with the distributive issues in risk. However, it does not help us to deal with the equally fundamental issue of which risks we should accept. This is a matter that transcends the limit between morality and rational self-interest. Of course the standard answer to this question is that we should apply expected utility theory. However, that theory has several weaknesses, some of which have been highlighted above. Another important weakness is its instability against the actual occurrence of a serious negative event that was included in the calculation. This can be seen by studying the post-accident argumentation after almost any accident. If the expected utility argumentation were followed to the end, then many accidents would be defended as consequences of a maximization of expected utility that is, in toto, beneficial. However, this type of reasoning is very rarely heard in practice. Seldom do we hear a company that was responsible for a deadly accident justify the loss of lives by saying that it was the result of a decision which, in terms of its total effects, produced far more good than harm. Instead, two other types of reactions are common. One of these is to regret one’s shortcomings and agree that one should have done more to prevent the accident. The other is to claim that somebody else was responsible for the accident. It should also be noted that accident investigation boards are instructed to answer the questions “What happened? Why did it happen? How can a similar event be avoided?”, not the question “Was the accident defensible in an expected utility calculation?”. Once a serious accident has happened, the application of expected utility maximization appears much less satisfactory than what it did

Risk and Safety in Technology

1099

before the accident. In this pragmatic sense, expected utility maximization is not a stable strategy. A framework for argumentation is needed that increases our ability to come up with risk decisions that we are capable of defending even if things do not go our way. Such a framework can be obtained by a systematizing a common type of arguments in everyday discussions about future possibilities, namely arguments that refer to how one might in the future come to evaluate the possible actions under consideration. These arguments are often stated in terms of predicted regret: “Do not do that. You may come to regret it.” This type of argument can be systematized into a procedure in which future developments are systematically identified, and decision alternatives are evaluated under each of these possible developments. Such hypothetical retrospection can be used as a means to achieve more well-considered social decisions in issues of risk. However, it cannot be adequately accounted for in terms of regret-avoidance. Psychologically, regret is often unavoidable for the simple reason that it may arise in response to information that was not available at the time of decision. Therefore, regret-avoidance has to be replaced by more carefully carved-out methods and criteria for hypothetical retrospection [Hansson, 2007]. For a simple example, consider a factory owner who has decided to install an expensive fire alarm system in a building that is used only temporarily. When the building is taken out of use, the fire alarm has yet never been activated. The owner may nevertheless consider the decision to install it to have been right, since at the time of decision other possible developments had to be considered in which the alarm would have been life-saving. This argument can be used, not only in actual retrospection, but also, in essentially the same way, in hypothetical retrospection before the decision. Alternatively, suppose that there is a fire in the building. The owner may then regret that he did not install a much more expensive but highly efficient sprinkler system. In spite of this regret he may consider the decision to have been correct since when he made it, he had to consider the alternative, much more probable development in which there was no fire, but the cost of the sprinklers would have made other investments impossible. Of course, this argument can be used in hypothetical retrospection just like the previous one. In this way, when we perform hypothetical retrospection from the perspective of a particular branch of future development, we can refer to each of the alternative branches and use it to develop either counterarguments or supportive arguments. In short, in each branch we can refer to all the others. Hypothetical retrospection can be developed into a precise procedure for collective deliberation on future risks and uncertainties [Godman and Hansson, 2009]. However, it can also be simplified to a risk manager’s version: “Make a decision that you can defend also if an accident happens.” In both cases, hypothetical retrospection aims at ensuring that whatever happens, the decision one makes will be morally acceptable (permissible) from the perspective of actual retrospection. Just as we can improve our moral decisions by considering them from the

1100

Sven Ove Hansson

perspective of other concerned individuals, we can also improve them by considering alternative future perspectives. BIBLIOGRAPHY [Ahteensuu, 2008] M. Ahteensuu. In Dubio Pro Natura? PhD thesis in philosophy, University of Turku 2008. [Beattie et al., 1998] J. Beattie, J. Covey, P. Dolan, L. Hopkins, M. Jones-Lee, G. Loomes, N. Pidgeon, A. Robinson, and A. Spencer. On the Contingent Valuation of Safety and the Safety of Contingent Valuation: Part 1–Caveat Investigator. Journal of Risk and Uncertainty 17:5-25, 1998. [Bollinger et al., 1996] R. E. Bollinger, D.G. Clark, A.M. Dowell III, R.M. Ewbank, D.C. Hendershot, W.K. Lutz, S.I. Meszaros, D.E. Park, and E.D. Wixom. Inherently Safer Chemical Processes – A Life Cycle Approach. Center for Chemical Process Safety of the American Institute of Chemical Engineers, New York, 1996. [Bondi, 1985] H. Bondi. Risk in perspective. In Risk, M.G. Cooper, ed., pp. 8-17, 1985. [Brinkmann et al., 2006] G. Brinkmann. J. Pirson, S. Ehster, M.T. Dominguez, L. Mansani, I. Coe, R.. Moormann, and W. Van der Mheen. Important viewpoints proposed for a safety approach of HTGR reactors in Europe. Final results of the EC-funded HTR-L project. Nuclear Engineering and Design 236:463-474, 2006. [Carlson, 1995] E. Carlson. Consequentialism Reconsidered. Kluwer, 1995. [Christensen-Szalanski and Bushyhead, 1981] J. J. J. Christensen-Szalanski and J.B. Bushyhead. Physicians’ use of probabilistic information in a real clinical setting. Journal of Experimental Psychology: Human Perception and Performance 7:928-935, 1981. [Clausen et al., 2006] J. Clausen, S.O. Hansson, and F. Nilsson. Generalizing the Safety Factor Approach. Reliability Engineering and System Safety 91:964-973, 2006. [Cohen, 1985] B. L. Cohen. Criteria for Technology Acceptability, Risk Analysis 5:1-3, 1985. [Copp, 1987] D. Copp. The Justice and Rationale of Cost-Benefit Analysis. Theory and Decision 23:65-87, 1987. [Ditlevsen, 1994] O. Ditlevsen. Distribution arbitrariness in structural reliability. In Proc. of ICOSSAR’93: Structural Safety & Reliability. G. Schu¨ eller, M. Shinozuka, and J. Yao, eds., pp. 1241-1247, 1994. [Duncan, 2000] J. M. Duncan. Factors of safety and reliability in geotechnical engineering. Journal of Geotechnical and Geoenvironmental Engineering 126:307-316, 2000. [Fischhoff et al., 1981] B. Fischhoff, S. Lichtenstein, P. Slovic, S.L. Derby, and R.L. Keeney. Acceptable Risk. Cambridge University Press, 1981. [Godman and Hansson, 2009] M. Godman and S.O. Hansson. European Public Advice on Nanobiotechnology – Four Convergence Seminars. Nanoethics, 2009. [Hansson, 1993] S. O. Hansson. The false promises of risk analysis. Ratio 6:16-26, 1993. [Hansson, 1999] S. O. Hansson. The Moral Significance of Indetectable Effects. Risk 10:101-108, 1999. [Hansson, 2002] S. O. Hansson. Uncertainties in the knowledge society. Social Science Journal 171:39-46, 2002. [Hansson, 2003] S. O. Hansson. Ethical criteria of risk acceptance. Erkenntnis 59:291-309, 2003. [Hansson, 2004] S. O. Hansson. Weighing Risks and Benefits. Topoi 23:145-152, 2004. [Hansson, 2006a] S. O. Hansson. Economic (ir)rationality in risk analysis. Economics and Philosophy, 22:231-241, 2006. [Hansson, 2006b] S. O. Hansson. Uncertainty and the Ethics of Clinical Trials. Theoretical Medicine and Bioethics 27: 149–167, 2006. [Hansson, 2007] S. O. Hansson. Hypothetical retrospection. Ethical Theory and Moral Practice 10:145-157, 2007. [Hansson, 2008] S. O. Hansson. From the Casino to the Jungle. Synthese, in press, 2008. [Hare, 1973] R. M. Hare. Rawls’s Theory of Justice. American Philosophical Quarterly 23:144155 and 241-252, 1973. [Harsanyi, 1975] J. C. Harsanyi. Can the maximin principle serve as a basis for morality – Critique of Rawls, J theory. American Political Science Review 69(2): 594-606, 1975.

Risk and Safety in Technology

1101

[Heinzerling, 2000] L. Heinzerling. The rights of statistical people. Harvard Environmental Law Review 24: 189-207, 2000. [Heinzerling, 2002] L. Heinzerling. Markets for Arsenic. Georgetown Law Journal 90:2311-2339, 2002. [Hoerl and Fallin, 1974] A. E. Hoerl. and H. K. Fallin. Reliability of Subjective Evaluations in a High Incentive Situation. Journal of the Royal Statistical Society 137, Part 2, 227-230, 1974. [Hynes and Vanmarcke, 1976] M. Hynes and E. Vanmarcke, Reliability of embankment performance predictions. Proceedings of the ASCE Engineering Mechanics Division, Specialty Conference, Waterloo, Ontario, Canada, University of Waterloo Press, 1976. [International Organization for Standardization, 2002] International Organization for Standardization. Risk Management – Vocabulary – Guidelines for use in standards, ISO/IEC Guide 73:2002, 2002. [Khan and Abbasi, 1998] F. I. Khan and S.A. Abbasi. Inherently safer design based on rapid risk analysis. Journal of Loss Prevention in the Process Industries 11:361-372, 1998. [Knoll, 1976] F. Knoll. Commentary on the basic philosophy and recent development of safety margins. Canadian Journal of Civil Engineering 3:409-416, 1976. [Knuth et al., 2003] B. A. Knuth, N.A. Connelly, J. Sheeshka, and J. Patterson. Weighing Health Benefit and Health Risk Information when Consuming Sport-Caught Fish. Risk Analysis 23:1185-1197, 2003. [Lichtenstein et al., 1982] S. Lichtenstein et al. Calibration of probabilities: The state of the art to 1980. In Judgment Under Uncertainty, Heuristics and Biases. Kahneman et al., eds., pp. 306-334, 1982. [Luce and Raiffa, 1957] R. D. Luce and H. Raiffa. Games and Decisions. Introduction and critical survey. Wiley, 1957. [Miller, 1988] C. O. Miller. System Safety. In Human Factors in Aviation. E. L. Wiener and D. C. Nagel, eds., pp. 53-80, Academic Press, 1988. [M¨ oller et al., 2006] N. M¨ oller, S.O. Hansson and M. Peterson. Safety is More Than the Antonym of Risk. Journal of Applied Philosophy 23(4):419-432, 2006. [Moses, 1997] F. Moses. Problems and prospects of reliability-based optimisation. Engineering Structures 19:293-301, 1997. [Murphy and Winkler, 1984] A. H. Murphy and R.L. Winkler. Probability forecasting in meteorology. Journal of the American Statistical Association 79:489-500, 1984. [National Research Council, 1983] National Research Council (NRC) Risk Assessment in the federal government: Managing the process. National Academies Press, 1983. [Nizick, 1974] R. Nozick, R. Anarchy, State, and Utopia. Basic Books 1974. [Otway, 1987] H. Otway. Experts, Risk Communication, and Democracy. Risk Analysis 7:125129, 1987. [Pat´ e-Cornell and Neu, 1985] M. E. Pat´ e-Cornell and J.E. Neu. Warning Systems and Defense Policy: A Reliability Model for the Command and Control of U.S. Nuclear Forces. Risk Analysis 5:121-138, 1985. [Porter et al., 1991] A. L. Porter et al. Forecasting and Management of Technology. John Wiley & Sons, 1991. [Raffensperger and Tickner, 1999] C. Raffensperger and J. Tickner, eds. Protecting Public Health and the Environment: Implementing the Precautionary Principle. Island Press, 1999. [Rahman et al., 2005] M. Rahman, A.-M. Heikkil¨ a, and M. Hurme. Comparison of inherent safety indices in process concept evaluation. Journal of Loss Prevention in the Process Industries 18:327-334, 2005. [Randall, 1976] F. A. Randall. The safety factor of structures in history. Professional Safety, January 1976:12-28. [Rechard, 1999] R. P. Rechard. Historical Relationship Between Performance Assessment for Radioactive Waste Disposal and Other Types of Risk Assessment. Risk Analysis 19(5):763807, 1999. [Rosenberg, 1995] N. Rosenberg. Why Technology Forecasts Often Fail. Futurist, July-August 1995: 16-21. [Royal Society, 1983] Royal Society. Risk Assessment. Report of a Royal Society Study Group. Royal Society, 1983. [Sandin, 1999] P. Sandin. Dimensions of the Precautionary Principle. Human and Ecological Risk Assessment 5: 889-907, 1999.

1102

Sven Ove Hansson

˚. Bergman, I. Brandt, L. Dencker, P. Eriks[Sandin et al., 2004] P. Sandin, B.-E. Bengtsson, A son, L. F¨ orlin, P. Larsson, A. Oskarsson, C. Rud´en, A. S¨ odergren, P. Woin, and S.O. Hansson. Precautionary Defaults – A New Strategy for Chemical Risk Management. Human and Ecological Risk Assessment 10(1):1-18, 2004. [Tench, 1985] W. Tench. Safety is No Accident. Collins, 1985. [Thompson, 1985] J. Thomson. Imposing Risk, in To Breathe Freely. M. Gibson, ed., pp. 124140. Rowman & Allanheld, 1985. [UNCED, 1993] UNCED. The Earth Summit: The United Nations Conference on Environment and Development, Rio De Janeiro 1992. Introduction and commentary by S.P. Johnson. Graham & Trotman, 1993. [Zhu, 1993] T. L. Zhu. A reliability-based safety factor for aircraft composite structures. Computers & Structures 48:745-748, 1993.

TECHNOLOGY ASSESSMENT: CONCEPTS AND METHODS

Armin Grunwald

1

INTRODUCTION

Technology Assessment (TA) constitutes a scientific and societal response to problems at the interface between technology and society. It has emerged against the background of various experiences pertaining to the unintended and often undesirable side effects of science, technology and technicisation which, in modern times, can sometimes assume extreme proportions. The types of challenges that have evolved for TA are these: that of integrating at an early stage in decision-making processes any available knowledge on the side effects, that of supporting the evaluation of the value of technologies and their impact, that of elaborating strategies to deal with the knowledge uncertainties that inevitably arise, and that of contributing to the constructive solving of societal conflicts on technology and problems concerning technological legitimisation. What characterises TA is its specific combination of knowledge production (concerning the development, consequences and conditions for implementing technology), the evaluation of this knowledge from a societal perspective, and the recommendations made to politics and society. TA is thus both interdisciplinary and transdisciplinary and in accordance with its research methods, it can be classified as a “post-normal science” [Funtowicz and Ravetz, 1993] and as one of the forms of “new production of knowledge” [Gibbons et al., 1994]. All the various questions regarding TA concepts, methodology and content are linked to philosophy. In terms of all the normative questions that have a bearing on technological evaluation and technological design, there are close ethics of technology ties [Grunwald, 1999], as well as links with the respective branches of applied ethics (e.g., bioethics, medical ethics, information ethics). Questions on the validity of the available knowledge are relevant to the philosophy of science, especially in conjunction with scientific controversy, the ratio of knowledge to non-knowledge, and the divergent interpretations of the societal implications of scientific knowledge (as currently, for instance, exemplified in neuroscience). Normative and epistemic questions (knowledge and values) are often interwoven, like for instance, when it comes to the application and consequences of the precautionary principle [Harremoes et al., 2002; Schomberg, 2005]. Many TA topics are, Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

1104

Armin Grunwald

furthermore, pertinent to the philosophy of technology or are anthropologically relevant, such as questions regarding the man-machine interface, the substitutability of human beings by robots, the increasing degree to which living beings are being penetrated by technology, or the “technical enhancement” of human beings [Roco and Bainbridge, 2002; Grunwald, 2007a]. An overview of TA is first given (Section 2). TA is introduced in a problemorientated fashion by presenting the societal needs it sets out to address. The historical background to TA is then sketched on the basis of the proposed and realized TA concepts and the spectrum of methods employed in TA. The central TA challenge lies in treating the normative dimensions of technology. An entire section (Section 3) is therefore devoted to this aspect. The final section (Section 4) gives an overview of the current TA developments and of the requirements for the foreseeable future. 2 TECHNOLOGY ASSESSMENT: AN OVERVIEW The term “Technology Assessment” (TA) is the most common collective designation of the systematic methods used to scientifically investigate the conditions for and the consequences of technology and technicising and to denote their societal evaluation. At first sight, entirely heterogeneous activities are subsumed under this name, such as the predicting of the consequences of technology, the communicating of risk, promoting innovation, improving the legitimacy of decisions on technology through increased participation [Joss and Belucci, 2002], mediating in technological conflicts, and observing sustainability. The problem met in defining TA consists in the fact that it is not a priori clear what the common denominator of such heterogeneous efforts should be. No consensual, unambiguous and selective definition of TA has yet been provided. As the emergence and development of TA are closely connected with specific situations arising at the interface between technology and society, these same situations form the central background to the introducing and clarifying of TA.

2.1 The historical origins of technology assessment TA arose from specific historical circumstances in the 1960s and 1970s. The US congressional representative Daddario is now held to be the coiner of the term and of the basic theory underlying TA [Bimber, 1996], which culminated in the creation of the Office of Technology Assessment (OTA) at Congress in 1972 [United States Senate, 1972]. The concrete background consisted in the asymmetrical access to technically and politically relevant information between the USA’s legislative and executive bodies. While the executive, thanks to the official apparatus at its command, was able to draw on practically any amount of information, parliament lagged far behind. This asymmetry was deemed to endanger the — highly important — balance of power between the legislative and the executive facets of technology-related issues. From this point of view the aim of legislative TA was to restore parity [Bimber, 1996].

Technology Assessment: Concepts and Methods

1105

Parallel to this very specific development, radical changes were taking place in intellectual and historical respects, which were to prove pivotal to TA. First and foremost, the optimistic belief in scientific and technical progress, which had predominated in the post Second World War period, came under pressure. The ambivalence of technology was a central theme in both the Critical Theory of the Frankfurt School (Marcuse, Habermas) and in the Western “bourgeois” criticism of technology (Freyer, Schelsky) with its dialectical view of technological progress: “the liberating force of technology — the instrumentalisation of things — turns into a fetter of liberation; the instrumentalisation of man” [Marcuse, 1966, p. 159]. At the same time, broad segments of Western society were deeply unsettled by the “Limits of Growth” [Meadows et al., 1972] which, for the first time, addressed the grave environmental problems perceived as a side effect of technology and technicisation, and by discussions on technical inventions in the military setting forecasting the possibility of a nuclear attack that would put an end to humanity. The optimistic pro-progress assumption that whatever was scientifically and technically new would definitely benefit the individual and society was questioned. As of the 1960s deepened insight into technological ambivalence led to a crisis of orientation in the way society dealt with science and technology. Without this crisis surrounding the optimistic belief in progress, TA would presumably never have developed or, more precisely, would never have extended beyond the modest confines of the above-mentioned US Congressional office. Furthermore, the legitimization problems linked to technologically relevant decisions have been crucial to the genesis of TA. Problems with side effects, the finiteness of resources and new ethical questions have all heightened decision-making complexity and have led to societal conflicts on the legitimacy of technology. The planning and decision-making procedures developed as early as the 1950s in the spirit of planning optimism [Camhis, 1979] turned out to be clearly unsuited to solving this problem. In addition, the technocratic and expertocratic character of these procedures became an issue in a society in which the populace and the media was starting to monitor democracy and transparency more closely [van Gunsteren, 1976]. Demands for a deliberative democracy [Barber, 1984] led to a climate in which it was particularly the critical aspects of scientific and technical progress that started being debated in the public arena. The move away from metaphysical and philosophical assumptions about technology also instigated the emergence of TA, a field that focuses on the criteria and means underscoring the concrete development of technology in concrete historical contexts, the conditions facilitating the malleability of technology in society, and the relevant constraints. In the post-metaphysical world [Habermas, 1988a], it is no longer a matter of humanity’s technology-driven liberation from work constraints (Marx, Bloch) or of humanity’s “salvation” thanks to engineering intervention (Dessauer), neither is it an issue of man’s deplored “one-dimensionality” in a technicised world (Marcuse), of the “antiquatedness of man” in sharp contrast to the technology he has developed (Anders) or of fears of a technologically-induced end to human history [Jonas, 1979]. It is more about the impact of technology

1106

Armin Grunwald

and the concrete design of specific technical innovations, for instance, in transportation, in information technology, in space flight and in medicine. TA does not concern itself with technology as such but rather with concrete technical products, processes, services, systems, and with their societal impacts and relevant general settings.1 These developments are reflected in the philosophy of technology where more emphasis is placed on empirical research [Kroes and Meijers, 1995]. The problems mentioned at the outset about the effect that parliamentary decision-making has on technology only give the “occasion” for the initiation of legislative TA facilities, not the deeper reasons for TA formulation which are rooted in the experience of ambivalence towards technical progress, in problems surrounding technological legitimacy in a society with increasing demands for participation, and in the need to concretise and contextualise technology evaluation in complex decision-making situations. The occurrence of TA is thus one of the very specific descriptors rendering our historical situation one that may be dubbed “reflective modernity” [Beck et al., 1996].

2.2

TA as a response to societal challenges

The social climate of the 1960s and 1970s led to a specific TA requirements profile, which is, to a large extent, still relevant today, though new expectations and requirements continue to emerge. 2.2.1

The mounting implications of science and technology

In the twentieth century, the importance of science and technology in almost all areas of society (touching on economic growth, health, the army, etc.) has grown dramatically. Concomitant with this increased significance, the consequences of science and technology for society and the environment have become increasingly serious. Examples are the increasing intervention in the natural environment as a result of economic activity and man’s increased interference — through scientific and technical progress — in his own social and moral traditions and ultimately also in his own biological constitution [Habermas, 2001]. Technological progress alters social traditions, fixed cultural habits, collective and individual identities and concepts of the self while calling into question traditional ethical norms. Decisions concerning the pursual or abandonment of various technological paths, regulations and innovation programs, new development plans, or the phasing-out of lines of technology often have far-reaching consequences for further development. They can influence competition in relation to economies or careers, trigger or change the direction of flows of raw materials and waste, influence power supplies and longterm security, create acceptance problems, fan the flames of technological conflict, challenge value systems, create new societal “states of mind” and even change 1 This contextualization is occasionally criticized on the grounds that TA delves too deeply into the details of technical development so losing sight of the “broader questions” relating to technology, society and the shaping of the future. In this process also the degree of critical distance needed could be lost.

Technology Assessment: Concepts and Methods

1107

human nature [Roco and Bainbridge, 2002]. New and emerging technologies are not only a means of realizing new technical functions they are also “indicators of the future” [Grunwald, 2006a], on the basis of which society arrives at an understanding of non-technical questions like those relating, for example, to changes in the conceptions of humanity or new societal orders. In this respect, there is close affinity between many TA problems and the great philosophical questions, even if the former concern themselves with the details of technical innovations. The considerably increased influence of science and technology earns such problems more attention both in politics and from the public point of view and they become the subject of critical reflection. This directly concerns technological side effects but increasingly also the entire direction of technological progress. TA has an important function when it comes to discussing and advising, in a knowledgebased and ethically reflective manner, the possibilities and/or necessities of the social shaping of technology [Yoshinaka et al., 2003], establishing informed democratic opinion [Fisher, 1990], creating a knowledge policy [Stehr, 2004], or encouraging sustainable development [Ludwig, 1997; Grunwald and Kopfm¨ uller, 2006].

2.2.2

Side effects and precaution problems

Since the 1960s the adverse effects of scientific and technical innovations have been considerable and some of them were of dramatic proportions: accidents in technical facilities (Chernobyl, Bhopal), threats to the natural environment (air and water pollution, ozone holes, climate change), negative health effects as in the asbestos case, social and cultural side effects (e.g., labour market problems caused by productivity gains) and the intentional abuse of technology (the attacks on the World Trade Centre). This list illustrates why many optimistic expectations relating to future technological progress have currently been abandoned. The rising range of negative effects in time and space, reaching even a “global” technological level, emphasises the relevance of all of this. In part, even the perception of technology has been dominated by a fear of apocalyptic threats to humanity’s continuity (for example, [Jonas, 1984]). Playing down the side effects by referring to them as “the price of technical progress” can cause people to really question the positive aspects of technology. This experience with such unexpected and in some cases serious impacts of technology is central to TA’s motivation. Indeed, in many cases, it would have been desirable to have been warned about the disasters in advance, either to prevent them, or to be in a position to undertake compensatory measures. This explains why the methodologically quite problematic term “early warning” with regard to technological impacts has always had a prominent place in TA discussions from the very beginning [Paschen and Petermann, 1991, p. 26]. The increasing complexity of technical systems, their diverse interlacing, and their connectivity with many areas of society increases the difficulties of being able to predict and consider the consequences of actions or decisions. This applies

1108

Armin Grunwald

on the one hand, for example, to the infrastructure technologies, particularly in the fields of transportation, energy, and water, which are closely allied to habits, consumption patterns, and societal institutions. On the other hand, due to the vast number of interfaces that have to be taken into consideration, the new crosssectional technologies such as nanotechnology tend to broaden the spectrum of the possible side effects that have to be included in decisions concerning these technologies thereby increasing the related uncertainty. This situation leads to a societal and political precautionary problem: how can a society which places its hopes and trust in innovation and progress, and must continue to do so in the future, protect itself from undesirable, possibly disastrous side effects, and how can it preventatively stockpile knowledge to cope with possible future adverse effects? Classic problems of this type are, for example, the use and release of new chemicals — the catastrophic history of asbestos use being a good example [Gee and Greenberg, 2002] —, dealing with genetically modified organisms, or the unknown consequences of the accumulation of non-degradable chemicals in the world’s oceans, especially in the polar regions (for further examples, cf. Harremoes et al. [2002]). In order to be able to cope rationally with these situations of little or no knowledge of certain of the effects of the use of technology, prospective precautionary research and corresponding procedures for societal risk management are required, for instance by implementing the precautionary principle [Schomberg, 2005]. Precautionary problems of this type are a classic field of TA. 2.2.3

The ethical questions of technical progress

For a long time, the question of whether technology had any morally relevant content and could, therefore, be a subject of ethical reflection at all was a controversial topic. Well into the 1990s, technology was held by many, in particular scientists and engineers, to be value free. Since then, the value content of technology has been revealed, and the normative backgrounds of decisions on technology (both in design and in the laboratory) have been recognized in numerous case studies and made the subject of reflection (e.g., [Winner, 1980; Mitcham, 1994; van de Poel, 2001; van Gorp, 2005]). The basis for this development is to view technology less as a set of abstract objects or procedures but more as embedded in societal processes and to take it seriously. Technology is not nature; it does not emerge of its own accord but is instead produced to satisfy goals and purposes. Technology is, then, always already embedded in societal intentions, problem diagnoses and action strategies. Because of the side effects mentioned above, the entire field of ethical questions of risk acceptance and acceptability comes into play. In this sense, there is no such thing as a “pure” technology, divorced from society. It has thus now been acknowledged that technology comprises values and is a legitimate object of responsibility in the normative sense (cf. for example, van Gorp and Grunwald [2007]). The moral criteria employed (that is to say whether something should, would, could, might or must be) clearly differ according to the group

Technology Assessment: Concepts and Methods

1109

concerned, be they manufacturers, operators, users, or those affected directly or indirectly. Tasks requiring ethical reflection present themselves precisely whenever the judgement of various actors leads to diverging results and makes moral conflicts manifest [Grunwald, 2000]. A number of serious ethical questions have been raised, especially as a result of innovations in the modern life sciences, and have also become the subject of public debate. These questions relate particularly to reproductive cloning, reproductive medicine, stem cell research and the “technical enhancement of human beings” [Roco and Bainbridge, 2002]. Nowadays, there is thus hardly any doubt that TA must also inevitably concern itself with normative questions which means that in this way it becomes closely connected to ethics [Grunwald, 1999]. 2.2.4

Technology conflicts and problems of legitimisation

Societal conflicts relating to science and technology are not unusual; they are inherent to any pluralistic society. Answers to questions about the desirability or acceptability of technology, about whether technological risks are acceptable or about where precisely the ethical limits of technology lie are generally controversial due to social pluralism, the differing degrees to which different groups in the modern world are affected by various technological impacts, diverging interests and people’s differing moral convictions. Images of the future, desires and fears, visions and scenarios are also usually contested [Brown et al., 2000]. Conflicts are characteristic of decisions in the field of technology, while consensus tends to constitute the exception. Making decisions in such conflict situations often results in problems of legitimisation because there will be winners (who profit from specific decisions) and losers. This is frequently the case when decisions must be made about the site of a technical facility such as a nuclear power plant, a waste disposal plant or a large chemical production plant. Depending on the selected location, people in the direct neighbourhood will have to accept more disadvantages than others. Problems of legitimisation always surface when the distribution of advantages and disadvantages is unequal. In view of the decades of experience with a number of very serious acceptance problems and certain grave conflicts over technology it has become clear that the question of legitimisation is obviously important. Many examples can be given such as: opposition to nuclear power, the problem of expanding airports, the problem of how to dispose of radioactive waste, the release of genetically modified plants, and regional and local conflicts on waste disposal sites, waste incineration plants, or the location of chemical processing facilities. In these areas, political decisions are frequently not accepted by those affected or by the general public, even though they are the result of democratic decision-making procedures. The differentiation of modern societies, their fragmentation into plural groups with different moral convictions, and the cultural heterogeneity increased by migration and globalization all make it difficult to achieve a general consensus on technology. As demonstrated above by the nuclear technology examples (atomic

1110

Armin Grunwald

reactors, reprocessing plants, the transportation and disposal of radioactive waste), conflicts on technology and its lack of acceptance in society have led to situations which virtually lead to a societal standstill. This is precisely where the danger lies: the escalation of conflict on technology can lead to a hardening of fundamentalist positions which, in turn, can be an obstruction to finding pragmatic solutions to problems and can sometimes almost even lead to civil war. The challenge to society consists in dealing with the conflicts in such a way that the resulting decisions are acknowledged as legitimate, even if they run counter to the interests, values, and preferences of some parties. In particular it is participative TA procedures that try to provide solutions to this problem (Section 2.4.2). The solving of problems allied to legitimisation and technology conflicts is complicated by a certain public mistrust of decisions made by experts that has been growing for decades. Frequently a situation arises in which expertise and counterexpertise conflict thus invalidating in the eyes of the public the expertise of scientific authorities. Scientists are not only — as their traditional self-understanding dictates — incorruptible advocates of objective knowledge, but they are also interested parties in their own cause, lobbyists for external interests, or committed citizens with political convictions, not all of which can always be kept clearly separated from their professional position. In addition to this the political system is perceived to be less and less of a trustee of citizens’ interests, and increasingly interested in its own gain. Methods for solving problems of legitimisation therefore basically involve more frequently integrating non-experts [Fischer, 1990]. New forms of legitimisation (through participative TA, [Joss and Belucci, 2002]; cf. Section 2.4.2 of this contribution) and solutions to specific problems in the communication between experts and non-experts [Bechmann and Hronsky, 2003] therefore belong to the spectrum of TA responsibility. 2.2.5

Economic difficulties and prerequisites for innovation

From the outset, TA has been an aspect of the national innovation system. If, in the initial phase, it was primarily a question of providing an early warning on technological risks, this was not so much done to hamper new technologies as to open up opportunities to avoid or overcome such risks by detecting them early on. The early detection of risks fits into the tradition of deploying the innovation potentials of science and technology as “well” as possible. For this reason, another TA topic that emerged early on was the early detection of technological opportunities so that the best possible use could be made of these benefits and so that the benefits and hazards could also be rationally determined. The search for opportunities and possible innovative applications of technology is an inseparable aspect of TA [Ayres et al., 1970; Smits and Leyten, 1991]. Since the 1990s, new challenges have arisen. In many national economies, serious economic problems have cropped up, which have led to mass unemployment and to the accompanying consequences for the social welfare systems. Increased innovativeness is said to play a key role in solving these problems. On the basis of

Technology Assessment: Concepts and Methods

1111

this analysis, new functions have been ascribed to TA within the scope of innovation research [Smits and Den Hertog, 2007]. Its basic premise is to involve TA in the design of innovative products and processes because innovation research has shown that scientific-technical inventions do not automatically lead to societally relevant and economically profitable innovations. The “supply” from science and technology and the societal “demand” do not always correspond. This means that more attention has to be paid to more pronouncedly orienting towards society’s needs within the scientific-technical system, the diffusion of innovations and the analysis of opportunities and constraints. The theoretical question as to how economic conditions contribute to the success or failure of technical innovations demonstrates that TA takes an active interest in the relevant societal background. Cultural and social questions are also seen as relevant factors for innovations. Including users in technology design, in order to better link technical proposals and consumer demands, should also be mentioned here [Smits and Den Hertog, 2007].

2.3

General characteristics and definition of TA

The above-mentioned facets of the diagnosis of societal developments in the past decades form the problem background against which TA was formulated, and the solution to which it is supposed to contribute. Depending on the context, corresponding societal expectations present themselves in a specific form, and show considerable heterogeneity. In spite of the diversity stimulated by this situation the general characteristics of TA can nonetheless be listed: • Orientation on Advice and Decision-Making: TA supports public opinion and public participation in decisions on science and technology. In this endeavour, it aims at embedding TA knowledge and orientations into the perspective of decision makers: TA knowledge is knowledge for those who are to be advised. Because decisions always affect the future, a reference to the future is always included. TA always functions ex ante with regard to decisions. • Side Effects: In TA, it is a matter of combining “comprehensive” decision support with the widest possible contemplation of the spectrum of foreseeable or presumable effects. Beyond classical decision theory, which establishes the relationship between goals and means according to the viewpoint of efficiency, TA turns its attention to unintentional side effects as a constitutive characteristic [Bechmann et al., 2007]. • Uncertainty and Risk : Orientation to the future and the problems posed by side effects often leads to considerable uncertainty regarding TA knowledge. TA therefore always has to do with providing decision-making support in conjunction with complex innovations under conditions of uncertainty. The impact of such decisions is difficult to predict.

1112

Armin Grunwald

• Value-Relatedness: The rationality of decisions not only depends on knowledge about the systems involved and of the available action-guiding knowledge, but also on the basic normative principles. The disclosure and analysis of the normative positions involved is therefore also an aspect of the TA advisory service (e.g., depending on ethical reflection or sustainability evaluations [Grunwald, 1999; Ludwig, 1997]). • Systemic Approach: TA aims at achieving a comprehensive view of the fields affected. Several perspectives, e.g., from different scientific disciplines, have to be integrated into a coherent picture. Specific attention is dedicated to the systemic interrelationships between the impact of technology in different societal areas. • Broad Understanding of Innovation: TA understands a broad notion linked to the term “innovation”. Beyond the mere technical understanding of innovations as new products or systems, TA contemplates social, political, and institutional innovations and does, in general, also consider socio-technical innovations. • Thinking in Alternatives: When working on concrete projects, TA does not confine itself to a certain technology but always operates in an open window of possible alternatives. Presumed inherent necessities are broken down so that leeway for structuring can be gained. In concrete processes, the question of whether the results desired could not also be realized in a different manner is always posed. Alternative options are thus also examined which are not based on technology, but concern the political planning measures. “Thinking in alternatives” has thus become a specific TA tradition. • Interdisciplinarity and Transdisciplinarity: TA concerns itself with complex societal problems that affect technological decisions and technological side effects. It does this on scientific grounds backed up by research. As a rule, such problems are worked on in an interdisciplinary or transdisciplinary manner. • Time Limitation: Deadlines for completion of the analyses and studies are also inextricably intertwined with the decision-making process. TA knowledge has to be available at certain times regardless of whether all desiderata for comprehensive and reliable knowledge can be satisfied. Without this pragmatic limit, TA’s claim to provide analyses that are as comprehensive possible could lead to never-ending stories. Now that the main characteristics have been listed we shall introduce TA in a problem-orientated fashion according to its societal responsibilities in the providing of specific knowledge and advisory services. We can draw on the existing definition of TA which states that: “Technology Assessment (TA) is a scientific, interactive, and communicative process which aims to contribute to the formation of public and political opinion on societal aspects of science and technology” [Decker and

Technology Assessment: Concepts and Methods

1113

Ladikas, 2004]. This definition stresses that TA contributes to problem-solving, but does not pretend to provide actual solutions. TA provides knowledge, orientation, or procedures on how to cope with certain problems at the interface between technology and society but it is neither able nor legitimized to solve these problems. Only society can do this, through its institutions and its decision-making processes. There is, therefore, a constitutive difference between advising and deciding. The definition given above also includes the attribute “societal” which specifies that the public and political sphere is the place for discussing and dealing with the relevant effects of technological impact. TA is concerned with the aspects of technology that have societal implications. Here the focus of TA dwells in the perspective of unintended side effects. Accidents, environmental impact, unintended side effects on social life (e.g., in employment) and other technological consequences that were neither intended nor anticipated are some of the most important issues in modern times. TA has also been set up as a societal means to enable such situations to be dealt with constructively while making use of scientific research [Bechmann et al., 2007]. Early warning, sustainable development, and the precautionary principle are relevant notions here. An international community has been formed around the concept of TA roughly sketched above [Rader, 2002; Vig and Paschen 1999]. Part of this community works in institutions explicitly devoted to TA (e.g., to provide advice on parliamentary policy) and its organizations (cf., for instance, the European Parliamentary Technology Assessment Network EPTA, www.eptanetwork.org), part of it is organized in networks (cf., e.g., the German-language network TA, www.netzwerk-ta.net), and another part converges on the fringes of disciplinary organizations and conferences, such as in sections of professional sociological or philosophical organizations, or in the social scientific STS Community (Science and Technology Studies), e.g., under the auspices of EASST (the European Association for the Study of Science and Technology), and of many IEEE (Institute of Electrical and Electronics Engineers) activities relating to the social implications of technology.

2.4

Concepts of technology assessment

Fulfilling TA’s above-mentioned responsibilities and satisfying the societal expectations behind those responsibilities requires an operable framework including different facets, typically research concepts, knowledge-dissemination models, task concepts for dedicated TA institutions, or ideas on public discourse and TA’s role within that. TA concepts exist at the uppermost level of TA operationalisation since they reduce the complexity of the entire collection of requirements to the focal points determined in each case. Throughout its history, TA has undergone a series of metamorphoses. Societal trends and research directions such as planning optimism or scepticism, positivism and value orientation, social constructivism and research into the genesis of technology, participation and civil society, loss of confidence in expert decisions, and concepts such as the Risk Society, the Network Society, and the Knowledge Soci-

1114

Armin Grunwald

ety, economic globalization, the discussion on the uncertainty of knowledge, new forms of knowledge production [Gibbons et al., 1994], and the guiding principle of sustainable development have all made their mark on TA. For more than 30 years some complementary, some competing, and some TA concepts that were adapted to varying requirements have been developed in this manner. The concepts of TA presented below are intended to provide the most varied impression possible of TA’s conceptual diversity. 2.4.1

The “classical” concept of TA

The classical concept of TA is an ex post facto construct. It does, in fact, incorporate aspects of the way in which TA was practised during its ”classical” phase in the 1970s, in the Office of Technology Assessment (OTA) at the US Congress [United States Senate, 1972; Bimber, 1996] but in many respects it is a later stylization and not an adequate historical reconstruction. Nonetheless, it is useful to recall the elements of this classical concept, particularly as delimitations and re-orientations can be more clearly described against this backdrop. The following (partially normative) six elements are deemed to be constitutive for the classical conceptualisation of TA: Positivism: TA in the classical sense is dominated by a positivistic understanding of science. It designates a method of producing “exact, comprehensive, and objective information on the technology, in order to facilitate the deciders’ effective societal commitment” [United States Senate, 1972]. In the foreground and elaborated for the deciders’ purposes is the description of what is technologically state- of- the- art and the presumed consequences thereof. Recommendations or independent judgements remain reserved for the political sphere; they are not the domain of TA. The OTA gives “no recommendations, what should be done, but rather...information about what could be done” [Gibbons, 1991, p. 27]. The positivistic legacy of TA that “OTA never takes a stand” [Williamson, 1994, p. 212] is derived from the postulate of science being value free (Weber). The classical concept corresponds in this manner to a decisionist division of labour between positivistic TA and the planning preserve of politics: TA provides purportedly value-free knowledge about technology and the impact of technology while the political system evaluates this knowledge and makes decisions. Etatism: In the classical view TA is exclusively perceived to provide advice to politics. This is founded on the assumption that the state has the authority to direct technology in a societally desired direction: the state can procure the necessary knowledge about impacts; it represents the public interest, as opposed to citizens’ preferences and interests, and it is the central planning authority empowered to actually implement intentions and programs of societal management. This etatist interpretation of the state is characteristic of the period of planning optimism [Camhis, 1979] when TA was established. This fixation on the state in the early phase of TA has since met with harsh criticism (e.g., [van Gunsteren, 1976]) which has motivated the development of more participatory TA approaches.

Technology Assessment: Concepts and Methods

1115

Comprehensiveness: TA in the classical sense aims at comprehensiveness with regard to the consequences of the technology to be studied. The hope is that a complete record of the effects of a technology will help society to avoid unpleasant surprises during its introduction and in the automation of processes. In a certain respect, this assumption is the legacy of former planning optimism. According to this view there must be complete knowledge of all the data on the problem to be decided — and a complete knowledge of all the side effects — in order to eliminate uncertainties. In earlier conceptions, people tried to fulfil this demand for completeness through system analysis [Paschen et al., 1978]: the side effects of technology are often the result of a systemically reticulated process with nonlinear cause-and-effect relationships and interactions which are difficult to discern. Quantification: In this approach there were also great expectations regarding the quantitative apprehensibility of the effects of technology. It was expected that systems theory would, in combination with the quantification of social regularities, prepare quantitative models of causal chains and laws of societal processes and, thus, “objectivise” them. This approach also harboured the expectation that the problem of subjectivity (or of lack of inter-subjectivity) in evaluations of the effects of technology could be solved by means of quantification (see Section 3.2 for limitations on this expectation). Prognosticism: TA in the classical interpretation was seen, above all else, as a prognostic determination of the impact of technology and as an early warning mechanism for technologically caused risks. In analogy to a prognosis based on natural systems, the laws of societal processes were to be discovered and used for quantitative prognoses, which should be as exact as possible. Trend extrapolations and assumptions about laws should therefore make it possible to extrapolate an empirically recorded series of relevant parameters into the future. Such prognostic knowledge should then enable the political system to react appropriately and promptly and, if the situation arose, to take countermeasures against hazards. Orientation towards experts: The classical concept of TA is orientated towards TA experts. They must provide the necessary knowledge and communicate with decision-makers by offering political advice. In contrast to the various models for participative TA (see below), classical TA is deemed to be focussed on experts, hence the coining of the sometimes-used term “expertocratic”. 2.4.2

Participative technology aAssessment

Since the very beginnings of TA, there has been repeated demand for participative orientation, frequently following normative ideas from the fields of deliberative democracy or discourse ethics [Barber, 1984; Habermas, 1988b; Renn and Webler, 1998]. Problems of evaluation were a driving force behind this demand since according to ideas derived from the theory of democracy (e.g. [Barber, 1984]), evaluation should not be left solely to the scientific experts (expertocracy) or to

1116

Armin Grunwald

the political deciders (decisionism). It is the task of participative TA to include societal groups — lobbyists, affected citizens, non-experts, and the public in general — in the process of evaluating technology and its consequences. In this manner, participative TA procedures are deemed to improve the practical and political legitimacy of decisions on technology [Paschen et al., 1978, p. 72]. Such TA is informed by science and experts and, in addition, by people and groups external to science and politics [Joss and Durant, 1995; Joss and Bellucci, 2002]. The demand that those affected participate in decisions on technology has been increasingly put into practice since the1980s, beginning in the smaller, traditionally discursive western and northern European nations, such as Denmark and the Netherlands. Participation has gained particular relevance, on the one hand, in many discussions on technological locating (e.g., airport expansion, waste disposal sites, chemical processing plants, final disposal sites for radioactive waste), in which the widespread NIMBY (Not In My Back Yard) problem leads to participation being emphatically needed [Renn and Webler, 1998]. On the other hand, participation became a constitutive feature of the so-called “Foresight” processes [Martin and Irvine, 1989] in which, for example, the agenda for research policies and for promoting technology, was set or visions for the development of certain regions were formulated [FOREN, 2001]. The participation of citizens and of those affected is believed to improve the knowledge as well as the values on which judgements are based and decisions are made. “Local knowledge”, with which experts and decision-makers are often not familiar, is to be used in order to achieve the broadest possible knowledge base and to substantiate decisions. This discernibly applies especially to local and regional technological problems, in particular, to questions of location. Furthermore, in a deliberative democracy, it is necessary to take the interests and values of all those participating and affected into consideration in the decision-making process. Participation should make it possible for decisions on technology to be accepted by a larger spectrum of society despite divergent normative convictions. In the end, this will also improve the robustness of such decisions and enhance their legitimacy [Joss and Belucci, 2002]. The participation in TA of those affected by technology is designed to improve the legitimacy of ensuing decisions and thus prevent conflict. The expectation is that when those affected have had the opportunity to present their arguments and to weigh them against those of their opponents, they are more likely to acknowledge the resulting decisions as legitimate and accept them, even if such decisions run counter to their own interests. For many, participative TA is also supposed to counteract the political disenchantment observed in many countries and “empower” those affected. The model of representative democracy, which is threatened by emaciation, is confronted here with a civil-societally renewed democracy [Barber, 1984]. These ambitious objectives are, however, hard to realize in practice [Grunwald, 2004b]. Not only representative democracy but also participatory TA is confronted with the problem of representation: only a few people can attend such meetings

Technology Assessment: Concepts and Methods

1117

but they should represent all the relevant groups. The willingness to engage in participatory TA varies according to the population group and correlates with the level of education. The relation between participatory processes and the usual democratic decision-making processes remains an unresolved issue in many countries and this endangers the relevance of participatory TA. 2.4.3

Constructive technology assessment (CTA)

The basic assumption of CTA, which was developed in the Netherlands [Schot, 1992], is that TA meets with difficult problems of implementation and effectiveness whenever it concerns itself with the impacts of a technology after the latter has been developed or is even already in use [Rip et al., 1995]. According to the Collingridge dilemma [Collingridge, 1980], once the impacts are relatively well-known, the chances of influencing them will significantly decrease. It would therefore be more effective to accompany the process of the development of a technology constructively (similarly to the idea of a “real time” TA, cf. [Guston and Sarewitz, 2002]). The origin of technological impact is traced back to the development phase of a technology so that dealing with the consequences of technology becomes a responsibility that already starts in the technology design phase. The theoretical background to CTA is the Social Construction of Technology (SCOT) program, which was also developed in the Netherlands and which has been elaborated in a number of case studies [Bijker et al., 1987; Rip et al., 1995]. According to this approach, the development of technology should be perceived as the result of societal processes of meaning giving and negotiation. Technology is “socially constructed” during these steps. CTA has pleaded for the early and broad participation of societal actors, including key economic players and for the establishment of a learning society that experiments with technology. In the normative respect, CTA builds on a basis of deliberative democracy in which a liberal picture of the state highlights self-organising processes in the marketplace. To this end, three processes have been proposed (according to [Schot and Rip, 1997, p. 257f.]): Technology Forcing: Influencing technological progress through the promotion of research and technology as well as through regulation is how the state can intervene in technology. The options are, however, restricted. CTA therefore also addresses other actors (banks and insurance companies, standards bodies and consumer organizations). Through their business and organizational policy, these institutions can directly intervene in certain technological innovations, for instance, by dispensing with chlorine chemistry, by investing in environmentally compatible manufacturing technology, or by developing social standards that are also valid for branches of a company located in developing nations. Strategic Niche Management: Governmental promotion of innovations should, according to CTA, be concerned with occupying “niches” in technology’s repertory. In these niches publicly sponsored technology can — if protected by subsidies — be developed, make use of processes of learning, gain acceptance, and finally —

1118

Armin Grunwald

it is hoped — maintain its own in free competition unaided by public support. This approach, in which the state directs technology close to the market, is especially relevant in fields reluctant to embark on innovation, such as infrastructure technologies. Successful implementation presupposes considerable learning processes and careful observation of developments either to avoid exposing “niche technology” to competition too early, thereby endangering its growth, or to prevent prolonged subsidies leading it to miss the moment of its marketability. Societal Dialogue on Technology: It is necessary to create the opportunities and structures for critical and open dialogue on technology. In the process, one has to go beyond the limits of scientific discourse and expert workshops to include the economy and the populace. This applies to technology forcing as well as to niche management. “Managing Technology in Society” [Rip et al., 1995] is possible only when these elements harmonise. 2.4.4

Leitbild assessment

In Germany, the concept of empirical technology shaping research developed in parallel with CTA [Dierkes et al., 1992; Weyer et al., 1997]. As in CTA, the paramount objective is to analyse the shaping of technology and its “enculturation” by society instead of reflecting on its impacts. The shaping and diffusion of technology are traced back to social processes of communication, networking and decision-making. TA accordingly consists of research into the social processes which contribute to technological design, analysing the “setscrews” for intervening in these processes and informing decision-makers on these findings. There is, in this concept, almost no further mention of technological impact; it is presumed that the unintended side effects could be completely or largely avoided by improving the process of technology shaping. Leitbild assessment [Dierkes et al., 1992] has made it clear that technology development often follows non-technological ideals. Leitbilder (“guiding visions”, cf. Grin [2000]) are often phrased in metaphors which are shared, implicitly or explicitly, by the relevant actors (e.g., the “paperless office”, “warfare without bloodshed”, or the “automobile city”). Research into such ideals has investigated in detail, empirically and hermeneutically, which mechanisms dominate this development, including linguistic analysis of the use of metaphors in engineering circles [Mambrey and Tepper, 2000]. The expectation is that through societal construction of the ideals shaping it, technology can be indirectly influenced in order to prevent any negative effects. These deliberations have led to a wealth of instructive case studies [Weyer et al., 1997], but they have not really been integrated into TA practice. The reason probably lies in the fact that strong assumptions are necessary for the transfer of knowledge gained ex post in case studies on TA problems, which are always inevitably concerned with the future. Leitbild assessment is a way of explaining the course of technology development ex post rather than by giving indications on

Technology Assessment: Concepts and Methods

1119

how to shape technology. Moreover, the sociological perspective has resulted in the neglecting of the normative dimension of technological shaping. 2.4.5

Innovation-orientated technology assessment

Embedding technology in society takes place by means of innovation. There are thus overlaps between TA and innovation research and in recent years the two fields have developed “innovation-orientated” TA concepts at their interface [Smits and Den Hertog, 2007]. Innovation research focuses on the analysis of completed and current innovation processes and is primarily interested in factors that are crucial to successful market penetration. Factors enabling and preventing innovative success are identified. The objective is to attain a better understanding of innovation processes and their influencing factors. With this knowledge, governmental research and technological policies, as well as industrial decisions on innovations can be supported. In this respect TA first contributes, by broadening the spectrum of influencing factors, by adding social and cultural elements. TA then examines — analogous to participative TA — the role of the users in innovation processes. In innovationorientated TA, a special role is assigned to the users whenever customer-orientated and social technology designs are at stake. In order to realize this objective, the users must be included in the early phases of technology development [Smits and Den Hertog, 2007]. The classical instruments of market research are inadequate for this purpose. Instead, users have to be integrated into deliberative and prospective processes of technology prognosis (foresight). In this respect, they can play very different roles. “Users can play a role as more or less active consumers, and modifiers, as domesticators, as designers, and, in fact, also as opponents of technological innovation. . . . High quality user-producer relations as well as possibilities for learning and experimenting are prerequisites for successful innovation processes” [Smits and Den Hertog, 2007, p. 49]. To this end one important function for TA is to identify the relevant actors in a certain field, to inform them and then, most importantly, to use discursive procedures to establish the users’ needs, visions, interests and values. It is then a question of integrating these findings into the process of technology development. Innovation-orientated TA should thus contribute to making the regional or national innovation systems more strongly orientated towards citizens’ and consumers’ needs [Smits and Den Hertog, 2007]. 2.4.6

Technology assessment and engineering ethics

In the engineering sciences, the challenges with which TA is confronted have been discussed as demands on the profession of engineers. The value dimension of technology has been shown in many case studies, especially in engineering design processes ([van de Poel, 2001; van Gorp, 2005]; cf. also the chapter on values and design by Ibo van de Poel, this Handbook, Part V). Decisions on technology design involve value judgements. There is, in other words, a close relationship between professional engineering ethics and the ethics of technology [Mitcham, 1994]. By

1120

Armin Grunwald

way of example, one can cite VDI guideline no. 3780 of the Association of German Engineers [VDI, 1991], which has become relatively widespread. It envisages a “Guide to Technology Assessment According to Individual and Social Ethical Aspects”. For engineers and in industry, assessments are to a certain extent part of their daily work. Evaluations play a central role whenever, for instance, a line of technology is judged to be promising or to lead to a dead end; whenever the chances for future products are assessed; whenever a choice between competing materials is made; or whenever a new production method is introduced to a company. Though evaluation may be commonplace in daily engineering practice, what is essentially new in this guideline for societal technological evaluation is its scope, which also includes the societally relevant dimensions of impacts as well as technical and economic factors. Technological evaluation should be conducted in line with societally acknowledged values. Eight central values forming the VDI “Value Octagon” have been identified: functional reliability, economic efficiency, prosperity, safety, health, environmental quality, personality development and social quality [VDI, 1991]. These values are thought to influence technical action and fall under the premiss [VDI, 1991, p. 7]: “It should be the objective of all technical action ... to secure and to improve human possibilities in life.” They are involved in technology development when observed by engineers in practice, that is to say, they are virtually built into the technology. Engineers or scientists should, on the basis of their knowledge and abilities, point the development of technology in the “right” direction by observing these values and avoiding undesirable developments. If this exceeds their authority or competence, engineers should take part in the corresponding procedures of technology evaluation.

2.5 Methods in TA Methods assume a central function in TA to fulfil its responsibilities in research, assessment or advice. The guaranteeing of the transparency, comprehensibility, and inter-subjectivity of TA results is primarily ensured, as in the classical scientific disciplines, by the ability to follow the materialisation of the results step by step as the method proceeds. The use of methods is closely allied to TA’s observance of quality standards [Decker and Ladikas, 2004]. TA requires specific methods or method sets which are tailored to the relevant assignments, backgrounds and actor constellations. In TA methods can be used to collect data, provide knowledge, organize TA-relevant communication, gain ideas on conflict management, uncover the normative structure of technology conflicts, establish scenarios on future developments or assess value structures. In order to operationalise TA activities in specific projects, a set of methods is available in the form of a “method toolbox” (see Decker and Ladikas [2004]). A first step in designing a TA project is to select appropriate methods and clarify their integration in a coherent mix relevant to the overall project goals and the specific environment. Often the specific goals of a TA project can only be attained by combining different methods or adopting new ones. The needs and expectations

Technology Assessment: Concepts and Methods

1121

of the respective beneficiaries will, of course, influence the set of methods chosen, because TA knowledge has to be “customised”. The project design also takes general TA quality criteria into account such as scientific reliability or interactive fairness. The project design is influenced by the institutional setting, the mission of the institution, its tradition or history and its formal status. Careful “situation appreciation” must therefore be carried out in advance to identify which methods are appropriate [B¨ utschi et al., 2004]. The methods applied in TA are research methods, interactive methods and communication methods [Decker and Ladikas, 2004]. Research methods are developed in disciplines pertaining to the sciences and humanities. They are applied to TA problems in order to collect data, to facilitate predictions, to do quantitative risk assessment, to allow for the identification of economic consequences, to investigate social values or acceptance problems and to do eco-balancing. This class of methods includes (1) modelling, systems analysis, risk analysis (cf. Section 3.3.1), material flow analysis (cf. Section 3.3.3) (to understand the sociotechnical system being investigated as well as to be able to assess the impacts of the political measures proposed); (2) trend extrapolation, simulation, scenario building (to create systematic knowledge in order to contemplate the future); (3) the Delphi method (to gather expert knowledge, especially on the assessment of future developments in science and technology); (4) expert interviews and expert discussion (to gain more insight into current situations but also to analyse scientific controversy and diverging assessment with respect to the arguments used); (5) discourse analysis, values research, ethics, and value tree analysis (for the evaluating and revealing of the argumentative landscape in normative respects). Interactive, participatory or dialogue methods are developed to organise social interaction in such a way as to facilitate conflict management, allow for conflict resolution, bring scientific expertise and citizens together, involve stakeholders in decision-making processes and mobilise citizens to shape society’s future. This class of methods includes (1) consensus conferences (to involve citizens in societal debate on science and technology in a systematic manner, according to a specific framework (cf. Sect. 3.3.5)); (2) expert hearings (to inform the public but also to confront experts with laymen’s views and with diverging expert judgement); (3) focus groups (to gain coherent views on a specific topic from a set of actors and citizens); (4) citizens’ juries (to assess measures and planning ideas with respect to the values and interests of stakeholders and interested parties); (5) scenario workshops and perspective workshops (to create drafts of the future in an interactive way). Communication should be seen as a two-way process. On the one hand communication methods are used to communicate the corporate image of a TA institute, the TA approach, the TA process and the TA product to the outside world so as to increase the impact of TA. On the other hand communication is important for enabling the TA institute to keep in touch with the outside world and thus

1122

Armin Grunwald

with reality. This class of methods includes (1) newsletter and focus magazines, perhaps including opinion articles (for creating awareness and pointing out critical issues); (2) science theatre and video presentations (to illustrate possible science and technology impacts on future society and everyday life); (3) websites, local questionnaires or debate forums (to facilitate or strengthen interactive communication at informal level); (4) networking and dialogue conferences (to promote mutual exchange and the distribution of the new ideas and issues to be considered). 3

NORMATIVITY AND VALUE ISSUES IN TA

It is no longer a point of controvery that technology development needs normative orientation because values and normative judgements enter into technological design (cf., for example, van de Poel [2001] and van Gorp [2005]) and technology development at many stages of the process thus determining the eventual societal implications of technology to a considerable extent. Normative judgements on technical options, technological impacts, or innovation potential with regard to societal desirability or acceptability are some of the many decisions which have to be made during technology development. Analysing such normative questions of technology and giving advice to society are some of the responsibilities of TA. However, the specific problems related to this type of advice must be carefully observed [Grunwald, 2003].

3.1 Normative judgements in TA practice The prospective assessment of technological impacts is an important part of TA projects where normative and evaluative considerations play a role but not the only role. These considerations also accompany TA processes in the definition phase, in the implementation and in impact assessment: Definition of the Task : TA topics do not arise “by themselves”. Many questions on technology and automation could be asked in various ways, e.g., from economic or social, cultural or political, or even environmental or psychological perspectives. Stem-cell research can be addressed from the medical angle of curing Alzheimer’s disease, or can be seen as a moral breach in the dike, gene therapy can be seen as a therapeutic instrument, or as a step towards a new form of eugenics, whatever the approach each uncovers completely different horizons of treatment and possible answers. The definition of the task is connected to a corresponding perception of the problem (e.g., with respect to the anticipated side effects). It is all bound up with priorities, perspectives, values, actors’ interests and occasionally there might even be a desire to conceal certain questions. It is relevant to see who defines the problem, which people, groups and societal subsystems are involved, and what interests they pursue. Topic determination is the result of evaluation and it is, therefore, politically relevant. For that reason, the participation of those affected and of “stakeholders” in the definition, description, and structuring of the problem must be taken into consideration, more to the point it is even absolutely

Technology Assessment: Concepts and Methods

1123

necessary in order to avoid coming up with answers which are completely irrelevant in social terms. At this point, tension occasionally arises between the scientific independence requirements of TA institutions [Grunwald, 2006b] and the topic determination dependency of clients, for example, parliaments [Vig and Paschen, 1999]. One of the responsibilities of TA is also to be critical of mainstream problem formulations and, in particular, to draw attention to aspects which have been neglected so far. Delimitation of the System: Since it is impossible to completely investigate the entire spectrum of technological impacts or the consequences and implications of a technology, the contours of a concrete TA project must be determined in detail. Before beginning a TA study, one has to decide what is of cognitive interest and what can be left out. This concerns, on the one hand, converting the subject in hand into a series of detailed questions and, on the other hand, demarcating the limits of the system to be examined in spatial, temporal and thematic terms. Taking the example of life cycle assessment (LCA, Section 3.3.3), the significance of this delimitation can be immediately seen. Even for a simple technical product, the chain of preliminary products and processes can take on quite considerable proportions and this is even more so with complex products, such as a washing machine or an automobile. In view of the limits of temporal and financial resources, decisions have to be made as to how far one wants to retrace the manufacturing chain, and which processes or material flows can be rejected as irrelevant. When this sort of decision is made, disputes often arise concerning the matter of the extent to which these system demarcations prejudice the subsequent results. Decisions of this type are decisions on relevance and the importance of the problem in hand. In terms of method they are, therefore, evaluations. Thematic demarcations of knowledge interest have an effect on the choice of scientific disciplines, and possibly also of the societal groups that are invited to participate. This is how the areas of knowledge, ranges of values and interests taken into consideration are determined — and these, too, are normative decisions about what is relevant and what is not [Decker and Ladikas, 2004; cf. Section 2.6.1]. TA has to determine what interaction or aspects of the area of study are relevant to analysis and to finding a solution. This is done according to the normative evaluation criteria used to distinguish important aspects from unimportant aspects and is often controversial. What is important for one actor may be unimportant or even detrimental to others. There is a risk involved in making such relevance judgements: they could later turn out to be unjustified. It could transpire that despite all the care taken important aspects are “forgotten” or fail to be adequately assessed. This normative dimension in the initiating phase of TA projects and processes is precarious because it often crucially and irreversibly influences further stages. Normative Aspects of the Methodical Approach: Certain TA project methods are not based exclusively on means-end rationality, that is to say, their likelihood of attaining the relevant aims pursued. Instead, normative considerations also come into play. By choosing quantitative methods, for example, one also accepts cer-

1124

Armin Grunwald

tain (normative) quantification rules. Although in many cases this may not be a problem, like in the quantitative recording of emissions by power plants, in other areas quantifications can be ethically questionable (Section 3.2.4). When choosing methods it is thus imperative to consider the relevant normative presuppositions. For example: are they adequate in that context, and accepted by those involved? This is analogous to scientific modelling which always involves normative preconceptions. For instance, in neoclassical economics we have the common concept of a homo oeconomicus, whose knowledge is comprehensive and who makes his decisions according to utility maximization. As soon as models are used in TA one therefore has to inquire into their normative assumptions, their adequacy, and their acceptability in the context in question. Evaluation of the State of Knowledge: Comprehending and evaluating the level of knowledge on the technology in question as well as establishing its operating conditions and foreseeable consequences is an integral part of TA. This is no trivial matter involving the mere gathering of available knowledge but rather an activity with its own normative challenges. First of all, there is usually no consensus on the acknowledged “status of knowledge” regarding a certain issue. Because the knowledge required for TA is not generally textbook knowledge but rather knowledge that has to be sought at the cutting edge of research there is often no consensus within the relevant scientific communities. Instead scientific controversies tend to become the order of the day. These may consist of different estimations of the reliability of certain stocks of knowledge; they may also derive from divergent opinions on the significance of these stocks of knowledge within the context of the particular TA problem in question. The interdisciplinary nature of TA knowledge complicates these judgements. Knowledge assessment thus forms an independent step in TA processes [Pereira et al., 2007]. The constituents of the “status of knowledge” which can be established as a consensus have to be determined and the scientific controversies have to be more closely investigated, both with regard to their epistemological and their normative origins. For this reason, the reflexive dimension of rationality [Decker and Grunwald, 2001] requires us, on the one hand, to reveal the uncertainties and controversies connected with the available knowledge. On the other hand, the difficulties that hinder the clear determination of the limits of knowledge in consensus have to be made transparent. TA includes, in this sense, epistemological considerations: the epistemological status of the stocks of knowledge used must be clear in order to preclude one-sided, exaggerated or arbitrary conclusions being drawn on the basis of knowledge which does not epistemologically support it. Prospective Evaluation of the Impact of Technology: The evaluation of the possible consequences of a technology is in itself the most prominent and most often discussed point and the stage when TA problems of evaluation arise. This relates to challenges such as the assessment of risks, the appraisal of expectations concerning benefits and often the need to weigh up the facts. The following types of assessment situations are common in TA practice:

Technology Assessment: Concepts and Methods

1125

• The consequences of a technology can be judged in relation to the technologically, societally, or politically determined and legitimised goals pursued. Wherever there are politically determined objectives there is always a clear normative basis for evaluation (which can naturally be questioned on a different level). In numerous sustainability strategies there are, for example, political target values (e.g., with regard to CO2 emissions) that can be used as evaluation criteria. • The evaluation of the effects of technology can include a study of the attainment of goals from the viewpoint of efficiency. Are there other ways of achieving the same goals with fewer side effects, fewer risks and at lower costs, etc.? • Such an evaluation can concern itself with the acceptance or acceptability of side effects. In this case, even the general rejection of a technology can be a topic, notwithstanding expectations surrounding possible benefits (as is often the case with genetically modified organisms). It could alternatively concern proposals for a moratorium or (as is more frequently the case) comparing the side effects that have to be accepted and the expected benefits. In any case, TA’s claim to transparency and comprehensibility makes it obligatory to disclose the respective assessment criteria (see Section 3.2.1). In that way citizens, politicians, or stakeholders can compare the premises of TA’s conclusions with their own values and either accept (for well-founded reasons), modify or reject them. This increases the transparency of the public debate because positions are established and conclusions are drawn in relation to the underlying premises and values.

3.2

Methodological challenges

TA’s methodological orientation aims to make it possible, even in the field of evaluations, to provide for the greatest possible amount of rationality, transparency and inter-subjectivity. The results of TA have to be protected from ideological suspicions and from being accused of being particularist or arbitrary. In this way specific methodological problems emerge, including the question of whether “objective” normative conclusions can be justified in the first place [Grunwald, 2003]. 3.2.1

The origin of normative criteria

Normative criteria are required to evaluate all the fields mentioned above. These can be derived and justified in conceptually divergent ways: • Decisionism: In the view governed by a strict “division of labour”, the normativity needed for societally relevant decisions is created directly and immediately through the political system [Schmitt, 1934]. It is therefore superfluous to advise political bodies. Such possible advising has to limit itself to

1126

Armin Grunwald

a representation of the factual situation and to the provision of descriptive knowledge. This position (cf. Section 2.4.1) will not be explored any further in the following sections because it is held to be obsolete for the reasons explained above. • Values Research: With social-scientific methods, the values currently prevailing in society can be empirically investigated (for the risk case see Slovic [1993]). These empirical results can then be used by politicians or engineers as a normative basis for technologically relevant decisions to pursue technological design in accordance with citizens’ values. • Participation: Evaluative criteria can be negotiated directly with those affected. Through participative procedures (cf. Sections 2.4.2 and 3.3), citizens with their values, preferences and interests, can be directly involved in the constitution of the evaluation criteria [Joss and Belucci, 2002]. • Philosophical Ethics: Normative ethics attempts to derive the criteria for judging alternative technical options from universal principles by taking, for instance, the categorical imperative or the utility maximization rule [Ferr´e, 1995; Mitcham, 1994; Beauchamps, 2001]. Precisely which of these approaches to including normative considerations should be brought into play remains controversial (e.g., Grunwald [1999]). The question of where the evaluative criteria should come from and how it can be justified leads to fundamental controversy between the normative approach of philosophical ethics and the empirical approach of social scientific values research. While ethics warns against a “naturalistic fallacy” [Moore 1905] and rejects the idea that an “ought” can be derived from an empirically observed “is”, values research investigates the values represented empirically in society and sets out to derive orientation from exactly those empirical observations. In this field of tension participation can be employed in various ways: participative procedures can be “informed” by research into values and by philosophical ethics. Procedures can alternatively be understood to be the implementation of discourse ethics. Discourse ethics and deliberative democracy [Habermas, 1988b] have been taken as a model for participatory TA [Renn and Webler, 1998]. With such an approach, no substantial values about acceptable or unacceptable technologies are assumed to exist but the recourse to discourse ethics suggests the presence of normative criteria indicating how the participation procedures should be organised. It is, for example, required that the processes be fair and transparent, that the participants commit themselves to providing arguments instead of to merely trying to persuade their opponents and that they are willing to question and to modify their own positions if there are good counterarguments. In this way, discourse ethics can offer orientation on the organisation of a “good” and just participative procedure [Skorupinski and Ott, 2000]. At present, the relationship between the descriptive approaches of values research and the normative approaches of philosophical ethics are held to be pre-

Technology Assessment: Concepts and Methods

1127

dominantly complementary and cooperative by many TA practitioners as well as by many practitioners in Applied Ethics. Accordingly, there are TA problems where one can derive the evaluative criteria empirically and other TA problems where that is not possible. The question that then arises is: When, in TA, is there an explicit need for reflection and when is empirical research sufficient? The answer to this question obviously depends on one’s understanding of ethics. Inasmuch as ethics is seen as a discipline that reflects on empirically existing moral conceptions and so relevant at the precise moment when conflicts arise between divergent moral conceptions [Grunwald, 2000], the decisive criterion becomes whether a moral conflict has to be dealt with or not in a given TA project. The following requirements have been proposed to operationalise this abstract criterion [Grunwald, 2005]: pragmatic completeness (the current normative framework has to cover all normative aspects of the decision to be made); local consistency (there must be a “sufficient” degree of consistency between the normative framework’s elements); non-ambiguity (between the relevant actors there must be sufficient agreement on the interpretation of the normative framework); acceptance (the normative framework must be accepted by those affected as the basis for the decision); and compliance (the normative framework has to be complied with in practice). If all these conditions are fulfilled there is neither moral conflict nor moral ambiguity and so there is no need for ethical reflection. The normative framework can be used by TA as a basis for normative evaluation without the need for further ethical reflection. In such situations, it is possible to carry out virtually descriptive TA, in which the normativity that has to be considered is not in itself an object of reflection but rather something that is gathered empirically from the prevailing political circumstances. This is especially true of standard design process situations [van Gorp, 2005]. It becomes problematic as soon as the scope of these criteria is transgressed. It is a serious challenge to TA to recognize this point at all. To do this, there must be corresponding “awareness” of and competence in making ethical judgements. 3.2.2

The possibility to generalise on evaluative judgements

In its advisory capacity to society and to politics TA operates in the public sphere and must work towards results that are valid beyond a subjective or particular level. The question is whether, to what extent, and under what circumstances assessments of technological impacts can be generalised. Can TA support judgements in a generalisable way, and in what methodologically secure manner can that be done? Can the evaluative aspects of TA just be left to societal negotiation processes and do they, therefore, depend on power differences? It is first of all indisputably true that TA cannot posit that normative postulates or societal values are valid, nor declare them to be binding. TA cannot, accordingly, substantially decide whether the development and use of a technology is acceptable, desirable or even imperative. TA can only concern itself conditionally with certain normative principles in order to propose methodologically secure conclusions on this basis.

1128

Armin Grunwald

It can propose “if-then” statements in the following syntax: “If one applies certain normative criteria, then this has the following consequences or implications for this technological issue ...”. The “if” antecedents cannot be declared valid by TA: that is society’s responsibility in its legitimized procedures and institutions, informed and orientated in a normative fashion through ethical deliberation and consultation [Grunwald, 2003]. It is this conditionally normative structure of evaluations that makes general and intersubjective statements on assessment problems possible. It makes it clear that evaluations are not simply a function of individual, subjective normative decisions but that there are in fact possibilities for scientific (generalised) evaluations. It is, thus, the task of TA to make this structure transparent and comprehensible. A political evaluation or decision is by no means anticipated or even obviated by this; it is still the responsibility of political or other societal opinion-forming and negotiating procedures to decide on the validity of the “if” clause. The “if-then” nexus must, however, be acknowledged as a scientific proposition that is accessible to scientific cognitive interest and to scientific method. In this manner, TA can contribute to not leaving the elaboration of the normative aspects of the evaluation basis to chance – in other words, to random constellations of actors or power relationships – but to rather improving the comprehensibility and transparency of societally relevant evaluations through systematic critical appraisal and through conditionally normative judgements [Grunwald, 2003]. 3.2.3

Multidimensional integration

The choice of technical solutions usually depends on a number of criteria (cf., for example, Section 2.5.6). These criteria, such as risks, costs or environmental aspects, are generally rather heterogeneous and in part incommensurable (cf. Ibo van de Poel’s chapter in this Handbook). Depending on the facts of the case, they carry varying weight when it comes to arriving at an overall evaluation and they can conflict. One particular challenge is, therefore, that of aggregating the evaluations according to specific criteria in order to provide a comprehensive evaluation that can form the basis of a decision. It is often impossible to achieve this by projecting the criteria onto a uniform quantitative scale (of, for instance, monetary values) in order to solve the problem by, for example, quantitatively maximising utility. In this way, conflicts on technology, problems of legitimisation and the inherent normative problems would merely be concealed in the underlying quantification procedures (Section 3.2.4). TA studies on sustainability aspects are especially challenging [Ludwig, 1997]. They are carried out with the help of life cycle assessments (LCA; cf. Section 3.3.3) in all the pertinent dimensions of sustainability: ecologically, economically and socially. Over the course of a life cycle, for instance, in the extraction of raw materials, transportation, processing, use and disposal, a wealth of diverse and incommensurable aspects relevant to sustainability come into play. A sustainability assessment would have to provide a complete balance of these very heterogeneous

Technology Assessment: Concepts and Methods

1129

factors. It would be an integration problem of considerable complexity. There is an extreme risk of providing arbitrary results in these multidimensional integrations, because methodologically secure integration is hardly possible in such a “jungle” of heterogeneous and possibly contradictory evaluations. 3.2.4

Limitations of quantitative methods

In many fields of science, quantification is the means chosen to make objective statements possible. Inasmuch as an acknowledged normative measurement theory and a correspondingly acknowledged quantification method exist “quantitative” can, under such conditions, be equated with “objective”. In the social sciences similar hopes are to some extent founded on methods of empirical social research. In this area such expectations also lead to criticism and to allusions to the fact that only selective knowledge of societal phenomena can be gained through quantification. Reference is then made to the dimensions of meaning, communication and understanding, etc. that resist quantitative compilation. Quantification is very popular among politicians and in public administration. These actors hope that quantification will enable the subjective questions of evaluation to be “objectivised”. The availability of numerical values serving, for instance, as evaluative notches on a ranking scale not only facilitates a practical approach to problems of evaluation but it also suggests a kind of objectivity: the evaluation is reduced to a mathematical operation. Criticism arises when one queries the actual significance of these “objective” statements. These quantified evaluation procedures are only objective and adequate under the condition that the “measurement rules” and the method of calculation of the evaluative figures are acknowledged as methods by those involved. In TA very diverse parameters are quantified. These include, on a level still very close to technology, the emissions of technical processes into various environmental areas (water, soil and air). In questions of evaluation, economic (monetary) quantifications of the expected benefits or detriment and, using the quantitative version of the risk concept, the probability of a possible adverse occurrence are some of the most frequently quantified dimensions. However, the degree of acceptance of or resistance to technologies in the population, or other representative survey results is also quantified, as are the results of Delphi-sample surveys. There are limitations to quantitative analysis though like, for instance, when data is not available or quantifying measures are disputed. The latter is encountered particularly frequently, and not just in compilations and evaluations of the social and cultural consequences of technology. Even quantifications of the effects of technology on the natural environment, for example, in the form of monetary values for damaged natural capital are controversial because the utility of such external effects is not estimated by means of a market-like supply-and-demand mechanism but only through market simulations, for example by the “willingness to pay” approach. Examples of such problems are questions concerning the monetary value of a rare species of toad or of a songbird in comparison to the expected

1130

Armin Grunwald

economic utility of building a road through their biotope. The value of subjective well-being or of an aesthetically pleasing landscape destroyed by the construction of an industrial park can also only be quantified with reservations, or not at all. A number of evaluation methods have been developed which, despite these problems, arrive at quantifications by using some dubious substitutive considerations. One of these is inquiry into the willingness to pay on the part of the persons affected. Those potentially affected are for instance asked in view of the possible loss of an aesthetically valuable landscape how much they would be willing to pay in order to preserve that landscape. On the basis of that method, the personal preferences of those concerned are transformed into monetary values. Methods of this type are controversial when compared to the methods employed in physics or chemistry. Attributing a monetary or utilitarian value to an impact of technology (to a benefit or damage) is not free of political and ethical questions (cf. van de Poel, this Handbook). The basis of quantifications in theoretical measurements is inseparable from preferences, values, norms, and their changes over the course of time, and this is what differentiates all social domains, not only economics, from the domain of the natural sciences. In the social domain quantifications are dependent on the normative assumptions that enter into the method of quantification. This is why in the field of technological impact quantification remains controversial and does not just simply supply the expected “objective” facts of the case. This is especially drastic when, for example, in the economic modelling of the effects of climate change monetary values taken from calculations in the insurance business are assigned to human lives. A quantitative assessment of human life and of the quality of human life obviously meets with ethical objections. These limitations do not render quantification obsolete in TA. In many cases, quantitative approaches are absolutely crucial to the development of assertions that will stand up to debate. In life cycle analysis (Section 3.3.3), quantification is conducted to assess the environmental impact of technology. This is vital to achieving an overall balance when faced with effects that to some extent compete. In appraising risk (Section 3.3.1), quantitative risk analysis is also often very helpful. Despite the problems already mentioned, the result of quantification is often beneficial but this does not mean that the results are acknowledged as objective by all parties. For example, the debate on the better environmental compatibility of non-returnable as opposed to returnable packaging cannot be decided on the basis of quantitative analysis: instead the dispute switches to how one could adequately quantify and how the limits of the system could be determined (Section 3.1.2). If the results of a quantitative evaluation in a technological conflict are not accepted by a given party, it is often not difficult to attack the quantification rules. The results of evaluations are dependent on the quantification methods chosen. For this reason, the normative aspects of quantification methods must be made transparent. Only then can the results of quantitative evaluations be interpreted appropriately and linked to qualitative content. Quantitative evaluations do not stand up “objectively” on their own. In TA they often depend on the manner

Technology Assessment: Concepts and Methods

1131

of quantification. Therefore they must be integrated into a transparent frame of interpretation and deliberation.

3.3

Assessment methods

For reasons of inter-subjective comprehensibility and transparency, TA evaluation must be conducted in a methodologically well-substantiated way. In individual cases this can be achieved through chains of painstaking justification like, for example, with the argumentation for a certain interpretation of a system’s limits under relevant conditions. However, when evaluating technological impact, there are scientific methods that have been developed for the further “objectivisation” of evaluation in the political and public spheres in which TA operates. In the following section some relevant methods are briefly presented in which the specific focus lies on the discussion of the normative aspects of these methods. 3.3.1

Risk assessment

One of the main reasons for the emergence of TA was because of the risks directly or indirectly caused by technology and its use. Any decisions made on technology are also simultaneously decisions about risks and they are, therefore, dependent on ex ante estimations of these risks and on a readiness to accept them. The mere fact that in the present we take decisions on future hazards and living conditions testifies to the considerable relevance of this subject while revealing its societal sensitivity. TA should and does contribute to the early signalling of risks and to how they should be dealt with (Section 2.3.2). In this respect, TA embraces elements of an “early warning” system. Dealing with technological risks has always been a facet of the development of technology. In order to meet safety standards and, for instance, to obtain public licensing, some proof has to be submitted. Technical risk analysis and risk evaluation methods were therefore developed. When risk is interpreted as the product of the probability of damage (i.e., the probability of the occurrence of an accident) and the extent of damage (expressed as a rule in monetary units), the assumption is that risk can be quantified and thus “objectified”. This procedure makes it possible to carry out risk-benefit analyses prior to decision making (cf. Hansson in this Handbook). These traditional procedures of risk assessment have, however, two intrinsic limitations (cf. Hansson’s chapter on risk in Part V of this Handbook and [ShraderFrechette, 1991]). Firstly, for many new technology risk analyses quantitative experience is lacking which means that the extent of damage cannot be properly quantified. If quantifications are nonetheless given, it is easy to dismiss them as arbitrary, subjective or ideological. In controversial fields of technology, such as nuclear power or genetic engineering, the expected objectivisation of technological risks to be achieved on the basis of irrefutable practical knowledge have not succeeded. Secondly, especially in the discussion about the hazards of nuclear power, it has emerged that this “objective” concept of risk was useless in cases of crisis

1132

Armin Grunwald

because the affected public was not very impressed by the “objective” numerical values. Although atomic energy, according to technical risk analysis criteria, did not seem to be problematic, society refused to accept it, notably because of the perceived risks. This was a reason for integrating the social-scientific and psychological approaches to risk into risk analysis while devoting attention to the phenomenon of risk communication [Slovic, 1993]. Philosophical ethics, by contrast, stresses the role of normative considerations in determining the acceptability of risks and formulated corresponding principles, such as the principle of pragmatic consistency [Gethmann and Mittelstraß, 1992]. According to this principle, the acceptability of technological risk is proportionate to the risks which someone voluntarily accepts when choosing a lifestyle (like, for example, that of engaging in risky sport). It is considered to be irrational to reject technological risks if they do not exceed the risks voluntarily accepted. This approach, however, fails on at least two grounds [Grunwald, 2005]: firstly, there is no objective and value-neutral way of comparing categorically different types of risk, for example risks of technologies that do not serve the same ends; secondly, the technologically induced risk would be additional to other risks so an extra step of agreed acceptance would be required in all cases (other philosophical approaches are analysed by Hansson in this handbook). What is completely different is Hans Jonas’ “imperative of responsibility” [1984] which advocates that the use of technology is to be avoided if it is conceivable that the perpetuation of humanity could be endangered by such technology (“priority of the negative prediction”, “heuristics of fear”). In this case, the judgement does not depend on probabilities of occurrence. This type of radical judgement of technological risks and similarly radical demands for relinquishment or withdrawal has not gained general acceptance. Standpoints like Jonas’ would, of necessity, lead to a complete standstill since one can, after all, imagine a catastrophe for practically every innovation. Arguments that give priority to negative prediction do not permit distinctions to be made between more and less risky undertakings. In view of the lack of knowledge about possible risks and to avoid being confined to a “wait and see” strategy, with all the dangers of catastrophe which that brings (cf., for instance, the history of asbestos, Gee and Greenberg [2002]), the precautionary principle has been introduced to European environmental legislation. It has been incorporated in 1992 in the Treaty on the European Union. The precautionary principle establishes a rationale for political action in case of highly uncertain knowledge and it substantially lowers the (threshold) level for action of governments. The following characterisation of the precautionary principle shows – in spite of the fact that it still does not cover all relevant aspects – the complex inherent structure of the precautionary principle: “Where, following an assessment of available scientific information, there is reasonable concern for the possibility of adverse effects but scientific uncertainty persists, measures based on the precautionary principle may be adopted, pending further scientific information for a more comprehensive risk assessment, without having to wait until the reality and seriousness of those adverse effects become fully apparent” [Schomberg 2005, p.

Technology Assessment: Concepts and Methods

1133

168]. At present this discussion is particularly centred on the possible toxicity of nano particles. The implementation of the precautionary principle requires a careful evaluation of the state of scientific knowledge and of the gaps in that knowledge, as well as a political decision about the level of protection required against potential risks. TA is concerned with providing advice about political action with regard to precautionary and uncertain problems. 3.3.2

Cost-benefit analysis

The assessment of technology or of measures for dealing with the impact of technology with regard to economic efficiency is a standard TA evaluation (especially in Health Technology Assessment, HTA). This especially affects the cost-benefit ratio of major public projects or the evaluation of the efficiency of research stimulation programs. Cost-benefit analysis is a managerial evaluation procedure which is occasionally also employed in TA (cf. Ibo van de Poel’s chapter in this Handbook for more detailed explanation). It attempts to quantify and balance all the pertinent decision data — the costs as well as the benefits — in monetary units. Although in this calculation, “external effects” such as risks to human health or to the environment can be taken into account, the corresponding damage must ultimately be expressed in terms of monetary units (cf. Section 3.2.4 for the problems). Technological projects have to be appraised early on with regard to their expected economic efficiency. This not only applies to technical products such as automobiles or mobile telephones but also indirectly to questions, for example, of traffic infrastructure, building construction or to large-scale technical projects such as dam construction. Cost calculations for technological products have to be made over their entire life cycle. They consist of the development costs (expenditure for the planning stage, the potentially necessary research work, design, the drafting and conducting of tests and, if necessary, the construction of a prototype followed by production testing), the manufacturing costs (production costs in the form of expenses for materials, energy and labour or staff employment costs, construction or adaptation of production facilities, quality control, preparation of manuals), the operating costs (energy and material requirements at plant level, expenses for monitoring, day-to-day operational tests, maintenance, repairs) and the waste disposal costs (possibly also the reserves needed for specific risks, for disposal as well as for the final deposition of spent fuel rods; provision for the realisation of liabilities for the taking back of, for instance, old automobiles or electrical equipment). 3.3.3

Life cycle analysis and ecological balances

Sustainability assessment technology [Ludwig, 1997] is not restricted to the operating life of a technology but extends to include the entire life cycle, including the input chains and disposal. The sustainability effects of a technological product can only be comprehended by means of a life cycle assessment (LCA). When evaluating technological impacts on the environment, the LCA approach has long since been established. Ecological balancing indicators of the environmental compatibility of,

1134

Armin Grunwald

for example, products or facilities make it possible to compare various alternatives and find optimum solutions according to environmental considerations. A process chain can highlight ecological weak points and pinpoint the priorities for necessary change. The norm DIN EN ISO 14040 “Eco-Management — Ecological Balance. Principles and General Requirements” has been formulated as a framework for carrying out eco-balances. Despite the numerous methodological difficulties, the field of environmental policy and the evaluation of environmentally relevant processes cannot be envisaged without ecological balance. A recent development is the idea of including economic and social aspects in sustainability evaluations. An ecological balance consists of the definition of its objectives, a resource balance, an impact balance and an evaluation. The definition of objectives includes determining the scope and goals of the investigation. The resource balance includes drawing up a material use and energy balance for each of the system’s individual processes, examining the processes with regard to meeting environmental standards and aggregating the resource balance for the entire product line. “Product line” should be understood to mean a representation of all the relevant processes in the life cycle from raw material depositing to waste disposal site. The inclusion of transportation processes and energy consumption details may also be important in this investigation; this is decisive in the dispute on whether non-returnable packaging materials are more environmentally compatible than returnable packaging materials. In the impact balance, the materials and energies consumed in the product line are determined in relation to environmental categories and are weighted accordingly. The result is then evaluated in relation to environmental compatibility. Ecological balances do not make it possible to ascertain absolute environmental compatibility; they merely enable comparisons. Comparisons made using this method must relate to products with the same specific purpose. The results are presented as aggregated data, in other words, they say nothing about real environmental effects in specific places at a specific time but present instead total environmental impacts over the entire life cycle. If these results are to be accepted in decision making, the ecological balances must conform to the usual methodological requirements of comprehensibility, transparency and consistency. If results are questioned, it must be possible to trace them back to the input information, assumed functional dependencies or premises. Agreement must first be reached on these parameters — typical of methods in TA — particularly with regard to the system limits to be observed (Section 3.1.2). 3.3.4

Decision-analytical methods

Decision-analytical methods are oriented towards the problem of the multidimensional integration of various evaluative criteria (Section 3.2.3, cf. also Ibo van de Poel, in this handbook). They are based on the evaluation of options according to various, initially separate, evaluation criteria and on their subsequent weighting and aggregation leading to a comprehensive evaluation. First, the (socio-)technical

Technology Assessment: Concepts and Methods

1135

options to be assessed have to be selected. The evaluation criteria according to which the options are to be assessed then have to be formulated. Sufficient knowledge of the characteristics and effects of the options concerned is required. These parameters include, for instance, risks, costs and the possible side effects but also the expected gain or loss of utility. These “impact dimensions” must be quantified in the form of utility values for each option according to the various evaluation criteria. Finally, weightings for the criteria chosen have to be agreed upon so that the aggregation of the respective utility values can be calculated to provide a total utility for each dimension. Assuming that all of the criteria are functionally independent of one another and not redundant the best possible option will be the one with the maximum amount of “total utility”. The total utility is the sum of the separate utilities, added up for all of the n criteria. The separate utilities, in their turn, are the products of the individual utility values multiplied by the weighting for the respective criterion. Within the scope of this utility analysis (or scoring method ) there are a number of different procedures such as the multi-criteria analysis or the multi-attributive analysis with further method refinement, for instance, with regard to the compilation and processing of the data. By means of fuzzy logic, attempts are made to accentuate “soft” and differentiated evaluations. Furthermore, minimum requirements can be laid down for each evaluation criterion; failure to meet such minimum requirement would then disqualify the option concerned, even if it had done well according to other criteria. In this manner, the reciprocal substitutability of positive and negative part evaluations can be restricted. The influence of the individual contributions on utility and the influence of the weighting criteria can be tested by means of sensitivity analyses so that the robustness of the results can be examined. When applying these methods, the results depend to a great extent on the original assumptions. Uncertainties and estimates necessarily replace well-founded knowledge. The results also depend to a great extent on the weightings: by varying the weighting they can be altered. The risk of ideological abuse is very high. In view of these considerations, the notion of calculating a “total utility” might be generally doubted (Ibo van de Poel, this handbook). The total utility is a highly aggregated construct which might be viewed as an artefact with almost arbitrary values depending mainly on the aggregation procedure. In view of these limitations utility analysis is not so much an approach to the algorithmic determination of an “optimum” problem solving option as an expedient for bringing about transparency in complex decision-making situations. It indicates the consequences entailed in assuming certain (positive or negative) utilities and weightings and is, therefore, of elucidatory as well as heuristic value. 3.3.5

Consensus conferences

Consensus conferences are among the best-known participative TA procedures (Section 2.4.2). They have their roots in approaches of deliberative democracy

1136

Armin Grunwald

and were first employed in countries with highly developed cultures of discussion and standards of deliberation and discourse. The fundamental issue is to consider the prerequisites required by a functioning democracy in which highly specialised expert knowledge is essential especially in questions of science and technological policy: “ ... prior to all decisions an open (free of domination) and informed debate by all concerned is required. This debate shall secure a possibility for all concerned uver, 1995, p. 45]. To this end, “an to express their opinions and to be heard” [Kl¨ informed debate” should be conducted “between the lay and the learned” (p. 46). This is done in the following way: “A consensus conference is a chaired public hearing with an audience from the public and with the active participation of 10 –15 lay people” [p. 47]. This type of meeting requires considerable preparation: the relevant questions are to be clarified beforehand and the experts and participants are to be chosen. Lay participants are sought through advertising in newspapers. A selection of those interested is made that roughly represents a cross-section of the population in terms of age, gender, and educational and occupational background. The sampling of the participants involves declared readiness to participate and selection according to criteria of representation. Random sampling is, to a great extent, excluded. Participants may not be experts or stakeholders. When preparing the consensus conference, great importance is attached to imparting factual and specialist knowledge to the participants. The actual consensus conference takes three days. First there is “relay-running by the experts”, then there is a “cross-examination of the experts” and finally, the “presentation of the final document” [Kl¨ uver, 1995, p. 49ff.]. The first step in the procedures serves to determine the acknowledged state of knowledge and to reveal divergences in the experts’ opinions. In the second phase the aim is to reveal the reasons for these divergences through “cross-examination”. At this stage, at the very latest, discussions will arise on normative presuppositions and implicit premises. This is the most important phase with regard to the guaranteeing of transparency. In Denmark, where consensus conferences were developed, they are established by law. Consensus conferences have covered a vast number of topics like, for uver, 1995] for example, on the matter of genetically modified food products (cf. [Kl¨ an overview). Some of them have even reverberated in parliamentary decisions: in 1987, after a consensus conference on the subject, parliament decided to no longer use public funding to sponsor genetic experiments on animals. These consensus conferences acted as a model for the Swiss “PubliForum” approach, operated by TA-Swiss. Experiments with international consensus conferences have also now been carried out in a multilingual European setting. The recent and ambitious “Meeting of the Minds” project concerned itself with the challenges of neuro science (see http://www.meetingmindseurope.org/).

Technology Assessment: Concepts and Methods

3.3.6

1137

Citizens’ juries and variations

In the “Citizens’ Juries” method, lay people are required to judge a technological decision-making problem according to “common sense”. The members of the jury act as an independent committee which pronounces a well-balanced recommendation committed to “public interest” after hearing expert witnesses, persons affected and the stakeholders. These approaches provide assessment and judgement involving independent citizens which serves as advice for decision-makers. The “Planning Cell” can be seen as a specific type of citizens’ jury. It was developed at the beginning of the 1970s and is, therefore, one of the earliest participative procedures. It is mainly employed in municipal decision-making processes, for instance, for urban and traffic planning. The basic idea is that 25 randomly chosen citizens make themselves so familiar with the problem in hand over a four-day period either collectively or in small groups that they can understand and judge the possible solutions. In order to attain a greater overview, a number of planning cells are organised to deal with the same problem. Their results are summarised in a citizens’ expertise group. It is expected that in that way, socially acceptable and practicable recommendations will be acquired that are in the public interest. On the level of the individual participant, a strikingly high planning cell “event value” is acknowledged that is to say, the impression of being included in processes relevant to decision-making and of thereby being taken seriously as a citizen. On the societal level, a move towards more learning ability and towards a recapture of the role of the sovereign by the citizens is hoped for. 3.3.7

Mediation and arbitration

Mediation and arbitration are negotiation-orientated procedures designed to peacefully and consensually settle conflicts with the help of a neutral party (mediator, arbitrator). They usually derive from existing conflicts which the disagreeing parties are unable to resolve constructively without external help. The common interest of the parties in conflict is presupposed in a consensual and extrajudicial agreement. According to the “Harvard Model” [Fisher and Ury, 1988], it is assumed that the deadlocked positions can be loosened by revealing the parties’ “real” interests before being transformed into “win-win” situations. Here, compensations agreed upon through negotiations play an important role. Mediation procedures can also be employed preventively, in order to avoid matters escalating. Attempts are now frequently made to gather the potentially conflicting parties round a table in the preparatory phase of decisions on, for example, where to locate technical facilities, in order to effect understanding with the specific opponent before taking measures to de-escalate impending conflicts. In the end, it is a question of establishing a situation in which both sides have advantages or can partially realize their objectives. The role of the mediator is to break down existing blockades in communication, initiate a process of settlement and supervise it. The conflict solution is not decided by the mediator but has to be discovered by the parties in conflict under the

1138

Armin Grunwald

mediator’s guidance. The requirements for a good moderator are: strict neutrality in the case in question, sufficient technical competence, a knowledge of the legal regulations and provisions, competence in dealing with groups and individuals, communicative skills and practical experience of moderating discourse, orientation to public interest and social respect. Since 1973, such procedures have been practised in different variants in the U.S. in relation to environmental issues and have to some extent become integrated into the law as an alternative to judicial conflict solutions. In TA technological conflicts involving a limited number of actors and a precisely defined problem seem to be the appropriate fields for implementing mediation procedures. Such conflicts particularly revolve around location problems focused on the just and acceptable distribution of risks, damage, and the utility of largescale industrial facilities such as airports, power plants, waste disposal sites, or chemical processing plants. Such NIMBY (“Not In My Back Yard”) problems are, as a rule, local or regional and tend to be characterized by a specific planned event, by extreme intervention in the life and environment of the local residents or by a mixture of various interests. 3.3.8

Vision assessment

Quite often, as with the emergence of nanotechnology, visions and metaphors mark the revolutionary advance of science in general and act as an important factor in societal debates. Such visions have not yet been analysed comprehensively by TA. Preliminary analysis already has shown that futuristic visions are ambivalent: they may cause fascination as well as concern and fear. The main argument for requiring early vision assessment is the importance of visions in actual debates, that is, both in the debate on the opportunities afforded by scientific and technological progress and in ongoing risk debates. To provide for more rationality, reflexivity and transparency in these debates, vision assessment should also consider values [Grunwald, 2006a; 2007a]. Vision assessment is a new TA tool that is not directed at the assessment of technologies but at the assessment of visions which are communicated in the societal environment of technology [Grin and Grunwald, 2000]. The fields of nanotechnology and all the other converging technologies are currently being subjected to broad discussion [Grunwald, 2006a; 2007a]. Vision assessment can be analytically divided into vision analysis – which is itself subdivided into a substantial aspect (what is the content of the respective vision?) and a pragmatic aspect (how is it used in concrete communication?), vision evaluation (how could the content of the vision be evaluated and judged?), and vision management (how should the people and groups affected deal with the visions?). Vision assessment includes normative elements, like the questions of how the cognitive aspects can be categorised, how they can be judged according to a degree of realisation or feasibility, according to plausibility and evidence [Pereira et al., 2007], and what status the normative aspects have, for example, relative

Technology Assessment: Concepts and Methods

1139

to established systems of values or to ethical standards. The general aim is to achieve a transparent disclosure of the relationship between knowledge and values, knowledge and the lack of it and the evaluation of these relationships and their implications. In particular, vision assessment should allow the various and, partly divergent normative aspects of visions of the future to directly confront each another. This can be achieved through ethical analysis and desk research. In addition, the stakeholders should discuss their differing judgements in workshops directly with each another in order to reveal their assumptions.

3.4

Normative backgrounds to assessment methods

The TA methods presented above differ in several respects: they are relevant at different stages in the TA processes, require different types of data, offer different types of knowledge, and (as will be discussed below) differ with respect to their normative premises. The various TA methods (or families of methods) are usually applied in specific situations and in the context of specific TA approaches. Approaches such as participative TA or innovation-orientated TA adopt a specific view on technology, on society or on decision-making procedures: • Cost-benefit analysis and MCDA are tied to the utilitarian decision-making calculus. They share essentials of utilitarianism like the reduction of different criteria to monetary values and the principle of maximising utility. This category also includes quantitative risk assessment aimed at minimising risk. • Life cycle analysis (LCA) relies, in part, on ecological ideas about the environmental compatibility of industrial or other economic processes. • Sustainability assessments bring the idea of (intergenerational and intragenerational) justice [Rawls, 1999] and equity into the arena of technological development [Grunwald and Kopfm¨ uller 2006] . • Types of participative TA, such as consensus conferences, usually work on the basis of normative ideas about deliberative democracy and discursive ethics [Habermas, 1988b; Renn and Webler, 1998], in which persons in positions of responsibility and interested citizens all share normative ideas, which are often very close to the ideas of civil society. • Mediation approaches work with “checks and balances” and aim at mediating diverging interests, for example, by creating compensation strategies without giving priority to ethical principles. Two essential points have to be recognised in each concrete TA process and in TA theory as well. First of all assessment methods are not, as has been shown, value free. Normative premises and presuppositions are usually involved in the selection of specific TA methods, whether directly or because the application of

1140

Armin Grunwald

a certain method is often related to normative and conceptual assumptions. For example, there is a close relationship between cost–benefit analysis and the utilitarian decision-making calculus. In order to meet the goals of TA and to avoid biases it is, therefore, indispensable to apply a high degree of reflectivity with respect to such normative elements of TA methods and to establish a maximum degree of transparency in this regard. Secondly, what can be learned from this analysis (and what has been supported by TA experience in the past decades) is modesty in terms of the expectation that TA should be able to reduce decisions about technologies and their societal environment to algorithm-like methodical procedures. In contrast to such expectations it has been shown that methods do involve normative aspects. By applying TA methods, various kinds of data can be collected, aggregated and evaluated for the purposes under consideration. Transparency can be strengthened and arguments can be supported by methodically guided research. But such activities cannot replace the very political and ethical nature of far-ranging technological decisions; they can only inform and orientate them. Decision-support systems — and TA may be seen as a specific kind of decision-support tool — they do not replace decisions but they rather support decision-making.

4 CURRENT DEVELOPMENTS AND FUTURE CHALLENGES TA is context dependent with regard to the various topics, target groups, backgrounds, and fields of technology. Changes in context (the general societal and political setting, the roles and constellations of the relevant actors, processes of opinion formation and of decision-making) therefore have direct effects on TA’s options for meeting its responsibilities. TA therefore has to observe the changes in its environment and react to them conceptually. In other words it has to actively reflect these changes in its own conceptual self-understanding. Current developments in societal, political and scientific contexts that are highly relevant to TA are: Globalization: Until recently, TA’s target group in technology, research, and innovation policies were primarily institutions within nationally or regionally orientated decision-making structures. Economic, but also political and technological globalization has changed this situation. The fact that the impacts of technology have no borders has long since been acknowledged. Globalization, however, also affects technological development, diffusion, and application. Technological design takes place today in worldwide networks. Examples are Open Source software and the Human Genome Project, or nano(bio)technology. The use and diffusion of technology is also becoming increasingly global. Electric power supply networks have long since grown beyond the political boundaries of national states. In the promotion as well as in the regulation of technology, important decisions are shifting to levels of higher aggregation, that is, from the national to the European level. The influence that regional “cultures” have on how technology is dealt with

Technology Assessment: Concepts and Methods

1141

is decreasing, just as the leeway of the classical national states is shrinking. TA has to find ways of dealing – conceptually and methodically, but also strategically – with globalization and with new constellations. If it does not, it will be threatened by provincialisation and loss of importance. TA is challenged to organise itself internationally, to conduct the corresponding knowledge transfer, to contribute to the development and use of new governance structures and to set cultural and intercultural TA on the agenda. TA has to operate more actively than it has up until now on a supranational and, if required, a global level, and advise a correspondingly multilevel policy in the scope of a “global governance”. The Knowledge Society: The methods of production, the access to and the distribution of the means of utilising knowledge are affected by the development of a “knowledge society”. Driven by the spread of information and communication technology, the importance of knowledge is growing in economic, social and political respects. Knowledge policy and knowledge management are becoming new societal domains [Stehr, 2004]. Actions and decisions will be increasingly substantiated and legitimised by scientific knowledge. At the same time, however, the founding of societal decisions on knowledge necessarily generates risks due to uncertainties of the knowledge, even to the potential self-endangerment of society. This exacerbates the situation of contingency in the human condition [Grunwald, 2007a]. Sustainable Development: The guiding principle of sustainability is that it demands a research and technology policy that fosters sustainability. For TA, this is significant in at least two respects: On the one hand much prospective knowledge on the consequences of new technological innovations for sustainability is needed, which (a) covers the entire life cycle of the technology and its components and (b) is not only ecological but also concerned with all of the dimensions of sustainability [Grunwald, 2007b]. On the other hand, this quite considerably increases the expectations placed on an “integrative” assessment of the impacts [Ludwig, 1997]. Backcasting approaches: In the last years backcasting approaches have regained importance, especially concerning sustainable development. For example, transformation management which currently is a frequently used notion, operates by defining desired futures and deriving measures and strategies which should be implemented today in order to reach the desired future states. Foresight Exercises: there have been many (technology) foresight exercises in the past 15 years (for definitions of foresight cf. [Coates, 1985] and [Martin and Irvine 1989]). In particular the European Union has supported many such exercises, mainly in the field of regional foresight [FOREN, 2001]. Foresight activities have a lot of parallels with TA but are more explorative, emphasise the social effects (such as mobilising people in a regional or building network) and do not focus on normative assessment. New Technologies: Changes and shifts of emphasis can be discerned in the characteristics of current scientific and technical innovations. It is no longer the tradi-

1142

Armin Grunwald

tional problems of large-scale facilities that is central but rather the development — as seen in nanotechnology and information and communication technology, all of which have culminated in the notion of “converging technologies” [Roco and Bainbridge, 2002] — leading to increasing integration and to the creation of ever more interfaces. As a result, decision-making processes are becoming increasingly complex. The future pace of technology is determined by the integration of developments from originally separated areas, rather than by individual innovations. The Importance of Social Questions: The role of technology in society is less and less determined by the technical feasibility of products, processes or systems. Much that is technically feasible and that has also been realised and brought onto the market, is founded on societal embedding (as innovation research has shown), on economic aspects, on a lack of societal acceptance or on insufficient adaptation to existing technology (like with Transrapid, for example). Customer acceptance, or the lack of it, occasionally leads to unexpected turns — as, for instance, with the question of genetically modified food products in Great Britain, and currently with the question of whether and when UMTS mobile phones will be accepted on the market. Here, new interfaces between innovation research, the cultural sciences and TA are being opened up. ELSI studies: In the last few years a new type of TA-related activity has emerged. ELSI or ELSA studies (ethical, legal and social implications/aspects) have been elaborated in some emerging fields of new technologies, mainly in the area of nanotechnology. Such activities are more selective in their scope than classical TA and they are often not directly aimed at decision makers but intend to broadly inform the interested public. In a methodological and normative sense, however, there are great similarities with established TA. The Future of Human Nature: Converging technologies from the fields of nanotechnology, biotechnology, information technology and the cognitive sciences (NBIC, cf. Roco and Bainbridge [2002]) will enable humankind to improve human performance, at individual as well as at collective level. Emerging ethical questions [Habermas, 2001] as well as the potential for innovation and advance will be prominent topics for TA in the next years. The history of TA can be recounted as a history of experimenting with concepts and of learning by testing or deducing from relevant conceptual debates. To date this might have been done rather sporadically and against the background of practical pressure. If that is so the time now seems to have come to take a look at the “whole” spectrum of TA and to develop a theory of TA which does not yet exist (tentative steps were taken in this direction in TATUP [2007]). A theory of TA can only be a theory of learning about TA and therefore a theory of reflection on TA on the basis of its relationship to practice.

Technology Assessment: Concepts and Methods

1143

BIBLIOGRAPHY [Ayres et al., 1970] R. U. Ayres, J. W. Carlson, and S. W. Simon. Technology Assessment and Policy-Making in the United States. New York, 1970. [Barber, 1984] B. R. Barber. Strong Democracy. Participatory Politics for a New Age. University of California Press, 1984. [Beauchamp, 2001] T. Beauchamp. Philosophical Ethics. An Introduction to Moral Philosophy, McGraw Hill, 2001. [Bechmann et al., 2007] G. Bechmann, M. Decker, U. Fiedeler, and B.-J. Krings. Technology Assessment in a Complex World. International Journal on Foresight and Innovation Policy 3, pp. 6-27, 2007. [Bechmann and Hronszky, 2003] G. Bechmann and I. Hronszky, eds. Expertise and Its Interfaces. Edition Sigma, 2003. [Beck, 1986] U. Beck. Risikogesellschaft. Auf dem Weg in eine andere Moderne. Suhrkamp, 1986. [Beck et al., 1996] U. Beck, A. Giddens, and S. Lash, eds. Reflexive Modernisierung. Eine Kontroverse. Frankfurt, 1996. [Bijker et al., 1987] W. E. Bijker, T. P. Hughes, and T. J. Pinch, eds. The Social Construction of Technological Systems. New Directions in the Sociology and History of Technological Systems. Cambridge (Mass.), 1987. [Bijker and Law, 1994] W. E. Bijker and J. Law, eds. Shaping Technology/Building Society. Cambridge (Mass.), 1994. [Bimber, 1996] B. A. Bimber. The Politics of Expertise in Congress: The Rise and fall of the Office of Technology Assessment. State University of New York Press, 1996. [B¨ utschi et al., 2004] D. B¨ utschi, R. Carius, M. Decker, S. Gram, A. Grunwald, P. Machleidt, S. Steyaert, and R. van Est. The Practice of TA; Science, Interaction, and Communication. In Bridges between Science, Society and Policy. Technology Assessment – Methods and Impacts, M. Decker and M. Ladikas, eds., pp. 12-55, Springer, 2004. [Camhis, 1979] M. Camhis. Planning Theory and Philosophy. London, 1979. [Coates, 1985] J. F. Coates. Foresight in Federal Government Policy Making, Futures Research Quarterly 1, pp. 13-18, 1985. [Collingridge, 1980] D. Collingridge. The Social Control of Technology. New York, 1980. [Daniels, 1996] N. Daniels. Justice and Justification. Reflective Equilibrium in Theory and Practice. Cambridge, Cambridge University Press, 1996. [Decker and Ladikas, 2004] M. Decker and M. Ladikas, eds. Bridges between Science, Society and Policy. Technology Assessment – Methods and Impact. Springer, 2004. [Decker and Grunwald, 2001] M. Decker and A. Grunwald. Rational Technology Assessment as Interdisciplinary Research. In Implementation and Limits of Interdisciplinarity in European Technology Assessment, M. Decker, ed, Springer, 2001. [Dierkes et al., 1992] M. Dierkes, U. Hoffmann, and L. Marz. Leitbild und Technik. Zur Entstehung und Steuerung technischer Innovationen, Campus, 1992. [Ferr´ e, 1995] F. Ferr´ e. Philosophy of Technology. The University of Georgia Press, 1995. [Fisher, 1990] F. Fisher. Technocracy and the Politics of Expertise. Newbury Park, London, 1990. [FOREN, 2001] FOREN. A Practical Guide to Regional Foresight. Institute for Prospective Technology Studies, Sevilla, 2001. [Funtowitz and Ravetz, 1993] S. Funtowitz and J. Ravetz. The Emergence of Post-Normal Science. In Science, Politics and Morality, R. von Schomberg, ed, London, 1993. [Gee and Greenberg, 2002] D. Gee and M. Greenberg. Asbestos: from ‘magic’ to malevolent mineral. In The Precautionary Principle in the 20 th century. Late Lessons from early warnings, P. Harremoes, D. Gee, M. MacGarvin, A. Stirling, J. Keys, B. Wynne, S. Guedes Vaz, eds, pp. 49-63, Sage, 2002. [Gethmann and Mittelstraß, 1992] C. F. Gethmann and J. Mittelstraß. Umweltstandards. Gaia, 1, pp. 16-25, 1992. [Gibbons et al., 1994] M. Gibbons, C. Limoges, H. Nowotny, S. Schwartzmann, P. Scott, and M. Trow. The New Production of Knowledge. Sage, 1994.

1144

Armin Grunwald

[Grin, 2000] J. Grin. Shaping 21st century society through vision assessment? Technology assessment to support judgement. In Vision Assessment: Shaping Technology in 21st Century Society. Towards a Repertoire for Technology Assessment, J. Grin and A. Grunwald, eds, pp. 9–30, Springer, 2000. [Grin and Grunwald, 2000] J. Grin and A. Grunwald, eds. Vision Assessment: Shaping Technology in 21st Century Society. Towards a Repertoire for Technology Assessment. Springer, 2000. [Grin et al., 1997] J. Grin, H. van de Graaf, and R. Hoppe. Technology Assessment through Interaction. Rathenau Institute, 1997. [Grundahl, 1995] J. Grundahl. The Danish Consensus Conference Model. In Public Participation in Science. S. Joss, J. Durant, eds., pp. 31-40, Chippenham 1995. [Grunwald, 1999] A. Grunwald. Technology Assessment or Ethics of Technology? Reflections on Technology Development between Social Sciences and Philosophy. Ethical Perspectives, 6, pp. 170–182, 1999. [Grunwald, 2000] A. Grunwald. Against Over-Estimating the Role of Ethics in Technology. Science and Engineering Ethics, 6, pp. 181–196, 2000. [Grunwald, 2002] A. Grunwald. Philosophy and the Concept of Technology. On the Anthropological Significance of Technology. In On Human Nature. Anthropological, Philosophical and Biological Foundations, A. Grunwald, M. Gutmann, E. Neumann-Held, eds., pp. 173–188, Springer, 2002. [Grunwald, 2003] A. Grunwald. Methodical Reconstruction of Ethical Advises. In Expertise and Its Interfaces, G. Bechmann and I. Hronszky, eds., pp. 103-124, edition sigma, 2003. [Grunwald, 2004a] A. Grunwald. Strategic knowledge for sustainable development: the need for reflexivity and learning at the interface between science and society. International Journal of Foresight and Innovation Policy, 1, pp. 150-167, 2004a. [Grunwald, 2004b] A. Grunwald. Participation as a means of enhancing the legitimacy of decisions on technology? A sceptical analysis. Poiesis & Praxis, 3, pp. 106-122, 2004b. [Grunwald, 2005] A. Grunwald. Zur Rolle von Akzeptanz und Akzeptabilit¨ at von Technik bei der Bew¨ altigung von Technikkonflikten. Technikfolgenabsch¨ atzung – Theorie und Praxis, 14, pp. 54-60, 2005. [Grunwald, 2006a] A. Grunwald. Nanotechnologie als Chiffre der Zukunft. In Nanotechnologien im Kontext, A. Nordmann, J. Schummer, A. Schwarz, eds., pp. 49-80, Akademische Verlagsgesellschaft 2006. [Grunwald, 2006b] A. Grunwald. Scientific independence as a constitutive part of parliamentary technology assessment. Science and Public Policy, 33, pp. 103-113, 2006. [Grunwald, 2007a] A. Grunwald. Converging Technologies: visions, increased contingencies of the conditio humana, and search for orientation. Futures, 39, pp. 380-392, 2007. [Grunwald, 2007b] A. Grunwald. Working Towards Sustainable Development in the Face of Uncertainty and Incomplete Knowledge. Journal of Environmental Policy & Planning Volume 9, Issue 3, p. 245–263, 2007. [Grunwald and Kopfm¨ uller, 2006] A. Grunwald and J. Kopfm¨ uller. Nachhaltigkeit. Campus, 2006. [van Gunsteren, 1976] H. R. van Gunsteren. The Quest for Control: A Critique of the RationalCentral-Rule Approach in Public Affairs. London, 1976. [Guston and Sarewitz, 2002] D. H. Guston and D. Sarewitz. Real-Time Technology Assessment. Technology & Culture 24, pp. 93-109, 2002. [Habermas, 1988a] J. Habermas. Nachmetaphysisches Denken. Suhrkamp, 1988a. [Habermas, 1988b] J. Habermas. Theorie des kommunikativen Handelns. Suhrkamp, 1988b. [Habermas, 2001] J. Habermas. Die Zukunft der Natur des Menschen. Suhrkamp, 2001. [Harremoes et al., 2002] P. Harremoes, D. Gee, M. MacGarvin, A. Stirling, J. Keys, B. Wynne, and S. Guedes Vaz, eds. The Precautionary Principle in the 20 th century. Late Lessons from early warnings. Sage, 2002. [Jonas, 1984] H. Jonas. Das Prinzip Verantwortung. Versuch einer Ethik f¨ ur die technologische Zivilisation. Suhrkamp, 1979. English: The Imperative of Responsibility, 1984. [Joss and Belucci, 2002] S. Joss and S. Belucci, eds. Participatory Technology Assessment – European Perspectives. Westminster University Press, 2002. [Joss and Durant, 1995] S. Joss and J. Durant, eds. Public Participation in Science. Chippenham, 1995.

Technology Assessment: Concepts and Methods

1145

[Kl¨ uver, 1995] L. Kl¨ uver. Consensus Conferences at the Danish Board of Technology. In Public Participation in Science, S. Joss, J. Durant, eds., pp. 41–52, Chippenham, 1995. [Kroes and Meijers, 2000] P. Kroes and A. Meijers, eds. The Empirical Turn in the Philosophy of Technology. Elsevier Science, 2000. [Latour, 1987] B. Latour. Science in Action. How to Follow Scientists and Engineers Through Society. Milton Keynes, 1987. [Levidow, 1999] Levidow, L. Democratizing Technology — or Technologizing Democracy? In Democratizing Technology. Theory and Practice of a Deliberative Technology Policy, R. von Schomberg, ed., pp. 51-70, Hengelo, 1999. [Lindblom, 1973] C. E. Lindblom. The Science of Muddling Through. In A Reader in Planning Theory, A. Faludi, ed., pp. 151-170, Oxford, 1973. [Ludwig, 1997] Ludwig, B.. The Concept of Technology Assessment – an entire Process to Sustainable Development. Sustainable Development, 5, pp. 111–117, 1997. [Mambrey and Tepper, 2000] P. Mambrey and A. Tepper. Technology Assessment as Metaphor Assessment — Visions guiding the development of information and communications technologies. In Vision Assessment: Shaping Technology in 21st Century Society. Towards a Repertoire for Technology Assessment, J. Grin and A. Grunwald, eds., pp. 33-52, Springer, 2000. [Marcuse, 1967] J. Marcuse. One-Dimensional Man. Boston, 1966 (Der eindimensionale Mensch. Neuwied, 1967). [Martin and Irvine, 1989] B. R. Martin and J. Irvine. Research Foresight: Priority-Setting in Science. Pinter Publishers, London, 1989. [Meadows et al., 1972] D. Meadows, D. Meadows, P. Milling, and E. Zahn. Limits to Growth. Report to the Club of Rome. Rome, 1972. [Mitcham, 1994] C. Mitcham. Thinking through Technology. Chicago, 1994. [Moore, 1903] G. E. Moore. Principia Ethica. Cambridge: Cambridge University Press, 1903. [Nowotny, 2003] H. Nowotny. Democratising expertise and socially robust knowledge. Science and Public Policy 30, pp. 151-156, 2003. [Paschen and Petermann, 1991] H. Paschen and T. Petermann. Technikfolgenabsch¨ atzung – ein strategisches Rahmenkonzept f¨ ur die Analyse und Bewertung von Technikfolgen. In: Technikfolgenabsch¨ atzung als Technikforschung und Politikberatung, T. Petermann, ed., pp. 19–42, 1991. [Paschen et al., 1978] H. Paschen, K. Gresser, and S. Conrad. Technology Assessment: Technologiefolgenabsch¨ atzung. Ziele, methodische und organisatorische Probleme, Anwendungen. Campus, Frankfurt, 1978. [Pereira et al., 2007] A. G. Pereira, R. von Schomberg, and S. Funtowicz. Foresight Knowledge Assessment. International Journal on Foresight and Innovation Policy 3, pp. 53-75, 2007. [Rader, 2002] M. Rader. Synthesis of Technology Assessment. In: Strategic Policy Intelligence: Current Trends, the State of Play and Perspectives. A. T¨ ubke, K. Ducatel, J.P. Gavigan, P. Moncada-Patern` o-Castello, eds., pp. 27-37, Sevilla, 2002. [Rawls, 1999] J. Rawls. A Theory of Justice. Cambridge (MA), The Belknap Press of Harvard University Press, 1999 [1971]. [Renn and Webler, 1998] O. Renn and T. Webler. Der kooperative Diskurs – Theoretische Grundlagen, Anforderungen, M¨ oglichkeiten. In: Abfallpolitik im kooperativen Diskurs. In: O. Renn, H. Kastenholz, P. Schild, U. Wilhelm, eds., pp. 3–103, Z¨ urich, 1998. [Rip et al., 1995] A. Rip, T. Misa, and J. Schot, eds. Managing Technology in Society. London, 1995. [Roco and Bainbridge, 2002] M. C. Roco and W. S. Bainbridge, eds. Converging Technologies for Improving Human Performance. Arlington, Virginia, 2002. [Schmitt, 1934] C. Schmitt. Politische Theologie. M¨ unchen/Leipzig, 1934. [von Schomberg, 1999] R. von Schomberg, ed. Democratizing Technology. Theory and Practice of a Deliberative Technology Policy. Hengelo, 1999. [Schot and Rip, 1997] J. Schot and A. Rip. The Past and Future of Constructive Technology Assessment. Technological Forecasting and Social Change 54, pp. 251–268, 1997. [Schot, 1999] J. Schot. Constructive Technology Assessment and Technology Dynamics: The Case of Clean Technologies. Science, Technology and Human Values 17, pp. 36–56, 1999. [Shrader-Frechette, 1991] K. S. Shrader-Frechette. Risk and Rationality. Philosophical Foundations for Populist Reforms. Berkeley, 1991.

1146

Armin Grunwald

[Skorupinski and Ott, 2000] B. Skorupinski and K. Ott. Ethik und Technikfolgenabsch¨ atzung. Z¨ urich, ETH-Verlag, 2000. [Slovic, 1993] P. Slovic. Perceived risk, trust, and democracy. Risk Analysis 13, pp. 675–682, 1993. [Smits et al., 1995] R. Smits, A. Leyten, and P. den Hertog. Technology Assessment and Technology Policy in Europe: New Concepts, New Goals, New Infrastructures. Policy Sciences, 28, pp. 272-299, 1995. [Smits and den Hertog, 2007] R. Smits and P. den Hertog. TA and the management of innovation in economy and society. International Journal on Foresight and Innovation Policy 3, pp. 28-52, 2007. [Smits and Leyten, 1991] R. Smits and J. Leyten. Technology Assessment. Watchdog or Tracker, Kerkebosch, Zeist, 1991. [Stehr, 2004] N. Stehr. The Governance of Knowledge. London, 2004. [TATUP, 2007] TATUP – Technikfolgenabsch¨ atzung. Theorie und Praxis. Special Issue “Auf dem Weg zu einer Theorie der Technikfolgenabsch¨ atzung”, 2007. [United States Senate, 1972] United States Senate. Technology Assessment Act of 1972. Report of the Committee on Rules and Administration, 13 Sept. 1972. Washington, D.C., 1972. [van de Poel, 2001] I. R. van de Poel. Investigating Ethical Issues in Engineering Design. Science and Engineering Ethics 7, pp. 429-446, 2001. [van Est and van Eijndhoven, 1999] R. van Est and J. van Eijndhoven. Parliamentary Technology Assessment at the Rathenau Institute. In: Handbuch Technikfolgenabsch¨ atzung, S. Br¨ ochler, G. Simonis, K. Sundermann, eds., pp. 427–436, Berlin, 1999. [van Gorp, 2005] A. van Gorp. Ethical Issues in Engineering Design; Safety and Sustainability. Simon Stevin Series in the Philosophy of Technology, Delft, 2005. [van Gorp and Grunwald, in preparation] A. van Gorp and A. Grunwald. Ethical Responsibilities of Engineers in Design Processes. Risks, Regulative Frameworks and Societal Division of Labour. In: S. Roeser and L. Asveld (eds.), in preparation. [VDI, 1991] VDI – Verein Deutscher Ingenieure. Technikbewertung. Richtlinie 3780. D¨ usseldorf, 1991. English version available at www.vdi.de [Vig and Paschen, 1999] N. Vig and J. Paschen, eds. Parliaments and Technology Assessment. The Development of Technology Assessment in Europe. Albany, USA, 1999. [von Schomberg, 2005] R. von Schomberg. The Precautionary Principle and its normative challenges. In: The Precautionary Principle and Public Policy Decision Making. E. Fisher, J. Jones, R. von Schomberg, eds., pp. 161-175, Cheltenham, UK and Northampton, MA, 2005. [Weyer et al., 1997] J. Weyer, U. Kirchner, L. Riedl, and J. F. K. Schmidt. Technik, die Gesellschaft schafft. Soziale Netzwerke als Ort der Technikgenese. Berlin, 1997. [Williamson, 1994] R. A. Williamson. Space Policy at the Office of Technoloy Assessment. In: Technikbeurteilung in der Raumfahrt. A. Grunwald, H. Sax, eds., pp. 211–225, Berlin, 1994. [Winner, 1982] L. Winner. Autonomous Technology. Technics-out-of-Control as a Theme in Political Thought. Cambridge, USA, 1982. [Winner, 1980] L. Winner. Do Artifacts have Politics? Daedalus 109, pp. 121-36, 1980. [Wynne, 1995] B. Wynne. Technology Assessment and Reflexive Social Learning: Observations from the Risk Field. In: Managing Technology in Society. A. Rip, T. Misa, J. Schot, eds., pp. 19-36, London 1995. [Yoshinaka et al., 2003] Y. Yoshinaka, C. Clausen, and A. Hansen. The Social Shaping of Technology: A New Space for Politics? In: Technikgestaltung – Wunsch oder Wirklichkeit? A. Grunwald, ed., pp. 117-131, Berlin, Springer, 2003.

THE INTERACTION OF ETHICS AND TECHNOLOGY IN HISTORICAL PERSPECTIVE Carl Mitcham and Adam Briggle With only minor stretching, ethics may be conceived as a technology-like science. Ethics is technical insofar as it involves specialized terminology and includes techniques for the making of human action; it is scientific in the sense of involving systematic reflection and critical analysis. Furthermore, as Caroline Whitbeck [1998] argues, there are strong parallels between problems in ethics and problems in engineering design. Thus independent of other considerations, it is appropriate that a handbook on the philosophy of technology and the engineering sciences should include a chapter on ethics and technology. There are other reasons as well. In the European tradition, ethics — or systematic philosophical reflection on human action and its norms — can be traced back to Socrates, and from its earliest manifestations has included multiple references to technics or the arts and crafts, in the straightforward senses of the skilled making and using artifacts. Since the Renaissance such making and using has become increasingly systematized as technology and engineering, about which there have arisen further and more extensive ethical discourses. Ethics in such contexts has been called out to discuss technology as manifested in everything from objects and activities and their combined expressions in material culture to forms of knowledge and intentions. In these different aspects, technology has also been given moral shape by professional ethics codes, consumer use behaviors, and political determinations. Technology has thus influenced the way humans conceive and evaluate their worlds and itself been influenced by such evaluations. There exist extended interactions between ethics and technology that have contributed to shifts in ethical understanding and in technological making, using, thinking, and willing. Given the breadth of such discussions, the present chapter will focus on the broad phenomenon of technology, while on occasion and as appropriate distinguishing notions of technique, technics, art, craft, invention, engineering, engineering science, and technoscience. (For more on these distinctions, see the chapter by Mitcham and Schatzberg in Part I of this Volume, “Defining Technology and the Engineering Sciences”.) Indeed, within the general field of reflective engagement marked out by the phrase “philosophy of technology,” ethical judgment has received more attention, especially insofar as popular discussions are concerned, than those from other branches of philosophy such as metaphysics, epistemology, Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

1148

Carl Mitcham and Adam Briggle

and aesthetics. Almost any theoretical assessment of technology is likely to elicit a question concerning practical implications, and it is the resulting ethics and technology engagements that are surveyed here. Following two sections that sketch a historical background, the chapter focuses on distinctly modern interactions between ethics and technology. The modern focus begins in section three by defining and illustrating three general schools of ethical reflection on technology found in European and American philosophy. Section four considers ways in which moral practice, informed variously by the schools of ethics and their approaches, shapes technology, working across professional, personal, and political levels. Section five considers briefly how technology can shape moral beliefs and practices by suggesting an interactive conception of technology and moral change. 1 PRE-MODERN ETHICS AND TECHNICS Plants and animals alter the world by selectively ingesting materials from the environment, transforming them, and excreting newly formed materials. For a few animals, however, their own existence depends crucially on altering the world in more determinate ways. Spiders spin webs; birds build nests; beavers construct dams; and chimpanzees fashion tools. For no animal, however, is the making and using of physical objects more crucial to their lives and livelihood than for human beings, who make and use clothing, shelter, utensils, tools, utilities, weapons, structures, cities, transport and communication systems, and more, all as part of their distinctive way of being in the world — a way of being that differentiates into multiple traditions of material culture. In their technologically advanced forms, human material cultures have become comparable to geological forces in their abilities to alter the environment. An early recognition of the defining feature of human making and using was classically expressed in the second chorus of Sophocles’ Antigone: Of many wonders, none is more wondrous than human beings. They cross the seas with the winds storming and swelling and roaring about them. ... Cunning are humans. Through mechanical contrivances they master the beasts of the field and those that roam the hills. The horse with the shaggy mane they hold and harness about the neck, and the strong bull of the mountains. And speech and wind-swift thought and the temperaments that go with political life

The Interaction of Ethics and Technology in Historical Perspective

1149

they have taught themselves, and how to avoid cold frost under the open sky, and the pulsating rain as well. ... Clever beyond all dreams are the mechanisms and technai of humans which bring forth at one time evil and at another good. When they honor the law and pursue the justice of the gods their cities stand, but dishonored are those whose reckless hearts make them join hands with evil. May we not think like them nor may such impious persons dwell in our house. (lines 332-375) What is special about humans is that they navigate oceans, harness animals, and even tame their own impulses so as to be able to live together and build houses to protect against the elements. Yet to the Greek mind technical skills remained subordinate to a moral order, with those who acted outside the lawful framework being justly excluded from the human community. From the beginning of this appreciation of technical skill and its achievements, techne was thus associated with the possibilities of human good and evil. Already in the Odyssey technique or craft was identified with praiseworthy skill (e.g., Odyssey V, 259) and blameworthy trickery (e.g., Odyssey IV, 455). Likewise in the Hebrew scriptures, technical mastery and technics appear, on the one hand, as necessity and perfection and, on the other, as temptation or corruption. Noah built the ark in accordance with directions from Yahweh as a vehicle for salvation from the flood (Genesis 6:14 ff.), but subsequent humans used their technical prowess to construct the Tower of Babel as a spiritual rebellion (Genesis 11:1 ff.). In the works of Plato (428-347) there emerged a more explicit ethical reflection on technics. In Socrates’ autobiographies (Apology 15a ff. and Phaedo 96a ff.) it is not nature but the ideas of goodness, greatness, and beauty that were the orienting themes of philosophical inquiry. The search for a full account of ethical experience called forth an appreciation of different levels of being and different forms of knowing appropriate to each — though the highest reality was ethical, the good itself, conceived as “beyond being” Republic 509a-b). According to Aristotle (384-322), however, philosophy originated when discourse about the gods was replaced with discourse about nature (compare, e.g., Metaphysics 983b29 and 988b27). In the Aristotelian tradition it is the study of nature, as cause of the distinguishing functional features of a species, that both constitutes natural science and provides insight into the telos or end of any instance of its kind. For Aristotle the various branches of philosophy themselves became distinguished, and ethics assumed the character of a systematic examination of ethos, as manifested in human customs or behavior. More than any other type of entity, humans have a nature that is open to and even requires further deter-

1150

Carl Mitcham and Adam Briggle

minations through behaviors that actualize inherent potencies. At the individual level these supplemental determinations were called character; at the social level, cultures and political regimes. Their multiplicity provoked systematic (that is, in the classical sense, scientific) analysis and assessment. Roman philosophers, carrying forward the Greek tradition, likewise examined the mores (Latin plural for ethos) of peoples, in what came to be called moral theory. Ethics and moral theory are but two terms for the same thing: systematic reflection on human conduct that seeks to understand the good for humans and thus serve as a basis for prudential guidance in human affairs. Although Cicero (106-43) did not explicitly include the arts, his assessment of the moral obligations associated with social “offices” is a formulation of role responsibility with general applicability. Role responsibility has served as a framework for understanding the moral obligations associated with the traditional forms of artisanship and the modern professions of medicine and engineering. During the Middle Ages these articulations of moral theory (science) and practice (technique) were subsumed within a framework of divine revelation. According to Thomas Aquinas (1224-1274), for example, the supernatural perspective allows Christians to provide more perfect insights into the ultimate nature of reality and the human good than was possible for pagans. What for Aristotle could be no more than the counsels of practical wisdom became for Thomas natural laws of human conduct, laws that gear down the cosmic order and are manifest in human reason as a “natural inclination to [their] proper act and end” (Summa theologiae I-II, q.91, a.2). Additionally, influenced by the revelation of humans as created in the image of a creator God, Christians began to take special interest in technology. The century before Thomas witnessed the writing of the first book on tools (Theophilus Presbyter’s De diversibus artibus), conceptualization of seven mechanical arts as complements to the seven liberal arts (in Hugh of St. Victor’s Didascalicon), and an argument for technics as a way to remedy the loses of the Fall (another idea from Hugh). A contemporary of Thomas, Roger Bacon (12141294), even began to promote the development of a techno-experimental science and to imagine the possibility of such technical inventions as microscopes, telescopes, steam ships, and airplanes. Despite the vast differences among such premodern thinkers, we can identify a fairly consistent view on the relations between ethics and technics. As Hans Jonas [1984] has argued, technics itself made no claim to high moral purpose. Unlike politics, virtue, or religion, for instance, technics was a quite limited aspect of human life — limited in power and effect. Both scripture and ancient political thought worked in a moral language of virtue, character, purpose, and discipline that instructed about proper human form and ends. They espoused a worldview where limits to the pursuit of technical intervention in self and world were crucial to self-perfection. The traditional forms of ethics thus tended to argue for restraint in the independent, progressive pursuit of science and technics. Of course, technical skill was valorized when pursued within such limits and toward worthy goals such as the preservation of life and community. But the lim-

The Interaction of Ethics and Technology in Historical Perspective

1151

its were all important, because technical activity can quickly be overextended and create wealth that undermines virtue, change that weakens social stability, and a will to power at odds with natural piety or human flourishing [see Mitcham 1994a, pp. 275ff.]. Moreover, in general, the premoderns judged artifacts to be less real than natural objects and technical knowledge as, correspondingly, on a lower level than other types of knowledge. Restatements of such premodern positions can be found in, for instance, the work of the neothomist Jacques Maritain (1882-1973), the Jewish scholar Leo Strauss (1899-1973), and the radical Catholic social critic Ivan Illich (1926-2002). Their message of limiting the pursuit of technics takes on a particularly contentious character when applied to medicine and agriculture. 2 MODERN ETHICS AND TECHNOLOGY Beginning in the 1500s, the modern period witnessed an emerging transformation in the understanding of ethics, one related to a transformation in science and technology themselves. The scientific understanding of nature came to focus no longer on the natures of different kinds of entities, but on laws that transcend all particulars and kinds. The knowledge thus produced contributed to the transformation of technics into technology. This transformation denotes a change in scale from small handcrafts to large machines and industrial systems and a shift from animate to inanimate energy sources. This offered a new level of power to control or reorder matter and energy for external ends. These external ends increasingly came to be understood in terms of this-world human autonomy and welfare. Technological science thus became the basis for a progressive technological activity that produced artifacts more systematically and in greater abundance than ever before. Thus, underlying the shift from technics to technology is also a fundamentally changed vision of the relationship between humanity and the order of things. This vision and its realization through technology were supported with ethical arguments by Niccol` o Machiavelli (1469-1527), Francis Bacon (1561-1626), and Ren´e Descartes (1596-1650). Human beings deserve to manage and transform the world. This vision, going well beyond that of the chorus from Antigone, in which humans discover their place in nature through technical activity, is one of turning away from reflective observation toward a knowledge that enables humans not just to operate within but to control and subdue their environment. Moreover, no longer content to aspire with spiritual longing for recovery of a prelapsarian paradise, Machiavelli’s new politics emphasized virtue as power while Descartes’ new science aimed for humans to become “masters and possessors of nature” (Discourse on Method, Part 6). Where the ethics of technics had been one of properly proscribed limits, the ethics of technology was envisioned as infinite progress. It was Bacon who most forcefully articulated the distinctively modern ethics of technology. In The Great Instauration (1620), on the basis of a moral vision of human beings as unjustly suffering in the state of nature — a vision supported by his creative deployment of Christian revelation — Bacon criticized Greek philosophy as a vanity of words and prayed for a new beginning in which natural

1152

Carl Mitcham and Adam Briggle

philosophy would pursue knowledge linked to power. “I would address one general admonition to all,” he wrote, that they consider what are the true ends of knowledge, and that they seek it not either for pleasure of the mind, or for contention, or for superiority to others, or for profit, or fame or [political] power, or any of these inferior things; but for the benefit and use of life.... I am labouring to lay the foundation, not of any sect or doctrine, but of human utility and power. [“Preface,” paragraphs 5 and 6] To this end his proposal for a new, methodical engagement of the mind with nature would be based in a history not only of nature free and at large (when she is left to her own course and does her work her own way), — such as that of the heavenly bodies, meteors, earth and sea, minerals, plants, animals, — but much more of nature under constraint and vexed; that is to say, when by art and the hand of man she is forced out of her natural state, and squeezed and moulded. Therefore I set down at length all experiments of the mechanical arts, of the operative part of the liberal arts, of the many crafts which have not yet grown into arts properly so called .... Nay (to say the plain truth) I do in fact (low and vulgar as men may think it) count more upon this part both for helps and safeguards than upon the other; seeing that the nature of things betrays itself more readily under the vexations of art than in its natural freedom. [“Plan of the Work,” paragraph 21] In his use of the term “art,” of course, Bacon means to reference technics if not technology. In further contrast to the ancients, for Bacon technical change is inherently beneficial because it enhances human welfare and autonomy. People suffer more from the elements than from other human beings; they should therefore work together to conquer nature through science and art. As historical proof for his position Bacon sites how the inventions of printing, gunpowder, and the compass have been of more benefit to humans than all previous political activity, philosophical debate, or theological argument [Novum organum I, 129]. In the following two centuries first the Enlightenment and then the Industrial Revolution flourished in conjunction with the progressive articulation of such ideas about how humans might, through a new linkage between science and technics, remake both the physical and human worlds to satisfy desires. The creation of modern economics as a theory that endorses the pursuit of individual material self-interests was arguably the single most influential promotion of this linkage. Ethics also began to be re-systemized and moved from a reliance on prudential guidance toward the formulation of rules for human conduct. Divides emerged among different rule-focused ethical systems, but the major approaches nevertheless agreed in trying to formulate ethical decision making processes that could be

The Interaction of Ethics and Technology in Historical Perspective

1153

practiced with competence and regularity on scales that would be able both to advance and to cope with the new powers of industrialization, technization, and globalization. The modern period thus witnessed the development of ethics as a science with a unique intensity and scope. The theoretical development of a new science of ethics emerged in different but related forms within the empiricist and rationalist traditions of modern European philosophy. In the empiricist tradition, as exemplified in the work of David Hume (1711-1776), morality was argued to be based in human sentiment, which gave to ethics a subjectivist cast. Elaborating this perspective Jeremy Bentham (17481832), John Stuart Mill (1806-1873), and their followers developed a utilitarian theory that understood morality as rules for the pursuit of happiness by maximizing pleasures and minimizing pains. Pragmatist ethics replaced happiness by a wider conception of projected ends but continued to stress instrumental effectiveness in their pursuit. In the rationalist tradition, by contrast, Jean-Jacques Rousseau (1712-1778) and Immanuel Kant (1724-1804), grounded morals in a noninstrumental rationality of inner consistency. Kant and his followers developed a deontological or duty-focused ethics in which moral behavior was assessed in terms of intentions and their universalizability. This appeal to rationality ascribed to ethical principles a certain non-empirical but nonetheless objective character analogous to that found in mathematical laws. Both traditions were at one, however, in struggling to deal with the ethical challenge created by the loss of nature as embodying a normative potency, an inner reaching for perfection, within and without human beings. Prior to the modern period, natural entities were understood as possessed of functional tendencies toward harmony with the orders of being. When they functioned well and thereby achieve such teloi (or fitting ends), then fire ascended, seeds grew into flowers and trees, animals matured and gave birth to offspring, human beings spoke with one another and made offerings to the gods. Furthermore, fire and trees and humans fit in with and were parts of greater natural orders. Since these harmonies or proportionalities are what constituted being itself, they were also good, which is simply the way that reality presented itself to and drew forth or perfected the appetite. For the premoderns, moral practice was thus oriented toward the perfection of human nature. By contrast, insofar as nature came to be seen as composed not of entities with natures to be realized, but as indifferent matter able to be used one way or another and modified at will, questions arose about the foundations of the good as an end to be pursued and the rightness of any means to be employed in such pursuit. The romantic, rear-guard defense of nature as an aesthetic phenomenon only succeeded in modifying the modern trajectory at its margins. Instead, modern ethics initially manifested a basic shift, once teleology was replaced by balances of material forces, from efforts to identify the good as a natural end to goods as desires or ideals. Stated another way, the moderns replaced “the good” with “goods” or “values.” The good in the premodern sense was understood as a standard, that of reality as an ordered whole transcending personal interests, that could serve as a

1154

Carl Mitcham and Adam Briggle

guide for the assessment of such interests. Modern values took whatever was the object of personal interest or the interest-producing entity itself as the standard. Efforts to make consequentialst and deontological systems truly scientific have been variously pursued. During the late nineteenth and early twentieth centuries pragmatism sought to integrate especially the empirical social sciences into assessments of what might be the most effective means to pursue ends of interest. In the first half of the twentieth century aspirations to more formal rigor led to the development of metaethics. Eschewing normative analysis, metaethics aspired to clarify the structure of ethical language and its distinctive logic. In its most radical form metaethics reduced the meaning of ethical statements to forms of emotional approval; in more moderate forms it simply disclosed the complexities of ethical judgments, sometimes seeking to rectify inconsistencies. During the middle of the century ethics took more operational form in the mathematics of game and decision theory, operations research, and risk-cost-benefit analysis. In the last third of the twentieth century, however, the inadequacies of the social sciences, metaethics, and the formalization of decision theoretic procedures in the face of substantive issues presented especially by the creation and use of technology brought about the development of applied ethics. One formerly metaethical philosopher interpreted this intellectual turn as a transition that “saved the life of ethics” [Toulmin, 1982]. In effect, this also proved the occasion for a revitalization of pragmatist ethics (see, e.g., [Keulartz et al., 2002]). 3

ETHICS REFLECTING ON TECHNOLOGY: THREE SCHOOLS

Beginning with the early twentieth century, ethics became increasingly engaged with technology across a wide range of issues. So extensive has been this engagement that any overview of ethics interactions with technology is compelled to adopt some kind of simplifying perspective in order to attempt an approximate coverage. For present purposes, developments will be described as taking place in three distinctive ethical contexts with consequentially different emphases. Two of the contexts had their roots and orientation in the rationalist European tradition and were manifested in what will be termed socio-critical and historico-cultural schools or approaches. The third was more empiricist in orientation and associated with what become known as the modern analytic practice of philosophy. But it is important to note that by the end of the century mergers from these three schools were the norm, so that the narrative here should be used primarily to stimulate appreciation of how different perspectives are coming to mutually influence one another especially in relation to technology. As Table 1 indicates, the ethical problem space was originally defined somewhat distinctly for each of the three schools. Socio-critical approaches were generally concerned with reforming economic and political structures associated with technology in order to better accord with an ideal of human freedom; historico-cultural philosophy addressed questions of the meaning of life; and analytic work sought clarity in conceptualization and argument. In both the socio-critical and historico-

The Interaction of Ethics and Technology in Historical Perspective

1155

Table 1. Three Schools of Ethical Reflection on Technology School Frame

Emphases

Roots

Socio-critical Technology as productive process with potential for human liberation Historically informed orientation and social reform Rationalism, Marxism, pragmatism, and neoliberalism

Historico-cultural Technology as way of being and perceiving and threat to authenticity Historically informed orientation and personal meaning Rationalism, phenomenology, and existentialism

Analytic Particular technologies in context as problems or solutions Isolated concepts and piecemeal problem solving Empiricism and utilitarianism

cultural schools, discussions of ethics and technology developed with reference to the kind of historical background sketched in sections one and two. In particular, both contended that modern technology uniquely transformed the human condition and that received moralities were inadequate to address the altered situation. Phenomena such as dehumanization or inauthenticity were seen as characteristic of the historical emergence of modern technology in general rather than associated with any particular kind of technology. Whether workers were engaged with steam engines, chemical processing plants, electronics, or nuclear power does not matter; in each case they were faced with existential antagonisms between their socio-political aspirations, lived experience, and material culture. Steam engines dwarfed human physical activities, chemical processes poisoned and polluted, electricity and magnetism escaped any immediate grasp by the human sensorium, nuclear power contained inconceivably destructive potential. But there were disagreements within and between these first two traditions about the particular ways in which technology has altered the human lifeworld and about the appropriate moral responses. Analytic ethics, by contrast, worked with more isolated and well-bounded problems and tended to reject the notion of modern technology as a fundamentally new mode of human experience and social order. For analytic philosophers it was not history but problems that are controlling. Finally, the analytic tradition at least initially maintained a boundary between facts and values. It accepted the scientific knowledge of facts as a cognitive paradigm, with values understood as expressions of non-cognitive human sentiments or interests that come into conflict and as such need to be clarified and adjudicated. By contrast, Marxism saw both science and morality as expressions of class interests, while phenomenology viewed scientific knowledge as a restricted if not diminished form of cognition. Additionally, in the socio-critical and historico-cultural traditions facts and values were argued to

1156

Carl Mitcham and Adam Briggle

both be phenomena and thus equally open to systematic analysis and criticism. Thus, neither of the first two schools poses any radical gap between facts and values. Attitudes toward an alleged fact/value distinction over the course of the century become much less a dividing line between the analytic and the other two schools but their approaches to the relationship between these two aspects of experience have remained modestly at odds. To explore such comparisons in greater detail, however, it is appropriate to proffer a more extended examination of these distinctive schools.

3.1 Socio-critical approaches The most influential figure in the socio-critical school has been Karl Marx (18181883), whose “critique of political economy” aimed to undermine what he saw as naive beliefs in the political benefits of industrial technology and associated economic structures. According to Marx, “the modern science of technology” undermined traditional skills and the satisfactions of craft production, placing workers under the control of large-scale, capitalist-owned factories in which labor functions became equal and interchangeable [Das Kapital I, 13]. This disturbed a traditional social ecology in which the “species essence” of material production was once directed to the general human welfare, a corruption that could be corrected only by means of a social revolution in ownership of the new technologies. The Marxist ethical assessment of industrial technology thus highlighted technological change as restructuring society such that prior economic orders were made obsolete. The focus of Marx’s critique was not on the quality of the emerging consumer society, but the maldistribution in power over production. His position was opposed to the attempts of both “utopian socialists” and liberal economists to manage the creative destruction of technological change. On one hand, Henri de Saint-Simon (1760-1825) in France called for a “New Christianity” to manage society through a technocratic linkage of scientists, artists, and industrialists, while Robert Owen (1771-1858) in England established cooperative worker-owned industrial model communities such as New Lanark. On the other hand, classical liberal economists conceived of production in terms of inputs and outputs organized by what Adam Smith (1723-1790) described as the “invisible hand” of the free market. For Marx, technocratic management was not enough, because ordering the technological society required more than just technical knowledge, and idealistic model communities were unable to transform entrenched techno-political institutions. Similarly, though liberal economists recognized the primacy of production over politics, they failed to appreciate that productive processes were always also social processes. Marx’s effort to liberate techno-economic powers from bourgeois class interests rested with a new analysis of production. It examined “how the instruments of labor are converted from tools into machines” and the way machines themselves tended to become organized into a system in which “the subject of labor goes through a connected series of detailed processes” [Das Kapital I, pt. 4, Chapter 13, Section 1]. In an economy where capitalists owned the means of production,

The Interaction of Ethics and Technology in Historical Perspective

1157

workers existed as wage slaves tied to specific mechanical routines. Only if the means of production were placed in worker hands would they be free in reality to “become accomplished in any branch [they wish],” to “do one thing today and another tomorrow” [Die deutsche Ideologie I, 1, a]. Only liberated from the capitalist mode of production would modern technology realize its ability to promote not only justice but also true human freedom. The reality was that technology made wealth and freedom possible for everyone, but its historical appearance realized this ideal only for the few capitalist owners. Subsequent economic developments that reduced extreme depredations of poverty without bringing about worker liberation in other forms, gave rise to what has become known as the Frankfurt School of social theory in the work of Max Horkheimer (1895-1973), Theodor Adorno (1903-1969), and Herbert Marcuse (1898-1979). Their work and that of such leading students as J¨ urgen Habermas, William Leiss, and Andrew Feenberg shifted critique from a focus on political economy to a questioning of the character of the Enlightenment, attributing the failure to realize the full liberating potential of technology not simply to economics but to culture. For Horkheimer and Adorno, the Enlightenment produced “instrumental reason” — without, however, providing a guideline in objective reason (theory) for how the new powers of reason were to be used. This led to the production of social orders dominated by the military and the “culture industry,” that is, brute force and entertainment. Rather than leading to an ever-larger conversation about goals and values, Enlightenment reason was reduced to instrumental thinking concerned with increasing the efficiency of means in order to achieve already given ends. Their dystopian conclusions were similar to sociologist Max Weber (1864-1920), who described the process of rationalization, or the increasing role of calculation and control in industrial democracies, as creating an “iron cage” of bureaucracy that stifled individual freedom. Marcuse rejected this pessimism, arguing that although technology was oppressive under capitalism, it might be otherwise under a different social order such as that foreshadowed by the student counterculture of the 1960s or the women’s liberation movement. In place of Marcuse’s apotheosis of the counterculture, Habermas developed a theory of communicative action as a formal guide for political and technical development. A number of other efforts that may be loosely associated with critical social theory focused directly on promoting the democratic participatory control of technology. Langdon Winner [1986] and Richard Sclove [1995], for instance, argue that technological artifacts even more than political institutions influence the way people lead their lives. Given the ethical principle that individuals should have a say over what affects them, technological decision making and design deserve to be subject to the same standards of public participation as political decision making. Such threats as computer invasions of privacy and technological transformations of the environment only reinforce the ethical principle of “No innovation without representation” [Goldman, 1992]. More positively, Feenberg [1995 and 2000] offers proposals for reconfiguring the diverse possibilities of technology. These include

1158

Carl Mitcham and Adam Briggle

action by which workers would recontextualize their labor, public recognition of the human significance of vocation, investment of aesthetic and ecological value in technological products, and the pursuit of voluntary collegial cooperation in work. In a complementary manner the institutional sociologist Thorstein Veblen (18571929) argued that engineering attitudes and technological achievements were being corrupted by pecuniary interests and the price system. Controlled by persons of limited perspective, the full potential of technology was being thwarted. Again, the full capabilities of technology for contributing to human welfare were not being realized. Such arguments all constitute attempts to re-enclose technology within a new social framework. Despite important variations, the main theme has remained: the liberating powers of technology can only be realized under the right social circumstances, which thus deserve careful and conscious restructuring. A diametrically opposed ethical argument for liberation of the full potential inherent in technology can be found in neoliberal entrepreneural and libertarian economics that became prominent in the latter third of the twentieth century, drawing inspiration especially from the thought of Frederich Hayek (1899-1992). The neoliberal revival and defense of a classical liberal economic perspective on technology relied on Hayek’s distinction between two fundamentally different types of human making. Material artifacts can be the result of conscious or intentional technological design. But such human constructions as language and the free market are “the result of human action but not of human design” and are not subject to direct intentional control [1967]. Indeed, for Hayek and others any attempt to control or manage technology in the name of equality will not only be counterproductive but may also require unethical restrictions on human freedom. Additionally, the libertarian political philosopher Robert Nozick [1974] considered the alleged general principle of “having a say over what affects you” and found it wanting. He defended as legitimate, innovation without representation, when the innovation is pursued in relation to one’s own primary goods, even though it may have secondary effects on others. Thus, classic liberal economic theory made an essentially moral argument for the liberation of science and technology from state control. Liberated from political control, industrial technology would enhance human freedom and welfare and, not incidentally, limit state power. Moderate versions of liberal economics, however, have recognized the need to complement liberty with order and equality, which often necessitates a modest amount of state intervention, although nothing as radical as a political revolution in the ownership of technological property. Finally, the pragmatist John Dewey (1859-1952) developed a different argument for complementing liberty with equality in the exercise of technology. For Dewey, the black box of technology was opened not so much as a social production process as one of cognition and practice. Human beings, like all organisms, exist in an environment in which they are trying to achieve specific ends or what Dewey calls “ends in view.” When frustrated in their efforts to achieve these ends, problematic situations arise that people have the potential to subject to conscious analysis and to inquire concerning possible responses — entertaining hypotheses about modify-

The Interaction of Ethics and Technology in Historical Perspective

1159

ing the ends to be pursued or improving the means to be utilized. Human beings then test the alternatives and undertake new forms of behavior on the basis of what they learn. As it becomes increasingly conscious and effective, this process of inquiry and improvement in human thought and action is, for Dewey, what is meant by technology. As argued by Larry Hickman [1990], Dewey’s notion of the free development of technology so construed should not be limited to industrial production but should be generalized and applied equally across the full range of human experience, from social and political affairs to art and religion.

3.2

Historico-cultural approaches

In the historico-cultural school of the assessment of technology, reflection was framed not in terms of technology as a productive process and its possibilities for human liberation, but in relation to science and technology as forms of consciousness. From this perspective, the primary danger has been argued to be some form of inauthenticity or bad faith — that is, failures on the part of individuals to recognize and accept the ways in which they are responsible for the lifeworld they create and their attitudes toward it. The attacks on the inauthenticity of Christian culture by Søren Kierkegaard (1813-1855) and the iconoclastic pronouncements of Friedrich Nietzsche (18441900) founded what has been called the existentialist movement, which involved a special ethical stance toward technology. Although he did not thematize it as such, from the perspective of Kierkegaard modern technology could be interpreted as a form of bad faith. For Nietzsche, technology might be described as genealogically rooted in a slavish morality that valued convenience and safety over a life of heroic reach and challenge. Nietzsche’s two-pronged attack on scientism and technological culture — scientism understood as the claim that science is the highest form of knowledge and technological culture as one in which massification and consumerism have conspired to corrupt nobility of achievement in art, music, and literature — exposed a nerve of doubt and resistance to the trajectory of technological progress. Science and technology threatened to trap humans in an impoverished existence that denied their deepest truths. As such Nietzsche has fertilized an ethical stance deeply at odds with that of Francis Bacon and his Enlightenment heirs, one at the root of a diversity of existential ethical engagements with technology. It was Edmund Husserl (1859-1938) who turned genealogical analysis into a method that he named phenomenology. Husserl refused to accept science on its own terms and argued that science was not self-explanatory. His description of the genesis or coming to be of scientific phenomena disclosed that more fundamental than science was a lived experience or lifeworld from which modern science itself was abstracted with the aid of technologicial instruments. The technoscientific world was argued to be a reduced or diminished form of the lifeworld on which it remained, often unknowingly, dependent. Husserlian phenomenology was concerned with disclosing the ongoing framework — that pre-existing familiarity with

1160

Carl Mitcham and Adam Briggle

the world — that made all human experience possible. As Maurice Merleau-Ponty noted, the return to the lifeworld or “the things themselves” was “from the beginning a disavowal of science” [Merleau-Ponty, 1945, p. ii]. Humans are not objects of biological, psychological, or sociological investigation. All knowledge, including scientific and technical knowledge, can be argued to be gained from a particular point of view or experience without which the symbols of science and engineering would be meaningless. Merleau-Ponty and others made more explicit Husserl’s implicit ethical critique of the modern scientific worldview and the lifeworld transformed by modern technology. Three other important contributors to development of the historico-cultural criticism of technology were Max Scheler (1874-1928), Jos´e Ortega y Gasset (18831955), and Martin Heidegger (1889-1976). According to Scheler, the historical transformation of the lifeworld was more than an economic or productive phenomenon; it was also the rise and dominance of a new “ethos of industrialism” even among technical workers themselves. Such an ethos (which is intermediary between moral principles and moral actions) exalted utility and instrumental values over vital and organic ones. This is a distortion not just of the economic order but of an axiological hierarchy — a distortion that calls for a cultural reformation. For Ortega, however, right within modern technology and the ethos of industrialization there arose a moral problem that cannot be addressed by means of either social revolution or cultural reform. Scientific engineering, in contrast to traditional craft technics, radically increased what can be done without any corresponding deepening of ideals about what should be done. In Ortega’s formulation of the issue: Previously human beings, struggling to achieve some vision of what it was to be human or a lifeworld, only acquired a particular technics in a form already embedded in an existing cultural project; as a result they only possessed a particular instance of what might be called technology in general. But with technology and the engineering sciences human beings possess technology in general disembedded from any particular cultural project. They are thus able to do almost anything prior to having any idea about what they really want to do. To address this problem, Ortega concluded his Meditaci´ on de la t´ecnica [1939] by suggesting the need to cultivate what he called “technics of the soul.” The suggestion is perhaps echoed in subsequent proposals by G¨ unther Anders [1961] for a moral education capable of matching the power of our imagination (vorsellen) with the technological expanded power of our abilities (herstellen). Heidegger was undoubtedly the most influential European philosopher to address the issue of ethics, science, and technology from the historico-cultural perspective, even though he rejected the discipline of ethics as such. In his “Die Frage nach der Technik” [1954] Heidegger undercut the distinction between science and technology, and argued that modern scientific technology — or what Bruno Latour [1987] would latter call technoscience — is not so much an ethos as a form of truth. This truth or knowledge reduces the world to Bestand or resources available for manipulation by a world-configuring, nihilistic destiny he calls Gestell. Heidegger seemed at once to make ethical reflection more necessary than ever before and to

The Interaction of Ethics and Technology in Historical Perspective

1161

destroy its very possibility. Less controversial is his thesis that science and technology are interpenetrating practices: the science of nuclear physics is as much the applied technology of cyclotrons and reactors as the technology of nuclear engineering is applied nuclear physics. To the extent this is the case, the ethics of science tends to merge with the ethics of technology. Two other thinkers who are not always identified as philosophers, but who have nevertheless contributed to the historico-cultural assessment of technology, were Lewis Mumford (1895-1990) and Jacques Ellul (1912-1994). As a social historian and cultural critic, Mumford argued that modern age technics (his term for technology) has transformed the making and using of artifacts into a complex sociotechnical system oriented almost exclusively toward power and control. This “mono-technics” was actually foreshadowed by premodern slave-labor “megamachines” for large-scale construction projects such as the Egyptian pyramids. In contrast, Mumford promoted the recovery of “bio-“ or “polytechnics” oriented toward a living multiplicity of human interests and activities, from religious ritual to aesthetic creativity and play. As a sociologist and theologian, Ellul distinguished technical operations from their distinctively modern unification in the technical phenomenon or la Technique (translatable as “technology”) [Ellul 1954]. The distinctive feature of this phenomenon, in Ellul’s view, is the effort to turn human activities toward the pursuit of some form of efficiency, that is, to assess all dimensions of culture in terms of an input-output analysis. As efficiency analysis comes to dominate in economics, in politics, and even in health care, education, and sports, technology takes on a semi-autonomous character that undermines human freedom. To counterbalance such technological determinism — that is, the making of decisions always with a concern for opportunity costs, risk-benefit analysis, or other forms of calculative rationality, the combination of which is in effect to enhance technological power — Ellul argued for an “ethics of non-power” that would encourage individuals voluntarily to delimit technicization, especially in their relationships with persons and with nature. One way to interpret Michel Foucault (1926-1984) and his controversial analyses of the disciplinization of modern life, from the insane asylum to biopolitics, is as providing a complementary perspective on other aspects of the same phenomenon with which Ellul was concerned. Finally, Albert Borgmann [1984] offers another critique and response to the technology-culture relationship in terms of human meaning or the good life. For Borgmann, technology takes shape as a ruling pattern of human experience that he calls the “device paradigm.” The conveniences of consumer culture offered by mass-produced devices are alluring but ultimately impoverished substitutes for engaging experiences with “focal things” or through “focal practices.” Although only briefly summarized here, Borgmann perhaps more than any other cultural critic enters into extended dialogue with diverse approaches, from economics and social criticism to political theory as well as systematic defenders of existing technological trajectories, and then deftly restates with renewed vigor a full spectrum of major concerns about the cultural consequences of technology. In Borgmann,

1162

Carl Mitcham and Adam Briggle

for instance, one can find echoes of Illich’s concept of a “counter productivity” in technological progress that undermines friendship, community, health, and other experiences central of meaningful lives, but presented in the framework other than Illich’s own slash-and-burn attack; for Borgmann it is always necessary to offer a positive vision that can motivate efforts to recapture missing aspects of a once vibrant lifeworld. In addition, one can find Borgmann himself echoed in worries by Leon Kass [2003] concerning consumer driven experiments in the pursuit of improved children, performance enhancement, ageless bodies, and psychological happiness. The holistic approach characteristic of the historico-cultural school is nevertheless dependent on some attempt to address particular issues, if only to illustrate more comprehensive claims. Borgmann’s contrast of the premodern focal hearth and the central heating system to illustrate the device paradigm is a case in point. Some theorists can be interpreted as heirs of the historico-cultural tradition, even when they rely more heavily on context-specific studies, an emphasis Hans Achterhuis [2001] labeled as the “empirical turn” in ethical reflection on technology. Practitioners often retain an interest in evaluating science and technology in terms of their broader human and cultural significance, but do so by relying on empirical case studies of specific artifacts and actual practices, which makes them open as well to analytic approaches (e.g., [Verbeek, 2005]). Those adopting the empirical turn in the contextual examination of technological artifacts, ask both how specific artifacts mediate or shape experience and how user behavior shapes the function or meaning of artifacts. For instance, Donna Haraway [1985] uses detailed studies to support her theory of the “cyborg,” a being whose lifeworld, self-interpretation, and social context are permeated by technology. By questioning normalized categories, she de-naturalizes (shows the contingency of what had seemed necessary) the power of those who define the categories, thus using cyborg theory to advance feminist, socialist, and anti-racist ethical ideals. Additionally, Diane Michelfelder [2000] argues that what is important for the lifeworld is the moral significance of material culture understood from the viewpoint of actual user experiences. A mass-produced box of wine may be a “device” for Borgmann, but if shared at a festive meal it may still function as a “focal thing.”

3.3 Analytic approaches I: background As they came to the fore during the mid-twentieth century, analytic engagements with technology tended to reflect two beliefs: that technology has not fundamentally altered the human condition and that received ethical theory is largely adequate — or, paradoxically, irrelevant — to address associated moral issues. While viewing technological development as continuous and progressive from stone tools to electronic computers, and by questioning any need for radically new moral theory, analytic approaches adopted a loose method focused on a dialogue between mid-level theory and concrete issues seeking to clarify divergent problems asso-

The Interaction of Ethics and Technology in Historical Perspective

1163

ciated with diverse technologies. Indeed, a progressive orientation can also be detected in the genial confidence with which problems were approached as meliorable if not solvable. Being the broadest of the three schools of ethical reflection, any attempt to survey analytic approaches is difficult. The only major individual who stands out as consistently engaged with technology was Bertrand Russell (1872-1970), who considered ethics a more public intellectual activity than one representative of professional philosophy. His contributions often tended toward political provocation, as in Why I Am Not a Christian [1927]. A life-long anti-authoritarianism and sympathy for the underprivileged led to both active protest against World War I (for which he served six months in jail) and a consistent skepticism about the social benefits of what he called scientific technique. Icarus, or the Future of Science [1924] was written in explicit criticism of the optimism found in J.B.S. Haldane’s Daedalus or Science and the Future [1924]. Although Russell defended scientific knowledge as an ultimate truth about the world with manifest human benefits in the technologies of food and consumer goods production and health care, he also emphasized the manifold ways such knowledge could be abused when under the control of totalitarian governments or those deficient in a relevant education [Russell, 1951]. For Russell the only hope was better democratic education in science and technology [see, e.g., Russell, 1958]. Most dramatically, Russell strongly opposed the development, use, and spread of nuclear weapons; in 1955, after the U.S. testing of the hydrogen bomb, he drafted a manifesto co-signed by Albert Einstein that called on scientists and engineers to take more public responsibility for their work — and thus stimulated the 1957 creation of an on-going series of Conferences on Science and World Affairs that became known as the Pugwash Movement (after the name of first meeting place, Pugwash, Nova Scotia). The question of technology and warfare was the leading edge of analytic ethical engagements with technology. Along with Russell, for instance, G.E.M. Anscombe (1919-2001) — a student of the other major figure of analytic philosophy, Ludwig Wittgenstein (1889-1951) — opposed British bombing policies in World War II (and in the 1960s endorsed the Catholic rejection of certain technological means of contraception). Other analysts brought their methods to bear on postwar nuclear deterrence policy, and from there analysis spread out across diverse categories of technological activity: chemical processing and manufacturing; biomedicine; information, communication, and media technologies; agricultural and biotechnologies; and nanotechnologies. (New categories on the horizon include emerging/converging technologies and synthetic biology.) Earl Winkler and Jerrold Coombs [1993] consolidate this diversity into three basic applied ethics domains — bioethics, business ethics, and environmental ethics — and argue that in each case problem generation was closely associated with advances in technology. Adding to these the neglected domain of computers and information technology yields four basic domains of applied (mostly analytic) ethical reflection on technology. As Winkler and Coombs additionally suggest, the analytic ethics of technology can be situated as part of a shift within philosophy itself from metaethics to ap-

1164

Carl Mitcham and Adam Briggle

plied ethics. One pivotal contribution to this transformation was work by Toulmin. For Toulmin [1958], ethical decision making could be usefully conceived as learning to appreciate and give good reasons for selected courses of action. He made reference to the reasonableness of drawing on standards of professional practice in fields such as engineering and the efforts of moralists to convert ethical possibilities into practical policies. As noted previously, this orientation toward practical utility was given enhanced articulation in an argument praising biomedical ethics for moving ethics away from abstractions toward situations, needs, and interests; emphasizing cases; and relating these cases to traditions of professional practice [Toulmin, 1982]. Appropriately enough, however, what constitutes applied analytic ethics has itself been subject to analysis, and thus became another contentious issue in need of conceptual clarification. Although the applied turn originated with efforts to bring ethical theory to bear on practical problems associated with technology, it was quickly discovered that abstract theory was seldom directly useful. In many cases it was irrelevant. What functioned better were mid-level principles that could enter into mutually informing dialogue with particular problems. Extending the movement toward the particular, Toulmin and others thus undertook to revive the tradition of bottom-up, case-based casuistry [Toulmin and Jonsen, 1988]. The theory behind such an approach argues for understanding moralities as social artifacts or institutions functioning as basic elements of a culture that must be appreciated as such — that is, in terms of the ends they serve and their historical evolution. Contextualism — and appreciation of the different contexts created by different technologies — thus becomes a primary pathway into analytic ethical reflection. Although each of the contextual domains of reflection in the analytic ethics of technology is marked by its more or less unique discourse, it is nevertheless possible for synoptic purposes to identify four overarching themes falling into two broad areas. In the spirit of the analytic school itself, however, these themes remain no more than weakly linked. The broad areas are those having to do with (1) issues of justice and equity and (2) autonomy and liberty. Certainly it is the case that modern ethics and politics have repeatedly manifested opposing arguments for the primacy of equity or fairness and associated notions of solidarity and community (John Rawls) versus a stress on individual rights, private property, and near anarchic liberty (Robert Nozick). With respect to the first nexus of values, it is also simply an observational fact that justice and equity issues concerning technological benefits and harms relate to such key concepts as (1a) human health, safety, information, privacy, and risk and (1b) human and non-human welfare in relation to concepts of environmental pollution, obligations to nature, and sustainability. With respect to the second value perspective, it is again a matter of fact more than logic that the relation between technology and personal autonomy or liberty has been discussed especially in relation to themes of (2a) technical professional or producer responsibility and misconduct and (2b) public consumer or citizen participation in contrast to technocratic expertise. On the basis of such observa-

The Interaction of Ethics and Technology in Historical Perspective

1165

tions, Table 2 makes an analytic effort to summarize the basic issues manifest in analytic progressivism, with attention drawn as well to some of the key positive and negative concepts at play in the different contexts. Table 2. Themes in Analytic Approaches of Technology Concept modality Area of concern General issue (1) (a) Humans (b) Non-humans General issue (2) (a) Technical professionals (b) Public consumers and citizens

3.4

Benefits and Goods

Harms, Actual or Potential

Justice and Equity in relation to Health and Safety Risk Information Loss of privacy Environment / Sustain- Pollution ability Autonomy and Liberty in relation to Responsibility / In- Fraud and Misconduct tegrity Participation Technocratic Expertise

Analytic approaches II: selective specifics

Historically, social justice issues arose in relation to the distribution of technological goods and services, and have exhibited a movement from concern for the fair distribution of positive benefits among humans to an emergent concern for negative impacts on animals and eventually the environment. Any number of analyses of overlapping efforts at conceptual clarification and application linkages related to health, safety, privacy, risk, environmental pollution, and sustainable development are characteristic of this theme cluster. Kristin Shrader-Frechette, for instance, has argued that virtually the whole question of ethics and technology can be subsumed under various aspects of risk analysis: how to define risk, how to evaluate technological uncertainties, threats to due process from non-compensatable risks, risk assessment methods, the determination of socially acceptable risk (or safety), and consent to risk. Stimulated by ethical questions concerning nuclear power and public policy [ShraderFrechette, 1980], she defends extending the concept of free and informed consent from medicine to technology in general [Shrader-Frechette, 1991]. Persons should be subject to technological risk only on the basis of intelligent personal assessment of those risks and choices not unduly constrained by economic pressures. Indeed, in the spirit of populism she maintains that laypersons are often more rational in their assessment of risks than experts (see Hansson’s chapter “Risk and Safety in Technology” in this Volume, which clarifies how questions of risk have expanded well beyond issues related to nuclear technology.)

1166

Carl Mitcham and Adam Briggle

As exemplified in the domain of environmental ethics, the analytic approach again began with efforts to clarify rather high-level concepts such as environment, nature, pollution, wilderness, ecological systems, and sustainability. Although initially reflecting a top-down emphasis on distinctions between anthropocentric and non-anthropocentric (or biocentric) theories, extrinsic and intrinsic values, and arguments for obligations to future generations and the rights of nature, the analytic ethics of technology and the environment has evolved toward more contextualized interests in working with ranchers, farmers, forest and park managers, and even tourists and urban consumers to clarify interactions between their evolving social moralities and trans-human environments transformed by human technology. An example of this “policy-turn” in environmental ethics can be found in Adam Briggle’s analysis of the controversies surrounding the proposed wind farm in Nantucket Sound [Briggle 2005]. Related bottom-up analyses have taken place in the domain of business ethics to reflect on the extent to which protection of an environmental commons and the ideal of sustainability might legitimately limit private ownership or entrepreneural technological actions. Taking analytic questioning into the context of computer and information technologies development, it is also useful to consider how access to information technologies should be facilitated under democratic capitalist structures. What are the parameters of intellectual property rights in software design, digitized information, and genetic engineering? From their technological beginnings such ethical issues of distributive justice have bedeviled welfare and public choice economics in relation to advanced communications technologies and called for more careful analysis than had previously been the case. Closely related to the issues of health, safety, risk, pollution, and privacy are interdisciplinary efforts at technology assessment (TA), which was pioneered in the 1970s in the United States but then became more firmly institutionalized in Europe in the following decades. (See also Grunwald’s chapter “Technology Assessment: Concepts and Methods” in this Volume.) Concern initially focused on technologies with unintended consequences which, if they had been appreciated, might have altered their economic adoption or utilization. What rapidly became apparent, however, was that many such non-desired (and even desirable) consequences were largely subject to probabilistic rather than to deterministic calculations, and that even the articulation of the probabilities was subject to value influences. This led to arguments for greater public participation in technical decision making and for the introduction of the perspectives of the social sciences and the humanities into science and technology policy [Frodeman and Mitcham, 2004]. A second general area of analytic work related to ethics and technology has focused on questions related to autonomy and liberty in relation especially to the exercise of technological power by technical professionals and the general public. Historically, advances in technology have both increased human freedom and diminished it; industrialization increased the power and influence of capitalists and reduced the autonomy of workers. According to one interpretation by Dewey, the idea of individual dignity that grew up in a Christian religious provenance was

The Interaction of Ethics and Technology in Historical Perspective

1167

given “a secular and worldly turn” by the Industrial Revolution [Dewey, 1930, p. 75]. For Dewey, however, a concept that once separated human beings from their social context and in secularized form thus promoted practices of innovation and entrepreneurship among mechanics such as James Watt needs, as a result of the new context created by their industrial prowess, to be given a more social interpretation. Indeed, one can detect such a trajectory from atomistic to socialized individualism in the history of utilitarianism from Bentham to Mill. With respect to the values of personal autonomy and liberty there have also emerged needs for conceptual clarification. In particular, there are a number of problematic dimensions to the exercise of such values by technical professionals or consumers. One general challenge is what David Collingridge [1980] called the dilemma of the social control of technology. In the early stages of a technology, when individuals might exercise some free control with relative ease, there is frequently insufficient knowledge to do so; by the time better understanding of the costs or risks is available, control has become difficult if not impossible and the exercise of personal autonomy has become highly constrained. David Rothman [1997] has described the same phenomenon in terms of an emergent technological imperative in the U.S. healthcare system and Daniel Callahan [2003] as a technoscientific research imperative that can mold human dignity so that citizens as patients are convinced against what would most likely be their better judgments to make excessive investments in scientific projects with little prospects for a costeffective return. In response to what has become known as the “Collingridge dilemma,” Collingridge argued for careful assessment and the adoption of technologies that are themselves explicitly designed for flexibility. Other efforts to reconstruct opportunities for the exercise of practical autonomy have included constructive technology assessment [Schot, 2001], real time technology assessment [Guston and Sarewitz, 2002], values sensitive design ([Friedman et al., 2006] and van de Poel’s chapter “Values in Engineering Design” in this Volume), and “midstream modulation of technology” research and development [Fisher et al., 2006] — all of which constitute proposals to overcome the dilemma by inserting ethical and political reflection at points that are neither too early nor too late. In the context of biomedicine there have likewise been complementary efforts to conceptualize the exact parameters of free and informed consent, and then to propose ways to institutionalize them. With regard to information technology, ethical analysis has attempted to clarify the meaning of privacy and security in the use of computerized databases and freedom of speech on the internet. Responsibility in the technical professions constitutes one of the most widely analyzed themes in the analytic tradition. Responsibility in this context assumes moral accountability in the formal sense (sufficient knowledge plus free agency) but seeks to outline guidelines for the exercise of a producer morality beyond economic self-interest or social demand. The aim has been to construct a bridge between the exceptional powers of new technologies and accepted societal values. As will be referenced in more detail in the following section, one common position is that sketched by Stephen Unger [1994] in relation to engineering: engineers have

1168

Carl Mitcham and Adam Briggle

a responsibility not only to their profession but to the general welfare. The formulation (or re-formulation) of codes of conduct for technical professionals has been one attempt to operationalize such notions of responsibility (see also Pritchard’s chapter “Professional Standards for Engineers” in part V of this Volume). Paul Durbin [1993] argues, however, that living up to the full measure of technoprofessional responsibility often requires techno-professionals to step outside their professional roles and take public action in the larger techno-social world. For instance, techno-professional responsibility to protect consumers and users of technical goods and services through the formulation and enforcement of health, safety, and advertizing standards is not possible without involvement on the part of technical professionals. Occasionally such involvement has included public protest, as when physicists during the 1950s and early 1960s lobbied for a world-wide ban on the atmospheric testing of nuclear weapons or computer scientists in the 1980s opposed funding of the U.S. Strategic Defense Initiative (see [Mitcham, 2003]). Extension and intensification of notions of producer responsibility and consumer autonomy easily shades into the issue of public participation in technical decision making, with its tensions between expertise and democracy. Responses to this problem are both practical and theoretical. An example from practice is the Center for Working Life established in the mid-1970s by the Swedish Parliament to allow Swedish workers to participate in the organization of work processes, especially as they are affected by scientific and technological change. The most well-developed theoretical analysis of the tension between technical expertise and democratic participation in decision making can be found in Robert Dahl [1985], who focused on the issue in relation to the control of nuclear weapons. In earlier work, Mumford [1967-1970] and Illich [1973] questioned the abilities of modern technological structures to facilitate social interaction, a topic that has been further pursued by Winner [1986] and Sclove [1995]. But in this area especially particular analytic analyses raise general or global criticisms reminiscent of phenomenological perspectives. 4

ETHICS SHAPING TECHNOLOGY

Adopting a distinction between ethics and morality — in which ethics constitutes a theoretical perspective on beliefs and practices, which by means of criticism in some measure also influences them — it can readily be argued that all three reflective approaches have informed the shaping of technology both directly and through the mediation of moral practices. Most directly, they have done so by expanding awareness, prompting critical thinking, clarifying concepts, as well as formulating and reinterpreting principles or guidelines for action. The fluidity of the ethics/morality distinction is nevertheless manifest in the fact that when such moral guidelines are institutionalized in professional practice they are commonly termed a kind of ethics. Such ethico-moral shaping of technology has occurred in at least three often overlapping spheres: professional, personal, and governmental. For analytic purposes, however, the three levels may be disaggregated.

The Interaction of Ethics and Technology in Historical Perspective

4.1

1169

Professional ethics shaping technology: biomedicine

As has previously been indicated, biomedical ethics and engineering ethics are two of the more prominent versions of applied ethics; in both fields, as well, ethical reflection has strongly engaged with the morality of professional life so as to influence technological practice. Indeed, so prominent is this with regard to biomedicine — which is not always fully recognized as a kind of technology — that biomedical ethics (also called bioethics) is the most intensively developed form of the applied ethics with the most technological influence. Yet precisely because the extensive literature on biomedical ethics does not always intersect with discourse on the ethics of technology and the engineering sciences, it deserves highlighting here. As Edmund Pellegrino [1993] has observed, a historico-cultural metamorphosis of medical ethics into bio- and biomedical ethics took place during the 1970s. Until this period, professional medical ethics remained within a 2500-year old Hippocratic tradition, more or less independent of professional philosophy. On the basis of an oath to help the sick without causing harm, not to cause abortions, to lead a pure life, not to perform surgery or have sexual relations with patients, and to preserve patient confidences, medical morality strongly informed a relatively autonomous community of technical practice. This premodern shaping of the physician-patient relationship also evidenced a stance of humility before nature that endorsed pursuit of a human-nature harmony. For Hippocrates, the aim of medicine was “preserving nature, not altering it” (Hippocrates, “Precepts,” 19), and the physician had an obligation to “refuse to treat those overwhelmed by disease, since in such cases medicine is powerless” (Hippocrates, “On techne,” 3). This ideal of working with nature found further expression in Aristotle’s distinction between cultivation and construction: that is, between the technai of agriculture, education, and medicine, which assist nature in the realization of qualities that would appear to some degree independently of human action, versus such technai as carpentry, which introduces into nature forms that would not appear without human intervention (see, e.g., Physics II, 1, 193a12-17; Politics VII, 17, 1337a2; and Oeconomica I, 1, 1342a26-1343b2). The notion of the physician as one who cultivated health with quite limited technical means was allied as well to a paternalistic not to say authoritarian model of a profession that limited patient autonomy. During the mid-twentieth century, the Hippocratic tradition was challenged by basic changes in society and in medical science and technology. World War II, for instance, produced not only the atomic bomb and a resultant need to develop ethical policies for its design and delimited use (e.g., fail safe triggers and deterrence theory) but also a dawning realization that expanding medical power (this was the first war in history where more soldiers died from combat than from infection and disease) might well call for new forms of moral guidance. As if to reinforce the point, the Nuremberg War Crimes Tribunals (1945-1949) disclosed the failure of the Hippocratic tradition in that some members of the medical profession in Germany abused their authority by conducting radically immoral human ex-

1170

Carl Mitcham and Adam Briggle

perimentations — a discovery that led to formulation of the Nuremberg Code to protect human subjects. In succeeding decades, advances in medical science (altered by alliances with psychology, molecular biology, and other life sciences) and in medical technology (e.g., the engineered invention of new means of birth control and abortion as well as heart-lung, dialysis, and other life-extending machines) transformed medicine into biomedicine and overwhelmed the ideal of cultivation in favor of systematic construction and control. New moral or professional ethical guidelines for the treatment of human subjects were initially imposed on biomedicine from without. But the biomedical community quickly made them its own, and between the 1960s and 1980s increasingly collaborated with applied ethicists to further re-envision technomedical practice. Physician Henry Beecher, for instance, in the Belmont Report, documented how medical researchers in the United States — in multiple less flagrant but nonetheless serious cases — pursued technoscientific knowledge via human subjects experimentation deeply deficient in respect for human dignity [Beecher, 1966]. In response, the biomedical profession itself, admittedly against some internal resistance, undertook to develop stronger protocols and to strengthen institutional mechanisms for their enforcement. In multiple instances expanding awareness of real practices prompted critical thinking and the ethical reshaping of technical practice. This ethical shaping of the technology of medicine took place in a series of overlapping stages. The first, during the 1960s and 1970s, featured broad philosophical reflections similar to those found in the historico-critical approach. For instance, Reiser [1978] examined how since the nineteenth century medical diagnosis technologies — from the thermometer and stethoscope through x-ray machines to electromagnetic resonance tomography — had increasingly diminished direct physician-patient contact and thus dehumanized medical practice. Involvement by theologians and various Christian religious traditions was another distinctive feature of this early period. In a second stage, during the 1980s and 1990s, the ethical shaping of biomedicine became both institutionalized and increasingly analytic. The so-called Georgetown University school of bioethics, for instance, developed a series of principles — nonmalfeasance, beneficence, respect for autonomy, and justice (see [Beauchamp and Childress, 1979]) — that were taught in large numbers of continuing education workshops for physicians. The second period also witnessed stimulation by case studies and, perhaps as a result, concentration on specific issues such as the redefinition of death (in the presence of heart-lung machines that could substitute indefinitely for a patient’s own failed organs) and guidelines for the proper practice of human cloning (after the 1997 announcement of the cloned sheep Dolly). The early 2000s witnessed emergence of a third stage, in reaction against the alleged narrowness of the second. Led by such scholars as Leon Kass there was an effort once again to enlarge bioethics to take on the big questions of human meaning and the good life in ways that echoed socio-critical concerns. As chair of the President’s Council on Bioethics in the first term of President George W. Bush, Kass [2003] questioned biotechnological aspirations to biomedically engineer better chil-

The Interaction of Ethics and Technology in Historical Perspective

1171

dren, enhanced performance, ageless bodies, and happy souls. In addition, in the name of defending human dignity, he helped formulate restrictions on the federal funding of human embryonic stem cell research. Indeed, in one quite remarkable instance of the ethical shaping of biomedical research technology, this limitation promoted development of stem cell technologies employing non-embryonic tissue. In opposition, transhumanists such as computer engineer and inventor Raymond Kurzweil [2005] enthusiastically endorsed the use of biomedical technologies for a wide array of enhancements, provided the initiative came from a bottom-up consumer base rather than top-down governmental decision. Biomedical change thus stimulated and was itself stimulated by ethical arguments between competing concepts and guiding principles.

4.2

Professional ethics shaping technology: engineering

Another instance in which ethical reflection strongly engaged professional life and in the process gave distinctive shape to technology can be found in engineering more broadly understood. Here the collaboration of technical professionals has been primarily with ethicists representing analytic approaches. From the last third of the twentieth century, as a result of unique social circumstances, this was especially the case in the United States, as engineers struggled for professional recognition in ways that included the development of professional codes of ethics — codes that dialectically reflected and helped mold engineering practice. Engineering as a distinctly modern discipline did not originate until the late 1700s, and it began by needing to play catch-up in growth as a profession. The classic professions of medicine, law, and theology were already well established social institutions that from the beginning engineering aspired to imitate. The prehistory of engineering can be traced back to military personnel who designed and operated “engines of war” and fortifications. One example of the emergence of engineering from its military roots took place in France when, in 1716, state service was given civilian but highly regimented form in the Corps des Ponts et ´ Chauss´ees, with subsequent establishment of the Ecole des Ponts et Chauss´ees (1747) for the more effective training of its leaders. This institution of higher ´ ´ education was followed by the Ecole des Mines (1783) and the Ecole Polytechnic (1794), the latter founded to support the French Revolution by Lazare Carnot and Gaspard Monge, two creators of the engineering sciences (see also [Didier, 1999]). A complementary emergence took place in England when, in the late 1770s, John Smeaton took the title “civil engineer” (as opposed to military engineer). It was Smeaton as well who organized an informal dining club as a kind of non-governmental organization called the Society of Civil Engineers (later called “Smeatonians”). The Society of Civil Engineers morphed in 1818 into the Institution of Civil Engineers, which in 1828 was granted a Royal Charter. The British model of non-governmental organization became the pattern for professional engineering organizations in North America. Historian Edwin Layton (1971), for instance, has described in detail how engineers in the United States, unlike physi-

1172

Carl Mitcham and Adam Briggle

cians, struggled with professional divisions into civil engineers, mechanical engineers, electrical engineers, and a host of other discipline and class delimited groups. The American Society of Civil Engineers (ASCE, founded in 1852) was an elitist organization often at odds with business interests. The American Institute of Mining Engineers (AIME, founded in 1871), by contrast, was more egalitarian and allied with business. Different mixes of autonomous professionalism and commercial pragmatism characterized such subsequent organizations as the American Society of Mechanical Engineers (ASME, 1880) and the American Institute of Electrical Engineers (AIEE, 1884). But none could escape the fact that most U.S. engineers were employees of large firms that benefitted from engineering fragmentation, in opposition to the professional autonomy enjoyed by self-employed physicians. In response to the forces of division there emerged a series of efforts to unify the professional engineering community, one aspect of which involved attempts to formulate professional ethics codes that might articulate a common engineering ideal of public service. The classic definition of the defining activity of the profession, that of the British engineer Thomas Tredgold (1788-1829), described engineering as “the art of directing the great sources of power in nature for the use and convenience of man.” But in comparison with the ideal of health that animates the practice of medicine, “use and convenience” was subject to determination more by employer or client than professional. One inadequate effort to escape such conceptual subordination with the articulation of an ideal that would justify engineering professional independence focused on efficiency, an approach promoted by the technocracy movement [Akin, 1977]. But efficiency as an engineering ideal has a complex history (see Alexander’s chapter “History of the Concept of Efficiency” in this Volume) and was problematic on two counts. It elevated technical expertise over public decision making and was therefore at odds with commitments to both democracy and the marketplace. Moreover, as a ratio of outputs over inputs, efficiency remained context dependent — thus still subject to multiple interpretations, depending on how inputs and outputs themselves were defined, with the relevant determination usually being made by non-engineers. To side step the technocracy dilemma, engineering ethics codes simply resorted to stressing a generalized public service ideal. The most common formulation became the statement that engineers have an obligation to hold paramount the protection of public safety, health, and welfare — or what is often called a “paramountcy clause” (see, e.g., the “Code of Ethics for Engineers” of the National Society of Professional Engineers, founded 1934). Initially engineering codes had highlighted professional loyalty — especially loyalty to a client or employer. For instance, the 1914 code of the ASME made the first duty of the engineer to be a “faithful agent or trustee” of some employing client or corporation. Although Michael Davis [2002] has contested a too literal reading of this requirement, the ASME Committee on Code of Ethics (1915) in a contemporaneous commentary emphasized “protection of a client’s or employer’s interests” as an engineer’s “first obligation.” At the same time, the ASME code counseled engineers “to assist

The Interaction of Ethics and Technology in Historical Perspective

1173

the public to a fair and correct general understanding of engineering matters.” Across the twentieth century such counsel, together with later commitment to the paramountcy clause, forced engineering educators repeatedly to confront the difficulties of communicating to engineers a broad conception of their professional responsibilities and best practices for public communication. Post World War II, with special vigor during the 1970s, engineering ethics codes in the United States became subjects of extended discussion and revision in order to address issues raised by the increased importance of engineering in the social order and public concern about a number of specifically technical disasters, including environmental problems implicating major engineering projects. Well-known disasters included two major DC-10 crashes (Paris in 1974 and Chicago in 1979) and a large number of fatal accidents with the Ford Pinto automobile (manufactured from 1971 to 1980), both associated with problematic engineering designs that companies refused to correct even though engineers had called them to attention. From the 1960s on debates have multiplied concerning the environmental impacts of U.S. Army Corps of Engineers projects, from dams on the Colorado River to Everglades wetlands management. Such experiences led to the profiling of “whistle blower” as a moral-technical hero who transgresses company loyalty to expose threats to public safety, health, or welfare. To this historical juncture, arguments regarding engineering ethics took place largely below the radar of professional philosophy. This is not to say that they lacked philosophical significance and as such could not be referenced to argue the ethical shaping of technology, only that they failed to engage the professional community of academic philosophers. In an effort to overcome this hiatus and to promote further conceptual clarification and principle formulation, the U.S. National Science Foundation thus undertook to fund collaborative research between philosophers and engineers to better analyze engineering ethics issues — such as those associated with whistle blowing, autonomy, and the “paramountcy clause” — and to develop appropriate materials for teaching engineering ethics. This led to efforts such as one by philosopher Michael Martin and engineer Roland Schinzinger [1983] to explore the extension of the biomedical principle of free and informed consent to engineering. When the Challenger shuttle disaster of 1986 further exposed weaknesses in engineering independence, it stimulated discussions that led by 2000 to explicit requirements for any accredited engineering curriculum to include the teaching of engineering ethics. Beyond this, Carl Mitcham [1994b] argued for a review of the case studies that had most contributed to the evolutionary trajectory of professional engineering consciousness implied a new obligation plus respicere for engineers to move beyond personal responsibility and to take into account more than the technical dimensions of their work; and Davis [1998] advanced a philosophically sophisticated analysis of engineering ethics codes arguing they be understood as analogous to technical standards and thus as functionally binding — that is, in effect, to integrate technology and ethics in professional practice.

1174

Carl Mitcham and Adam Briggle

Although the ethical shaping of engineering in the United States took place absent dialogue with discussions in other countries, the problems with which North American engineers had to deal could not help but transcended national boundaries. From the late 1980s on, engineering ethics outside the United States progressively provided complementary analyses while profiting as well from U.S. developments. To cite a spectrum of examples: In the ethics codes of Canadian and Australian engineering societies it is possible to find variations on the nongovernmental organization model for professional engineering that originated in the British Isles, while state sponsored promotions of engineering professionalism in Latin American countries shared some of the approaches found in the French model. In Germany engineering ethics, influenced by efforts to compensate for the complicity of engineers with National Socialism, came to exhibit a much more systematic and philosophical form (see [Mitcham and Huning, 1994]). Some developing countries, such as the Dominican Republic, have used engineering ethics codes to criticize persistent patterns of corruption. In Hong Kong codes were created to buttress autonomy in anticipation of the reversion of a colonial outpost to governance by China. In three transnational professional engineering associations — the Pan American Federation of Engineering Societies (UPADI, founded 1949), the European Federation of National Engineering Associations (FEANI, founded 1951), and the World Federation of Engineering Organizations (WFEO, founded 1968) — can be found some of the strongest efforts to promote environmental responsibility as elements in engineering ethics. (For further review with a documentary collection of codes, see ]Mitcham, 2005].) Finally, it is important to note that the ethical shaping of engineering is itself shaped by the historical and social contexts of different engineering cultures. Such recognition, stimulated by the development of interdisciplinary scholarship in engineering studies (see [Downey and Lucena, 1995]) draws on the approaches to ethical reflection present in the socio-critical and historio-cultural schools of ethics and technology. As has been argued in a case-study comparison of engineering ethics in France, Germany, and Japan, different issues can influence the kind of interest engineers and engineering educators take in engineering ethics. A key variable is the relationship between the identities of engineers, e.g., what it means to be an engineer and who counts as an engineer, and the responsibilities of engineering work, including technical responsibilities. The contents of this relationship have varied significantly over time and from place to place around the world. As a result, when one inquires into who has counted as engineers, and what has counted as engineering knowledge and engineering responsibilities at different times and places, the relatively straightforward questions . . . become significantly variable in meaning and attract remarkably diverse answers [Downey et al., 2003, p. 465]. With regard to the cases at hand, engineering ethics is of little interest in France because of the integration of engineering and civil service. In Germany engineering

The Interaction of Ethics and Technology in Historical Perspective

1175

ethics has been integrated into a broad philosophical reflection on engineering. And in Japan, a rise of interest in engineering ethics can be linked to a decline in the extent to which corporations no longer function as “households.” In a globalizing world it must still be recognized that similar challenges can nevertheless have “variable significance and manifestations [depending on] how these challenges are internalized” [Downey et al., 2003, p. 482], even when the interpretations of particular internalizations can be contested.

4.3

Personal shaping: consumers as producers

Professional ethics has become integral to the practice of biomedicine and engineering and thus given them historically distinctive characters. Never before in the history of these professions have ethics and philosophy been so influential on their moral codes and thereby on the technical professions themselves, their practices and products. Equally important, however, is the degree to which not just professionals but consumers and their moral concerns, also molded to some degree by popular ethical reflections, have exerted subtle influence over the shapes of technological processes and products. Patients have themselves influenced the ethics of biomedicine and consumer users, through their approvals and their rejections, have modulated the mix of engineered products. To appreciate the moral shaping of technology, it is not enough to consider a physical object, technical process, or intended function as conceived and designed within the technical community. When an artifact emerges from the laboratory, it shifts from being a predictable and insular entity in a controlled context and becomes simply one more element in a complex, uncontrolled, and interactive social network. When a television is turned on, a series of predictable electromagnetic processes occur that leads to the generation of an image. Outside the functioning of the electronic device itself, however, little if anything is predictable, because further attributes derive not from physical laws but from the socio-cultural networks into which it is deployed. What is being broadcast to whom, when, and where? What activities do viewers forego in order to watch television, and how does this impact the character of society and quality of life? Even more important, what do viewers make of what they watch? For the cultural anthropologist Michel de Certeau, watching television is misunderstood if seen only in terms of passive consumption. Watching and the inevitable reacting to television constitutes a second order productivity — a first order being that of engineering the artifact and creating the programmatic content, the second order taking place with the always creative receptivity of the viewer. According to Certeau, the analysis of the images broadcast by television (of representations) and of the time passed sitting in front of them (a behavior) must be complemented by the study of what the cultural consumer “makes” or “does” [fabrique] during these hours and with these images. The same

1176

Carl Mitcham and Adam Briggle

goes for the using of ... the products purchased in the supermarket. [Certeau, 1980, p. 11] For Certeau it is necessary to distinguish between the strategies of primary production that create powerful institutions and the tactics that ordinary people use in everyday life to carve out niches for themselves within those institutions, as when workers subtly appropriate the workplace to make things for themselves or renters repaint apartments in non-approved colors, in both cases transforming the technological milieux presented to them in predetermined shapes. As consumers, operators, and citizens, people make choices about the technologies they use and how they use them, and these choices influence as well the behavior of markets and governments. Consumers may choose, for example, to purchase a hybrid vehicle, a sports utility vehicle, or forego a personal automobile in favor of using the bus or train. Parents make decisions about what media content and which communication technologies are appropriate for their children. Acting as citizens, on the basis of their own informal ethical reflections, people lobby their democratic representatives to pursue one energy production strategy or another. Consumers and citizens, either consciously or not, weigh risks, costs, and benefits to form judgments that guide their uses of artifacts and systems. In a world in which the family, religion, and other traditional structures of popular morality have become attenuated, while the stakes of making and using artifacts have only become increased, common experience becomes increasingly ethicized. The ethics/morality difference itself diminishes as people are forced to think for themselves. For Charles Taylor [2007] this is the key feature of the modern secular age. Morality ceases somewhat to be morality and, just as is the case with professional life, reaches out and becomes receptive to philosophy so that quotidian experience becomes infused with ethical reflection. Adapting the suggestive argument of Lorenzo Magnani [2007], in the technological world morality might almost be said to have a duty to become ethics. That ordinary individuals thus reflecting on their beliefs and behaviors have a degree of leeway in shaping technology is significant. This is especially so in terms of assigning responsibility for accidents and failures. It is particular drivers, for example, who start automobiles and drive them carefully or not down city streets. Although engineers design and develop the use plans for cars, these are like seeds that sprout and flourish differently under different conditions. Drivers are the ones who provide the conditions that “bring cars to life.” When an intoxicated driver has an accident, the default assumption is that the driver is responsible, although if the accident resulted from a steering wheel coming loose from the steering column because of a design flaw in the pen connector, the engineer or manufacturer might be charged with responsibility. Lines between designers and users, however, are not always clear. This is especially the case with open source software and “share and share alike” software licenses that allow users to alter code or contribute content collectively. Such developments have blurred distinctions between the technical producer and consumer, giving rise of the concept of the “prosumer” (producerconsumer).

The Interaction of Ethics and Technology in Historical Perspective

1177

A key question raised by the consumer shaping of technology thus concerns the level of freedom and responsibility that people really have in working out their personal technological existences. What is the range and quality of decisions they can make? Robust answers will depend on context, the interpretation of which will nevertheless easily be influenced by general beliefs about the neutrality of technology. The neutrality thesis as a principle of the popular ethics of technology argues that insofar as technologies express values these are values of effectiveness with respect to a given function (see also Radder’s chapter “Why Technologies Are Inherently Normative” in this Volume). Artifacts are otherwise neutral with respect to the wider practices and contexts in which they are deployed. They are objects that can be put to good or bad uses by good or bad people, because there are always multiple ways they can be used. As the saying goes, “Cars don’t drive themselves.” There is obviously some truth to this thesis, but the situation is more complex than it would seem to admit; a society with cars is different than one without. When an automobile sits in the driveway, it takes a special act of the will for its owner to walk five blocks to the store. A simple decision to purchase groceries would likely take on a quite different trajectory than the same decision absent the presence of the car. Although artifacts and systems do not possess agency in the standard sense, they do structure the human lifeworld so as to transform situations and options available to their inhabitants. Another aspect of technological non-neutrality has been conceptualized in the notion of a technological “script.” Like a dramatic script, a technological one prescribes behavior to some extent while allowing actors to make diverse interpretations in their performances [Bijker and Law, 1992]. Another related concept is that of “value suitabilities.” A given technology may well be “more suitable for certain activities and more readily [supporting of] certain values while rendering other activities and values more difficult to realize” [Friedman et al., 2006, p. 351]. At the same time, the culture in which technological scripts exist will make their own contributions to use. Although “Cars don’t drive themselves,” the people in one country with cars may be strongly influenced by a culture that nevertheless inhibits their use by taxes, road design, or more informal social expectations. The degree to which a person’s morality, even when ethicalized, will be able to influence the shape of technology may be somewhat marginal. Is it possible that Certeau romanticized the productivity of the consumer? Technological cultures “confront people born into them not as something they may freely choose to adopt if they wish, but as an imposed given imbued with great inertia” [McGinn 1991, p. 75]. After all, most people in many parts of the advanced technological world cannot choose to live without cars, television, the internet, or related artifacts — or in cultures that enact these artifacts in particular ways. The question of freedom must confront the fact that the human subject is always already a subject-in-the-technoworld. From such a historico-cultural perspective, individuals can only achieve a free relation to technology once they realize the extent to which their world and consciousness are technologically me-

1178

Carl Mitcham and Adam Briggle

diated. To achieve a free relation with technology is as difficult as achieving a free relationship with one’s parents or one’s religion. Some people do it, but not without effort. Even the simple truth, often argued with an analytic approach, that technologies which open doors almost always close others, is difficult to appreciate at the existential level. Thinking about how individuals or groups of non-technical consumers shape the technological world in which they live raises anew the issue of responsibility for unintended consequences and externalities such as air pollution, climate change, species extinction, or groundwater depletion. Can any individual user be considered morally accountable for such phenomena? Such collective problems emerge as a result of individuals each fulfilling quite restricted social roles and associated responsibilities. Problems result from the sum total of millions of people performing simple actions, from driving cars to watching television, none of which alone may have any significant effects. Although individuals may be able to mitigate these problems to some degree, and could even be argued to have a meta-role responsibility to bring ethical reflection to bear in technological societies, in the absence of dramatic events, not everyone can be expected to practice what may be experienced as supererogatory virtues. More importantly, individuals acting on their own are fundamentally restricted in terms of the changes they can effect. Problems are systemic, that is, are the result of entire technological networks such as transport and industrial agriculture. Beyond marginal refinements, individuals as such cannot alter the systems they have created. Serious change requires collective or political action.

4.4 Political shaping: regulation and promotion Since the late-nineteenth century, nation states have matched the increasing complexity of material culture with a growing bureaucracy. Technologies create both expected and unexpected health, safety, environmental, and socioeconomic benefits, risks, and responsibilities. Governments grow and adapt in efforts to defend and apply received moral traditions to new situations. The relationship between law and ethics, however, is not a straightforward matter of application. Though many laws are a direct social embodiment of some ethical principal (e.g., laws protecting minors from sexual predators on the internet), some laws are motivated by pragmatic needs rather than directly by ethics (e.g., laws that coordinate government agencies). Furthermore, laws can be unjust or otherwise unethical, at least from the perspective of certain moral theories. This is clearly the case, for example, with laws permitting slavery and discrimination based on race and gender. Laws and regulatory agencies together with promotional policies shape technology in ways that are often more potent and direct than either technical professionals or consumer users. Although historio-cultural reflections sometimes interpret the emergence of such bureaucratic agencies as themselves expressions of modern

The Interaction of Ethics and Technology in Historical Perspective

1179

technology, analytic and socio-critical reflections are more likely to see them as ways morality and associated ethical analyses influence technology. It is thus analytic approaches, together with modest contributions from sociocritical approaches, that have been most engaged with political and legal institutions. The law, as enacted by legislatures (statutory law) or executive agencies (administrative law) and then interpreted by the courts, is the primary mechanism for the political shaping of technology. Legal activities and the policies they manifest fall into the two broad categories: down-stream regulation and up-stream policy guidance. Regulation involves the creation of standards — for construction (building codes), foods and drugs (health and quality standards), transport (operating criteria), environmental protection, advertising, etc. — that rely crucially on scientific and technical knowledge [Jasanoff, 1995]. Regulation seeks both to reduce risks ex ante, before they are imposed, and ex poste, often through litigation after risk exposures have occurred. Guidance policies in turn seek to directly fund or provide indirect tax and related incentives for one technology over another, thus influencing the types and mixes of technological products, processes, and systems. In undertaking such activities, the law plays a critical role in fostering innovation and selective development, sometimes focused on special regions or groups, and in further distributing the public and private goods benefits of technology. Feenberg [2002] explores the notion of “technical code” to demonstrate how governmental regulation can shape technology at basic levels. Technical codes reveal how technical parameters are socially constructed. For example, by 1852, 5,000 steamboat passengers in the U.S. had died as a result of boiler explosions. The U.S. Congress awarded its first federal grant to do technical research on the problem and then created an agency that mandated technical changes such as thicker walls and safety valves. Boiler design was shaped by social and political judgments about safety; ethics was literally “cast in iron.” (Illustrating Certeau’s user creativity, however, steamboat captains often disabled safety values in order to run at higher pressures and make better time.) The same political negotiation is at work in the adoption of environmental standards, such as fuel efficiency requirements for automobiles, emission restrictions on power plants, or extended manufacturer responsibility. As values such as safety, sustainability, or justice become part of technical codes, they are treated as intrinsic features of the artifacts. They cease to be broken out as the specific “price” that an otherwise pure technical rationality must pay. Upholding these standards eventually becomes the law, not a “trade-off” with efficiency. The regulation of technology raises a host of ethical issues. In analytic outline, it is possible to identify five main tensions that continuously occur: (a) conflicting interests between regulating agency and regulated industry; (b) relative benefits of regulation versus deregulation and market solutions; (c) tradeoffs between values such as safety and cost, security and free speech, or profit and environmental preservation; (d) jurisdictional disputes regarding the power to regulate, especially in transnational situations; and (e) disputes about which principles or goals ought to guide regulation. The last is particularly notable in debates concerning the

1180

Carl Mitcham and Adam Briggle

meaning and desirability of the precautionary principle as a guide to the governance of technologies. With regard to promotion, governments use a wide variety of mechanisms to encourage or direct technological innovation. Indeed, regulation often spurs innovation, as with automobile emissions standards, which forced automakers to develop and adopt new technologies. Other mechanisms include: (a) research and development; (b) subsidies; (c) loan guarantees for companies developing technologies — such as nuclear power plants — that require massive amounts of capital; (d) technology transfer legislation that promotes the flow of government-funded inventions to the private sector or the flow of technologies in industrial nations to developing nations; and (e) intellectual property laws that give inventors a time-limited exclusive right to commercially exploit the output of their work. Adjudication of the tensions involved in regulation or decision making with regard to the mechanisms of promotion always involve moral judgments — judgments which are again increasingly enhanced by critical ethical reflection. Of particular salience have been questions concerning the just distribution of scarce resources and allocations of authority to manage technological change. Since such decisions and reflection upon them are made by individuals, ethico-moral shaping of technology at the political level inevitably implicates as well the personal and professional spheres. What for analytic purposes is separated, in the technolifeworld is a complex whole that can also be thought as a technological shaping of ethics. 5 TECHNOLOGY SHAPING ETHICS Thus far the focus has been on how ethics has assessed and shaped technology — across historical periods; in different philosophical schools; and through moralities operative at the levels of professional life, consumer behavior, and political governance. Technologies have been considered primarily as objects and processes for ethical reflection and intervention. But at several points it has also been suggested that the interaction between ethics and technology could on occasion be reciprocal. If technology can shape society, which is not open to doubt, then why not as well the cultures, morals, and ideas that help make up a society? For example, the developments of birth control and in vitro fertilization (IVF) technologies in the 1980s were met initially with widespread moral and religious objections that nevertheless moderated over time. Yet that technology might shape morality or ethics is more problematic than the ethical shaping of technology, because ethics is viewed by many philosophers as well as the public at large as an autonomous dimension of culture — perhaps even the most autonomous dimension. Indeed, for a modern philosopher such as Kant, to think of morality in heteronomous terms as determined by something other than itself is to fail to recognize what ethics really is. In thinking about the possible technological shaping of ethics it may thus be useful to consider how even from a Kantian perspective modern science, a close

The Interaction of Ethics and Technology in Historical Perspective

1181

associate of technology, can be admitted to have exercised a shaping influence on ethics. According to the Kantian ethical theorist Christine Korsgaard [1996], modern moral and political philosophy can be read as a series of responses to what she calls the modern scientific worldview. Prior to the rise of modernity, form and value were taken as more real than experienced fact. The form of a thing, even when not fully present in some particular, pointed toward and constituted its perfection. With modern science, however, the world came to be thought in terms of matter and energy, indifferent stuff acting according to universal laws. Under such circumstances, Korsgaard argues, “If the real and the good are no longer one, value must find its way into the world somehow. Form must be imposed on the world of matter” [1996, p. 5]. Modern science thus set the stage for and helped shape the rise of subsequent ethical theory. Hobbes and Samuel von Pufendorf (1632-1694) were among the early philosophers to recognize the ethical challenge posed by the scientific worldview and in the process to become agents through which this worldview began to shape ethics. Both made the ethical argument that since morals are not to be found in nature, they must be created by human decisions. Kant himself, of course, sought to ground morality in the inherent rationality of practical decision making itself rather than simply in decision makers. But all three ethical theorists, insofar as they undertook to respond to a scientific worldview, allowed in ethics for a subtle shaping by science. This indirect scientific shaping of ethics has perhaps been more widely appreciated than technological shaping — and yet the former surely suggests the latter. More boldly stated, Enlightenment philosophers attempted to use modern science to reshape ethics through its cognitive products (facts and theories) and its distinctive methods (experimentation and quantitative analysis). Scientific enlightenment sought to use factual knowledge to dispel illusion and myth, as in Galilean astronomy and Darwinian evolution, and to provide new forms of moral analysis, as with utilitarianism. Taking such shaping as a suggestive template, technology can in like manner be thought of as able to influence ethics in four ways: (a) technology creates new moral issues and questions; (b) technology requires adjustments in morally significant concepts; (c) technology may require new moral theories; and (d) technology and technological concepts change our moral self-image and visions of the good life. The first mode of shaping is perhaps the most obvious. When technologies change or new ones are invented a culture can become “maladjusted” or exhibit what sociologist William Fielding Ogburn [1964] called “cultural lag” (see also [Toffler, 1970]). After the introduction of the automobile, for example, it took time for the culture of roadway design to catch up, and new habits and expectations on the part of drivers had to be cultivated, eventually constituting a new morality operative within the transportation system. The widespread adoption of mobile phones has created new moral questions about the etiquette of their use in public spaces. The moral questions and issues involved in global climate change would not have arisen absent modern technologies.

1182

Carl Mitcham and Adam Briggle

Second, technological change can challenge the adequacy of morally significant concepts, perhaps demanding reformulation of those concepts. Advances in lifesupport technologies during the 1970s called into question traditional definitions of death in terms of cardiac or pulmonary arrest and led to another in terms of brain functioning. Advances in ambient intelligence, surveillance, and genetic technologies similarly challenge received notions of privacy and personal property. The environmental impacts of industrial technologies encouraged reconsideration of the concept of wilderness—what was once threat came to be conceived, under conditions of harried technological affluence, as sanctuary, and subsequently as itself threatened by industrial pollution. Similar industrial impacts led to the creation of endangered species as a new legal and moral concept and environment as a good to be protected. The rise of computer-mediated relationships has promoted new interpretations of community and friendship (e.g., [Briggle 2008]) while computer simulation raises new questions about reality and its normativity [e.g., Borgmann, 1999]. Enhancement technologies suggest new considerations of human nature as they open doors to emergent post- or trans-humanist ideals [Bostrom 2005]. Indeed, modern technology has long been used as a source of images for thinking about human nature. In the 1600s, Hobbes asked, “For what is the Heart, but a Spring; and the Nerves, but so many Strings; and the Joynts, but so many Wheeles. . . ?” [Leviathan, introduction] In some cases, technology expands the scope of moral concepts, reshaping the sense of the normal. For example, oil and electricity, once objects of desires only among the wealthy few, have become needs. They are necessary for survival for all in a society of mass affluence due to the technologies that require their input. From this perspective, technological development imposes new needs and higher levels of consumption on supposedly underdeveloped peoples. Suddenly, a villager needs a bus ticket, rental housing, utilities, and schooling [Escobar, 1995]. Technology not only expands needs but also rights claims [McGinn, 1991]. For example, the right to life was traditionally understood in negative terms — as an entitlement not to be deprived of life or physical integrity. But in the presence of life-preserving technologies, this right tends to take on expansive form — as an entitlement to be provided with whatever medical treatment is necessary to sustain life. In a similar fashion, clean water, vaccinations, and other goods and services become entitlements as technologies develop that can make them readily available. Third, technology may not just destabilize or engender significant moral concepts but actually call for the development of entirely new moral theories. As noted above, there is disagreement on this point. Analytic philosophers tend to argue that existing theories are adequate, while some representatives of the other traditions disagree. Jonas [1984], for example, argues that modern technology creates the need for a new ethics in terms of theory and practice. Premodern ethics could allow technics to remain in the background as a marginal aspect of life. During the modern period technology entered the foreground of human experience at precisely the same time modern science undermined natural teleology and the notion of a

The Interaction of Ethics and Technology in Historical Perspective

1183

stable human nature as a guide to its wise use. As a result, technology became “restless” in serving desires and creating needs. A “new conception of duties and rights” is required to take account of the global and intergenerational impacts of technology. Jonas saw modern technology as having “introduced actions of such novel scale, objects, and consequences that the framework of former ethics can no longer contain them” [1984, p. 6]. In place of consequentialism or deontology, Jonas proposes a new ethical imperative: “Act so that the effects of your action are compatible with the permanence of genuine human life” [1984, p. 11]. There are a number of other efforts to formulate ethical principles that would catch up or accommodate the technological transformation of the lifeworld. Mitcham [1994b], for example, has argued that engineers especially have a duty plus respicere, to take more things into account. Magnani [2007], more generally, has argued that in the presence of technology acting with knowledge becomes a moral duty. Lawrence Schmidt and Scott Marratto [2008] echo Jonas in arguing that there is no consistent ethical framework to deal with the long-range negative consequences of certain technological developments. They propose a post-liberal theory that rejects the ideology of progress in favor of caution and limitation. Fourth, technology and technological concepts can alter our moral self-image. Technology not only introduces new material products and processes, it also conditions how these new realities will be conceptualized and evaluated. Humans are not just confronted with new needs, for example, but come to understand and assess themselves as beings in possession of needs who calculate their satisfaction. Humans come to conceive themselves, their goals, and their world partly through and in terms of technology. In this regard, techniques of writing and reading are especially important. According to Walter Ong, for oral peoples judgment “bears in on the individual from outside, not from within” [1982, p. 55]. By contrast, the literate mind tends to stand abstracted from the concreteness of lived experience, which fosters greater introspective self-judgment. Illich [1993] further explored the importance of writing and reading techniques for patterning self and reality. In addition, Michel Foucault [1988] argued that different cultures utilize technics to engender different notions and experiences of the human self. Sherry Turkle [1995] argued that computers have ushered in a novel self-understanding. Nicholas Carr [2008] has picked up on this theme to argue that the internet alters the way humans read and think. Whereas books foster and demand the discipline to follow a sustained argument or narrative, the internet promotes a style of reading that puts efficiency and immediacy above all else. This alters the self, because humans are not only what they read but how they read (see [Wolf, 2007]). Indeed, the kind of media used in reading and writing actually shapes neural circuits and thus the kind of thinking self that emerges. But the human self is configured not just through and in terms of media. Another example is medicalization, or the process whereby certain features of human life come to be defined and treated as medical conditions. With the advent of new medical techniques, certain conditions become diseases or disorders rather than,

1184

Carl Mitcham and Adam Briggle

for example, curses or personality traits. This fosters a new and normatively charged self-conception. Some argue that the uncritical expansion of a medicalizing mindset — considering all occasions of sadness, anger, or regret as medical conditions — threatens to picture the human solely in biomedical terms rather than in psychic, spiritual, or moral ones [President’s Council on Bioethics, 2003]. Medical terminology has also fostered a self-conceptualization in terms of risk. Illich argued that this is disembodying, because thinking in terms of risk is “an invitation to intensive self-algorithmization . . . reducing myself entirely to misplaced concreteness by projecting myself on a curve” [Illich, 2005, p. 210]. A similar point could be made about the extension of the ideal of efficiency from the technical sphere into the lifeworld. In technological societies, activities that consume considerable amounts of time become targets for efficiency enhancement: thus, the rise of food processors, microwaves, word processors, household cleaning appliances, and electronic communication in place of handwritten letters [McGinn, 1991]. Yet there are dangers in understanding a family meal or correspondence between friends in terms of efficiency. The patterning of the self by technology suggests that technology will also shape how that self conceives of flourishing or a good life. For example, television does not just satisfy preferences, but shapes and engenders preferences. Similarly, leisure — a component of the good life — is understood now in ways that are fundamentally shaped by technology [McGinn, 1991]. Industrialization has compartmentalized life into structurally differentiated spheres of activity, including work and leisure. Leisure has since taken on its own evolving values and forms in which technologies play a central role. The good life in a materially affluent world is widely conceived of in terms of the consumption and comfort afforded by technologies, which brings along its opposite, namely, increased anxiety about death and risks to physical well-being. This notion of the good life was not in ascendency in medieval European cultures that valued honor, bravery, and risk-taking above comfort and convenience. It is worth further considering the role of technology in such historical moral transformations. Finally, the technological configuring of the good life raises two important questions that deserve further scrutiny. One is conceptual: Is “television-watching” or any new technology-based vision of the good life a new type of answer to the good life or simply a new token or species within a timeless category (e.g., passive entertainment)? That is, does technology create fundamentally new values or does it just shift — expand or contract — pre-existing categories of human valuing and activity? A second concerns whether values or schemes of preferences simply change and adapt to a changing technological lifeworld. If so, does this preclude normative appraisals of past and future worlds from our bounded perspective in the present? Are not the citizens of Brave New World (1932) leading debased lives, even though their schemes of preferences are perfectly adapted to a particular set of technological constructions?

The Interaction of Ethics and Technology in Historical Perspective

6

1185

CONCLUSION: ETHICS AND TECHNOLOGY INTERACTING

A summary statement of the thesis of this chapter might be that ethics and technology have, since the beginning of the modern period, increasingly influenced one another. One way to reiterate and take large-scale measure of such mutual influence is to observe how over the course of its modern development technology has been associated in the moral imagination with an overlapping series of images, from machines and industrial factories to computers and the internet, each of which has been an influence on and been subject to critical ethical reflection. Just as different images have tended selectively to engage and sometimes to reinforce different ethical judgments, ethical reflection has likewise tended to pick up and highlight differential images of technology. Parsing the public images or synecdoches that the term “technology” often evokes — along with the cloud of associations that cannot help but be present as well in philosophical thought — it is possible to identify at least seven broad types. At the dawn of the popular recognition of modern technology as a distinctive phenomenon technology was easily identified with (1) machines and industrial factories, both of which also connoted power, as was even more specifically represented by the steam engine. The fact that human beings were the creators of such powerful machines could not help but promote a heroic vision of the human. Within a short period of time technology also became imagined in terms of (2) stores and homes well stocked with items denoting wealth and affluence; the Crystal Palace and the 1851 world fair exhibition of industrial products is perhaps an even more specific classical image that promoted a different notion of humans as consumers of mass produced goods and services. (3) Tall buildings and bridges have again suggested a heroic vision of the human associated with urbanization and technologically reconstructed spatio-temporal habitats, often contested as at once humanizing and dehumanizing. Electric lights, which are central to such reconstruction, are promiscuous metaphors in this regard, being associated at once with extending human action into the darkness and subjecting them to more exposure than they can always bear. Re-emphasizing the notion of dynamism inherent in the image of industrial machines is another set of images associated with (4) transport by steam ships, railroad trains, automobiles, airplanes, and space exploration probes. Complementing transport are (5) communication technologies such as the telegraph, radio, telephone, and computer all of which are stationary or enclose yet dynamic electronic processes. Both would seem to endorse and be endorsed by dynamic moralities: in the one case an affirmation of physical change from place to place, in the other of rapid information acquisition and network connections, respectively. Finally there are the conflicting images associated with leisure and with warfare. In the realm of (6) leisure technologies are motion pictures and film, television, and video games that signify the unification of technology and the production of human happiness. Yet although entertainment, the connotation of which cannot help but be predominately positive, is also subject to opposing ethical judgments

1186

Carl Mitcham and Adam Briggle

in which such technologies serve to distract humans from more seriousness purpose and activity. To this same category might be added healthcare technologies as imaged by a host of medical instruments and devices from the stethoscope and cardiac monitor to the artificial heart. How could one not but be morally approving of the level of human health made possible by modern technology? But even this has been contested with arguments, however much they are themselves contested, concerning high costs (economic and psychological) if not hubris. By contrast, (7) military technologies are primarily imaged in terms of explosions, tanks, and bombs — with the predominant connection being one in which happiness is replaced by pain and suffering. Apologists for military technologies as providing defensive security have to struggle almost as much as critics of medical technologies in order to make their case. These popular images have played out differentially weighted roles and distinctive issues in the three schools of ethical reflection on technology. For example, in the perspective of the socio-critical approach, machines and industrial factories are argued to be instruments of powerful elites from which they somehow need to be freed in order to realize their true liberating potential, whereas in a historicocultural perspective an emphasis is placed on how such technologies constitute a historical transformation of unprecedented character. In the analytic perspective, by contrast, there is a tendency to see industrial machines as simply more complex tools, with different kinds of machines perhaps raising problems that call for philosophical reflection, without any need to pursue some kind of comprehensive assessment of technics and technology as a whole. Table 3 ventures a simplified These images, simply summary of such a spectrum of different roles and issues. as images, cannot avoid emphasizing the physical dimensions of technology. By contrast, some images call more attention than others to the dynamism of technological processes. Still others could be more easily interpreted to draw out the epistemic dimensions of technology and the ways it is linked with science, or the extent to which technology can occasionally be supported by and support some aspirations (such as desires for power and control) over others (such as contemplative awareness). Additionally, such a conceptual map might also be used to revisit some of the different ways that ethics shapes technology and technology shapes ethics — again observing that some images tend to be more supportive of or resistant to one perspective than another. In a car culture it is relatively easy to make automobiles safer; in such a culture it is less easy to make places for bicycles and pedestrians. In a culture infused with technological making, using, and the engineering sciences there are also strong pressures to give ethical understanding and analysis forms that reflect or are compatible with such dominant phenomena. Reflection on the interactive shaping of ethics and technology cannot help but raise questions about the degree to which ethics is truly able to shape or influence technology. To what extent is the ethical shaping merely marginal or decorative? To what degree can it be substantive? Or do such questions lack meaning, insofar as it becomes progressively difficult to conceive of humans separate from their

The Interaction of Ethics and Technology in Historical Perspective

1187

Table 3. Ethics and Technology Interactions: Images and Interpretations Popular images of technology (1) Machines and industrial factories (2) Shopping marts, clothes, money (3) Tall buildings and bridges

(4) Steam boats, trains, cars, and airplanes (5) Telegraph, radio, telephone, computers, internet (6) Movies, TV, videos games — and medical technologies

(7) Explosions, tanks, and bombs

Socio-critical reflection Ruling class power, working class oppression Distraction from liberating potential Alienating vs possibilities for humanizing architecture Must not be restricted to the wealthy classes Critical of broadcast control but not distributed networks Need for more democratic participation — and patient consent

Historico-cultural reflection Industrial Revolution

Caused by ruling class control and mistaken ideologies

Reveals inherent destructiveness

Consumer culture undermines creativity Urbanization replaces community with society Travel uproots people, alienates from place Enhanced technical means without enhanced content Mass culture is a cultural decline — and an addiction to physical well being

Analytic reflection Productive process, creativity Issues of distributive justice Can exemplify creativity and technical beauty Need to promote safety through regulation Raise issues of privacy and equal access New forms of art and entertainment are being created — post-humanist possibilities Dangerous risks to be moderated and restrained

1188

Carl Mitcham and Adam Briggle

technologies? Maybe the question of whether or not ethics influences technology is practically analogous to questions about whether there is an external world or other minds: If one begins with doubt about the reality of the external world or other minds, it is difficult to see how one can ever prove their existence. The external world and other minds are structured into and presumed by the very thought that would try to consider their absence. Nevertheless, given the extent to which images can be used selectively to promote arguments for ethics shaping technology or technology shaping ethics — at multiple levels and in multifarious ways — it can only be concluded that the interaction of ethics and technology must also make alliance with other philosophical engagements and arguments. Although in philosophy of technology ethics may have received quantitative pride of place, philosophical pursuit of the ethics of technology cannot finally be sustained without considering such questions as how to define technology, the ontology of technological objects, the structures of technological action, and the epistemic dimensions of technology. In the end it is necessary to bring philosophy as a whole to bear to help make reasoned judgments about which images are more adequate than others — and to what extent some approaches (or parts thereof) to understanding ethics and technology interactions might be more adequate than others. BIBLIOGRAPHY All references to classical or standard works are handled with standard page references or to textual divisions that are independent of any particular edition and thus not included here. All quotations not originally in English are in our own or highly adapted translations. [Achterhuis, 2001] H. Achterhuis, ed. American Philosophy of Technology: The Empirical Turn. R. Crease, trans. Bloomington, IN: Indiana University Press, 2001. [Akin, 1977] W. E. Akin. Technocracy and the American Dream: The Technocrat Movement, 1900-1941. Berkeley: University of California Press, 1977. [Anders, 1961] G. Anders. Off Limits f¨ ur das Gewissen. Reibek bei Hamburg: Rowohlt, 1961. [Beauchamp and Childress, 2001] T. L. Beauchamp and J. F. Childress. Principles of Biomedical Ethics. New York: Oxford University Press, 1979. (Fifth edition, 2001.) [Beecher, 1966] H. K. Beecher. Ethics and Clinical Research. New England Journal of Medicine, vol. 274, no. 24 (June 16), pp. 1354-1360, 1966. [Bijker and Law, 1992] W. Bijker and J. Law, eds. Shaping Technology/ Building Society: Studies in Sociotechnical Change. Cambridge, MA: MIT Press, 1992. [Borgmann, 1984] A. Borgmann. Technology and the Character of Contemporary Life. Chicago: University of Chicago Press, 1984. [Borgmann, 1999] A. Borgmann. Holding onto Reality: The Nature of Information at the Turn of the Millennium. Chicago: University of Chicago Press, 1999. [Bostrom, 2005] N. Bostrom. In Defense of Posthuman Dignity. Bioethics 19, 202-214, 2005. [Briggle, 2005] A. Briggle. Visions of Nantucket: The Aesthetics and Policy of Wind Power. Environmental Philosophy, vol. 2, no. 1, pp. 54-67, 2005. [Briggle, 2008] A. Briggle. Real Friends: How the Internet can Foster Friendship. Ethics and Information Technology, vol. 10, no. 1, pp. 71-79, 2008. [Callahan, 2003] D. Callahan. What Price Better Health? Hazards of the Research Imperative. Berkeley: University of California Press, 2003. [Carr, 2008] N. Carr. Is Google making us Stupid? Atlantic Monthly (July/August), no. 21, pp. 75-80, 2008. [Certeau, 1980] M. de Certeau. L’Invention du quotidien, 2 vols. Vol.1, Arts de faire; vol. 2, editions, 1980. Habitier, cuisiner. Paris: Union g´en´ erale d’´

The Interaction of Ethics and Technology in Historical Perspective

1189

[Collingridge, 1980] D. Collingridge. The Social Control of Technology. New York: Palgrave Macmillan, 1980. [Dahl, 1985] R. Dahl. On Democracy. New Haven, CT: Yale University Press, 1985. [Davis, 1998] M. Davis. Thinking Like an Engineer: Studies in the Ethics of a Profession. New York: Oxford University Press, 1998. [Davis, 2002] M. Davis. Profession, Code and Ethics. Burlington, VT: Ashgate, 2002. [Dewey, 1930] J. Dewey. Individualism Old and New. New York: G.P. Putnam’s. in John Dewey, The Later Works, 1925-1953, vol. 5: 1929-1930 (Carbondale, IL: Southern Illinois University Press, 1984), pp. 41-123, 1930. [Didier, 1999] C. Didier. Engineering Ethics in France: A Historical Perspective. Technology in Society, vol. 21, no. 4 (November), pp. 471-486, 1999. [Downey and Lucena, 1995] G. L. Downey and J. C. Lucena. Engineering Cultures. In Sheila Jasanoff, G.E. Markle, J.C. Petersen and T. Pinch, eds., Handbook of Science and Technology Studies (Thousand Oaks, CA: Sage), pp. 167-188, 1995. [Downey et al., 2007] G. L. Downey, J. C. Lucena, and C. Mitcham. Engineering Ethics and Identity: Emerging Initiatives in Comparative Perspective. Science and Engineering Ethics, vol. 13, no. 4 (December), pp. 463-487, 2007. [Durbin, 1993] P. T. Durbin. Social Responsibility in Science, Technology, and Medicine. Bethlehem, PA: Lehigh University Press, 1993. [Ellul, 1964] J. Ellul. La Technique, ou l’enjeu du si` ecle. Paris: Colin, 1954. English version: The Technological Society, trans. John Wilkinson (New York: Knopf, 1964). [Escobar, 1995] A. Escobar. Encountering Development: The Making and Unmaking of the Third World. Princeton, NJ: Princeton University Press, 1995. [Feenberg, 1995] A. Feenberg. Alternative Modernity: The Technical Turn in Philosophy and Social Theory. Berkeley: University of California Press, 1995. [Feenberg, 2000] A. Feenberg. From Essentialism to Constructivism: Philosophy of Technology at the Crossroads. In Technology and the Good Life? Eric Higgs, Andrew Light, and David Strong, eds., pp. 294-315, Chicago: University of Chicago Press, 2000. [Feenberg, 2002] A. Feenberg. Transforming Technology: A Critical Theory Revisited. New York: Oxford University Press, 2002. [Fisher et al., 2006] E. Fisher, R. L. Mahajan, . and C. Mitcham. Midstream Modulation of Technology: Governance from Within. Bulletin of Science, Technology, and Society, vol. 26, no. 6 (December), pp. 485-496, 2006. [Foucault, 1988] M. Foucault. Technologies of the Self: A Seminar with Michel Foucault. L. Martin, H. Gutman, and P. Hutton, eds., Amherst, MA: University of Massachusetts Press, 1988. [Friedman et al., 2006] B. Friedman, P. Kahn, Jr., and A. Borning. Value Sensitive Design and Information Systems. In P. Zhang and D. Galletta, eds., Human-Computer Interaction and Management Information Systems: Foundations, M.E. Sharpe, London, pp. 348-372, 2006. [Fordeman and Mitcham, 2004] R. Frodeman and C. Mitcham, eds. Toward a Philosophy of Science Policy: Approaches and Issues. Theme issue, Philosophy Today, vol. 48, no. 5 (supplement), 2004. [Goldman, 1992] S. L. Goldman. No Innovation without Representation: Technological Action in a Democratic Society. In Stephen H. Cutcliffe et al., eds., New Worlds New Technologies, New Issues (Bentham, PA: Lehigh University Press), pp. 148-190, 1992. [Guston and Sarewitz, 2002] D. Guston and D. Sarewitz. Real Time Technology Assessment. Technology in Society, vol. 24, nos. 1-2 (January-April), pp. 93-109, 2002. [Haldane, 1924] J. B. S. Haldane. Daedalus, or, Science and the Future. London: Kegan Paul, 1924. [Haraway, 1985] D. Haraway. A Manifesto for Cyborgs: Science, Technology, and Socialist Feminism in the 1980s. Socialist Review, vol. 15, no. 2, pp. 65-108, 1985. [Hayek, 1967] F. A. Hayek. The Results of Human Action but not of Human Design. in F.A. Hayek, Studies in Philosophy, Politics and Economics (Chicago: University of Chicago Press), pp. 96-105, 1967. [Heidegger, 1954] M. Heidegger. Die Frage nach der Technik, in Vortr¨ age und Aufs¨ atze (Pfullingen: Neske). English version: The Question Concerning Technology, in The Question concerning Technology and Other Essays, W. Lovitt, trans. (New York: Harper and Row, 1954). [Hickman, 1990] L. A. Hickman. John Dewey’s Pragmatic Technology. Bloomington, IN: Indiana University Press, 1990.

1190

Carl Mitcham and Adam Briggle

[Illich, 1973] I. Illich. Tools for Conviviality. New York: Harper and Row, 1973. [Illich, 1993] I. Illich. In the Vineyard of the Text: A Commentary to Hugh’s Didascalicon. Chicago: University of Chicago Press, 1993. [Illich, 2005] I. Illich. The Rivers North of the Future: The Testament of Ivan Illich. D. Cayley, ed. Toronto: House of Anansi Press, 2005. [Jasanoff, 1995] S. Jasanoff. Science at the Bar: Law, Science, and Technology in America. Cambridge, MA: Harvard University Press, 1995. [Jonas, 1984] H. Jonas. The Imperative of Responsibility: In Search of an Ethics for the Technological Age. H. Jonas and D. Herr, trans. Chicago: University of Chicago Press, 1984. [Kass, 2003] L. Kass. Beyond Therapy: Biotechnology and the Pursuit of Happiness. A Report by the President’s Council on Bioethics, Washington, DC: Government Printing Office, 2003. [Keulartz et al., 2002] J. Keulartz, M. Korthals, M. Schermer, and T. Swierstra, eds. Pragmatist Ethics for a Technological Culture. Dordrecht: Kluwer, 2002. [Korsgaard, 1996] C. Korsgaard. The Sources of Normativity. Cambridge: Cambridge University Press, 1996. [Kurzweil, 2005] R. Kurzweil. The Singularity Is Near: When Humans Transcend Biology. New York: Viking, 2005. [Latour, 1987] B. Latour. Science in Action: How to Follow Scientists and Engineers Through Society. Cambridge, MA: Harvard University Press, 1987. [Layton, 1971] E. T. Layton. The Revolt of the Engineers: Social Responsibilities and the American Engineering Profession. Cleveland, OH: Case Western Reserve University Press, 1971. [Magnani, 2007] L. Magnani. Morality in a Technological World: Knowledge as Duty. Cambridge, UK: Cambridge University Press, 2007. [Martin and Schinzinger, 2005] M. W. Martin and R. Schinzinger. Ethics in Engineering. New York: McGraw-Hill, 1983. (Fourth edition, 2005.) [McGinn, 1991] R. McGinn. Science, Technology, and Society. New York: Prentice Hall, 1991. [Merleau-Ponty, 1945] M. Merleau-Ponty. Ph´ enom´ enologie de la perception. Paris: Gallimard, 1945. [Michelfelder, 2000] D. Michelfelder. Technological Ethics in a Different Voice. in E. Higgs, A. Light, and D. Strong, eds. Technology and the Good Life? (Chicago: University of Chicago Press), pp. 219-233, 2000. [Mitcham, 1994a] C. Mitcham. Thinking through Technology: The Path between Engineering and Philosophy. Chicago: University of Chicago Press, 1994. [Mitcham, 1994b] C. Mitcham. Engineering Design Research and Social Responsibility. in K.S. Shrader-Frechette, Research Ethics (Totowa, NJ: Rowman and Littlefield), pp. 153-168, 1994. [Mitcham, 2003] C. Mitcham. Professional Idealism among Scientists and Engineers: A Neglected Tradition in STS Studies. Technology in Society, vol. 25, no. 2 (April), pp. 249-262, 2003. [Mitcham, 2005] C. Mitcham. Ethics Codes in Professional Engineering: Overview and Comparisons. In C. Mitcham, ed., Encyclopedia of Science, Technology, and Ethics (Detroit: Macmillan Reference), vol. 4, pp. 2176-2182, followed with code documents, pp. 2183-2260, 2005. [Mitcham and Huning, 1993] C. Mitcham and A. Huning. The Historical and Philosophical Development of Engineering Ethics in Germany. Technology in Society, vol. 15, no. 4 (November), pp. 427-439, 1993. [Mumford, 1967-70] L. Mumford. The Myth of the Machine. 2 vols. Vol 1, Technics and Human Development. Vol. 2, The Pentagon of Power, 1967-1970. [Nozick, 1974] R. Nozick. Anarchy, State, and Utopia. New York: Basic Books, 1974. [Ogburn, 1964] W. F. Ogburn. Cultural Lag as Theory. in On Culture and Social Change, ed. O.D. Duncan (Chicago: University of Chicago Press, 1964). [Ong, 1982] W. J. Ong. Orality and Literacy: The Technologizing of the World. New York: Routledge, 1982. [Ortega y Gasset, 1939] J. Ortega y Gasset. Meditaci´ on de la t´ ecnica. in Obras completas (first edition, Madrid: Revista de Occidente, 1945-1947), vol. 5, pp. 317-375, 1939. [Pellegrino, 1993] E. D. Pellegrino. The Metamorphosis of Medical Ethics: A 30-Year Retrospective. Journal of the American Medical Association, vol. 269, no. 9 (March 3), pp. 1158-1162, 1993.

The Interaction of Ethics and Technology in Historical Perspective

1191

[President’s Council on Bioethics, 2003] President’s Council on Bioethics. Beyond Therapy: Biotechnology and the Pursuit of Happiness. Washington, DC: U.S. Government Printing Office, 2003. [Reiser, 1978] S. J. Reiser. Medicine and the Reign of Technology. Cambridge: Cambridge University Press, 1978. [Rothman, 1997] D. J. Rothman. Beginnings Count: The Technological Imperative in American Health Care. New York: Oxford University Press, 1997. [Russell, 1924] B. Russell. Icarus, or The Future of Science. London: Kegan Paul, 1924. [Russell, 1951] B. Russell. The Impact of Science on Society. New York: Columbia University Press, 1951. [Russell, 1958] B. Russell. The Divorce between Science and ‘Culture’. Address on receipt of the Kalinga Prize for the Popularization of Science, UNESCO Courier, vol. 49 (February), 1958. [Schmidt and Marratto, 2008] L. E. Schmidt with S. Marratto. The End of Ethics in a Technological Society. Montreal and Kingston: McGill-Queen’s University Press, 2008. [Schot, 2001] J. Schot. Constructive Technology Assessment as Reflexive Technology Politics. in P. Goujon and B.H. Dubreuil, eds., Technology and Ethics: A European Quest for Responsible Engineering (Leuven, Belgium: Peeters), pp. 239-249, 2001. [Sclove, 1995] R. Sclove. Democracy and Technology. New York: Guilford Press, 1995. [Schrader-Frechette, 1980] K. S. Shrader-Frechette. Nuclear Power and Public Policy. Boston: D. Reidel, 1980. [Schrader-Frechette, 1991] K. S. Shrader-Frechette. Risk and Rationality: Philosophical Foundations for Populist Reforms. Berkeley, CA: University of California Press, 1991. [Taylor, 2007] C. Taylor. A Secular Age. Cambridge, MA: Harvard University Press, 2007. [Toulmin, 1958] S. Toulmin. The Uses of Argument. Cambridge: Cambridge University Press, 1958. [Toulmin, 1982] S. Toulmin. How Medicine Saved the Life of Ethics. Perspectives in Biology and Medicine, vol. 2, no 4, pp. 736-750, 1982. [Toulmin, 1988] S. Toulmin and A. R. Jonsen. The Abuse of Casuistry: A History of Moral Reasoning. Berkeley: University of California Press, 1988. [Turkle, 1995] S. Turkle. Life on the Screen: Identity in the Age of the Internet. New York: Simon and Schuster, 1995. [Unger, 1994] S. S. Unger. Controlling Technology: Ethics and the Responsible Engineer. Malden, MA: Wiley-Interscience, 1994. [Verbeek, 2005] P.-P. Verbeek. What Things Do: Philosophical Reflections on Technology, Agency, and Design. University Park, PA: Pennsylvania State University Press, 2005. [Whitbeck, 1998] C. Whitbeck. Ethics in Engineering, Practice and Research. Cambridge: Cambridge University Press, 1998. [Winner, 1986] L. Winner. The Whale and the Reactor: A Search for Limits in an Age of High Technology. Chicago: University of Chicago Press, 1986. [Wolf, 2007] M. Wolf. Proust and the Squid: The Story and Science of the Reading Brain. New York: Harper, 2007.

This page intentionally left blank

Part VI

Philosophical Issues in Engineering Disciplines

This page intentionally left blank

INTRODUCTION TO PART VI

Sven Ove Hansson, associate editor In discussions on the philosophy of technology and engineering it is often useful to make comparisons with the philosophy of science that has a longer history and a more extensive literature. The philosophy of science can be divided into two major parts: On the one hand we have general philosophy of science that deals with the issues that are common to the different sciences. On the other hand we have the philosophy of the individual sciences, such as the philosophy of physics, the philosophy of biology, etc. A similar division can be made of the philosophy of technology. However, as was noted by Banse and Grunwald, the proportion of research in the philosophy of technology that is devoted to one of the specific engineering sciences is much lower than the proportion of research in the philosophy of science that is devoted to one of the specific sciences. It should also be recognized that the division of technology into technologies is somewhat less clear than that of natural science into the specific natural sciences. The conventional division of engineering science into subdisciplines is ruled by the “primacy of practice” [Banse and Grunwald, this volume Part I]. The creation of specialties such as mechanical engineering, civil engineering, bioengineering, etc. largely coincides with the establishment of new educational programmes for engineers. It is the result of complex social processes in which the organization of the labour market for engineers has a large influence. The resulting subdivision of engineering is somewhat unsystematic. Some branches of engineering such as electrical and chemical engineering are defined according to the methods or the physical principles that they use. Others such as medical technology and safety engineering are defined according to their social purpose. Not surprisingly, there are many overlaps among the specialties of engineering and consequently among the disciplines of engineering science. The issues in the philosophy of science that are specific for only one of the sciences tend to be of three major types. First, many discipline-specific concepts are in need of philosophical clarification. That applies for instance to the biological concept of a species and the chemical concept of a functional group. Secondly, many scientific theories give rise to metaphysical or epistemological questions. Biological evolution, quantum mechanics, and relativity theory have each given rise to an extensive literature on such topics. Thirdly, the implications of scientific knowledge on our worldviews are mostly discipline-specific. They are exemplified by discussions on human genetics, the origin of the human species, and behavioural neurosciences. Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

1196

Sven Ove Hansson, associate editor

In the philosophy of technology and the engineering sciences, all these three categories of issues are present. To this a fourth type of discipline-specific issues should be added. Technology does not only produce knowledge; its primary task is to produce new artifacts and put them to effective use for practical purposes. Therefore, the social, cultural, and ethical effects of the creation and use of such artifacts are important topics in philosophical reflection on the technological subdisciplines. (True, such discussions have also arisen in relation to scientific discoveries, but they have then been focused on potential technological applications of these discoveries, such as nuclear weapons and human cloning.) These four categories are of course to a large extent overlapping, but it is useful to distinguish between them in order to see the breadth of the issues that will be treated in the final part of this handbook. Each of its five chapters deals with the philosophical issues in an engineering discipline: architecture (Christian Illies and Nicholas Ray), agricultural technology (Paul Thomson), medical technology (Sven Ove Hansson), biotechnology (Henk van den Belt), and computing and information technology (Philip Brey and Johnny Hartz Søraker). The four categories of philosophical issues are all amply represented in these chapters. The need for clarification of discipline-specific concepts will be abundantly exemplified in the chapters that follow. Some of these concepts are indeed part of the foundations of the discipline in question. Hence, the definition of biotechnology is far from clear, and this is an important definition since it has implications for the regulation of technological practices. Similarly, the notion of a computation is subject to in-depth debates among computer scientists. Other basic concepts in computer science that are in need of clarification include those of data and information. In architecture, the concepts of function and utility have been subject to heated debates. Medical technology has provided new methods for the early detection of propensities to a disease long before the disease becomes manifest, and this has led to new problems for the definition of health and disease. Important discipline-specific epistemological and metaphysical issues will be found in at least two of these disciplines. In computer science, the ontology of virtual reality is in urgent need of philosophical clarification. The same applies to the nature of a computer simulation, its relation to a natural experiment and its role as a source of knowledge. In biotechnology the distinction between living and non-living matter, as well as that between natural and artificial objects, comes to the fore in new ways. The achievements of the engineering sciences sometimes have implications for our worldviews, i.e. the ways in which we conceive ourselves and our relations to the world in which we live. Hence, some authors fear that current developments in biotechnology will lead to a transgression of what they see as a boundary between the natural and the unnatural. Others claim that there is no such boundary or that there is no danger in crossing it. In medical technology, the use of increasingly advanced methods for genetic diagnosis has been claimed to lead to a different view on ourselves as persons, and perhaps even have impact on how we view the issue of free will. Medical technology has also frequently been a culprit in debates on

Introduction to Part VI

1197

medicalization, i.e. the transfer to medicine of problems that we previously treated as existential or as parts of the normal conditions of human life. In computer science, attempts at creating artificial intelligence have given rise to anticipatory discussions on what human life would look like in a possible future where we live alongside with artificial agents that are in some respects more capable and more perspicacious than what we are ourselves. A less drastic, but more immediately relevant worldview issue is how our views on human societies and communities will be impacted by the new communication patterns that develop with the Internet. Finally, in principle all technologies have social, cultural, and ethical implications that are in need of careful reflection. In architecture, the social effects of different approaches to urban planning are constantly under discussion. In biotechnology, much of the philosophical discussion has focused on the specific types of (potential) risks that these technologies may be associated with. The possible risks emanating from the agricultural use of genetically modified organisms are a case in point. Furthermore, the use of Genetic Use Restriction Technology in agriculture may have a major impact on power relations in particular in developing countries. In medical technology recent discussions have focused on the ethical aspects of enhancement, i.e. treatments that improve human functioning to levels above the normal, thus going beyond the traditional medical aim of restoring human functioning to normal levels. The ethical implications of computer and information technology are far-reaching enough to have given rise to a whole new subdiscipline of applied ethics, namely computer ethics. It deals for instance with the issues of privacy, information control, and intellectual property that computer and information technology gives rise to. The five technologies that are treated in this final part of the Handbook have been selected for the significant philosophical literature that is available about each of them. A couple of these chapters are the first comprehensive summary of the philosophy of the respective discipline. Many other engineering disciplines give rise to important philosophical issues. This applies for instance to structural mechanics [Clausen et al., 2006], chemical engineering [Van Brakel, 2000; Schummer, 2001], nanotechnology [Schummer and Baird, 2006], systems engineering [Hughes and Hughes, 2000; Ferris and Cook, 2006] and military technology [Fichtelberg, 2006; Haws, 2006]. The philosophy of engineering disciplines is still for the most part an unexplored territory. Judging by the expeditions that have been made into a few parts of this vast terrain, the prospects for its further investigation are promising.

BIBLIOGRAPHY [Clausen et al., 2006] J. Clausen, S. O. Hansson, and F. Nilsson. Generalizing the Safety Factor Approach. Reliability Engineering and System Safety, 91, 964–973, 2006. [Ferris and Cook, 2006] T. L. J. Ferris and S. C. Cook. The necessity for a phenomenological and analytic philosophy of systems engineering. Proceedings of the 36th CIE Conference on Computers and Industrial Engineering, Taipei, Taiwan, 4079–4089, 2006. [Fichtelberg, 2006] A. Fichtelberg. Applying the Rules of Just War Theory to Engineers in the Arms Industry. Science and Engineering Ethics, 12, 685–700, 2006.

1198

Sven Ove Hansson, associate editor

[Haws, 2006] D. R. Haws. Engineering the Just War: Examination of an Approach to Teaching Engineering Ethics. Science and Engineering Ethics, 12, 365–372, 2006. [Hughes and Hughes, 2000] A. C. Hughes and T. P. Hughes, eds. Systems, Experts, and Computers: The Systems Approach in Management and Engineering, World War II and After. Dibner Institute studies in the history of science and technology, Cambridge, Mass., 2000. [Schummer, 2001] J. Schummer. Ethics of Chemical Synthesis. Hyle: An International Journal for the Philosophy of Chemistry, 7, 103–124, 2001. [Schummer and Baird, 2006] J. Schummer and D. Baird, eds. Nanotechnology Challenges: Implications for Philosophy, Ethics and Society. World Scientific Publishing, Singapore, 2006 [Van Brakel, 2000] J. Van Brakel. Modeling in Chemical Engineering. Hyle: An International Journal for the Philosophy of Chemistry, 6, 101–116, 2000.

PHILOSOPHY OF ARCHITECTURE Christian Illies and Nicholas Ray

1

1.1

INTRODUCTION

Two roles of a philosophy of architecture

This paper considers the relationship between Western philosophy and architecture. We discuss philosophy’s contribution to architecture and the way in which we can reflect philosophically about architecture, that is to say principally about individual buildings rather than cities.1 This double role of philosophy is unusual in the field of Technology and Engineering Sciences: a philosophical investigation would more often look critically at the process of gaisning new knowledge or designing and producing something, or it would try to understand or evaluate types of artefacts (for example computers). In contrast, the practice of architecture is already deeply involved in and shaped by philosophical reflection and ideas. Architecture is guided by architectural theory, an ongoing discourse that bridges between philosophy and the practical discipline. Thus when philosophy looks at architectural artefacts or production, it investigates something that is itself partly philosophical, at least in the widest sense of the word.

1.2

Why architecture needs theory

That architectural practice is closely related to theory has been acknowledged for a long time. The Roman architect Vitruvius (ca. 80/70 - 25 BC), whose writing has exerted an unequalled influence on European architecture, at least since the Renaissance, argued for an inseparable link between the two: Architecture is a science arising out of many other sciences, and adorned with much and varied learning; by the help of which a judgment is formed of those works which are the result of other arts. Practice and theory are its parents. Practice is the frequent and continued contemplation of the mode of executing any given work, or of the mere 1 The city, rather than architecture per se, has been a fruitful area for philosophical reflection, from Jerusalem as a metaphor through St Augustine, and from Plato’s polis to the writings of Derrida, Habermas and Alexander Mitscherlich. But it would not be possible to do justice to this topic in a single chapter. For the complex issues raised by cities see Meagher [2008].

Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

1200

Christian Illies and Nicholas Ray

operation of the hands, for the conversion of the material in the best and readiest way. Theory is the result of that reasoning which demonstrates and explains that the material wrought has been so converted as to answer the end proposed. Wherefore the mere practical architect is not able to assign sufficient reasons for the forms he adopts; and the theoretic architect also fails, grasping the shadow instead of the real (umbram non rem persecuti videntur ). He who is theoretic as well as practical, is therefore doubly armed; able not only to prove the propriety of his design, but equally so to carry it into execution. [Vitruvius, 2001, Chapter 1,1 - 1,2] Vitruvius suggests that one of the characteristics of architecture is the role that theory plays for architects: they should be knowledgeable in many sciences and reflect upon what they do. Amongst other disciplines, Vitruvius wants the architect to be “acquainted with history, informed on the principles of natural and moral philosophy, somewhat of a musician, not ignorant of the sciences both of law and physic”[Vitruvius, 2001, Chapter 1.3]. And Vitruvius seems to suggest that these qualities are also present in the building: only if theory has guided the architect in the designing and building process, will the product be proper architecture. Vitruvius’ concept of “theory” is not identical to current understanding, of course, nor would such an emphasis on theory be unquestioned today.2 Some have suggested that architecture could be a natural product of meeting society’s needs, pointing to widely-admired “vernacular” buildings, and to the apparently inevitable way in which certain great buildings evolved within a tradition.3 But in a self-conscious society with its complex demands it is unlikely that building can be such a natural activity (see, for example, [Rapoport, 1969]). Others warn against overburdening buildings with inappropriate social theories, when the criteria for judgement should be aesthetic (See, for example, [Scruton, 2000]). The Swiss architect Le Corbusier (1887-1965), who is often cited as an architect fixated by technology (“The house is a machine for living in” [Le Corbusier, 1946, p.89, p. 112], a statement frequently quoted out of context) believed that architecture is distinguished from buildings, or other engineered artefacts, not by theory’s contribution but by the fact that it moves us emotionally:

2 Most commentators regard Vitruvius as a muddled thinker, whose writings are an eclectic collection of the work of others. Nevertheless he had a concern for practical inventiveness, and a profound respect for traditions that had been inherited from the Greeks. His convoluted prose is the despair of modern translators, just as it was for Alberti in the Renaissance, who even suggested at one time that it would have been better if his Ten Books had never survived. See [Vitruvius. 2001]. 3 As William Lethaby (1857-1931) put it: “a Gothic cathedral may be compared to a great cargo-ship which has to attain to a balance between speed and safety. The church and the ship were both designed in the same way by a slow perfecting of parts; all was effort acting on custom, beauty was mastery, fitness, size with economy of material.” [Lethaby, 1955, p. 158]. Historians have answered by indicating the highly self-conscious ways in which Gothic architecture developed stylistically. See [Panofsky, 1957].

Philosophy of Architecture

1201

Finally, it will be a delight to talk of ARCHITECTURE after so many grain-stores, workshops, machines and skyscrapers. ARCHITECTURE is a thing of art, phenomenon of the emotions, lying outside questions of construction and beyond them. The purpose of construction is to HOLD THINGS TOGETHER; of architecture TO MOVE US. [Le Corbusier, 1946, p. 23]4 But all these cases do not show that there can be architecture without some, possibly implicit theory. Le Corbusier’s claim, for example, also implies a philosophical position — one that distinguishes “architecture” as having some special qualities that can move us emotionally, and involving some special skills that architects, as opposed to mere builders, would possess. There are strong reasons for thinking that architectural practice necessarily entails a theoretical position even if it is explicitly denied — at least if it is the result of conscious decisions. Whereas there may be one or few right (in the sense of most efficient or functional) answer(s) to an engineering programme in other areas, once the parameters and goals have been correctly defined, there will always be more than one possible answer to an architectural problem. Thus architecture demands that conscious decisions are made between different forms and ideas — and therefore questions of judgement and interpretation are intrinsic to the discipline. Further, architecture is nearly always embedded in a culture, its ideas and ideals, its language of forms and its tradition, even if it is critical about it. The practical answers architecture gives to a problem (and the very formulation of the problem) will have to reflect this context. And reflection on a cultural context and on one’s embeddedness is theory.5 Thus as long as architecture is conscious, it will be accompanied by architectural theory — including theories about the exclusion of architectural theory, such as the romantic longings for a pre-theoretical vernacular way of building. As a public activity that stands at the heart of cultural life, architecture has also to address fundamental issues like questions of right and wrong in ethical matters, problems of the good life, or about meaning and its expression. It is here that architectural theory meets philosophical reflection, and all of its areas may be involved: logic, epistemology, ethics, aesthetics, philosophical anthropology and metaphysics, to name just the most important.

1.3

How philosophical theories intersect with architecture — some examples

Before outlining some of the philosophical themes that are relevant for the production and the appreciation of architecture, here are several examples that reveal how deeply architecture is intertwined with philosophical questions. There are, for instance, ethical issues that begin even before the planning or design of a building: The architect must decide for whom he builds at all: Rem 4 The

emphasis is characteristic of this book, which had a powerful influence as propaganda. as we shall argue below, such reflection can be seen as an essential component of responsible “practice”. See also [Sch¨ on, 1983]. 5 And,

1202

Christian Illies and Nicholas Ray

Koolhaas (of OMA, the Office of Metropolitan Architecture) is the architect of the Educatorium of the University of Utrecht — certainly a worthy cause. But what about him accepting the commission to build the new headquarters for the Central Chinese Television (CCTV) Station, a d600m project in Beijing? Arguably, China’s national TV station has played a supportive role within a regime that has a long record of human rights violation, and it has been suggested that architects should refuse to associate themselves with such a commission.6 Certainly, whether architects choose to accept a request will often depend on the personal position that they hold. The Finnish architect Alvar Aalto, who regarded himself as something of an anarchist politically, built simultaneously for his wealthy Swedish-speaking patrons the Gullichsens and for the Finnish communist party — it is implied that architecture thereby in some way transcends the conditions of patronage to provide a good in itself.7 Ethical problems in architecture have at certain periods been more closely allied to architectural form and the problem of style. The problem only emerges when there are major choices to be made, so it was as a response to nineteenth-century eclecticism that A. W. N. Pugin (1812-1852) mounted a campaign to judge architecture on moral grounds [Pugin, 1843]. He argued that Gothic architecture was more “truthful” and “honest” in expressing the structure of a building, in employing materials in a way that was appropriate for the climate of northern Europe, and in its symbolisation of a Christian society. Such arguments are taken up by the influential critic John Ruskin (1819-1900), and appear in the twentieth century (but with different terminology and justifying a different set of forms) to support the claim that International Modernism, by purging itself of all superfluous embellishment, was the only honest way of building in a modern world. Critics have called this “the ethical fallacy”, arguing that there is confusion here between moral and aesthetic questions.8 It is clear that developments of science affect the creation of new forms of architecture, though to what extent remains the focus for debate in every period. In medieval times, structural understanding was conditioned by the use of geometry, ad quadratum and ad triangulatum, which was both convenient, and carried iconological significance. By the Renaissance, however, Bramante’s report on the vaulting of the crossing of the cathedral of Milan could treat the issue of statics foremost, whilst also considering issues of stylistic continuity. It is only in the seventeenth century that we encounter calculation of trusses beginning to determine the architectural form of roofs — it is no accident that the architect Christopher Wren was a member of the Royal Society, the pre-eminent scientific institute of its time. Yet the background to the possibility of this change is metaphysical and brought about by the new understanding of humans and the cosmos in the period of the Renaissance, which Jacob Burkhardt famously described as the “rise of 6 See

for example “Bauen f¨ ur Despoten?” in Der Spiegel Spezial, Nr.4, S. 84-87, 2002. a brief introduction to Aalto’s life and work see Ray [2005]. 8 For a detailed discussion of this issue see [Watkin, 1977]. An earlier refutation occurs in [Scott, 1980]. 7 For

Philosophy of Architecture

1203

the individual” in The Civilization of the Renaissance in Italy (1860). Following philosophers like Pico della Mirandola, the central tenet of the new anthropology reads: God has made humans as the masters of a world, a world that is intelligible and well-ordered and guided by natural laws. This contrasted with the medieval view which had emphasized the dependency of humans on his unique abilities. The new worldview found an expression in art’s return to antiquity (and was also nurtured by what scholars and artists found there). In architecture, the rediscovery and use of anthropomorphic, classical members with the fully developed architrave show the new pride of being humans, thereby making the form of the human body the supportive element of edifices. Also the understanding of nature as an ordered cosmos is mirrored in architectural changes, like the Renaissance ideal of geometrical urban structures with spatial centralisation. A programmatic manifestation of the new spirit is the design of the ideal city “Sforzinda” by Filarete (1400-1469), which he planned for the Italian despot Ludovico Sforza of Milan. It is star-shaped, highly centralized, and streets and squares are defined by buildings which consist of the same stereometric units. For Renaissance artists, architecture is meant to follow an understanding of man and nature, but we might ask from our current situation which understanding it should use as its starting point, and whether it is an ideal at all to build in harmony with a certain anthropology or cosmology. In an architectural debate between Christopher Alexander (born 1936) and Peter Eisenman (born 1932), Eisenman pleads for building in disharmonious, incongruent ways: “Alternative views of the world might suggest that it is not wholeness that will evoke our truest feelings and that it is precisely the wholeness of the anthropocentric world that it might be the presence of absence, that is, the nonwhole, the fragment which might produce a condition that would more closely approximate our innate feelings today.”9 We might also ask if there is something “more” to (good) architecture — something that cannot be grasped by theory or measured precisely? In his books Christopher Alexander speaks about the “life” that good buildings should have and that architects can create when they have a feeling for it [Alexander, 2002]; others have talked about the generation of “place” rather than “space”, or of a location that allows us to “dwell”. “The nature of building is letting dwell”, as the German philosopher Martin Heidegger (1889-1976) put it; he linked dwelling to a form of human existence that is in harmony with “Being”. He observed such harmony in the “dwelling” of farmers in the Black Forest and that is why their buildings are good buildings: “Only if we are capable of dwelling, only then can we build” [Heidegger, 1951, p. 362].

1.4

Short overview

In what follows, we will examine the ways in which philosophy intersects with architecture; in particular we will look at the way philosophy has been relevant for the reflection upon and for the production of architecture. 9 http://www.katarxis3.com/Alexander

Eisenman Debate.htm (23.3.2008).

1204

Christian Illies and Nicholas Ray

In the next section, we investigate how philosophical ideas have shaped architecture and its theory. We focus on different, often oppositional approaches: Plato’s philosophy of ideal forms and its role for Renaissance architecture is our first case, followed by an analysis of the ideas that have given birth to architectural Historicism in the 19th century. Heidegger and his importance for critical regionalism and new urbanism is discussed and, finally, Postmodernism and other post-metaphysical philosophies and their importance for architecture. In the third section we investigate ways in which architecture can be looked at ethically. There are several, rather different ways that can be distinguished — and given that some of them seem of high importance for our life as human beings, we find it surprising that philosophers of ethics have given so little attention to architecture. We move on to aesthetics in the fourth section, asking how do different aesthetic theories shape architecture and our understanding of it? Among the fundamental questions raised in this section are: what are the key categories for an appreciation of architecture? And how is the relationship between buildings as functional artefacts and aesthetic objects to be understood? Finally, in the fifth section, we shift the focus towards the activity of design: the work of three twentieth and twenty-first century designers (Kahn, Koolhaas, and Aalto) is examined in more detail. We analyse how the production of architecture, as well as its reception, may indicate a philosophical stance. Sometimes this may be explicit, since the architects are clearly claiming a position.10 Sometimes the architects’ intentions are not stated, or if they are they seem to be negated by the experience of the work itself, and then critics and historians tend to interpret the artefacts in the light of the prevailing cultural context. The chapter ends with a short reflection on what it has achieved, and suggests some directions for the future of philosophy of architecture. 2 PHILOSOPHICAL IDEAS AND WORLDVIEWS

2.1

Ideas and clich´es

Architectural theory is nurtured by philosophical ideas; the concerns and questions that move people at a certain time as much as their visions and worldviews are mirrored in their buildings: Architecture provides functional and technical solutions but is also a practical answer to philosophical questions. Yet to account for philosophy’s role for architecture one faces a profound difficulty: Architects have often read philosophers in ways that scholars find very problematic. Because architectural debates are embedded in wider reflection of a culture, they are shaped as much by clich´es as by ideas that are current at any time. Misunderstandings may also be influential. As an example, Spinoza, though 10 Of course there is the danger of what has been called an “intentional fallacy”. Even if architects intend certain meanings, these might not be evident in the actual design. See [Wimsatt and Beardsley, 1954].

Philosophy of Architecture

1205

his philosophy is in fact an extreme form of rationalism, became one of the cultural heroes of romanticism. A further complication stems from the often muddled use of concepts within architectural theory, that is neither congruent within the debate of the discipline, nor in harmony with philosophical usage. People simply mean very different things when they use words such as “function” or “deconstruction”.11 We will therefore look at philosophers and philosophical ideas in the way that they were interpreted and utilized by architects, independent from whether their reading should be seen as a correct or even plausible interpretation. Let us first turn to Plato whose importance can hardly be overestimated: His vision of a world of ideas or forms behind or above empirical reality inspired architecture for centuries. Building became a quest for the manifestation of such Platonic ideals.

2.2

Plato’s philosophy of ideal forms

Plato famously argued for the possibility of grasping “ideas” or “forms”, that is a space- and time-less realm of subject-independent principles (or universals) that ultimately provide the structure and character of our world (or the particulars). Among the most important of these ideas are goodness and beauty, but there are also mathematical ideas (like geometric forms) and ideas of empirical objects (such as of a tree, a horse, or a house). For Plato, these ideas are more perfect and real than the empirical world, because they are the paradigmatic models of all that is. A famous example is the circle: all circles that we see or draw are necessarily imperfect to some degree, but they are circles by being approximations to the mathematical idea of a perfect circle. Without having this idea, we would not regard these imperfect instantiations as circles. Plato refers to Socrates’ seemingly paradoxical remark that he who sees merely with his eyes is blind. For Plato, the invisible world is the most intelligible while the visible world is obscure and least knowable. Plato’s position is most famously captured in the Allegory of the Cave in his Republic [Plato, 1945, 7.514a ff]. Here Plato compares our situation with that of people who live in a dark cave and merely see the shadows of things that are projected by a fire at the wall of the cave; these shadows are the empirical world that we mistake for the true reality. Our task is to break through this ignorance — Plato talks about us being chained to chairs in front of the shadow-theatre-wall — and to climb up to the true sunlight outside the cave, namely the ideas. Those who reach this light will have objective knowledge about how things really are. Although Plato’s discussions in the dialogues are often about epistemological issues, his epistemology was embedded in an ontology — and this position became more important for the architectural tradition. For Plato, it seems, only these forms truly exist while empirical things are of a secondary order. Plato was, with regard to all reality, including mathematics, aesthetics and ethics, an “Idealist Objectivist”, because forms, that is the (true) objective reality, were ultimately seen as non-material. 11 See

[Forty, 2000].

1206

Christian Illies and Nicholas Ray

For Plato, these forms have also a normative (or ethical) side to them. By being models for empirical reality, they also show us how things ideally should be. His analysis of the craftsperson illustrates this point. A carpenter who constructs a house needs to know the idea of a house. That gives him the direction towards the perfect house; it will tell him how a house ought to be built, and how the materials ought to be used and arranged. To update the analogy: by knowing the idea, he knows the stresses and the strains and, most importantly, the right proportions. The idea of the house is a kind of blueprint to follow, and only if we approximate it will we have a good house (or any house at all, since a bad house will collapse). Ideas tell us not only what there is, but also what we should strive to realize. This ideal is captured by Plato’s term “imitation” (mimesis) that became crucial in the aesthetic theory of his pupil Aristotle: If the ideas or forms are superior and paradigmatic, artworks should try to imitate these ideals — sometimes also called the “essence” or “nature” of things. This principle became most influential, though its concretisation differed widely due to varying understanding of what these ideas or forms are and what it means to imitate them.12 The influence on architecture of Platonism (and Neo-platonism, the views of followers of Plato from the third century BC onwards, and evident also in the middle ages) can be seen most clearly in the Renaissance. This was the time when Plato’s work was rediscovered, and it was incorporated into a new interpretation of Christianity. Leon Battista Alberti (1404-1472) wrote an influential book on architecture, De Re Aedificatoria, modelled on Vitruvius’ ten books, but with a much more clearly argued thesis [Alberti, 1988].13 In relation to geometry, he claimed that pure forms, like the circle and the square, were closest to the divine; they should therefore be reserved for religious buildings, while less important buildings, such as houses, could have a more casual and pragmatic form. Thus geometry serves to reinforce our sense of propriety. The claim of Le Corbusier, explaining the affect that architecture (as opposed to mere building) can have on people, and arguing that “pure forms are beautiful forms”, not only suggests that decoration is redundant in the twentieth century but echoes Alberti’s prescriptions (albeit aesthetically rather than in their religious sense), and eventually a platonic view that actual artefacts aspire to an ideal form.14 This eschewal of decoration was a particularly twentieth-century phenomenon, however: for Alberti decoration was an important aspect of decor or “appropriateness”. In Book 6, chapter 2, he makes a clear distinction between ornament, which consists of the correct use of the clas12 For

the different meanings of nature see [Lovejoy, 1927]. is a mistake to see Alberti’s thinking as purely Platonic or Neo-Platonic, however: his theory incorporates aspects of Aristotelian thinking, and as Erwin Panofsky points out, his definition of beauty “renounces any metaphysical explanation of the beautiful”. See [Panofsky, 1968]. 14 Le Corbusier did not follow Alberti’s prescriptions in relation to religious architecture, however: his pilgrimage church at Ronchamp used a complex free form, whereas in Vers une Architecture he illustrates simple housing schemes based on a pure cube. In the nineteenth century the ideas of gestalt psychology had provided a further argument for pure geometries: they could be perceived more easily and were thus “stronger” figures. 13 It

Philosophy of Architecture

1207

sical orders in their proper proportions, and beauty, which he famously defined as “that reasoned harmony of all the parts within a body, so that nothing may be added, taken away, or altered, but for the worse”: Ornament may be defined as a form of auxiliary light and complement to beauty. From this follows, I believe, that beauty is some inherent property, to be found suffused all through the body of that which may be called beautiful; whereas ornament, rather than being inherent, has the character of something attached or additional. Beauty is thus the essential idea, whereas ornament is an embellishment. Alberti returns to these distinctions in Book 9. But when we make judgements on beauty, he argues, we do not merely follow our fancies, because reasoning is involved. There are three components: “number, what we might call outline, and position”. Composition is the art of bringing these components together, and when they are perfectly harmonious they represent what Alberti called concinnitas. Alberti also makes practical suggestions on how to achieve the “reasoned harmony of all the parts”, and some have pointed to how his own buildings illustrate them.15 In Book 7, chapter 15, he states that “for arched colonnades quadrangular columns are required”. This is not only for technical reasons (though these can be overcome), but also because Alberti understands an arched opening to be a break in a wall, a kind of curved beam. It is therefore logical that its supports should be pieces of wall, that is to say quadrangular, whereas the applied ornament of a pediment would use engaged columns. Alberti’s own Tempio Malatestiana, the transformation of the 13th century San Francesco at Rimini for Sigismondo Malatesta, constructed from 1450 but never completed, illustrates the theory. The powerful blank arcade at the side has massive piers, and turns around to the west front of the church; the entablature above is supported by elegant Corinthian halfcolumns. The well-proportioned arched wall architecture, obeying the principle of concinnitas, could be said to represent beauty, and the applied columns are the appropriate ornament. Both the proportioning of the parts and the application of decoration serve to ennoble the building and transform it into a tempio (with the neo-platonic associations that suggests), which is suitable to act as a mausoleum for the patron and his wife.

2.3

Historicism in philosophy and architecture

An architecture of association, rather than one derived from the authority of the ancients, appears for the first time in the mid eighteenth century. Its first manifestations are in garden buildings, and the influence is as much literary as archi15 In particular Rudolf Wittkower. The interpretation of Alberti’s theory of the colonnade and wall follows that advocated in his Architectural Principles in the Age of Humanism, of 1949. Others have argued that Alberti never refers to contemporary buildings, so that the principles he advocates are not to be understood as prescriptions. His book is certainly not a “primer” for use by practitioners, in the way that the later treatise by Sebastiano Serlio was. See [Kruft, 1994].

1208

Christian Illies and Nicholas Ray

tectural. By the middle to the end of the nineteenth century, all over Europe we find an eclectic Historicism. The style of the time involves the revival of numerous historical styles, beginning with a regained appreciation of the Gothic, but soon turning to other epochs. A dominant example of such a turning away from classicism was the new parliament building in London, by Barry and Pugin from 1840, at least in its decoration. Internationally, architects used architectural motifs from the antique or from the exotic east, such as India, China or Japan, often employing different styles to the facade and to interiors. We find (neo-) gothic cathedrals (everywhere), (neo-) baroque opera-houses (e.g. Paris), factories that looked like mosques (e.g. Dresden), or train stations with renaissance campaniles (e.g. Erfurt). Historicism is striking for its seemingly random combination of different style elements from one or more epochs — when Leo von Klenze designed the K¨ onigsplatz as the new centre of Munich, he built a 1:1-replica of the Propylaeum (from the Athenian Acropolis) with two side-towers in an Egyptian style. The term “Historicism”, however, has different meanings. In his essay “Three kinds of Historicism”, Alan Colquhoun mentions three dominant definitions [Colquhoun, 1985, p. 202]: firstly “the theory that all socio-cultural phenomena are historically determined”, secondly “a concern for the institutions and traditions of the past”, and finally “the use of historical forms”. While the third definition covers the artistic (and architectural) style of the nineteenth century, the first definition refers to the philosophical ideas that have shaped this style — namely the idea that all historical phenomena have a unique or singular character because they are not expressions of timeless principles but determined by the situation and context.16 People began to study different epochs and cultures in order to understand them in their own terms rather than searching for the allegedly universal principles behind them. The other side of this conception is to understand the world, or nature, as being in a state of permanent change — unstable rather than fixed or permanent. In order to understand how philosophical and architectural Historicism are connected, we should look more closely at the philosophical ideas behind Historicism, at least as it was generally understood. The new approach seemed to be supported by Kant’s critical epistemology, according to which we have no direct awareness of pure sense-data, but select and shape them; we apprehend sense-data only as unified and structured by a priori categories of the mind. Rather than being a passive receiver of information, about how the world is, the mind plays an active and creative role in the process of cognition. Consequently, all knowledge is seen as perspectival in character; it is not a discovery of the objective reality (let alone of the essence of, or ideas behind, reality) so much as a social or individual creation. Historicism can also be seen to follow Hegel’s analysis of history, and the discovery, by him and others, that the categories and ordering principles of the mind have changed with the succession of cultures and epochs: the historicity of the human mind. When we look at the past, we find not one interpretation of the world 16 For this meaning of Historicism see [Meinecke, 1946]. A good discussion of the history of the term can be found in [Mandelbaum, 1967, IV, 22-25].

Philosophy of Architecture

1209

but instead a variety of perspectives or worldviews. Thus Historiography should be about an “objective and exhaustive examination of facts” and the “attempt to penetrate the essential spirit of the country or period being studied” — that means to understand a time in its own terms [Colquhoun, 1985, p. 204]. Hegel, however, was still presupposing a (dialectic) teleology as an internal logic of the developments of the diverse cultures and epochs; later historicists have given up any such assumption, explaining changes in terms of the adjustment of men and groups to life under different historical and social conditions. This is accompanied by a rejection of any evaluation; no period is inferior or superior to another — “every age is next to God”, as Leopold von Ranke famously said. This philosophical position offers an explanation of the architectural historicist combination of different style elements: an artistic style was seen as an expression of a time, its culture and values, and no longer of eternal truth. As a result, no style was better than another; they all merely mirrored different cultures with their aesthetic ideals. Thus classicism was no longer privileged over other styles, instead styles became a matter of association or the architect chose the style that seemed suitable for particular purposes: a church might be Gothic, the style of medieval Christianity, whereas a library, to house classic texts from antiquity, could itself be classical. And King Ludwig I of Bavaria (1786 — 1868) wanted his capital Munich to be seen as a second Athens. One of the more radical consequences of 19th century perspectivism is the claim that all truth is relative. Each epoch or culture, historicism claims, develops its own view of the world in its totality because of the different values, categories and presuppositions upon which its cognition is based. But if these categories are themselves essentially variable and arbitrary, then all claims to an objective grasp of reality (let alone absolute knowledge) are baseless in principle — each epoch and culture has its own truth. As Colquhoun writes, the new insight was that “each culture could adhere only to its own notion of the true and the false, through values that were immanent in particular social and institutional forms.” [Colquhoun, 1985, p. 204] The position was developed comprehensively by Friedrich Nietzsche, for whom we literally make reality, and make our own truths. The only supratemporal, or “absolute” truth that might be found is this very insight, namely that no absolute truths about the nature of things are possible. The idea of the relativity of all knowledge and worldviews is characteristic of modernity: Nietzsche blamed his time for being unable or unwilling to accept this insight and draw the inevitable conclusions from it. Nietzsche’s critique of reason as the main tool for understanding and mastering reality has also had a profound influence. He famously made this point when he proposed the rehabilitation of the irrational ecstatic element in culture that he found exemplified in Dionysius, the god of intoxication. He contrasted Dionysian with Apollonian characteristics and saw both as sources of cultural production in Greek antiquity [Nietzsche, 1872]. Rational thought is Apollonian in character since it is structured and makes analytic distinctions; the Dionysian instinct, on the other hand, is characterized by irrationality, violence and chaotic emo-

1210

Christian Illies and Nicholas Ray

tions, but also by creative enthusiasm and exuberance. Nietzsche points out that from Socrates onward the Apollonian had dominated Western culture and thought and demands that we rediscover the “dark” side of art — as exemplified by the Gesamtkunstwerk of Richard Wagner, who was hailed as a saviour by Nietzsche, at least for a short time. Nietzsche values art very highly not because it gives us access to a higher ideal world, but because, believing existence to be absurd, he thought art, uniquely, had the capacity to assist in rendering “the terror and horror of existence” bearable.17 This, however, could only be achieved by a radically renewed art of an individual genius who opposes the taste of his time and bravely walks his own way. Nietzsche, although philosophically a historicist, attacked much of historicist culture. Nietzsche influenced many architects (and artists) at the end of the nineteenth century and beyond, especially of the Art Nouveau movement. Henri van der Velde (1863-1957) refers directly to Nietzsche as a prime source of inspiration for his work. He followed his sharp critique of bourgeois culture and was inspired by Nietzsche’s vision of renewing European culture through an art that, very much like a Gesamtkunstwerk, unites form and content, art and daily life, public and private identity. Van der Velde happily built the Nietzsche archive in Weimar (1902-03) and planned a Nietzsche-Stadion (1911-1913). In America, aspects of Nietzsche’s writings inspired the prose of Louis Sullivan (1856-1924), whose Autobiography of an Idea and 1886 Essay on Inspiration owes as much to German romanticism as to the poetry of Walt Whitman.18 It is particularly telling that he talks about himself in the third person: Louis is portrayed as a heroic figure who is destined to overcome the petty obstructions placed in his way in pursuit of his art. Another architect who consistently referred to himself in the third person was Le Corbusier — born Charles-Edouard Jeanneret: in 1920 he founded the magazine L’Esprit Nouveau with the painter Am´ed´ee Ozenfant and adopted his pseudonym. Le Corbusier’s thinking is complex, influenced by his voluminous reading, but we know that at the impressionable age of twenty-two he visited Paris for the first time and devoured Renan’s Life of Christ and Nietzsche’s Thus spake Zarathustra. According to William Curtis, “this was a time of turmoil in which he wavered between certainty of his Olympian role and deep self-doubts.” [Curtis, 1986, p. 29]19 In fact, wherever architects see themselves as the heroic advocates of new ways of building, or new ways of interpreting society (as Le Corbusier certainly was

17 It is obvious that much of this anticipates Sigmund Freud’s suggestion that the irrational and libido-driven subconscious is decisive for human actions, especially as a prime source of artistic production. Yet for Freud it is mainly the sublimation of primitive instincts that finds its outlet in human creativity, not the direct expression of the libidinous impulse. A second difference is that for Freud the subconscious is a possible object of science since it follows rules and principles that are universal to all mankind and that can be discovered empirically. (That is why Traumdeutung can be a scientific endeavour.) 18 See [Kruft, 1994]. 19 Charles Jencks’ Le Corbusier and the Tragic View of Architecture makes a case that Nietzsche is fundamental to Le Corbusier’s thinking.

Philosophy of Architecture

1211

¨ to) they tend to invoke the Nietzschean idea of the Ubermensch who transcends his circumstances to achieve the superhuman.

2.4

Heidegger’s critique of the subjective-objective divide

Historicism focused on the view that socio-cultural phenomena stem from, and can be explained by, a specific historic context. Martin Heidegger went a step further by historicising the relationship of humans to nature and all that is, including ourselves — or in his terms: our relationship to Being. The way we relate to Being, Heidegger argues, is developing, or more precisely deteriorating over time. It is a history of an increasing gap between the subject and the object, between autonomous, world-less humans and a world deprived of all value that becomes something we see merely in functional terms. Modern humans lost their place in Being (thus nature, and the world) and consequently find no more meaning in their existence. And, most importantly Heidegger would see the two opposed ways of looking at the world we have described, namely a more subject- and a more object-centred approach, as just part of this process — and something to overcome rather than embrace or take as a given framework. For Heidegger, the origin of this process has to be seen in ancient Greece where Western philosophy began. By raising the Socratic question of rational justification, man began to move himself away from, and place himself against, Being — a process that leads to modern science and technology as the ultimate attempt to dominate this world, with its associated positivistic philosophy, and this is an attempt that Heidegger regards as totalitarian in its aspiration. This critique explains an important aspect of his philosophy, namely a prevailing anti-rationalism. Precisely because of the negative effects of Western philosophy, he is critical of the ideal of reason that, according to Heidegger, lies at its heart and culminates in the Enlightenment.20 His own philosophy is designed to explore new manners of reflection that overcome this alienation. Heidegger aims at nothing less than regaining a unity of man with Being through a radically different way of thinking. For the committed Heideggerian the whole ambition of the present chapter, which describes various theoretical attitudes in relation to architecture, is inevitably circumscribed by post-enlightenment pre-conceptions about what theory is. Heidegger’s resulting methodological innovations are the cause of considerable animosity. While some argue that he opens a new way to think more deeply (and to understand Being more adequately), by departing from the traditional manner of reasoning, others oppose Heidegger for exactly this reason. His way of presenting his ideas does not allow for the normal standards of critical control or debate. It is held against him that instead of presenting arguments, he merely “reveals” some ideas he happens to have, or refers to other visionary prophets and poets like H¨olderlin. 20 A distinction should be made between different notions of reason, which it has been suggested Heidegger does not acknowledge sufficiently in his readings of Western philosophy. His critique is directed against what has sometimes been called instrumental reason, but it leaves out richer notions of reason, like Kant’s Vernunft, that are able to identify ends as much as means.

1212

Christian Illies and Nicholas Ray

Based upon this general account of an increasing alienation of humans and nature (or of Being), Heidegger developed a highly influential critique of modern technology. His central point is that we must go beyond the traditional view of seeing technical artefacts as mere instruments that support humans in achieving their goals. For Heidegger technology is itself a metaphysical problem, because it is the culmination of the very form of reasoning that has forgotten what Being is. Technology is the embodiment of a wrong philosophy. The essence of technology — Heidegger belongs to the phenomenological school that asks for the essence of things — is to confront (stellen) nature, to functionalize it and to reduce it to a mere object of manipulation. Ultimately all world becomes an artefact. What does this mean for architecture? For Heidegger, this negative development includes the way we build and comes to its peak in technocratic Modernism, the merely functional interpretation of buildings being one of its clearest expressions. Modern architecture seems just part and parcel of the ongoing alienation of man from Being. To understand a house as merely “a machine for living in” is for Heidegger an “absurdity” that shows all too well the “groundlessness that dominates today’s thinking and understanding”.21 In his 1951 lecture Building Dwelling Thinking — arguably one of the most influential texts for the twentieth century philosophy of architecture — Heidegger outlined what architecture should be like in order to go beyond the malaise of modernity, or, more precisely, to return to a lost unity of man and Being. In Heidegger’s words, man has lost the ability to dwell and to build for dwelling, not mere “living”: “The essence of building is letting dwell” [Heidegger, 1951, p. 361]. And dwelling is seen as a form of being in this lost unity that is not opposed to Being — and which includes the right way of thinking (thus the title of the lecture). Heidegger posits four elements — the “fourfold” — as essential to good dwelling, for bringing “dwelling to the fullness of its nature”, namely “earth, sky, divinities and mortals” [Heidegger, 1951, p. 351]. Basically, all four elements stand for a relation with Being that does not suffer from the split of modern subjectivism. The “earth” means to have a right, non-functional relation to plants, animals or water, the “sky” expresses our relation to light and the passage of time. The “divinities” are understood as unseen yet inherent beings in the world around us that give it its own meaning. By “the mortals” Heidegger means humans and their ability to reflect upon their death; only humans who fully understand their mortality have found an access to Being — and only they can dwell if they place themselves in relation to the earth and sky, and if they realise that the divinities reveal themselves through and in the world. Heidegger’s influence on twentieth century architectural reflection and architecture has been enormous.22 And he rightly reminds us that how we build cannot be reduced to functional considerations or the applications of bourgeois decoration — building is altogether a “deeper” issue. Such a philosophical awareness had become rare in the first half of the twentieth century, a time dominated by special21 These 22 For

remarks by Heidegger are from a 1929-1930 lecture course, cited in [Wigley, 1992]. a more extended discussion of this influence see the excellent article by Woessner [2003].

Philosophy of Architecture

1213

ized architectural theory and European Modernity. Most importantly, Heidegger pointed to the deep link between our ideas and thoughts, the way we live and the manner in which we build. The demands, for example, that architecture should be sensitive towards different cultures and their ways of life influenced the critic Kenneth Frampton (born 1930), who based his appeal for a Critical Regionalism directly on Heidegger’s work [Frampton, 1983]: the primary thesis being that we should build with more sensitivity to a place, its culture and context. Already, in his widely-read Modern Architecture, a Critical History, Frampton had used Heidegger in the final paragraph. He had been praising the buildings of Alvar Aalto, as an example of work which, in comparison to much twentieth century design is better in photographs than actuality, and concludes: Against his inspiring achievement, the present tendency of modern building to be devoid of content, to be reduced, so to speak, through the way it is built, returns us to the Heideggerian challenge that building, dwelling, cultivating and being were once indivisible. [Frampton, 1980] The thinking of the Norwegian architectural theorist and historian, Christian Norberg-Schulz, illustrates the continuing influence of Heidegger’s thought in the late twentieth century. He changed his position from that of his Intentions in Architecture, which is a version of Alberti, translating the Vitruvian firmitas, commoditas and venustas into technics, function and form [Norberg-Schulz, 1966]. In his later books he moves towards an existential theory (Existence, Space and Architectur [Norberg-Schulz, 1971]) and with Genius Loci: Towards a Phenomenology of Architecture [Norberg-Schulz, 1990], Norberg-Schultz made the concept of place a major criterion, and not something that can be subsumed into the other Vitruvian categories, explicitly invoking the authority of Heidegger. Other architectural theorists who have developed a Heideggerian position include Karsten Harries and Dalibor Vesely [Harries, 1975; 1997; Vesely, 2004]. Practising architects may be more influenced by teachers who have their own practice, however. One such is the Finnish architect, Juhani Pallasmaa (1936 -). As a young man, under the influence of the rationalist architect, Aulis Blomstedt, he had been critical of the apparently intuitive and irrational work of his compatriot Alvar Aalto. But in the 1960s he moved towards a phenomenological position, indebted to Merleau-Ponty and Heidegger: architectural creation is not a matter merely of the mind, it involves the whole person. In a lecture in Helsinki, he quotes a statement by the sculptor, Henry Moore: This is what the sculptor must do. He must strive continually to think of, and use, form in its full spatial completeness. He gets the solid shape, as it were, inside his head — he thinks of it, whatever its size, as if he were holding it completely enclosed in the hollow of his hand. He mentally visualizes a complex form from all round itself; he knows while he looks at one side what the other side is like; he identifies himself with its center of gravity, its mass, its weight; he realizes its

1214

Christian Illies and Nicholas Ray

volume, and the space that the shape displaces in the air. [Moore, 1966; cited in Pallasmaa, 2005] Architects should conceive of buildings in the same way. Pallasmaa justifies his advocacy of an empathetic method of design by reference to Heidegger: No wonder Martin Heidegger writes of the thinking hand: ‘The hand is infinitely different from all the grasping organs [. . . ] Every motion of the hand in every one of its works carries itself through the element of thinking, every bearing of the hand bears itself in that element. All the work of the hand is rooted in thinking.’ [Heidegger, 1938, p.357; cited in Pallasmaa, 2005] It is significant that it is easier to analyse the experience of architecture in Heideggerian or post-Heideggerian terms than to translate such thinking into prescriptions for design. Poetically thinking about shelter and its meaning does not tell architects directly how they can improve their work, and much critical writing merely illustrates buildings that are moving examples from the past, by architects who have probably never heard of Heidegger. But Heidegger’s critique of an extreme technocratic functionalism surely deserves attention: by common consent modern architecture often neglected human needs as well as aesthetic conventions. However, Heidegger’s approach is also the object of much critique. Surely his position leads towards a nostalgic romanticism well beyond the needs of the seven billion people that want a decent place to live. And others have argued that Heidegger’s thinking in relation to architecture fails to account for social and economic realities even in privileged societies [Leach, 2002]. Although he expressly rejects any nostalgia23 and claims not to want to return to traditional Black Forest farmhouses,24 it is striking that all his examples of successful dwelling stem from a pre-twentieth century era. Similarly, his strict condemnation of all functional reasoning can hardly convince.25 But more profoundly, we might critique Heidegger for some inconsistency in his critical account of the decline of European culture. According to his diagnosis we are at the end of a sad story of forgetting Being — but then it is neither clear how good dwelling and a right relationship to the fourfold should be possible 23 See this remark: “The flight into tradition, out of a combination of humility and presumption, can bring about nothing in itself other than self-deception and blindness in relation to the historical moment”, in “The Age of the World Picture” [Heidegger, 1938]. 24 See [Woessner, 2003]. 25 Since the details of the “Heidegger case” became widely known in 1987, his involvement with Nazism has been much discussed. It remains an open question to what extent his philosophical ideas and insights are discredited thereby. It is also a point of debate, whether his architectural philosophy has been shaped by the Blut and Boden ideology of the Third Reich. This influence has been argued for by Doreen Massey [1994] amongst others, while Julian Young [2002], for example, pleads for an independence of Heidegger’s philosophy; he sees it as uncompromised by his political involvement. Yet whether or not there is a link between his ideas and Nazi-ideology, his architectural philosophy is certainly detachable from any such alleged connection: all those who developed Heidegger’s ideas of dwelling further, have done so without any link to Nazism.

Philosophy of Architecture

1215

at all in our times, nor where Heidegger’s own place in the European history of ideas should be. (Why, after so much decline, is it suddenly possible to have all the direct insights that he claims to have?) Unsurprisingly Heidegger’s critique of particular phenomena (for example of the technocratic thinking endemic in modernity) has influenced philosophers and architects more profoundly than his general metaphysics of history.26

2.5

The post-metaphysical age

Anyone thinking and writing about architecture in the twentieth and twenty-first centuries has had to come to terms with the fact that the meanings that used to be embedded within the fabric of cities and their buildings have disappeared. If a Gothic cathedral was once a “sermon in stone”, what can be deduced about our society from the major constructions of our own times? The iconography of a complex of buildings such as Le Corbusier’s capitol at the new Indian city of Chandigarh was a self-consciously personal one. He included a vast “open hand” sculpture, and inscriptions describing the cycle of the sun and moon, but the buildings themselves have forms that refer more generally either to the idea of the technology of a “new” nation (the parliament building’s profile is based on power station cooling towers) or are fashioned in response to technical problems, such as the creation of shade on the facades of the buildings, which is achieved by means of the “brise-soleil” on many of the buildings, or the vast canopy over the law courts. Admirers believe that the sculptural forms that result have universal relevance and argue that they refer at a deeper level to primary existential experience. The work of Mircea Eliade [1971] would suggest that the very act of establishing a site in the landscape and the creation of a place under the sun can carry transcultural embedded meaning, and Christian Norberg-Schulz’s Existence, Space and Architecture similarly holds out this hope [Norberg-Schulz, 1971]. But others are less optimistic. The American planner Melvin Webber (19202006) coined the expression “non-place urban realm” in 1965 to refer to the fact that previous understandings of urban fabric had been made redundant in an age of rapid transportation and communication: community was no longer defined by propinquity so that fantasies about reconstructing the city along the models of Europe were pure nostalgia. And in a series of well-documented conferences entitled ANY, a number of architects and critics in the closing years of the twentieth century debated the loss of traditional architectural certainties: not only its material solidity, but also its reference to the specifics of location, stable definition of function, hierarchies of use, and modes of formal signification were all called into question. The final conference was addressed by Rem Koolhaas, who presented two projects: a new headquarters for Universal City in Los Angeles and four new Prada stores. As the critic of the New York Times described it: 26 Which

he would not call “metaphysics”.

1216

Christian Illies and Nicholas Ray

The modern aesthetic was theoretically objective. It evoked the rational mechanics of technology and industrial production. A Koolhaas design may employ modern forms and materials — glass, grids, sculptural abstraction — but does not organize these elements into a neutral, functionalist container. There are subjectivisms along the way. Floors morph into ramps, spirals into squares; escalators expand space as if through a camera lens. Reason is contorted by desire. We seem always to be slipping through some porous membrane, back and forth between semipublic and semiprivate space, as if no fixed boundaries divide environment from self. [Muschamp, 2000] There is a further critique of traditional (rational) metaphysics that has been of particular importance for architecture: Post-modernism, understood as a loosely connected set of ideas and cultural (architectural, artistic and philosophical) trends that lack a clear tenet, and consciously does so. The American architect Robert Venturi suggested that the times require a shift of attitude: At the beginning of this century you could be Bernard “Sure” —you could be a very strong artist and take unambiguous stands. The good guys and the bad guys — it was obvious who they were. Now, I think, intelligent people are no longer that sure of simple answers and drastic actions, and this is reflected in the fact that there are inevitable contradictions and ambiguities in the work.27 Thus Post-modernism rejects all “great stories” and coherent overall worldviews. One could describe it as scepticism that has been turned against itself: if there are no universal truths to be found, then even this judgement — namely that there is no universal truth — cannot be a universal truth.28 What we call “truth” is then not much more than the views of any one time, as for example Michel Foucault and Richard Rorty argue, resulting from both a conscious struggle for power and influence, and sub-conscious needs and desires. Hans Blumenberg (1920-1996) and Peter Sloterdijk (born 1947) suggest that these “truths” fulfil certain social and psychological functions, for example stabilising cultures and communities — but that there is no truth in any more profound, objective sense. The Western idea of reason as a unifying faculty, universal to all mankind, is seen as an essentialist illusion: the enlightenment optimism about rationality, and its further development by modernist functionalism, gets itself “deconstructed” — it is merely the apotheosis of the hegemonic ambition of Western culture and should be abandoned. Consequently, Post-modernism emphasises contradiction, ambiguity, diversity, and interconnectedness; to seek order, structure, let alone meaning in this world is precisely to succumb to the rejected point of view of a universal reason.29 This explains why for Post-modernism there seems no reason to take the departure 27 Radio

interview quoted in [Games, 1985, p. 31]. is an epistemological problem of which na¨ıve postmodernist authors seem be unaware. 29 In order to function as a building, however, any Post-Modern architecture requires technical knowledge and is based upon the (at least implicit) trust that this knowledge is universally 28 This

Philosophy of Architecture

1217

from objective truth (or values, or meaning) particularly seriously: “If sub specie aeternitatis there is no reason to believe that anything matters, then that does not matter either, and we can approach our absurd lives with irony instead of heroism or despair” [Nagel, 1979, p. 23]. While previous forms of nihilism gave much weight to the ‘heroic’ or superhuman task of overcoming the grand illusions and of creating one’s own values or meaning, for Post-modernism even this endeavour is regarded as vain. Nihilism becomes an ironic play — Post-modernism promises a light approach towards a world and existence without gravity and goal; it shows the bearable lightness of being. A clear target for architectural post-modernists was the orthodoxy of CIAM, the Congr`es Internationaux des Architectes Modernes, a group of avant-garde architects. In August 1933 the fourth and most famous meeting of CIAM took place, on the steamship Patras II voyaging to Athens, beginning and ending in Marseilles. The meeting was dominated by the powerful persona and polemic of Le Corbusier. Out of it was to emerge the Charter of Athens, though it was not published until ten years later, a devastatingly reductive prescription for the development of cities, distinguishing the functions of work, leisure, and recreation, connected by systems of circulation, as problems to be “solved” by rational town planning. Paragraph 16 for example runs: Structures built along transportation routes and around their intersections are detrimental to habitation because of noise, dust and noxious gases. Having eliminated the street, “high buildings, set far apart from one another, must free the ground for broad verdant areas” (paragraph 29). “The practice of using styles of the past”, we are told in paragraph 70, “on aesthetic pretexts for new structures in historic areas has harmful consequences. Neither the continuation of such practices nor the introduction of such initiatives will be tolerated in any form.” [Eardley, 1973] Such dogmatic certainty about the inadequacy of traditional forms of planning and architecture called into question the whole concept of “rational” architecture and planning. Post-modernist architecture advocated an unapologetically diverse aesthetics, for example by using traditional elements, historical forms, or surface ornaments from different styles. More obviously, Post-modernism gives up the ideal of any one point of reference. There is no ideal of unity left, no order to be discovered. If the world is ultimately diverse and ambiguous, then this requires an architecture that is based upon such ideas. While Historicism still hoped for a rational answer to the question of style, there is no longer a right answer to be expected — it is left to the aesthetic taste of the architect. Even the concept of a coherent style seems to suggest too much applicable. And the search for an architecture that expresses the diversity and ambiguity of the world seems to suggest some universal normative ideal — namely that art should be an expression of the way things are (in the Post-Modern argument, diverse and contradictory).

1218

Christian Illies and Nicholas Ray

consistency in an ultimately pluralistic world; that is why Post-modernism does not revive a single style but only fragments thereof. The general rejection of all universal values, moral or aesthetic, is mirrored in aesthetic productions: Robert Venturi, amongst others, seeks to avoid traditional evaluations. In his 1972 book Learning from Las Vegas, he celebrates the ordinary and common architecture of, for example, a shopping mall [Venturi and Scott-Brown, 1972]. Venturi argues for buildings that quite explicitly display their symbolism. Paradoxically, under modernist orthodoxy where decoration was prohibited, the whole building was distorted in order to create a symbol.30 But the syntax of this symbolism was so arcane that the buildings were often illiterate in their own terms and unintelligible to the layman, merely huge pieces of sculptural decoration. Venturi, using the precedent of the Las Vegas strip, advocates, in most instances, a “billboard” architecture where the functions were accommodated in quite ordinary sheds and the signs were applied up in front where it showed. This is the procedure, he claims, in mannerist and baroque churches and on the west fronts of gothic cathedrals. But sheds and some applied decoration are all that late twentieth-century budgets will afford in any case. The prescription is therefore both populist (in adopting the ordinary pattern of the Las Vegas strip), and ironic, to those who can read the irony. And the reason for the irony, ultimately, is the context in which architects of the late twentieth century inevitably find themselves. As Venturi explained in his earlier Complexity and Contradiction in Architecture, society no longer values architecture; it spends its big money on armaments, or technological gadgets: The architect who would accept his role as combiner of significant old clich´es — valid banalities in new contexts — as his condition within a society that directs its best efforts, its big money and its elegant technologies elsewhere, can ironically express in this indirect way a true concern for society’s inverted scale of values. [Venturi, 2006]

3

ETHICS

3.1 Ethical reflection on architecture

31

Put in most general terms, Ethics is that branch of philosophy which deals with distinctions between right and wrong in human affairs, in particular with the evaluation of human intentions and character traits, of actions and their consequences. But Ethics can also investigate products of human activities like institutions, artworks or technical artefacts. 30 This is one way of seeing Le Corbusier’s buildings at Chandigarh, described above. Few can be found to defend Le Corbusier’s city plans, but Venturi in fact admires the complexity and inherent contradictions of Le Corbusier’s architecture: it is na¨ıve and unskilful modernism that he is chiefly concerned to attack. 31 This part is partly based upon (and takes material from) [Illies, 2008].

Philosophy of Architecture

1219

In order to investigate the ethical side of architecture more closely, we must be aware of two important ethical distinctions. Neither is specific to the Ethics of Architecture, but of general importance for Ethics in general — and often confused in texts on architectural Ethics. We need to differentiate between, one the one hand, facts about buildings and, on the other hand, guiding ideas — values, norms, standards or the like (either ones that are held by a culture or person, or those that we subscribe to ourselves and thus apply for an ethical evaluation). “Facts” in architecture include the material used (for example: stained glass), the energy required for heating, or the style in which a building is designed. All of this can be described without giving an evaluation of a building — but it will have to be its basis. Ethics thus has a descriptive side to it, when it records the norms and values of a culture or tradition. And Ethics can be a normative discipline that evaluates phenomena and tells people what to do. Both aspects are of profound relevance for architecture. On the one hand, we can describe the values or norms that have guided architects or patrons and that have shaped their buildings. We can say, for example, that Bruno Taut (1880-1938) built the “Glass House” for the Werkbund’s 1914 exhibition with walls out of stained glass in order to inspire a reformed society that shows “translucency” — that is the mutual openness of all humans. (To make this statement neither implies that we share his ideal nor that we believe in the transformational force of stained glass.) On the other hand, people (including architects), institutions (such as building firms) and even artefacts (such as a bridge) can be evaluated from an ethical perspective. We make normative judgements, for example, when we ask whether a building is sustainable (under the assumption that the long-term protection of the environment is morally good) or whether bribes have been paid to get a building contract (under the assumption that such bribery is immoral even in cases when it is not illegal).

3.2

Six relevant ethical areas

We can distinguish the following six areas that seem relevant, of which only the first refers to the building process, and all others to effects of the finished structure: • issues of professional behaviour and interaction during the planning, designing and construction phase • the morally acceptable or unacceptable nature of the function and use of a building • the impact of the building on nature (broadly under the heading of what we would call “green issues”) • the impact on the health and safety of those who use the buildings • the psychological influence on human behaviour, individually and collectively, that the building promotes

1220

Christian Illies and Nicholas Ray

• the furnishing of cultural or symbolic meanings, by choices of form, materials, colours, and aesthetic style [Rapoport, 1990]. This is the most complex of the six areas identified. We will consider these in turn. 3.2.1

Professional behaviour

This area of ethical relevance is probably the dominant topic of much traditional Ethics; some would go so far to define Ethics as a system that regulates interactions amongst humans. And it seems that architecture as a profession is particularly vulnerable to unethical behaviour, mainly because so much money and so many different people are involved in any building process. There is pressure at all levels and many stakeholders are trying to exercise influence on architects. However, hardly any of the ethical rules that professional bodies seek to impose are specific for architects. The professional code of the Dutch Royal Society of Architects, for example, requires an architect not to break confidentiality (paragraph 3.4.), and not to talk negatively about someone else without informing him about it (paragraph 4.2.). Both demands seem right in general for any interaction amongst humans.32 3.2.2

Function and use of the building

What is the building designed for — or what is it used for presently? From a moral point of view, not all usages are equally good. We will, for example, rightly criticize the function of the former Berlin Wall, and expect architects to refuse to be involved in its design. Other cases are more ambivalent: architects are often involved in the design of supermarkets, and when they arrive on the fringe of a small town local traders are affected and campaigns might be mounted to prevent the new buildings gaining permission. Architects are party to this process (they might be involved on either side). More fundamentally, we expect buildings to function well — to be fit for their purpose — and architects who design buildings which look good but do not work are behaving more like sculptors. But we have to recognise that the intended function is not necessarily the function which the building realizes; Don Ihde rightly calls this mistaken conclusion the “designer fallacy” [Ihde, 2008]. It might be that over a period of time the original function is no longer required (as in many castles), or that it has changed its use contingently. This can even be the case in a very short time period, however. Zaha Hadid was commissioned to provide a Fire Station at Weil am Rhein in 1993 by the Vitra group, but the building’s purpose was swiftly made obsolete by the construction of another building and it has been used ever since for the display of furniture; it performs this function in addition to enhancing the “brand” for which it was designed, which was equally important 32 See

also [Spector, 2005].

Philosophy of Architecture

1221

for her clients.33 Evaluating the function can therefore have two objects, that originally intended and the actual realised use of a building. 3.2.3

Impact on nature

We live in a world that has been severely and often irreversibly damaged by technical intrusions, so that the green issue is quite rightly one of the most discussed questions in relation to architecture. Buildings have an enormous environmental impact. They are responsible for 2/5th of the world’s energy consumption (and in the United States 1/6th of all energy is used for air-conditioning).34 In 1999, according to the United Nations Environment Programme, construction activities were estimated to contribute over 35% of total global carbon dioxide emissions — more than any other industrial activity.35 The demand that our building activities should respect the carrying capacity of ecosystems is essentially ethical: the selection of the site, the building process, and the final product should not contribute to the destruction of the environment and should make minimal use of limited natural resources. Some claim that this moral imperative should over-ride all other requirements for architecture: it should not matter that a building is less comfortable for its occupants than it could be, provided it is more responsible environmentally. But views differ on precisely how this value is expressed: Pepper distinguishes between “eco-centric environmentalists” and “technocentric environmentalists”, where the first group harbours ecological ideals (like recycling building materials) in order to reduce the ecological footprints of buildings, the second group strives for “smart”, or highly efficient technical solutions [Pepper, 1984]. Each group profoundly disagrees on what an ideal green architecture would look like. Yet there might be even more options; in a recent paper, Simon Guy and Graham Farmer list as many as six different “logics of green buildings” or understandings of the “environment” and of how we should live in it [Guy and Farmer, 2000]. 3.2.4

Impact on health and safety of individual users

But buildings are occupied and used by people; it is quite generally seen as a moral demand that we build with a maximum concern for the health of users and occupants. A bad sanitation system, for example, threatens the health of the inhabitants. Negative effects can also be less obvious: the lack of natural light can lead to emotional problems (Seasonal Affective Disorder). A proportion of the research in schools of architecture is devoted to making measurements of what constitutes a good environment for people. Though there are cultural differences 33 See

www.galinsky.com/buildings/vitraf. from the U.S. Department of Energy (http://www.eartheasy.com/article global warming.htm, 23.3.2008). 35 http://www.unep.or.jp/ietc/focus/EnergyCities1.asp See also [Fox, 2000; Williamson and Radford, 2000]. 34 Data

1222

Christian Illies and Nicholas Ray

in different societies, it is possible to calculate aspects of comfort, in terms of temperature, humidity and rates of ventilation for example, and establish the range within which we would expect the performance of buildings to fall; if their buildings fail to meet these standards we would hold the architects responsible. And progressively we may be able to measure how a building’s technical performance affects people’s well-being. 3.2.5

Influence on human behaviour

Yet a detailed explanation for how architecture affects us will not by itself answer the ethical issues. What is the morally right way to deal with human needs? It seems plausible that buildings must satisfy at least those needs that are essential for human well-being and whose frustration makes human beings sick. But it is by no means obvious whether buildings should satisfy all needs that we have or only some of them, or which needs under which circumstances. Especially when we look at psychological needs this becomes obvious. Sometimes, it might be morally better to modify or guide some needs than to give them much space; universal desire-satisfaction has been seen as ideal by very few ethicists. Furthermore, oversensitivity to psychological effects of architecture might lead to idyllic holiday villages for tourists and other forms of kitsch, understood as overly sentimental, pretentious design, calculated to have popular appeal, if we are too close to the actual wishes and dreams of people.36 The village of Portmeirion in North Wales, built over a fifty-year period from 1925 by Clough Williams-Ellis (1883-1978), would be an example: while it pleases us on holiday, it is unlikely to do so as a permanent environment. Kitsch satisfies certain sentiments but does so in a partial manner. What it seems to be lacking is challenge: we might argue that truly satisfying artefacts must also demand something from the observer.37 Without falling into the trap of na¨ıve determinism, it seems that buildings can influence the behaviour of people in morally relevant ways. In 1951, Yamasaki, the architect of the New York World Trade Centre, designed the Pruitt-Igoe housing project for the socially disadvantaged. The 11-storey buildings which totalled 2870 apartments were heralded for their innovations, but their structures seemed to invite vandalism and crime so that no one wanted to live there. The complex got demolished after just 20 years. Architectural designs are not only influenced by a certain idea about how human beings live and what they do, buildings can also suggest ideals to their occupants: an obvious way is to make certain actions easy and others difficult or impossible. “Space syntax” research has indicated that not only the distance between starting point and target, but also the street layout in a city, determines the road routes people take; they prefer ways that provide visual experiences [Hillier and Lida, 36 See

[Posener, 1978, p. 164]. the effects on human behaviour are often very close to the effects on psychological well-being: Whether people are happy or unhappy will influence what they are likely to do; it might have been the cold and impersonal effect of some modern social housing project that has encouraged vandalism. 37 Obviously,

Philosophy of Architecture

1223

2005]. And even highly private decisions like family planning seem influenced by our housing conditions [Rzchtarikova and Akkerman, 2003]. These psychological effects became a topic only after the failure of many well-intended public housing projects like Pruitt-Igoe and resulted in a new science, namely architectural psychology.38 The architect and city planner Oscar Newman (1935-2004) claimed that this is a general phenomenon: he had observed in a study in New York that high-rise apartment buildings occupied by many people show a higher crime rate than lower buildings, because their residents tended to show a greater personal responsibility for their place. Based upon this research, he developed the concept of “defensible space” suggesting a form of crime prevention (and increased public health) through community design.39 Building design seems to discourage some kinds of behaviour and encourage others. Small, well-lit rooms with comfortable furniture, for example, can support social exchange in residential accommodation for old people. In 1957, the psychiatrist Humphry Osmond (1917-2004) labelled this quality “sociopetality” and characterised it as “that quality which encourages, fosters and even enforces the development of stable interpersonal relationships such as are found in small, face-to-face groups” ([Osmond, 1957], cited in [Lipman, 2003, p. 192]). However, to which extent architectural forms are responsible (or even to be blamed) for the behaviour of the inhabitants remains controversial. The sociologist Alice Coleman, argues for a strong influence of urban structures [Coleman, 1985], but others have seen more important social factors at work that merely interact with the physical ones. Bill Hillier and others argued that many of Coleman’s results were statistical artefacts and that the same forms might have been perfectly suitable for different inhabitants [Hillier, 1986]. But even if there were agreement on the factual question whether buildings can shape the life of people in such profound ways, there remains the questions of ideals and values: how do we evaluate any influence they may have on human behaviour? Again, it depends on the presupposed moral standards. If, for example, individual holism is seen as a moral ideal, as it is by Charles Taylor, then built environments which work against this should be criticized [Taylor, 1991, chapter 10]. More recent ideas of city planning, in favour of mixed uses rather then the zoning of different functions proposed in the 1933 Charter of Athens, seem a much more positive approach. However, “deconstructive” philosophers see a fragmented life as the authentic form of existence; and thus they should find nothing morally offensive in the alienating effect of a city which obeys the economic rules of the twenty-first century. Two competing ideals of social life and behaviour are evident: an independent, autonomous life of the individual versus a community-oriented life. The first might 38 For a general overview see Mikellides, http://bejlt.brookes.ac.uk/article/architectural psychology 19692007/. 39 http://www.defensiblespace.com/start.htm. and Newman [1973]. It should be added that the well-documented physical and mental illnesses associated with poorly designed social housing projects are often caused primarily by economic and social deprivation, the impoverished quality of the architecture merely illustrating the problem and inevitably compounding it.

1224

Christian Illies and Nicholas Ray

be exemplified in Mies van der Rohe’s designs of a series of court houses; while the interior of each is open-plan, they are all separated from each other by high garden-walls. In contrast to this, the principles of “New Urbanism” are based upon the ideal of a strong community life where people should meet and interact. Independently from the concrete ideals that guide architecture in the fifth area, it is also questionable whether architecture should guide human behaviour at all. We might reject it morally because we think that architecture should also take our freedom and independence seriously. Thus architectural paternalism is especially to be avoided — but we should not be driven to the opposite but equally unacceptable response of simply providing unlimited choice. If, on the other hand, cities are built in accordance with abstract Utopian ideas like those of the Bauhaus, then architecture may fail because it is too demanding; it ignores what people are capable of. A morally good architecture, it would seem, should take humans as they are, but should encourage them to grow beyond that. 3.2.6

Cultural or symbolic meaning

Not only the architecture parlante of the Revolutionary period in France, but architecture in general expresses something and thereby communicates ideas. Buildings can speak to us in a unique and powerful way and thus contribute to the process of arriving at new insights. “All architecture is a gesture”, Wittgenstein writes, “Not every purposive movement of the human body is a gesture. And no more is every building designed for a purpose architecture” [Wittgenstein, 1980, p. 42]. Often the message of this “gesture” is about morally relevant themes. Buildings might encourage certain values in us, or might inspire us to overcome false ideals. Frank Lloyd Wright (1867-1959), for example, designed most of his so-called Prairie Houses around a fire-place or hearth that was pivotal, formally and symbolically, to the plan. This fire-place served not merely as physical gathering area, but served to express the family life and its values, especially unity, harmony with nature and the simple life. Architecture’s functional nature has often been seen to oppose the idea of the building as an expressive artefact [Meyer, 1980, p. 34]. But why should it be impossible to “say something”, that is to express an idea artistically, while realizing a given purpose? It seems that we can appreciate the aesthetic qualities of a house while (happily) living in it. Certainly, if the conditions within which the artefact is created do not leave any room for self-expression, where economic constraints ensure only a very limited range of forms for example, the criticism holds; not because something cannot be functional and an artwork in principle, but because in this instance the artisan has little or no freedom in the making of artistic choices. But to the extent that the creator is free, the (technical) craft turns into an art; and in most cases the creator will have some freedom. Even a clearly defined function can be realised in different forms. Great architects were able to understand the function of the building as a theme for their work that they “talked” about in the language of architecture. Brahms’ Requiem and the Taj Ma-

Philosophy of Architecture

1225

hal are both ingenious artworks that reflect upon death and its meaning for our life. To comprehend a work of art and its message is not merely a form (probably a deficient form) of intellectual insight (as rationalists like Baumgarten claimed); it is an experience sui generis, a unique form of becoming aware of something, which is different from rational understanding and from sensual pleasure. Since art can have a stronger motivational force than (by their nature abstract) logical arguments, art has a specific moral significance. What, then, could be the morally relevant message of buildings qua art or artefact? Architectural messages can be morally relevant in many ways. In its symbolism a building can make us aware, for example, of the possibility of bridging between different cultures by exemplifying what Wolfgang Welsch has called “Transculturalism” [Welsch, 1996]. Buildings can also extol virtues or ideals — Bergilez and Genard have argued that the moral contribution of minimalist architecture is much needed today because this architecture expresses and promotes ideals such as “simplicit´e, d´epouillement, s´er´enit´e, int´eriorit´e” that can function as a critique of the omnipresent consumerist and economical worldview [Bergilez and Genard, 2004]. But whereas it is generally acknowledged that buildings should be safe and no threat to human health or life, and few would argue against the expectations for a green or sustainable architecture, there’s little consent on whether there are cultural traditions that should (in a moral sense) be acknowledged in the architectural style of a region — and also, whether this is an ethical question at all, or merely an aesthetic choice. It should be added that, first, it is obvious that to understand the message of a building some prior knowledge of the “language of architecture”, as John Summerson has put it, is in most cases essential [Summerson, 1966]. That some knowledge is presupposed is not surprising, and, indeed, is the case throughout the arts. Secondly, the language — and thus meaning — of an architectural message can change over time; the meaning we attach to an artefact is inherently unstable. After all, meaning is neither simply created by the architect’s intention nor by the physical features of the building; it is the way the community has reason to understand the artefact. Sometimes the changes of meaning over time can be dramatic. The Cit´e de la Muette near Paris was a public housing estate from the 1930s, but got transformed into a transit camp for jews during the German occupation. Later proposals to restore its housing function were rejected because of the meaning attached to it [Kroes and Primus, forthcoming]. Ceaucescu’s palace in Bucharest provides an interesting case of a meaning which changed in the other direction, as it were. It is bigger than Versailles but not quite as big as the Pentagon, built at the end of a “Rajpath” longer and wider than New Delhi’s, that was ruthlessly sliced through the fabric of the finest quarter of the old city. The destruction and vast building programme were carried out with record speed and efficiency. Surely this was a symbol of everything that is evil. But during the construction some 700 draughtsmen had been employed and countless craftspeople — in fact the construction, by using only Romanian materials, had encouraged a revivifying of craft traditions. And following the deposition of Ceaucescu parts of the huge

1226

Christian Illies and Nicholas Ray

building, named the “people’s palace”, have been hired out for weddings so that it is now regarded by some with pride and even affection.40

3.3 Ethics and ethos Following the reflections by Heidegger on the right way of dwelling, Karsten Harries has argued for a more holistically “ethical” role of architecture. Humans are faced with a severe displacement in the modern world, Harries argues, that might be remedied by a renewed architecture. On a practical level, intimacy and belonging to a place are destroyed by modern technology like television, by which “the faraway and the nearby are equally brought into our living rooms, but only as pictures from which the observer is excluded” [Harries, 1975, p. 14]. But without distance, there is no intimacy, and without intimacy no place where one is at home: “When all places count the same we cannot place ourselves and become displaced persons.” This displacement finds further support on a more theoretical level. Harries argues that the ideal of objectivity behind science and technology “transforms man from an embodied self into a pure thinking object”, and is but another strong reason for the general displacement of humans. It is here that Harries hopes for a renewed architecture that creates an environment that will “give shape to our activities” so that we will be able to live a different life; it should bring individuals back to encounters with others and thus to a community life. This is, according to Harries, the most profound task of architecture: “To build is to help to decide how man is to dwell on the earth or indeed whether he is to dwell on it at all rather than drift aimlessly across it” [Harries, 1975, p. 15]. Harries calls this the Ethics of architecture, because he sees it as closely connected to ethos, that is the characteristic mode of being of individuals and communities. Thus an ethical architecture is not supposed to follow moral rules or principles but to enable a new way of living individually and as a community with others; architecture should serve a common ethos. Although this Ethics of architecture claims to be different from (and not comparable with) more traditional approaches, it seems that this ideal of Ethics can be expressed in terms of the six areas above. Architecture assumes the task of opening our understanding and respect for nature and other humans (and our relationship to them) not merely on a discursive level, but rather by providing certain experiences. This admittedly holistic ideal includes mainly aspects of the fifth area (positive experience of a new possibility) and the sixth area (a better understanding of what it means to live and dwell).

3.4

Towards an ethics of architecture

An ethically satisfying architecture, it seems, would have to answer many different ethical demands, or at least offer a convincing trade-off. Given this complex challenge, it is rather surprising that we do not have any well-developed Ethics 40 For

an extended discussion of this and other examples see [Leach, 1999].

Philosophy of Architecture

1227

of Architecture. Building professionals often appear to have neglected most of these issues, or at least have not reflected upon them. And many architects see their profession as a branch of aesthetics rather than ethics [Fisher, 2000], some even dismissing ethical demands altogether (as apologists for Post-modernism and Deconstructivism sometimes have done). And if there are good-willed ethical gestures, they are often perfunctory (such as the ubiquitous token references to “sustainability”), or are made from a political [Illies, 2005] or simplistically utilitarian perspective [Spector, 2001]. This is not to say that there are not architects who deal with green issues in a serious way — a general awareness is certainly growing. Given its importance, it is surprising that moral philosophy has also paid almost no attention to the built environment. We find practical recommendations, mainly about “green” building, but no comprehensive ethical theory. That is not to say that ethical questions about architecture have not been frequently raised; the negative consequences of modern urban developments which were discussed by Lewis Mumford in the late 1930s, and the critique we have discussed from a phenomenological perspective illustrated by the work of Karsten Harries [1997]. Nevertheless, in 2000 Warwick Fox claimed that any Ethics of the built environment was “still in its infancy” [Fox, 2007]. In an account of “developments in the field of ethics of the built environment”, Fox identified only ten philosophical contributions to such an Ethics [Fox, 2000, p. 3f]. Since then, however, the topic has been given more attention by philosophers as well as architects [Spector 2001; Ray, 2005b]. What are the reasons for this difficulty in developing an Ethics of the built environment? Besides the fact that many architects simply underestimate the problems and think that they can be solved with common sense, there are three principal reasons why it is so difficult to develop an Ethics of architecture: its inherent complexity, the difficulty of finding a clear definition, and an uncertainty about categories, particularly its intersection with aesthetics. 3.4.1

The inherent complexity of architectural problems

As pointed out above, the built environment raises compound issues, because it is a highly complex activity, involving decisions, activities and reflections in very different areas that are governed by frequently heterogeneous requirements. The demand for sustainable architecture might, for example, clash with cultural values or the urgent need for cheaply available shelter. What is so special is architecture’s inclusion of aesthetic issues and the balance required between demands of very different types: any architectural “answer” will have to balance aesthetic, technical, cultural, ecological, legal, political, and economical demands. This is something that can be found in very few other fields — even if they are also of high complexity like Medical Ethics. Moreover, architecture has to provide a solution that works with the highest technical precision. Thus architects make judgements between unquantifiable and

1228

Christian Illies and Nicholas Ray

quantifiable goods, while being fully aware of the (often quantifiable) implications. This creates tensions that can grow into true dilemmas, because we do not have an obvious methodology as to how we should create a balance between them.41 (And it is not even clear whether “balancing” is the right approach.) 3.4.2

The difficulty of finding a clear definition of architectural ethics

Although architecture raises so many problems, it is by no means clear whether there are problems so particular for architecture that they allow us to talk about a specific ethics. Many of architecture’s moral challenges can be explored within the context of other ethics, for example environmental ethics (the second and partly the third of the issues raised above) or professional ethics (the first of those issues). We referred earlier to the problem of building for a TV station that contributes to a morally questionable government; it seems reasonable to apply here the same criteria for architects that we would apply for any businessman who thinks about a deal with a company. If that company is involved in severe violations of human rights, then one should not co-operate with it — unless there is reasonable hope that the engagement will improve the situation of those who suffer in some way. What might be specific about an Ethics of architecture? On the one hand, the clash between cultural sensitivity and economical constraints, between aesthetics and green imperatives and the like cannot be classed merely under environmental or professional ethics. It is simply too complex — and this might be a good reason to ask for a specialised ethics for this area. On the other hand, we could focus on the problems that are specific for architecture. Fox finds it in: what we might call a building’s ‘design fit’, that is, the extent to which a building fits with its natural, social, and built contexts when considered purely in terms of its design rather than in terms of its actual physical impact or even the preferences that people might have in regard to it. [Fox, 2008] This comes close to suggesting that the uniqueness of Ethics applied to architecture lies in architecture giving in some way an aesthetic answer to complex moral problems. To be sure, that can also be asked from art if it is engaged in moral issues — but it will hardly ever reach the complexity of issues to be found in the built environment.42 3.4.3

The uncertainty about categories

Traditionally, ethics was focussed on human interactions, and it developed concepts, norms and values to be applied for this task. It is only recently (most 41 Only if one understands ethics in a wider sense, would all of this be covered by such a methodology; it would need to achieve an overall balance between such different requirements as sustainability, efficiency, positive psychological influences, or sensitivity to the stylistic context, as well as the demands of economy and society. This can itself be seen as an ethical task. 42 Nevertheless, some would claim that architecture, in providing in some way a vision of a better life, can itself furnish ethical ideals — see [Harries, 1997].

Philosophy of Architecture

1229

obviously triggered through ecological and social problems) that ethics has turned systematically to new fields like technical artefacts.43 As a consequence, traditional ethics is conceptually inadequate for buildings, cities, or the like.44 Critics of the usage of “truth” in architecture (as employed by Ruskin in The Seven Lamps of Architecture, and before him Pugin) have pointed to the arbitrariness of the meaning given to “truth”, and of the understanding that it can be achieved. As was pointed out in the introductory section, this also reveals a confusion between aesthetics and ethics, one that continued into the twentieth century, when propagandists for Modernism claimed that the style was the only truthful expression contemporary society and its technology. Decoration was therefore proscribed, and even the application of paint to a material like concrete was thought to devalue the quality of the material itself — ideally it should be left “fair-face” to express its character truthfully. It was for this reason that Reyner Banham’s book on the “New Brutalism” from the 1950’s was sub-titled “Ethic or Aesthetic?” [Banham, 1966]. So how shall we build? Although some values of an Ethics of Architecture seem particularly plausible, a normative theory is still needed that provides welljustified ideals, values or goods for the different areas — and a theory that makes suggestions on how to deal with conflicting demands, both ethical and other, in specific cases. It is not clear yet what this theory might look like. 4

AESTHETICS

We turn now to Aesthetics, a philosophical sub-discipline of prime importance for architecture. In an extraordinarily influential, but passing remark in De architectura, Vitruvius mentions “beauty” (venustas) as one of the three components of architecture, along with usefulness (utilitas) and technical integrity (firmitas) — and for almost 2000 years this triad has been the Leitmotiv of architectural theory and philosophical reflection on architecture [Vitruvius, 2001, Book 1]. Obviously, it has also played a major role for the practical building activity. Ever since antiquity, we find architectural beauty being aspired to — and admired: When he visited Babylon that had been re-erected by King Nebuchadnezzar II (c 630-562 BC), the historian Herodotus remarked that it “surpasses in splendour any city in the known world” [Herodotus, 1987, Book I, paragraph 178]. Yet although aesthetic values seem so important for architecture, there is little agreement about the theoretical or conceptual framework that might guide the building process aesthetically, or help to analyse architecture in aesthetic terms; 43 It is revealing that Technology Assessment is a relatively new term (coined in 1966) as much as a new science. And Environmental Impact Assessment, a formal process used to predict the environmental consequences of a development project, was not introduced as a planning and decision making tool before the late 1960s (in the United States in the National Environmental Policy Act of 1969). 44 Values like “honesty” help at most in the first area; that is why professional ethics is the best developed sub-discipline of the built environment. We should also recall that the notion of the polis played a crucial role in Plato’s ethics (see footnote 1 above).

1230

Christian Illies and Nicholas Ray

the debate mirrors controversies that we find in all areas of philosophy. Moreover, architecture seems more difficult to grasp adequately than other artefacts. Building is a complex process in which many different people and skills are involved; the building’s shape and form are determined by the architect(s), the craftsmen, the patron, and other stakeholders, but also by its place (the nature of its site) and urban context (neighbouring buildings); and this shape might change over time. What makes it even more complicated is that a building (unlike a painting or piece of music) is fulfilling a certain function and having some utility. As Roger Scruton remarks: “Hence, although we may have reason to think that we sometimes treat buildings as aesthetic objects, it does not follow that in appreciating them as buildings we are appreciating them aesthetically” [Scruton, 1979]. In what follows, we will approach the aesthetic aspect of architecture by looking at different alternatives and debates that can be found in the tradition. The first will be in relation to theory; there is a fundamental difference in whether we take aesthetics to be about some objective values that can be discovered in a rational process or whether it is about subjective responses to certain sensual experiences. After that, we turn to the search for key concepts or categories for the aesthetic appreciation of architecture: here the 18th century debate on the sublime, the beautiful and the picturesque will serve as an example, but we will also talk about empathy. Thirdly, we turn briefly to the central problem of aesthetics and utility that we have just mentioned: is architecture capable of being judged aesthetically at all?

4.1

The problem of theory

The empiricist philosophy of art and beauty is most evident with British philosophers like Shaftesbury, Hutcheson, Hume and Burke. Empiricist aesthetics sees aesthetic judgements fundamentally as expressions of subjective states of mind or feelings that are shaped and conditioned by experiences, culture, or other contingent causes. It denies that beauty is a quality inherent in some object. Certainly a subjective response is still seen as a reaction to certain formal features of the object, but that response is something over and above a mere perception of its properties. Thus, for subjectivism, the idea of universal rules or standards is pointless. The popular version of aesthetic subjectivism is the wide-spread view that de gustibus non est disputandum — we cannot, and indeed should not argue about matters of taste: they are personal. But we are all aware that the process of development control, whereby planning committees are advised by panels of experts and come to conclusions as to what buildings are to be approved and what are to be prevented, involves aesthetic judgements about the buildings under consideration. And there is naturally plenty of dispute. In contrast to the empirical approach, there is a (Continental) rationalist or cognitive aesthetics. Alexander Gottlieb Baumgarten (1714 - 1762), one of the founding fathers of Aesthetics as a discipline, defined aesthetics as the “science” of sensory cognition [Baumgarten, 1750, § 1]. Although he subscribed to a different

Philosophy of Architecture

1231

understanding of aesthetics (in which he included all sense of perception) than our own would be, he initiated a pursuit for a rationalist aesthetics that would aim at universal insights and general rules. For Baumgarten and the continental tradition, in particular Immanuel Kant, “beauty” is not to be understood in terms of a physiological process or merely by means of a subjective response but has more to do with rational ideas, and objective standards. But what might these standards be? We have seen that within architecture, Leon Battista Alberti presupposed a cognitive aesthetic when he claimed that beauty was mainly a matter of the right proportion — that is the relationship of parts to a whole and to one another. “In all perfectly beautiful objects there is found the opposition of one part to another and a reciprocal balance”, as John Ruskin puts it more poetically [Ruskin, 1878-79]. And since Greek antiquity, the Golden Ratio,45 especially in form of the Golden Rectangle, has been considered as an objective standard providing the right balance for the formal composition of Western architecture.46 This ratio was seen as derived from the (idealised) human body, as the mathematician Pythagoras (560-480 BC) had already tried to show.47 The main argument in favour of aesthetic subjectivism are the different conceptions of beauty, both within a culture and between different cultures. In answer, cognitivists point to the surprising agreement about what are architectural masterpieces in very different cultures — the Taj Mahal or Chartres Cathedral are almost universally praised.48 Notably, even the empiricist David Hume wrote that taste is “far fewer liable to the revolutions of chance and fashion than these pretended decisions of science” [Hume, 1757, paragraph 26]. Further, cognitivists reason that we can, after all, argue successfully about the qualities of artworks, for example why Elgar’s Enigma Variations or Mies van der Rohe’s Pavilion in Barcelona are — or should be — admired by musicians or architects. Such rational discussion seems only comprehensible if we presuppose that there are some (minimal) standards as common points of reference. There are also attempts to find a middle ground between a strict cognitivist aesthetics and an understanding of aesthetics as being about subjective responses, for example pragmatist approaches. In his architectural aesthetics, Julian Roberts uses the analogy of British case law (rather than Roman Law), and argues that aesthetic judgment, though it should be informed, would more properly be like 45 I.e., a proportion of a+b = a = ϕ. For a discussion of proportion in architecture, see a b [Scholfield, 1958]. 46 The Golden Ratio also plays an important role in non-Western architecture, for example the Mosque from Kairouan (Tunisia) from 670 AD (see [Boussora and Mazouz, 2004]). 47 Other rules or standards have been suggested by artists and philosophers such as William Hogarth in his The Analysis of Beauty (1753) or Joshua Reynolds in his Seventh Discourse on Art (1797). And Dalibor Vesely [2004] has argued that in the Baroque period ideas of proportion were derived from a world view, rather than being merely an instrumentalised process. 48 Almost — Peter Eisenman, for example, does not think much of Chartres: “I think it is a boring building. . . . In fact, I have gone to Chartres a number of times to eat in the restaurant across the street — had a 1934 red Mersault wine, which was exquisite — I never went into the cathedral.” (See the debate with Christopher Alexander, already referred to: http://www.katarxis3.com/Alexander Eisenman Debate.htm.)

1232

Christian Illies and Nicholas Ray

a “trial by jury” than an oracular judgment issued by persons with a special training [Roberts, 2005]. The historian Peter Collins, tends to agree, distinguishing between principles and laws and quoting a famous case from 1761 (Hamilton versus Meades) that negotiations should depend on rules that are the result of “the dictates of common sense drawn from the truth of the case” [Collins, 1971]. Although pragmatic approaches have been of major importance in architectural practice (many aesthetic decisions in urban planning are made by commissions of lay-people), it might be objected that they face the conceptual difficulties of many other “bottom-up” approaches.49

4.2 Key concepts for the aesthetic appreciation of architecture What are the basic categories for an aesthetics of architecture? An important examination of how people might be affected by architecture (and by natural landscape) was undertaken in eighteenth century England, just at the time that aesthetics was born as a philosophical discipline. One of the most influential theorists was Edmund Burke, the author of A Philosophical Enquiry into the Origin of our Idea of the Sublime and the Beautiful, written in 1757-59. As his title suggests, Burke aimed to distinguish between two qualities that we might appreciate aesthetically. In the Enquiry, this is how Burke defines the sublime: whatever is fitted in any sort to excite the ideas of pain and danger, that is to say whatever is any sort terrible, or is conversant about terrible objects, or operates in a manner analogous to terror, is in a sense sublime; that is it is productive of the strongest emotion which the mind is capable of feeling. I say the strongest emotion, because I am satisfied the ideas of pain are much more powerful than those which enter on the part of pleasure. [Burke, 1759, Part I, Section VII, p. 36] In moving on to describe beauty, Burke turns to the question of sexual reproduction to illustrate how, in the act of love, in humans, a brute animal passion is replaced by “a mixed passion which we call love”, and the object of this mixed passion is beauty [Burke, 1759, Part I, Section X, p. 39]. We must expect to find combinations of the qualities of the sublime and the beautiful in every work of art, but Burke avoids the naming of any third intermediate term — he is content to describe differences [Burke, 1759, part III, Section XXVII, pp. 113-114]. The aesthetic distinctions which Burke draws between the sublime (huge, rough, awesome, natural, overwhelming) and the beautiful (petite, smooth, highly fashioned and seductive) can be read psychoanalytically, economically and politically, and, since Part V is concerned with words, the Enquiry has also proved fertile ground for 49 In brief the difficulties are as follows: how does one judge the differing opinions? Is the judgement of any lay person accepted — or do they have to be specially trained or knowledgeable or sensitive? It seems the “bottom-up approach” simply postpones the problem of whether a cognitivist or subjectivist account is more suitable to make an adequate judgement in the “trial”.

Philosophy of Architecture

1233

twentieth century scholars concerned with rhetoric and the workings of language itself.50 The appreciation of some contemporary architectural work (such as Rem Koolhaas’ “Large” and “Extra-Large” buildings, discussed below) would fall into the category of the sublime: we would be moved and awed by their scale, and maybe enjoy the pain of their brutal acceptance of the given conditions. The enjoyment of beauty becomes a private matter, confined to the erotics of the bedroom, since the condition of Modernism is such that an un-mediated, un-ironic public enjoyment of beauty is no longer possible. Later in the eighteenth century, writers sought the middle term that Burke avoided, which would combine elements of the sublime and the beautiful. William Gilpin’s Three Essays: On Picturesque Beauty; on Picturesque travel; and On Sketching Landscape, published in 1792, and Uvedale Price’s ”Essays on the Picturesque as compared with the Sublime and the Beautiful” of 1810 both advocate the concept of the Picturesque. The Picturesque shares with the sublime something of its roughness. And it “may be great or small” writes Gilpin, “but since it so depends on the character of boundaries, can never be infinite”. In other words the Picturesque is always framed. The symmetry and perfection of the Beautiful must be defaced in order for it to become picturesque. Gilpin recommends taking a mallet to Tintern Abbey: “we must beat down one half of it, deface the other, and throw the mutilated members around in heaps” [Gilpin, 1792; cited in Punter, 1994, p. 235].51 The picturesque does not, therefore, participate in the dangerously erotic character of the beautiful: the figure in a picturesque landscape is more likely to be a decrepit hag than a young woman. “In real life, I fancy”, wrote Price, “the most picturesque old woman, however her admirer may ogle her on that account, is perfectly safe from his caresses” [Price, 1842]. The picturesque is in fact a melancholic view of beauty — the pleasure of ruins lying both in the satisfying effect to the eye of their composition and the sense that they communicate of their being fragments of an ideal world that we can no longer achieve. There are several examples of nineteenth-century architects who anticipated the decay of their buildings in the way in which they represented them. One such would be John Soane, whose proposals for the Bank of England were illustrated in ruins before they were even constructed. He was a friend of the artist Henri Fuseli, who shared his deep pessimism about the state of the arts, which he regarded as debased in comparison to the vision of antiquity held by Winckelmann and Flaxman. Fuseli’s position, and thus we may infer Soane’s, is summed up in his famous etching of 1778-9 The Artist moved by the grandeur of ancient ruins, a poignant evocation of his melancholic attitude to the past. And this attitude survives into the twentieth century. Alvar Aalto, whose Baker House is considered below, avoided symmetry in his buildings, and frequently seems to suggest fragmentary ruins in the vicinity of his own work (notably in his summer 50 For a close reading of both the Enquiry and Burke’s later book the Reflections on the Revolution in France, see [Furniss, 1993]. 51 An excellent general introduction to the picturesque is [Macarthur, 2007].

1234

Christian Illies and Nicholas Ray

house at Muuratsalo). He also evokes an ancient precedent by representing an outdoor amphitheatre in the landscape at the Jutland art gallery, and at his own office at Tiilim¨ aki, which is eroded at its edges as if it has been there for many centuries. How we view architecture need not be mediated by eighteenth century theory, however, and might be understood at a more profound, psychological, level. Firstly, our appreciation of architecture is not merely visual. The historian Heinrich W¨ olfflin has already been mentioned. His 1886 Prolegomena to a Psychology of Architecture refers to Goethe’s remark that “we ought to sense the effect of a beautiful room, even if we were led through it blindfolded” in his argument that “the architectural impression, far from being some kind of “reckoning by the eye”, is essentially based on direct bodily feeling.”52 Architecture is not just a visual art, but something we “feel” with all our senses. Arguments about style often use linguistic analogies — architecture is a language which is “spoken” grammatically or practised illiterately; the syntactically uneducated “vernacular” is compared to the sophisticated grammatical work of educated architects — but as an experience architecture is actually more primitive and basic. Perhaps the first “architectural” experience is that of birth — a passage from dark to light, from enclosed and protected to open and unprotected, from within the comfort of the womb to the exposure to the world. The remark by the literary critic Cyril Connolly that architecture was a “womb with a view” was not only a witty reference to a novel by E.M. Forster, but refers to the distinction of experience that we all may have as children when we construct our first shelters. We make a “tree house” up in the branches, and hollow out a cave between its roots. Of the great psychoanalysts in the twentieth century, Freud was comparatively uninterested in architecture, though his pupil Melanie Klein affected art historical theory through her erstwhile patient Adrian Stokes. Stokes’ descriptions of the distinction between “carved” and “modelled” forms, relate psychoanalytical mental states to the way in which sculpture and architecture is fashioned. The complex (some would say muddled) but influential thinking of Carl Jung is a more common underpinning to the way in which people often express their feelings about buildings. Jung conceived a theory of archetypes — typical images and their associations that transcend cultural boundaries — and some have tried to relate what we find pleasing in buildings (or in city environments) to those archetypes.53 There is an element of Platonic idealism in Jung’s notion, since it would imply that satisfactory architectural images referred back to some archetypal form which is inherently satisfying on an instinctual level. If aesthetic appreciation goes beyond the visual, perhaps it is more empathetic. At its most basic, an empathetic theory of architectural understanding implies that, on the analogy of an empathy with other people, in perceiving a building 52 For Goethe’s important influence on Nineteenth Century aesthetic theory see [Mallgrave and Ikonomou, 1994]. 53 An idiosyncratic example is [Edwards, 2003], which sets out to examine “a chance remark by a friend that aesthetics, traditionally an aspect of philosophy, is properly an aspect of psychology”.

Philosophy of Architecture

1235

we both imagine ourselves inhabiting the spaces, and in some way put ourselves in the place of the elements of the construction. In everyday practice, architects might be expected to exercise empathy — to imagine themselves in the situation for which they are designing (Aalto claimed that his Paimio sanitorium design was influenced by his own experience in hospital when he was forced to gaze at the ceiling for a long period of time), but the real issue is how someone who experiences the building can be expected to exercise empathy. One of the characteristics of the classical orders was that they were characterised by gender: Doric was the sturdiest and most masculine, Ionic was feminine, as was Corinthian — but a more attenuated and graceful representation of the gender. Sometimes the Doric order could be substituted by statues — these are Atlantids. And taking the place of the Corinthian order, most famously in the Erechtheum in Athens, would be Caryatids. In aesthetic thinking the idea is associated with Theodor Lipps and Robert Vischer, who coined the term Einf¨ uhlung (“in-feeling” or “feeling-into”) in his 1873 doctoral thesis.54 In perceiving a building we therefore expect the strongest part of the composition to be where the heaviest loads are taken, nearest the ground. The use of “rustication” in classical architecture responds to that expectation, just as pilasters attached to a masonry wall dramatise the pattern of load as well as “ordering” the fa¸cade dimensionally. Modern Architecture, which rejoices in dramatic cantilevers and tends to have slender columns (or piloti ) at ground level arguably achieves much of its aesthetic effect precisely because it subverts those expectations.

4.3

Aesthetics and utility

As mentioned above, it has sometimes been argued that architecture should be seen as primarily functional — utility (utilitas) being, besides firmness (firmitas), another of the three basic qualities that Vitruvius introduces. Firmness refers to it being solid and lasting, while utility means the function of the building (e.g. a temple) and its general usefulness of providing protection for its users etc. For Vitruvius, and classic philosophy in general, this triad would be seen as a unity; the good and the beautiful seemed necessarily linked or at least their harmony the ideal to strive for. (Thus the ideal man was seen as having moral and aesthetic qualities, he was handsome and brave — Kalos kai agathos, as the tradition since Herodotus called it). This ideal of harmony still guides Alberti in his already quoted definition of beauty: “that reasoned harmony of all the parts within a body, so that nothing may be added, taken away, or altered, but for the worse” [Alberti, 1998]. But in the influential teaching of the Bauhaus, at least at some stages, function was accorded the primary role. The director during the period between 1928 and 1930, Hannes Meyer (1889-1954), wrote that: 54 It was titled On the Optical Sense of Form: A Contribution to Aesthetics. See [Mallgrave and Ikonomou, 1994].

1236

Christian Illies and Nicholas Ray

Building is a technical, not an aesthetic process, and time and again the artistic composition of a house has contradicted its practical function. Planned in ideal and basic terms, our house will become a piece of machinery.55 Unlike Le Corbusier, therefore, who, as we have seen, while claiming that a house was a “machine for living in” was clear that architects were artists who manipulated the functional ingredients of mere building to create architecture, Meyer sees no place for art at all. In his 1928 essay, Bauen, he claimed “all art is composition and thus anti-functional”.56 However, others have always been more doubtful about the unity of good, that is firm and useful, and beautiful architecture. Each area took on a more autonomous status — and thus reflection began about how to bring the two together. That the beautiful is distinguished from the good by “not being useful” was stated in the enlightenment by Karl Philipp Moritz [Moritz, 1788, p. 11]. He derived this idea from the Abb´e Charles Batteux who introduced an influential distinction between the fine and the useful arts (giving architecture a special position between the two) [1746]. And Kant famously stressed the difference between the good and the useful when he demanded that aesthetic judgements should be without selfinterest, abstracted from utilitarian concerns [Kant, 1790]. This line of reflection led to the aesthetic ideal of a l’art pour l’art: artworks should not be tainted by any functionality. “All art is quite useless“ said Oscar Wilde [Wilde, 1990, p. 17]. Applied to architecture, this would mean that only the decorative elements of a building would contribute to its aesthetic character. The shape and form of a building may be determined by their function, but ornaments give the freedom of artistic expression. In an architectural textbook from 1865 we read that architecture is “nothing more or less than the art of ornamented and ornamental construction“ (cited in [Gauldie, 1969, p. 3]). Yet, as we have seen whether there should be any ornament in addition to the structure became a much debated issue in the twentieth century, when the “truthfulness” of decoration was challenged. In a certain way, this functionalist ideal recaptures the classic kalos kai agathos by bringing beauty and goodness together again. However, the lead is taken by the good (for which read function) because simply by being functional things are seen as beautiful — this is not quite the classic type of unity. There are two fundamental reasons why this attempt in the twentieth century to reconcile utility and beauty are unconvincing. On the one side, the functional demands will always be underdetermined. Any building has a multitude of functions so that any design decision involves choosing which function to privilege. Even the mechanistic definition of a house as “a machine for living in” implies the “housing” of many different activities. We require shelter, entrances and exists, sanitary installations and noise protection, but also fresh air, and the right space for the things we want to do in a dwelling. But beyond this, there are many more 55 “Die

Neue Welt”, in Das Werk 13, 1926, in [Meyer, 1980], cited in [Kruft, 1994, p. 386]. Bauhaus 2, 1928, in [Meyer, 1980], cited in [Kruft, 1994, p. 386]. For the Bauhaus and its teaching see [Whitford, 1984]. 56 “Bauen”,in

Philosophy of Architecture

1237

functions that we need to satisfy: the ethical questions referred to earlier such as being “green”, or improving (or at least not impeding) people’s psychological well-being by paying, for example, due regard to their life-style and traditions. Any architect will have to make many functional trade-offs — and the resulting buildings will look very different, even if they have the same function. Heidegger was surely right in bringing “living” and “thinking” together: any building design is connected with a certain attitude to how we are to live. Secondly, even if function could be precisely determined, the form that buildings that meet that function is by no means determined, which is obvious when we consider the wide variety of styles that have been justified on the grounds of their being functional. These range from gothic cathedrals (according to Pugin and Ruskin) to the City for Three Million People that Le Corbusier advocated. Forms cannot be deduced in an algorithmic process. We have different materials (stone, concrete, metal, wood), we have different forms (round, square columns, walls) and even styles that could all fulfil the same function in a good way. What we regard as an adequate form to house a function is as much determined by our expectations and traditions. Arthur Schopenhauer (in empathetic vein) remarks that columns must be designed bigger than it is necessary for the static function because only then are we “absolutely reassured” that they will not break [Schopenhauer, 1859, Vol. 2, p. 390]. It might come as no surprise that the debate about form and function has become less vivid in the last years. It has proved impossible to reach an uncontroversial agreement on the connection between beauty and utility. Disagreement remains on all levels: what is the appropriate theory on which to base a view on? Is it ‘objectively’ right to try to make the form follow a function? If so, which function? And what of beauty? Can there be a beauty merely arising out of meeting functional needs? And, how are we to make design decisions in face of a plurality of heterogenous demands and values (moral values, economical values, functional demands)? Architectural aesthetics seems to inherit and unite the philosophical problems of many, if not all normative theories. But if architecture mirrors many of the debates and tensions of philosophy, could it in itself provide some kind of mediation between apparently irreconcilable positions? In the past it has been thought of as an embodiment of philosophical insights. “Humanity has lost its dignity; but Art has rescued it and preserved it in significant stone”, Friedrich Schiller famously wrote [Schiller, 1795, 9th letter]. Such mediation would be most evident in the scale and style of a building in an urban setting: a building or set of buildings might mediate between a set of aesthetic ideas and create a kind of synthesis. In the nineteenth century, a time, as we have seen, of stylistic eclecticism, certain architects, like Karl Friedrich Schinkel, thought very hard about the meanings that could be attached to particular styles.57 His Neues Museum in Berlin, for example, combines a classic Greek 57 See [Carter, 1981]: “Schinkel lived during a period of transition, a period when the conventions of the Baroque could no longer be accepted and a variety of new tasks arising from the social and industrial revolutions demanded new solutions. The self-conscious attitude vis-` a-vis

1238

Christian Illies and Nicholas Ray

fa¸cade with elements of the bourgeois architecture of nineteenth century Berlin, and thus speaks about the possibility of harmonizing a modern culture with classic ideals. In more recent practice, buildings which are clearly contemporary in expression and are placed in an historic context in such a way as to enhance it, and allow us to re-read it differently, may be appreciated similarly. Norman Foster’s gallery at Nˆımes, opposite the Roman Maison Carr´e, has been interpreted in this way, as has the Piano and Rogers design of the Centre Pompidou in Paris. In his influential Collage City, the critic Colin Rowe suggested that the quality of a great city like Rome could be explained precisely because it represented a “collision of utopias”, rather than a fully executed single vision — implying that this mediating ability of architecture was one of its most important cultural contributions, and that architects might proceed by a process of bricolage [Rowe and Koetter, 1978]. Naturally we can find such an ability to mediate or synthesize in many other areas of artistic activity. But architecture serves so many needs and functions at once that its answers appear to be more developed and richer than those of other arts. In particular, architecture must always provide a (practical) answer of how to integrate functional and other claims. As in other areas, whether the philosophical difficulties it throws up are seen as essentially problematic, or a possible contribution to reconciling profound difficulties in our culture and thinking, will depend on the position different people adopt.

5

PHILOSOPHICAL POSITIONS ILLUSTRATED IN ARCHITECTURAL PRACTICE

Three architects have been chosen to illustrate a variety of philosophical positions. Two practised in the twentieth century; the third continues to practice. Recent work has been selected for discussion, not because designs from earlier centuries do not illustrate a theory with equal clarity, but because the work of more contemporary designers is more likely to be relevant to the developing practice in the twenty-first century. It would be difficult to find practitioners whose work could be assigned exclusively to a particular position — even the most idealistic architects have to accommodate themselves to the circumstances of patronage in order to succeed in building anything. The position of many of them is difficult to unravel. The case of Le Corbusier has been mentioned already: his statements seem to exhibit a mechanistic positivism, but at the same time he stresses the primacy of art, ¨ and regarded himself as something of a Nietzschean Ubermensch. We must also recall that these architects are not philosophers, although they may have read quite the past promoted by archaeological investigation and historical speculation encouraged the notion of a new style appropriate to a new age, but the complexity of the new situation and the need for immediate action made a complete return to first principles impractical. While one waited for the new style to emerge, an eclectic approach could offer a temporary resolution. . . .If similar forms could express such a variety of ideas, was it also possible that a variety of forms could express similar ideas? That meaning was not intrinsic in the forms but rather attached to them by tacit agreement and confirmed by the specific context?”

Philosophy of Architecture

1239

widely. The positions that they may claim that their buildings reflect could prove to be quite mistaken. But architects’ misinterpretations of history have sometimes proved fruitful. Palladio’s reconstructions of Roman houses would be an example: he believed, quite wrongly, that the houses of the rich were dignified by pediments. His own architecture, and the books he published illustrating it, had widespread influence and laid the foundations for “classical” domestic architecture throughout the western world. In a similar way, how architects interpret and make use of philosophical ideas in their practice, while it may not be technically correct, can have surprising repercussions.

5.1

Louis Kahn

The American architect Louis Kahn (1901-1974) represents one of the clearest examples of the position of the architect as platonic idealist. Although, as we swiftly discover with all the most interesting people, his background was complex and his cultural influences enormously wide, his idealist position is clear in what he built, in the procedures he adopted during design, in how he described his design process to students, and in what he wrote more generally about architecture [Benedikt, 1992; McCarter, 2005]. The sequence of five designs undertaken by Kahn’s office between 1959 and 1967 for the Unitarian Church in Rochester, New Jersey, serves as an example of his design process. Kahn distinguished between what he called “form” and “design”: “Form doesn’t have shape or dimension. It simply has a kind of existence will”,58 he explained.: Form is impersonal. Design belongs to the designer. Design is a circumstantial act, how much money there is available, the site, the client, the extent of knowledge. Form has nothing to do with circumstantial conditions. In architecture, it characterises a harmony of spaces good for a certain activity of man. [Kahn, 1991, p. 113] Kahn spoke frequently of “the institutions of man”. It was the primary task of the architect to represent those institutions, whether museum, library, gallery, or school. His starting point, almost invariably, was with a simplified diagram which placed the primary element centrally and clustered subsidiary spaces around it. Kahn thus made a distinction between “served” and “servant” spaces. Geometry was used to reinforce the difference: the central space would be a clearly perceptible figure — a square or a circle, while the subsidiary spaces would adopt a more “circumstantial” geometry. Kahn therefore subscribes to Alberti’s suggestion that a hierarchy of forms can help to symbolise the importance of different spaces. Kahn’s earliest charcoal sketches assume such simplified geometries, but are far from precise in themselves — he is trying to get to the essence of the idea, and often it takes the work of many months, and the patient labour of his architectural 58 Panel

discussion, 1960, reprinted in [Kahn, 1961], and [Kahn, 1991, pp. 112-120].

1240

Christian Illies and Nicholas Ray

assistants to discover, as it were, the final pattern in which the idea will realise itself. The drawings are in fact a process of meditation: One may say that architecture is the thoughtful making of spaces. It is not the filling of areas prescribed by the client. It is the creating of spaces that evoke a feeling of appropriate use. [Kahn, 1991, p. 116] In the case of Rochester Unitarian Church, the idea remains consistent through numerous iterations: the body of the church will be central, and around it cluster the several rooms that serve it — vestries, offices, porches and lobbies, cloakrooms, and boiler house, as well as the classrooms, for Kahn saw these as “serving” the central sanctuary. His clients, however, appeared to have favoured an arrangement similar to Frank Lloyd Wright’s Unity Temple in Chicago, of 1906, which both in its brief and in its geometry presented a compelling precedent for the scheme. It is Kahn’s persistence in holding on to his original concept, the “form”, which gives the project much of its power. Some versions have a more dramatic expression of the central structure, especially the first with its truncated dome, reflecting Kahn’s interest in the work of the engineer Buckminster Fuller. The body of the church is sometimes square, sometimes rectangular. Each version has an arcade around the church as an intermediate space, of greater or lesser elaboration; the first version, which is the most “ideal” geometrically, has an ambulatory as well as an arcade. Over an extended period of design and redesign, eventually a final version emerges, which reconciles the prescriptions of the client, and budgetary constraints, with the nature of the institution that it is the task of the architect to define: “The final design does not correspond to the first design though the form held” [Kahn, 1991, p. 116]. The process of design in which Kahn engaged acted as a powerful critique of the somewhat positivistic myth of functionalism evident in the proclamations of CIAM and propounded by Walter Gropius and his followers. For them, building forms should arise spontaneously as a result of solving functional problems by means of modern technology; the resulting plan tended to be asymmetrical, and the structure lightweight.59 Kahn’s emphasis on the continuity of a conceptual idea harks back to the procedures of nineteenth century Beaux Arts architects, who, in the celebrated competitions for the Prix de Rome, first came up with an esquisse, a small sketch, which represented the primary design intentions, next proceeded to a poch´e, a pocket-sized set of drawings within which the design principles were embedded, and only in the third stage developed the detailed and beautiful drawings of the final projet which was submitted to the jury. The architectural language of the Beaux Arts was well-established — it was the fully developed apparatus of 59 Robert Venturi criticised Walter Gropius for perverting the Vitruvian trinity by suggesting that utilitas + firmitas = venustas, rather than seeing these three terms as equal participants in the definition of architecture. The teaching at the Bauhaus, where Gropius was Director from 1919 — 1927 was not as straightforwardly “functionalist” as has been often assumed (see [Rykwert, 1982]), but later, at the Harvard Graduate School of Design, where Gropius moved in 1937, the teaching programmes emphasised functional considerations as the only generators of architectural form (see [Herdeg, 1983]).

Philosophy of Architecture

1241

Figure 1. Six versions of the Rochester New Jersey church, by Louis Kahn. The church as realised is shown bottom right. It was later enlarged by Kahn to the west by the addition of a school hall and further classrooms. (Louis I. Kahn Collection, University of Pennsylvania Historical and Museum Commission.)

1242

Christian Illies and Nicholas Ray

classicism — and the technology was predominantly traditional, though innovative architects such as Labrouste had successfully introduced cast iron. Kahn, despite his preference for weighty materials (concrete rather than steel, for instance) did not eschew twentieth century technique, but saw it as part of the task of the architect to develop a language for building which would express the nature of the institution in the most direct way. Kahn therefore organised building services carefully, to ensure their subservience, and spoke, poetically, of “asking” the materials he employed how they wanted to be used: If you talk to a brick and ask it what it likes it’ll say it likes an arch. And you say to it, look arches are expensive and you can always use a concrete lintel to take the place of an arch. And the brick says, I know it’s expensive and I’m afraid it cannot be built these days, but if you ask me what I like it’s still an arch.60 Just as in his conceptual design, so in the development of the project towards realisation, the universal nature of a world of forms, or materials, is stressed, over the particularities of the situation. The material has its own desire to return to its origins, as it were, to behave in the way it wants to behave.

5.2 Rem Koolhaas Commentators on the work of the Dutch architect Rem Koolhaas are united in seeing his early career as a film-maker as a key to his architectural position.61 He had spent his childhood in the Dutch East Indies and intended a career in journalism and film-making, before studying architecture in London at the Architectural Association School. In Ren´e Daalder’s satirical film The White Slave (1969), on the script for which Koolhaas worked, a woman says “listening to this music your brother and I dreamt of a better world”. And the answer is: “A better world. . . it has not come to much”. Like Koolhaas’ architecture, the film is sophisticated, ironic, formally skilful and founded on a bleakly realist philosophy. In 1995, Koolhaas’s work to date, through his practice, the Office of Metropolitan Architecture (OMA) was recorded in a volume entitled S, M, L, XL, and the projects and essays described therein are used below as the key to understanding Koolhaas’s position [Koolhaas, 1995]. S, M, L, XL stands for “small, medium, large, extra-large” — the descriptions used by supermarkets for their undergarments; indeed pictures of men’s underpants crop up in the pages of S, M, L, XL, which is typographically and visually witty and inventive. One of the theses of the book is that sheer scale (rather than architectural meanings, or the formal ordering devices traditionally used in architectural design) determines the conditions for architecture today. We will treat selected projects, and the issues they raise in the order in which they are published. 60 Interview 61 See,

in House and Garden, October 1972, reprinted in [Kahn, 1991, p. 196]. for example, the essay on Koolhaas in [Moneo, 2004].

Philosophy of Architecture

1243

1244

Christian Illies and Nicholas Ray

Figure 2. Three projects by Rem Koolhaas (this and the previous page): section, showing ship railing and marine metaphors, and aerial view of the Villa Dall’Ava, perspective of the Morgan Bank proposal and diagrams of the Congrexpo at Lille. (Illustrations from the Office of Metropolitan Architecture.)

Philosophy of Architecture

1245

Villa Dall’Ava is a house in a suburb of Paris, completed in 1991. The reader is introduced to the brief for the house by means of an apparent film-script. Koolhaas has received a letter from the client requesting a meeting: . . . He would pick me up at Charles de Gaulle Airport. When I came out, there was an enormous scandal: someone was trying to kill a policeman. It turned out to be him. The policeman had asked him to move, but since he was waiting for his architect he had tried to run over the policeman. [Koolhaas, 1995, p. 133] Like many of Koolhaas’s buildings, Villa Dall’Ava is a commentary on the heroic optimism of the twentieth century, but seen from a melancholic perspective. The house is close to two of Le Corbusier’s famous villas. Its ribbon-windowed elevation clearly refers to the fenˆetre en longeur which constitutes one of Le Corbusier’s famous “five points” of the new architecture. But the walls, instead of being formed of a machine-like white render, are made of rust-coloured corrugated metal, and the columns, or piloti, are angled, not vertical. The house celebrates the Corbusian “essential joys” and has a swimming pool on the roof, but the white protective pipe railings originally proposed have been removed in favour of orange “temporary” fencing. Le Corbusier had promoted the analogy between steamships and buildings: ships with their decks and promenades, repetitive cabin structures and general air of purposiveness represented the paradigm of an architecture which would speak its function. In S, M, L, XL Koolhaas reproduces draft drawings of Villa Dall’Ava on which he has scribbled in red biro: “ I hate the ship metaphor. Railings are very hard to do without resurrecting the ocean liner from the 20’s” [Koolhaas, 1995, p. 180]. The completed house is illustrated by photographs presented as if they were stills from a movie, many taken at night. People appear as shadows or reflections; there is a giraffe in the garden. A 1983 essay reproduced in M, the second section of S, M, L, XL, is entitled “Typical Plan” [Koolhaas, 1995, p. 135-350]. It celebrates “zero-degree architecture, architecture stripped of all traces of uniqueness and specificity”. The product of the “new world”, “Typical Plan” answers the programmatic needs of business — neutral, artificial, repetitive, an example of utilitarianism “refined as a sensuous science of co-ordination” so that the architecture “transcends the practical to emerge in a rarified existential domain of pure objectivity”. The Typical Plan “is to the office population what graph paper is to the mathematical curve”, but of course it has attracted criticism, especially in Europe: “Suddenly the graph blamed the graph paper for its lack of character. . . . Nietzsche lost out to Sociology 101”. The sociology class are shown the alienating effect of modern office life, compared to the “cottage industries” of old, for which the grids of the office plan provide graphic evidence. Maybe architects should soften and humanise the environment in some way? Koolhaas has no time for such sentimentality: the typical office plan represents par excellence the conditions of late capitalism and there is nothing that architects can do about that — attempting to conceal the fact by

1246

Christian Illies and Nicholas Ray

“humane” design is both futile and fundamentally dishonest. Koolhaas illustrates a version of “Typical Plan” with his competition entry for the Morgan Bank in Amsterdam which ruthlessly imports the benefits as well as the disadvantages of the model: “. . . abstract office space, its dimensions chosen to enable a maximum number of permutations, introducing, in Holland, unusual (and ultimately unwelcome) depth.” The Congrexpo (Grand Palais) at Lille, is one of Koolhaas’s “Large” buildings, completed in 1994, and sitting within an “Extra-large” masterplan for Euralille, for which OMA had been selected as planners in 1989. The scale of the total proposals, as a new city (“the centre of gravity for a virtual community of 50 million Western Europeans”) on the periphery of the old city, means that any attempt to relate to the former urban context is doomed. The real context in fact is travel at the scale of the continent, not the city on the perimeter of which Eurolille happens to sit. Koolhaas acknowledges that his proposal to build over the TGV tracks is an investment in symbolism — at an additional cost of between 8 and 10%, but judged by his clients to be worth it [Koolhaas, 1995, p. 1070]. The Congrexpo itself has three components, a 5,000-seat concert hall, a conference centre with three auditoria, and a 20,000 square metre exposition hall. His assembly of the parts is “scandalously simple”: they are jammed together on an enormous sloping plane of concrete and under a single unifying roof. The client, Jean-Paul Baietto, director of Euralille, particularly appreciated the skills and approach of Koolhaas and his team. In such a context, the audacity to provide complexity of programme within extreme simplicity (or arbitrariness) of form is precisely what is required: Koolhaas has established “a dynamique d’enfer, a dynamic from hell” [Koolhaas, 1995, p. 1208]. Koolhaas goes some way towards explaining his attitude in the following passage from S, M, L, XL. All the certainties of architecture no longer prevail, in the conditions of the late twentieth and early twenty-first century city. The previous generation were “making sandcastles. Now we swim in the sea that swept them away.” To survive, he writes: . . . urbanism will have to imagine a new newness. Liberated from its atavistic duties, urbanism redefined as a way of operating on the inevitable will attack architecture, invade its trenches, drive it from its bastions, undermine its certainties, explode its limits, ridicule its preoccupations with matter and substance, destroy its traditions, smoke out its practitioners. The seeming failure of the urban offers an exceptional opportunity, a pretext for Nietzschean frivolity. We have to imagine 1,001 other concepts of city; we have to take insane risks; we have to dare to be utterly uncritical; we have to swallow deeply and bestow forgiveness left and right. The certainty of failure has to be our laughing gas/oxygen; modernization our most potent drug. Since we are not responsible, we have to become irresponsible. [Koolhaas, 1995]

Philosophy of Architecture

1247

Koolhaas is not alone in the general bleakness (or realism) of his vision; his work has been influenced by the Dutch writer and artist Armando (Herman Dirk van Dodeweerd), and the author Willem Frederik Hermans (1921-1995). Armando writes Do not take morality or interpretation of reality as your starting point, but emphasise the given. Accept reality as it is in any way... result: Authenticity! The artist is no longer artist but the cold rational eye. [Armando, 1964] It is clear Koolhaas’s approach has much in common with the “Dionysian” aspect of Nietzsche’s philosophy. But there are also aspects of “Apollonian” Nietzschism in his work — the surreal empty photographs of some of the spaces in the Villa Dall’Avra, and his “House for Two Friends” of 1988. Koolhaas’s belief that it is irresponsible not to acknowledge the brutal conditions of modernity is everywhere apparent; we have no choice but to despair at the condition of architecture, though in the hands of a hero there is an opportunity for sublimating an ironic or hyperenergetic response into something that might approach poetry.

5.3

Alvar Aalto

The work of the Finnish architect Alvar Aalto, and in particular his building known as Baker House, is selected to illustrate a pragmatic, or synthetic approach. Baker House is a dormitory for the Massachusetts Institute of Technology, designed in 1946 while Aalto held a visiting professorship there. The site runs along the north side of the Charles River and from the very start Aalto’s plans seek to find ways of maximising the view of the river for every student. Early sketches show clusters of rooms facing south, and because a simple single-sided slab would not contain sufficient rooms, several ways are examined of increasing the density: by parallel blocks in echelon, by fan-shaped ends, and by the “giant gentle polygon”, resolving itself into a sinuous curve, that was finally adopted. This gives a wide variety of room shapes, with the advantage that most rooms look along the river east or west rather than just straight across it. His presentation to the building committee post-rationalised the final form by comparing it to still more possible patterns, parallel blocks at right angles to the river, for example, which might be efficient but fail to take advantage of the site. As in so many examples of Aalto’s work, the formal solution is intuitive, only achieved after many free-hand sketched alternatives, some of which might seem inherently implausible, but the design, once reached, is then subject to scrutiny in quite measurable ways. A singleloaded corridor was undoubtedly more expensive than a regular central corridor arrangement with rooms both sides, but provided Aalto could achieve a cost per bed-space of $5,300 it could be justified. Baker House reflects many of Aalto’s social convictions and formal strategies. Firstly, the sets of rooms can be seen as an illustration of what Aalto might mean by his call for “flexible standardisation”. Each cell is essentially identical, but because of the shaped curve on plan 22 different room shapes are created on a typical

1248

Christian Illies and Nicholas Ray

Figure 3. Sketch drawings, ground floor plan and rear elevation of Aalto’s Baker House. (The Alvar Aalto Museum Drawings Collection; photo from the private collection of Nicholas Ray.)

Philosophy of Architecture

1249

floor of 43 rooms. This meant numerous special details, for the built-in furniture in each room for example, but from a basically similar vocabulary. The general notion of the wave-form seems like a huge enlargement of the experimental wood sculptures, on which Aalto had been engaged with Otto Korhonen since the late 1920s, or a recapitulation on the urban scale of the 1938 New York Fair interior, his only previous work in the United States. Aalto built the free form of the rooms in a “rustic” brick — indeed he went to particular lengths to find dark red bricks that were rough-textured and included clinkers — but clad the orthogonal main common room in limestone. It is treated as a calm and static space, in contrast to the dynamic of the climbing stairs. During the same period, Walter Gropius, with his Cambridge-based practice TAC, was building a group of student residences off Harvard Yard to a similar brief. Though it did not have the benefit of the Charles River bordering it, the site is a challenging one, at the end of a sequence of open courts formed by buildings by architects of the stature of H.H. Richardson and McKim Mead and White. Gropius fails to do justice to either the potential of the programme or the opportunities of the site. Gropius seems to subscribe to a positivistic attitude to function in that he answers every measurable problem, but then allows himself some relaxation in a bit of whimsy. But his definition of the functional was not wide enough. A reductive modernist approach, which accepts without demur the economics of the double-sided corridor, leads to a series of similar blocks which, however, are disposed without regard to orientation. Their form is “softened” in the expression of the service areas, by projecting them slightly and sometimes angling their flank walls. Despite the pleasant well-treed spaces the intended picturesque effect is weak and the articulation of the buildings by distinguishing the bathrooms less convincing than Aalto’s stress on the special character of the common areas. In fact, by addressing more completely understood programmatic issues (such as the provision of spaces for internal meeting) directly, Aalto succeeds in inventing a surprisingly powerful form. The requirement, for security reasons, for a single main entrance, leads to the dramatic and apparently unprecedented pair of cantilevered staircases which rise up from a first floor landing over the central entrance, and create what appears from the distance to be a giant inverted pediment. These great naturally-lit stairs perform a social function, because instead of rigorously separating the “social” areas from the areas devoted to circulation, Aalto acknowledges the social function of a staircase and its landings by providing widened areas to allow for informal meetings and conversation; each is differently sized because of the differing positions reached on the long straight flight. The diagonal placing of the limestone-clad ground floor social space might at first sight be taken for a merely formal gesture, until the immediate context is examined. It is accounted for by the angle of the pre-existing approach path from the rest of the campus, which slices right through the block at ground floor level and ends in the doubleheight dining area. In contrast to the behaviour of Louis Kahn, in starting with an ideal form which is then modified by circumstances to fit issues of programme, site and budgetary

1250

Christian Illies and Nicholas Ray

constraint, Aalto’s procedure is to begin with precisely those particularities. But the forms that appear to have arrived intuitively in Aalto’s imagination were such that they performed well when subjected to testing under a wide range of measurable criteria. If their generation was not “rational”, they could be shown to be neither wilful nor arbitrary, unlike the signature forms of some “artist” architects of a later generation. Aalto was in fact sceptical that architecture was an art at all, and according to his biographer G¨ oran Schildt believed the art of building was an art only in the sense that medicine and cooking were arts. He conceived it as a humanistic activity based on technical knowledge which can only be pursued by people with a capacity for creative synthesis.62 This attitude provides a way of explaining Aalto’s constant pre-occupation with the specifics, of site and context, of materials, and of the study of the poetics of particular functions like washing one’s hands or reading a book. It would also account for his well-known scepticism about the mathematics of modular co-ordination: there was nothing special in mathematics except its usefulness. Politically, Aalto’s distancing himself from all systems of belief could be seen the same light. If Aalto was no idealist, it is also clear that he did not subscribe to Nietzschean despair. While Aalto may have shared Nietzsche’s view of the preeminence of the artist in society, his concern for everyday comfort and convenience (for “the little man” as he used to describe it) hardly squares with that philosophy. Aalto was closer to an alternative, more optimistic, version of Nominalism, that we described above as “sceptical idealism”, which holds that though aesthetic structures are not mimetic of an ideal world, they can assist in improvements: meanings which used to be supported by a symbolic language relating to a “higher order” can to some extent be re-established as an allegory. Aalto described himself as a “positive sceptic” and outlined his position in an address to the Jyv¨ askyla Lyc´ee in 1958: The much-discussed sceptical world view is in reality a necessary condition for anyone who would like to make a cultural contribution. This is of course dependent on scepticism’s transformation into a positive phenomenon, an unwillingness to ‘move with the stream’. On a higher level scepticism is transformed into its apparent opposite, to love with a critical sensibility. It is a love that lasts, as it rests on a critically tested foundation. It can result in such a love for the little man that it functions as a kind of guardian when our era’s mechanized life style threatens to strangle the individual and the organically harmonious life. [Schildt, 1998, pp. 15-17; Ray, 2005a, p. 187] Thus for Aalto art was not the imitation of transcendental structures, nor their despairing rejection, but the affirmation that human constructions are none the less 62 For the fullest account of Aalto’s work and life see the three volume biography by Schildt [1984, 1986, 1989].

Philosophy of Architecture

1251

real for being merely human. Art becomes (among other things) the affirmation of aspirations that are utopian but secular. As a pragmatic architect, Aalto knew that buildings only got built by the tricky political processes, at which he was adept, of charming his clients and influencing local officials. They were realised using the technologies that were available and in the face of the hard facts of climatic conditions and budgetary constraints. But a building was not just the solution to a mechanical set of problems: an act of invention was required, and sometimes the practice of building could result in an artefact that others would recognise as architecture. 6 CONCLUDING REMARKS The reader who has followed us thus far is likely to have encountered more questions than answers. Philosophy has certainly influenced architecture in many ways — and there are even more ways to look philosophically at architecture. In relation to ethics, we concluded that a normative theory that could account for the diverse nature of the discipline of architecture is still difficult to imagine. The culmination of the discussion on aesthetics suggested that contemplating the problems of architecture, so far from assisting in clarifying philosophical problems, seemed only to reveal them. The positions evident in the three case studies suggest that practising architects do indeed think in very different ways about their discipline, but even if their thought reflects the principles they have absorbed in reading, as well as lessons from practice itself, that does not appear to resolve the many issues we have raised. If we have succeeded in illustrating the way in which the discipline of architecture focuses on philosophical problems in a particularly acute way, that may serve part of our purpose. But it would be disappointing if that is all that has been achieved. More positively we could claim that this essay has concentrated on the description of ways in which ‘thinking’ shapes our ‘building’. Thinking, (by which we mean philosophy) also helps to understand architecture better by providing theories and categories by which we can conceptualise and analyse architectural phenomena. Architectural forms do not come randomly: how we build is in many ways a response to different or even opposed philosophical approaches. The different philosophical approaches — actual, or culturally determined — are heterogeneous and sometimes even opposed to each other on the fundamental level of theory. One could take one side in this debate — adopt a positivist position, for example, and suggest that all the unknowables will eventually become measurable, and therefore the apparently deeper problems will go away. One can say that we only need to be true to the phenomena, as phenomenologists do, though that may be at the expense of the respect that we consider due to the advance of science per se. In the face of the brutal facts of an unequal world apparently hurtling towards ecological disaster, a melancholic realism is compelling to others. It is the measure of the philosophical nature of these conflicting positions that they none of them can be shown to be absolutely false — at any rate if they were proved

1252

Christian Illies and Nicholas Ray

or disproved they would cease to be philosophical. The acceptance of the many unsolved tensions that philosophy of architecture faces, on the other hand, may lead to a positive view of the position of architecture, which maintains an element of idealism, as follows. Architecture itself can be regarded as a way of overcoming philosophical tensions by suggesting practical possibilities, namely designs, that appear to bridge between rival theories and approaches. Design is in fact the core human discipline, being the only activity that properly involves the imaginative conception of ideas, leading to artefacts that are realised as actual constructions in the world. Thus architects are, uniquely, in a position to fashion buildings and spaces that, at least in the interpretation of some, can offer resolutions to the dichotomies and tensions that are endemic to philosophy. ACKNOWLEDGEMENTS We would like to acknowledge the generous support of CRASSH (the Centre for Research in the Arts, Social Sciences and Humanities) at the University of Cambridge, and of Eindhoven University of Technology, that made this collaborative article possible. BIBLIOGRAPHY [Alberti, 1988] L. B. Alberti. De re aedificatoria ‘On the art of building in ten books’. translated by Joseph Rykwert, Neil Leach, and Robert Tavernor. MIT Press, 1988. [Alexander, 2002] C. Alexander. The Phenomenon of Life (vol. I of The Nature of Order). Center for Environmental Structure, Berkeley, 2002. [van Armando, 1989] H. D. van D. Armando. Een internationale primeur. In De nieuwe Stijl, Sjoerd van Faassen and Hans Sleutelaar, eds. De Bezige Bij, 1989. [Banham, 1966] R. Banham. The New Brutalism: Ethic or Aesthetic? Architectural Press, 1966. [Batteux, 1746] C. Batteux, Les beaux arts r´ eduits ` a un mˆ eme principe, 1746. [Baumgarten, 1961] A. G. Baumgarten. Aesthetica [Vol. 1 1750, Vol. 2 1758]. Olms, 1961. [Benedikt, 1992] M. Benedikt. Deconstructing the Kimbell: an Essay on Meaning and Architecture. Lumen Inc., 1992. [Bergilez and Genard, 2004] J. D. Bergilez and J. L. Genard. Minimalisme architectural: quand l’´ ethique s’inscrit dans le style. In Intervalles 1 : Minimalism(e)s. Li` ege, CIPA, 2004. Retrieved October 20, 2007, from www.cipa.ulg.ac.be/intervalles1/contents.htm. [Buossora and Mazouz, 2004] K. Boussora and S. Mazouz. The Use of the Golden Section in the Great Mosque of Kairouan. In Nexus Network Journal, 6, 2004. [Brawne, 1992] M. Brawne. From Idea to Building. Butterworth Heinemann Ltd, 1992. [Burchardt, 1878] J. Burckhardt. The Civilization of the Renaissance in Italy. Translated by S. G. C. Middlemore, 1878. Republished by Penguin, 1990. [Burke, 1990] E. Burke. A Philosophical Enquiry Into the Origin of Our Ideas of the Sublime and the Beautiful [1759]. ed. Adam Phillips. Oxford University Press, 1990. [Carter, 1981] R. Carter. Karl Friedrich Schinkel: The last Great Architect. In Collection of Architectural Designs including those designs which have been executed and objects whose execution was intended by Karl Friedrich Schinkel, Carter, Rand, ed. Exedra Books Incorporated, 1981. Retreived September 5, 2008, from http://www.tc.umn.edu/∼peikx001/rcessay.htm. [Coleman, 1990] A. M. Coleman. Utopia on Trial: Vision and Reality in Planned Housing. Hilary Shipman Ltd, 1990.

Philosophy of Architecture

1253

[Collins, 1971] P. Collins. Architectural Judgement. University of Toronto Press, 1971. [Colquhoun, 1985] A. Colquhoun. Essays in Architectural Criticism. MIT Press, 1985. [Curtis, 1982] W. Curtis. Modern Architecture Since 1900. Phaidon Press, 1982. [Curtis, 1986] W. Curtis. Le Corbusier: Ideas and Forms. Phaidon Press, 1986. [Davies, 1994] S. Davies. Is Architecture Art?. In Philosophy and Architecture. M. H. Mitias, ed., pp. 31-47. Rodopi, 1994. [Dewey, 1988] J. Dewey. Human Nature and Conduct. In The Collected Works of John Dewey (Middle Works Vol. 14), Boydston, Jo Ann, ed., pp. 21-32. Southern Illinois University Press, 1988. [Dutch Royal Society of Architects, 2007] Dutch Royal Society of Architects. Gedragsregels van de Koninklijke Maatschappij tot Bevordering der Bouwkunst Bond van Nederlandse Architecten. 1998. Retrieved August 25, 2007, from http://www.bna.nl. [Eardely, 1973] A. Eardley, (trans.) The Charter of Athens. Grossman, 1973. [Edwards, 2003] A. Edwards. Images of Eden, an enquiry into the psychology of architecture. The Skylark Press, 2003. [Eliade, 1971] M. Eliade. The Myth of the Eternal Return. Bollingen, 1971. [Fisher, 2000] S. Fisher. How to think about the ethics of architecture. In Ethics and the built environment, Fox, W., ed., pp. 170-182. Routledge, 2000. [Forty, 2000] A. Forty. Words and Buildings: A Vocabulary of Modern Architecture. Thames and Hudson, 2000. [Fox, 2000] W. Fox, ed. Ethics and the Built Environment. Routledge, 2000. [Fox, 2007] W. Fox. A Theory of General Ethics, Human Relationships, Nature and the Built Environment. MIT Press, 2007. [Fox, 2008] W. Fox. Architecture ethics. In A Companion to Philosophy of Technology, Berg, J.-K. and Hendricks, V. F., eds. Blackwell, 2008. [Frampton, 1980] K. Frampton. Modern Architecture, a Critical History. Thames and Hudson, 1980. [Frampton, 1983] K. Frampton. Towards a Critical Regionalism: Six Points for an Architecture of Resistance. In The Anti-Aesthetic, Foster, Hal, ed. Bay Press, 1983. [Furniss, 1993] T. Furniss. Edmund Burke’s Aesthetic Ideology. Cambridge University Press, 1993. [Gadamer, 1983] H.-G. Gadamer. What is Practice? In Reason in the Age of Science [Vernunft im Zeitalter der Wissenschaft, 1976]. MIT Press, 1983. [Gadamer, 1989] H.-G. Gadamer. Truth and Method [Wahrheit und Methode, 1960]. 2nd revised edition. trans. J. Weinsheimer and D.G.Marshall. Crossroad, 1989. [Gauldie, 1969] W. S. Gauldie. Architecture (Appreciation of Arts). Oxford University Press, 1969. [Games, 1985] S. Games. Behind the Fa¸cade. BBC Arial Books, 1985. [Giedion, 1962] S. Giedion. Space, Time and Architecture. The Charles Eliot Norton Lectures 1938-9. Oxford University Press, 1962. [Gilpin, 1792] W. Gilpin. Three Essays: On Picturesque Beauty; on Picturesque travel; and On Sketching Landscape. London, 1792. [Gombrich, 1979] E. Gombrich. The Story of Art. Phaidon, 1979. [Guy and Farmer, 2000] S. Guy and G. Farmer. Contested constructions. The competing logics of green buildings and ethics. In Ethics and the Built Environment, Fox, Warwick, ed., pp. 73- 87. Routledge, 2000. [Harries, 1975] K. Harries. The ethical function of architecture. In Journal of Architectural Education, 29/1, pp. 14-15, 1975. [Harries, 1997] K. Harries. The ethical function of architecture, MIT Press, 1997. [Heidegger, 1977] M. Heidegger. The Age of the World Picture [Die Zeit des Weltbildes, 1938]. In The Question concerning Technology and Other Essays. trans. W. Lovett. Harper and Row, 1977. [Heidegger, 1993] M. Heidegger. Building, Dwelling, Thinking [Bauen, Wohnen, Denken, 1951]. In Basic Writings, Krell, David Farrell, ed., pp. 347-363. Routledge, 1993. [Heisenberg, 1958] W. Heisenberg. The Physicist’s Conception of Nature. Hutchinson, 1958. [Herdeg, 1983] K. Herdeg. The Decorated Diagram: Harvard Architecture and the failure of the Bauhaus Legacy. MIT Press, 1983. [Herodotus, 1987] Herodotus. The History. University of Chicago Press, 1987. [Hillier, 1986] B. Hillier. City of Alice’s Dreams. Architect’s Journal, Vol. 39-41, 1986.

1254

Christian Illies and Nicholas Ray

[Hillier and Iida, 2005] B. Hillier and S. Iida. Network and psychological effects in urban movement. In Proceedings of Spatial Information Theory: International Conference, COSIT, Cohn, A.G. and Mark, D.M., eds., pp. 475-490. Springer, 2005. [Hume, 1965] D. Hume. Of the Standard of Taste [1757] and Other Essays. The Library of Liberal Arts, 1965. [Husserl, 1970] E. Husserl. The Crisis of European Sciences and Transcendental Phenomenology. Northwestern University Press, 1970. [Ihde, 2008] D. Ihde. The Designer Fallacy and Technological Imagination, In Philosophy and Design: From Engineering to Architecture, Kroes, P. and Vermaas, P.E. and Light, A. and Moore, S. A., eds., pp. 51-59. Springer, 2008. ¨ [Illies, 2005] C. Illies. Architektur als Kunst. Zeitschrift f¨ ur Asthetik und Allgemeine Kunstwissenschaft, 50/1, pp. 57-76, 2005. [Illies, 2008] C. Illies. The moral relevance of architecture. IAPS Bulletin, 31, pp. 3-6, 2008. [Jencks, 1975] C. Jencks. Le Corbusier and the Tragic View of Architecture. Allen Lane, 1975. [Kahn, 1961] L. I. Kahn. Form and Design. Architectural Design XXXI, 4, pp. 130-149, 1961. [Kahn, 1991] L. I. Kahn. Writings, Lectures, Interviews. Rizzoli, 1991. [Kant, 1962] I. Kant. Critique of Judgement [Kritik der Urteilskraft, 1790]. transl. James Creed Meredith. Clarendon Press, 1962. [Koolhaas and Mau, 1995] R. Koolhaas B. Mau. S,M,L,XL. Jennifer Sigler, ed. 010 Publishers, 1995. [Kris, 1953] E. Kris. Psychoanalytic Explorations in Art. Allen & Unwin, 1953. [Kroes and Primus, forthcoming] P. Kroes and H. Primus. Technical artefacts as physical and social constructions: the case of Cit´ e de la Muette. Routledge. (forthcoming) [Kruft, 1994] H.-W. Kruft. A History of Architectural Theory from Vitruvius to the Present. Princeton Architectural Press, 1994. [Le Corbusier, 1946] Le Corbusier. Towards a New Architecture. Transl. Frederick Etchells. The Architectural Press, 1946. (originally published as Vers Une Architecture. Editions Cr`es, 1923.) [Leach, 1997] N. Leach, ed. Rethinking Architecture: a reader in cultural theory. Routledge, 1997. [Leach, 1999] N. Leach. Architecture and Revolution:Contemporary Perspectives on Central and Eastern Europe. Routledge, 1999. [Leach, 2002] N. Leach. Forget Heidegger. In Designing for a Digital World, Leach, Neil, ed., pp. 21-30. Wiley, 2002. [Lethaby, 1955] W. Lethaby. Architecture: An Introduction to the History and Theory of the Art of Building. Oxford University Press, 1955. [Lipman, 2003] A. Lipman. Architecture on My Mind: Critical Readings in Design. Unisa Press, 2003. [Lovejoy, 1927] A. O. Lovejoy. ‘Nature’ as Aesthetic Norm. In Modern Language Notes, 42/7, pp. 444-450, 1927. [Macarthur, 2007] J. Macarthur. The Picturesque: Architecture, disgust, and other irregularities. London and New York, Routledge, 2007. [Mallgrave, 1994] H. F. Mallgrave and E. Ikonomou, eds. Empathy, Form and Space, Problems in German Aesthetics 1873-1893. Getty Center for the History of Art and the Humanities, 1994. [Mandelbaum, 1967] M. Mandelbaum. Historicism. In Encyclopedia of Philosophy. Macmillan and The Free Press, 1967. [Massey, 1994] D. Massey. Space, place, and gender. University of Minnesota Press, 1994. [McCarter, 2005] R. McCarter. Louis Kahn. Phaidon Press, 2005. [Meagher, 2008] S. M. Meagher. Philosophy and the City: Classic to Contemporary Writings. State University of New York Press, 2008. [Meinecke, 1946] F. Meinecke. Die Entstehung des Historismus. Leibniz Verlag, 1946. [Meyer, 1980] H. Meyer. Bauen und Gesellschaft. Schriften, Briefe, Projekte. Verlag der Kunst, 1980. [Mikellides, 2008] B. Mikellides. Architectural Psychology 1969-2003, Theory, Practise and Education. Brookes Journal of Learning and Teaching, Vol. 2, Issue 2, 2007. Retrieved September 5 2008 from http://bejlt.brookes.ac.uk/article/architectural psychology 19692007. [Modiano, 1994] R. Modiano. The legacy of the Picturesque. In The Politics of the Picturesque, Copley, S. and Garside, P., eds. Cambridge University Press, 1994.

Philosophy of Architecture

1255

[Moneo, 2004] R. Moneo. Theoretical Anxiety and Design Strategies. MIT Press, 2004. [Moore, 1966] H. Moore. The sculptor speaks. In Henry Moore on Sculpture, James, Philip, ed., pp. 62-64. MacDonald, 1966. ¨ [Moritz, 1788] K. P. Moritz. Uber die bildende Nachahmung des Sch¨ onen. Braunschweig, 1788. [Muschamp, 2000] H. Muschamp. New York Times, 18th June 2000. [Nagel, 1979] T. Nagel. Mortal Questions. Cambridge University Press, 1979. [Newman, 1973] O. Newman. Defensible Space: Crime Prevention through Urban Design. Macmillan, 1973. [Nietzsche, 2000] F. Nietzsche. The Birth of Tragedy from the Spirit of Music [Die Geburt der Trag¨ odie aus dem Geiste der Musik, 1872]. Oxford University Press, 2000. [Norberg-Schulz, 1966] C. Norberg-Schulz. Intentions in Architecture. MIT Press, 1966. [Norberg-Schulz, 1971] C. Norberg-Schulz. Existence, Space and Architecture. Praeger, 1971. [Norberg-Schulz, 1990] C. Norberg-Schulz. Genius Loci: Towards a Phenomenology of Architecture, Rizzoli, 1990. [Osmond, 1957] H. Osmond. Function as the Basis of Psychiatric Ward Design., Mental Hospitals, 8, p. 23-29, 1957. [Pallasmaa, 2005] J. Pallasmaa. Touching the World — architecture hapticity and the emancipation of the eye, speech on the occasion of receiving an honorary doctorate at the University of Helsinki, retrieved on: www.enhsa.net/downloads/2005proceedings/06pallasmaa.pdf [Panofsky, 1957] E. Panofsky. Gothic Architecture and Scholasticism. Meridian Books, 1957. [Panofsky, 1968] E. Panofsky. Idea, a concept in art theory. University of South Carolina Press, 1968. [Pepper, 1984] D. Pepper. The Roots of Modern Environmentalism. Routledge, 1984. [Pevsner, 1943] N. Pevsner. An Outline of European Architecture. Penguin, 1943. [Plato, 1945] Plato. Republic. Trans. F. M. Cornford. Oxford University Press, 1945. [Popper, 1959] K. Popper. The Logic of Scientific Discovery. Basic Books, 1959. [Posener, 1978] J. Posener. Eine Architektur f¨ ur das Gl¨ uck? In Was ist Gl¨ uck?, Carl Friedrich von Simens Stiftung, ed., pp. 149-170. Deutscher Taschenbuchverlag, 1978. [Price, 1842] U. Price. An Essay on the Picturesque, as compared with the Sublime and the Beautiful. In Sir Uvedale Price on the Picturesque. Lauder, Dick, ed. Edinburgh, 1842. [Pugin, 1843] A. W. N. Pugin. An Apology for the Revival of Christian Architecture in England. John Weale, 1843. [Punter, 1994] D. Punter. The Picturesque and the Sublime: two worldscapes. In The Politics of the Picturesque, Copley, S. and Garside, P., eds., Cambridge University Press, 1994. [Ramroth, 2006] W. G. Ramroth. Pragmatism and Modern Architecture. McFarland, 2006. [Rapoport, 1969] A. Rapoport. House Form and Culture. Prentice-Hall, 1969. [Rapoport, 1990] A. Rapoport. The Meaning of the Built Environment: A Non-Verbal Communication Approach. University of Arizona Press, 1990. [Ray, 2005a] N. Ray. Alvar Aalto. Yale University Press, 2005. [Ray, 2005b] N. Ray, ed. Architecture and its Ethical Dilemmas. Taylor & Francis, 2005. [Roberts, 2005] J. Roberts. Architecture, Morality and Taste. In Architecture and its Ethical Dilemmas, Ray, Nicholas, ed. Taylor & Francis, 2005. [Rowe and Koetter, 1978] C. Rowe and F. Koetter. Collage City. MIT Press, 1978. [Ruskin, 1878-79] J. Ruskin. Modern Painters (5 Vol.). John Wiley, 1878-79. [Rykwert, 1982] J. Rykwert. The Dark Side of the Bauhaus. In The Necessity of Artifice. Rizzoli, 1982. [Rzchtarikova and Akkerman, 2003] J. Rzchtarikova and A. Akkerman. Trajectories of fertility and household composition in the demographic profile of the Czech Republic. Population and Environment, 24 (3), pp. 225-254, 2003. [Scott, 1980] G. Scott. The Architecture of Humanism. Architectural Press, 1980. ¨ [Schiller, 1983] F. Schiller. Letters upon the Æsthetic Education of Man [Uber die ¨ asthetische Erziehung des Menschen, 1795]. Clarendon Press, 1983. [Schildt, 1984-89] G. Schildt. Alvar Aalto: the Early Years; the Decisive Years, the Mature Years. Rizzoli, 1984, 1986, 1989. [Schildt, 1998] G. Schildt, ed. Alvar Aalto in his own words. Rizzoli International, 1998. [Scholfield, 1958] P. H. Scholfield. The Theory of Proportion in Architecture. Cambridge University Press, 1958. [Sch¨ on, 1983] D. A. Sch¨ on. The Reflective Practitioner: How Professionals think in action. Basic Books, 1983.

1256

Christian Illies and Nicholas Ray

[Schopenhauer, 1849] A. Schopenhauer. Die Welt als Wille und Vorstellung. [1859] Berlin und Wien, 1924. [Scruton, 1979] R. Scruton. The Aesthetics of Architecture. Princeton University Press, 1979. [Scruton, 2000] R. Scruton. After Modernism. City Journal, Spring 2000. [Simon, 1969] H. A. Simon. The Sciences of the Artificial. MIT Press, 1969. [Spector, 2001] T. Spector. The Ethical Architect: The Dilemma of Contemporary Practice. Princeton Architectural Press, 2001. [Spector, 2004] T. Spector. Pragmatism for Architects. Contemporary Pragmatism, Vol. 1/1, pp. 133-, 149, 2004. [Spector, 2005] T. Spector. Codes of Ethics and Coercion. In Ray, Nicolas, ed., Architecture and its Ethical Dilemmas, Taylor & Francis, 2005. [Stokes, 2002] A. Stokes. The Quattro Cento and The Stones of Rimini. First combined edition, Ashgate Publishers Ld, Aldershot, 2002. [Sullivan, 1886] L. Sullivan. Essay on Inspiration. Inland Architect, 8:2, pp. 61–64, 1886. [Sullivan, 1924] L. Sullivan. Autobiography of an Idea. Pess of the American Institute of Architects, Inc, New York City, 1924. Republished by Reprint Services Corp, 1991. [Summerson, 1966] J. Summerson. The Classical Language of Architecture. MIT Press, 1966. [Taylor, 1991] C. Taylor. The Malaise of Modernity. Anansi, 1991. [Tillich, 1952] P. Tillich. Being and Love. In Moral Principles in Action, Anshen, R. N, ed., pp. 661-672. Harper, 1952. [Van Eyck, 1999] A. Van Eyck. Works 1944-99. Birkhouser Verlag AG, 1999. [Venturi, 2006] R. Venturi. Complexity and Contradiction in Architecture. New York, Museum of Modern Art, 2006. [Venturi and Scott-Brown, 1972] R. Venturi and D. Scott-Brown. Learning from Las Vegas. MIT Press, 1972. [Vesely, 2004] D. Vesely. Architecture in the Age of Divided Representation. MIT Press, 2004. [Vischer, 1994] R. Vischer. On the Optical Sense of Form: a Contribution to Aesthetics. In Empathy, Form, and Space: Problems in German Aesthetics, 1873-1893, Getty Center for the History of Art and the Humanities, ed. University of Chicago Press, 1994. [Vitruvius, 2001] Vitruvius. Ten Books on Architecture. ed. Ingrid D Rowland. Cambridge University Press, 2001. [Wasserman et al., 2000] B. Wasserman, P. J. Sullivan, and G. Palermo. Ethics and the Practice of Architecture. John Wiley, 2000. [Watkin, 1977] D. Watkin. Morality and Architecture: the development of a theme in architectural history and theory from the Gothic Revival to the Modern Movement. Clarendon Press, 1977. ¨ [Welsch, 1996] W. Welsch. Grenzg¨ ange der Asthetik. Reclam, 1996. [Whitford, 1984] F. Whitford. Bauhaus. Thames and Hudson, 1984. [Wigley, 1992] M. Wigley. Heidegger’s House: the violence of the domestic. In Columbia Documents of Architecture and Theory, D 1, 1992. [Wilde, 1990] O. Wilde. The Picture of Dorian Gray. In The Complete Works of Oscar Wilde, pp. 17-167. Enderby, 1990. [Williams, 1981] B. Williams. Moral Luck, philosophical papers 1973-1980. Cambridge University Press, 1981. [Williamson and Radford, 2000] T. Williamson and A. Radford. Building, Global Warming and Ethics. In Ethics and the Built Environment. Fox, Warwick, ed., pp. 57-73. Routledge, 2000. [Wimsatt, 1954] W. K. Wimsatt, Jr. and M. C. Beardsley. The intentional fallacy. In The Verbal Icon: Studies in the Meaning of Poetry. University of Kentucky Press, 1954. [Wittgenstein, 1980] L. Wittgenstein. Culture and Value [Vermischte Bemerkungen, 1977]. ed. G.H. von Wright and Heikki Nyman, trans. Peter Winch. Blackwell, 1980. [Wittkower, 1949] R. Wittkower. Architectural Principles in the Age of Humanism. The Warburg Institute, University of London, 1949. [Woessner, 2003] M. Woessner. Ethics, architecture and Heidegger. ‘Building Dwelling Thinking’ in an American context. City, Vol.7/1, pp. 22-44, 2003. [W¨ olfflin, 1886] H. W¨ olfflin. Prolegomena to a Psychology of Architecture, PhD Thesis, University of Munich, 1886. [W¨ olfflin, 1950] H. W¨ olfflin. Principles of Art History, The Problem of the Development of Style in Later Art. trans. M. Hottinger. Dover Publications, 1950. [Young, 2002] J. Young. Heidegger’s Later Philosophy. Cambridge University Press, 2002.

PHILOSOPHY OF AGRICULTURAL TECHNOLOGY Paul Thompson Mapping the uncertain and shifting boundary between nature and artifice is a Sysiphusian task for the philosophy of technology, and a poignant one in the case of agriculture. Farming is a primal form of technology, yet the farm is, for many, a paradigm of nature. Perspective and background can situate agriculture more firmly on one side of this divide or the other. Modern food shoppers perusing a display of fresh fruits and vegetables are scarcely aware that the objects arrayed before them are technological artifacts, products of decades (if not centuries) of R&D dedicated to the adjustment and adaptation of plant characteristics such as ripening, flowering, response to soil and climate, as well as, of course, uniform appearance and flavor. Some agronomists have argued that Zea mays (more colloquially “maize” or “corn”) was the (perhaps serendipitous) invention of ancient farming peoples in what is now Mexico. Modern varieties are wholly dependent upon human intervention for their biological reproduction [Goodman, 1988]. Agricultural scientists — and to a significant degree farmers themselves — are thus inclined to see agriculture itself as a form of technology. It is natural for them to see different cultivars, each having distinct characteristics that make them better or worse for different climates, soils and farming systems, in instrumental terms. In fact, they are likely to refer to a farming strategy such as “no till” (to control soil erosion) or crop rotation (alternating cereals and legumes to maintain soil fertility) as a technology. Seeds themselves will be described as technology, though interestingly animals and livestock breeds will not. It was, in part, this instrumental attitude toward the objects and methods with which farming is accomplished that made it easy for people involved in agriculture to see genetic engineering as a straightforward logical development of what they had been doing all along. Modern agricultural science raises problems for the philosophy of technology that deserve study in their own right. Green revolution crops, for example, have been tied to debates over population growth and issues in environmental justice, while there has been widespread moral protest over genetic engineering and so-called “genetically modified organisms” (or GMOs). But of broader interest, perhaps, is the way that a focus on agriculture brings many standard components of the European philosophical canon into philosophy of technology. Such a focus requires a broad understanding of agriculture. Agriculture is always connected with the production of food, but the production of fiber (cotton, wool, silk and linen) for clothing and other products is also an agricultural activity. Farming Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

1258

Paul Thompson

quintessentially follows the cycle of seasons, but food and fiber are also derived from tree and vine crops that are not sown or even harvested every year. As a research and administrative matter, fisheries and forestry are typically included in the mandate of scientific institutes or government ministries responsible for agriculture. The domain of technological application referenced by the word ‘agriculture’ is thus large and may include forms of renewable natural resource exploitation and management that we do not normally associate with the word ‘farming’. Despite reasonably held presumptions of urbanites, I will assume that agriculture is an inherently technological activity. Beyond obvious farming tools (such as scythes, tractors, and harvesters), the technological base of agriculture includes cropping systems, the grading systems for meats and grains that structures commodity markets, and of course the plants and animals themselves, artifacts of both farmer selection and scientific manipulation. 1

AGRICULTURE THROUGH THE HISTORY OF PHILOSOPHY

While much recent philosophy of technology has been impressed with its own recentness, agriculture enjoys a long and rich philosophical tradition. Indeed, Hesiod’s Works and Days, the epic pre-philosophical poem that predates what is often taken to be the birth of philosophy in ancient Greece extols human occupation and purpose in agrarian terms. Greek philosophy grappled with notions of polity and the good life that presuppose a society organized around citizens whose fundamental interest in governance revolves around an agrarian household structure. Victor Davis Hanson writes that the hoi mesoi, the focus of Aristotle’s Politics and the mainstay of the phalanx, were freehold farmers. He argues that their willingness to engage in this uniquely Greek form of military tactics was based on their shared interest in protecting their lands. Their independence and mindset was the product of farming a diverse set of tree and vine crops that could be tended on a year round basis by a relatively small household unit, but that also involved substantial lifetime investment of labor that could be wiped out overnight by invading armies. The entire structure of Greek political life reflects its agricultural technology [Hanson, 1995]. In contrast, large scale agricultures in China, Egypt and Mesopotamia relied on annual crops that used elaborate irrigation systems and a large menial labor force that was deployed at harvest. They produced highly stratified hierarchical societies, and no corresponding sense of personal freedom, familial responsibility or community reciprocity. This view of Greek agriculture did not originate with Hanson. Aristotle himself may not have held this view, but Xenophon’s Oeconomicus is clearly a treatise on agrarian households that defends the form of gender division and household organization that emerges most naturally under Greek agricultural technology. Furthermore the agrarian interpretation of Greek thought and culture is an explicit component in Hegel’s Philosophy of History. Here, the progress of Spirit marks a critical turn with the Greeks because the topography and climate of Greece are ill

Philosophy of Agricultural Technology

1259

suited to the plantation systems that were the basis of Oriental civilization. The mountains and valleys of Greece lent themselves to smaller administrative units, but the soils were also well suited for grapevines and olive trees. This allowed the Greeks to develop an agriculture that could be managed by a familial household, but it also required reliable protection against marauders. Thus these independent households saw in each other common interests that gave rise to the notion of a Greek personality type as well as to the idea of citizenship. The first inkling of a notion (for Hegel the notion) of freedom was the result. If we understand agriculture as a form of technology — as an organized and considered means for accomplishing ends — then it is plausible to see the philosophy of technology as occupying a central role in Hegel’s thought, if not Aristotle’s, and indeed as virtually ubiquitous in the ancient world. The theme is that the material practice of bringing forth the necessities of life permeates culture and mentality. The quotidian tasks by which a people make their daily lives support and determine mentality and cultural identity. They fix particular conceptualizations of self-interest, occupation and way of life through the repetitive performance of subsistence tasks. From this underpinning of material activity flow further determinations of sociability, community and virtue. Whether or not we want to endorse this technologically deterministic picture of human nature and the fundamental philosophical problems of human existence, it is one that has been articulated by a distinguished list of poets and philosophers. Other names from the past include Virgil, Montesquieu and Adam Smith. Albert Borgmann is the foremost contemporary philosopher of technology to have followed this approach, especially in Real American Ethics where he emphasizes the culture of the table [Borgmann, 2006]. 2

THE POLITICAL ECONOMY OF AGRARIAN TRANSITION

In addition to the prominent role that Hegel and Aristotle gave to agriculture, the long transition from agrarian to industrial society also provides the backdrop to many political ideas throughout the modern period. John Locke’s theory of property, for example, makes reference to the practice of enclosure whereby the lands dedicated to common use (hence the term “commoners”) were literally walled off and dedicated to more intensive forms of crop and livestock production. Locke writes: [He] that encloses land, and has a greater plenty of the conveniences of life from ten acres, than he could have from an hundred left to nature, may truly be said to give ninety acres to mankind: For his labour now supplies him with provisions out ten acres, which were but the product of an hundred lying in common. [Locke, 1690, pp. 24-23] This passage is significant for several reasons. One is Locke’s contrast between enclosed land and “nature”. Common lands were not wilderness areas, but had been both cultivated and used for grazing livestock for a very long time when

1260

Paul Thompson

Locke wrote in 1690. Locke’s phrase “left to nature” does not imply pristine or undeveloped land. Enclosure was itself a technical component in the transition away from a social form in which the vast majority of the human population was daily occupied in agriculture, with well over half engaged in field level farming or animal husbandry. In Locke’s time, the practice of allowing all livestock to graze on temporarily idled fields was on the decline, as new crop rotations were eliminating the long periods of fallowing that had been needed to restore soil fertility. But it was difficult to reach agreement among those whose right to graze would be challenged by the new system [Mazoyer and Roudart, 2006]. Enclosure and private ownership was thus emblematic of an individual’s freedom to control the use of land. Today, the farming population of industrialized nations never exceeds 5%, and has dropped to less than 2% in the United States and Western Europe. This remarkable change in our way of life tracks the entire history of political thought for the last 500 years. Historians such as E. P. Thompson and Karl Polanyi have traced how the seeds of radical discontent over industrialization were sown in earlier agricultural displacements such as enclosure or the Scottish Highland clearances. The Marxist critique of capitalism is seen by some as an attempt to combine the intense sociability and mutual concern of the agrarian world with the productivity of industrial technology [Schmitt, 1987]. Locke’s passage on enclosure is also notable for the way that it utilizes a utilitarian form of argument to rationalize technical change. First it is important to see that Locke is talking about technical change. This was the view of Locke’s contemporary Arthur Young, who argued that the division of common lands into strips made it infeasible to utilize the best tools then available, including the best livestock breeds. Common tenure also was not conducive to improvements in drainage or terracing [Allen, 1994]. Given this presumption, Locke is making an argument that weighs costs and benefits expected to accrue in virtue of a change in technology. The argument continues to be used to support agricultural science and intensification of production in the developing world and in defense of genetically engineered crops. Whatever harm may accrue to the poor or to the losers in the transition to a new agricultural technology, this harm is alleged to be offset by the fact that increased production of agricultural commodities is a “gift” to mankind as a whole. Jeffrey Burkhardt has offered an extended critique of this oft repeated refrain, arguing that in simply presuming the priority of increased productivity, the advocates of new agricultural technology do not even acknowledge the legitimacy of claims based on equality or rights [Burkhardt, 1998; 2001]. Karl Kautsky’s critique of the Lockean/utilitarian argument became a central plank in the platform of socialist land reform. Kautsky noted that the effect of innovation in agricultural production methods was to alienate some aspect of the production process from the embedded nexus of land and labor power that had been characteristic of the agrarian world. The returns from increased efficiencies in agricultural production would flow to the capitalists who controlled the newly alienated good. Thus benefits of more productive livestock go to the breeders,

Philosophy of Agricultural Technology

1261

not pastoralists, and benefits from farm machinery go to farm implement manufacturers, rather than farmers [Kautsky, 1899]. Decades later, similar scenarios would play out in conjunction with the development of hybrid corn and other highyielding varieties, making fortunes for seed innovators such as Henry C. Wallace [Fitzgerald, 1993]. Much of the controversy over biotechnology replays Kautsky’s original critique, as will be discussed below. The theme of embedded production resurfaced in 20th century political economy, especially in the debate between critical theory and neo-classical economics. Marx himself had argued that the terms with which 19th century political economists developed their analyses reflected the mentality of early capitalism and industrialism. Critical theorists such as Marcel Mauss argued that the neoclassical portrayal of rationality in terms of self-interested maximization of returns from any activity or transaction was, in fact, inconsistent with the mode of production and distribution in pre-capitalist societies [Mauss, 1954]. E. P. Thompson characterized manorial Britain as a “moral economy” in which villagers operated with a felt entitlement to the grain growing in surrounding fields. It was, he argued, only when a new technological infrastructure of improved roads and canals made it feasible to transport grain from town to town in search of higher prices that grain took the form of an alienable, tradable good; that is, a commodity. The mentality of profit seeking and trading for maximum return was, he argued, also a product of this new technical infrastructure [Thompson, 1972]. Thus fairly standard elements in the repertoire of libertarian or utilitarian defenses of capitalism and private property, on the one hand, or of Marxism and critical theory, on the other, have grounding in the philosophy of agricultural technology. Lockean individualism and its view of the rationale for private property refer back to the fetters of commonly held land, while the utilitarian argument for seeing total social benefits as outweighing costs has an early instance in Locke’s defense of enclosure. On the other side of the coin, theorists of the left have celebrated the embeddedness of rural village life and decried commodification, a phenomenon that has an early historical occurrence in the way that large scale trading of food grains was made possible by technological advances in the distribution system for grain and livestock. 3 THE GREEN REVOLUTION The expression “Green Revolution,” may have originated in a speech given by United States Agency for International Development (USAID) Administrator William Gaud in 1968 [Gaud, 1968]. Gaud’s intent was to draw a contrast to a “Red” revolution in developing countries, allegedly being promulgated by the Soviet Union and the People’s Republic of China. He was referring to an orchestrated program of research and development for increased agricultural production in the developing world that had been financed originally by the Rockefeller Foundation and Ford Foundation of the United States. By 1968, this funding was already being supplemented by foreign aid from many Western developed nations. These efforts were eventually coordinated and administered by the Consultative

1262

Paul Thompson

Group on International Agricultural Research (CGIAR), which continues to oversee productivity oriented R & D at international research centers that emphasize specific crops, regions or topical areas. It is, however, wholly appropriate to use the term ‘revolution’ here in the sense originally promulgated by Thomas Kuhn and applied specifically to agriculture by Mazoyer and Roudart, who describe a series of agricultural revolutions in which the complex of plants, cultivation techniques and irrigation methods are reconfigured in response to ecological pressures in the form of resource depletion and degradation or population growth [Kuhn, 1957; Mazoyer and Roudart, 2006]. The Green Revolution displaced agricultural systems that utilized a genetically diverse mix of crops to minimize risks that were associated with their generally low productivity (as measured in terms of calories available for human consumption per year of human labor). In order to minimize the chance of a disastrous year, poor farmers developed farming strategies with a mix of crops, varieties and production styles to ensure the chance of producing yields adequate for subsistence even in years of drought, flood or pestilence [Scott, 1977]. By the mid 20th century, developed nations were controlling this risk through grain storage and international trade. But the economies of many nations did not permit their populations to participate in the emerging global system of agricultural trade. The Green Revolution was thus conceived as a program to first ensure food security in developing countries by increasing domestic productivity of cereal crops to levels that would support rapidly growing urban (hence non-farming) populations, and second to build wealth that would permit these countries to enter the international trading system. The chief elements in early CGIAR centers stressed dwarf varieties of rice and wheat. They are plants that respond well to fertilizers. Fertilizing many of the traditional varieties developing country farmers were growing led to too much growth in the stalk, causing lodging. The plants would fall over and perish in wind or rain. Dwarf varieties convert a higher proportion of the energy contained in fertilizers to edible grains, boosting per acre yields considerably. In this respect, the Green Revolution was a specific application of a much broader scientific paradigm that began to take shape in the pioneering work of Justus Liebig (1803-1873) in the 19th century. Insights into the chemistry of soils and plant or animal metabolism were systematically deployed to increase the output of consumable protein as a function of land area, in the case of plants, or feeds, in the case of animals. Agricultural scientists’ continued success in increasing yields contradicts the famous prophecy of Thomas Malthus (1766-1834), who argued that human population growth would increase exponentially, while agricultural yields would be challenged to grow at an arithmetic rate. Malthus believed that the only responses to this underlying tension lay in the high death rates associated with warfare and the diseases of extreme poverty, or in other vices (such as prostitution) that lowered the birth rate. His belief that human population growth is both an irresistible natural force and a source of unmitigated evil was resurrected in the 1960s by ecologists such as Garrett Hardin (1915-2003).

Philosophy of Agricultural Technology

1263

Prominent Green Revolution scientists such as Norman Borlaug also endorsed programs to control population growth, but rejected the extreme neo-Malthusianism of Hardin. Borlaug argued that with adequate support for agricultural science, yields would continue to outpace population growth [Borlaug, 1983]. Over the forty years since the debate was enjoined, it is Borlaug that appears to have been borne out by events. Yields have continued to increase steadily. But revolutions are seldom bloodless. Economist Willard Cochrane describes the “technology treadmill” that accompanies technological innovation in agriculture. The market price for grains reflects the average cost of production, and early adopters of new technology produce at below average cost per unit. As such, they enjoy significant profits. But as many farmers adopt the new technology, market prices fall to reflect the new average. Farmers’ gains are temporary; the enduring economic benefits go to consumers and to the middlemen who distribute, process and retail food [Cochrane, 1983]. What is more, late adopters are producing above the average price, and suffer losses that, given the tenuous economic circumstances under which farming has often occurred, cause bankruptcy. Their assets (land and equipment) are purchased by early adopting farmers with cash to spare. So improvements in agricultural technology spawn a continuing cycle of fewer and larger farms, while those displaced from farming must enter the non-farm workforce at the very bottom rung. The Green Revolution did not escape the treadmill. Worse, because dwarf varieties required fertilizer, returns to farmers had to be shared with fertilizer suppliers, an instance of the phenomenon described by Kautsky. Worse still, the reduction in diversity of the agricultural system created new vulnerabilities to insect pests, further shifting farmer profits to an emerging pesticide supply industry in developing countries. In a final coup de grˆ ace, developing country farmers become dependent on developed world technology suppliers, who reap the principal economic benefits of Green Revolution technology. These social and ecological dimensions have led many to question the sustainability of yield increases associated with Green Revolution technologies in a debate that both revives and also expands upon the earlier dispute between Hardin and Borlaug. 4 DEBATING THE GREEN REVOLUTION This critique of the Green Revolution was originally mounted in the 1970’s. Keith Griffin, Michael Perelman, and Kenneth Dahlberg argued the points noted above and also pointed out how Green Revolution development strategies had not only failed to take advantage of village institutions for organizing and sharing the benefits of work, but had contributed significantly to the erosion of those institutions. Environmental problems also ensued [Griffin, 1974; Perelman, 1976; Dahlberg, 1979]. Their critique spawned a literature assessing and evaluating the Green Revolution and the CGIAR (see [Dalrymple, 1986; Lipton and Longhurst, 1988]). A recent summary concludes that developing country consumers received nutritional benefits and reduced the proportion of their income spent on food, especially

1264

Paul Thompson

among the poor. Benefits among farmers were uneven and much less widespread [Evenson and Gollin, 2003]. This debate over the Green Revolution is relevant to the philosophy of technology in at least three important respects. First, the debate involved moral claims that were thinly veiled behind the positivist rhetoric in which it was conducted. On the one hand, the putative beneficiaries of the Green Revolution — poor peasant farmers in developing countries — were arguably its victims. There was thus, at a minimum, an element of dissembling and self-deception often associated with technological optimism present in the way that the Green Revolution was represented to the broader public. On the other hand, defenders of the Green Revolution could reply that net benefits to the poor had exceeded harms. Even if those beneficiaries were in urban areas and purchased most of their food, the fact that poor people spend a far greater share of their income on food means that poor people benefit disproportionately from lower food prices. There was thus a utilitarian argument with strong egalitarian credentials to rationalize the turmoil that Green Revolution technology created in rural areas. Second, the Green Revolution debate epitomizes longstanding philosophical disagreements about the role of technology in historical processes of development. While defenders could frame a basically progressive narrative within the parameters of a utilitarian conception of human welfare, critics saw a dialectical opposition of domination and resistance, with technology emerging as a weapon of the powerful wielded against the weak. Finally, the fact that this debate occurred within the agricultural universities and research institutes where technology was being developed created a nascent critical consciousness among agricultural researchers. Other debates, especially the controversy over agricultural pesticides sparked by Rachel Carson’s Silent Spring from 1962, reinforced this trend. Far more than technology developers in many fields, the crop scientists, entomologists, geneticists and agronomists in agricultural research institutes began to practice a form of deliberative technology assessment in their research planning. This rudimentary critical consciousness endured a trial by fire in the debate over recombinant methods of plant and animal transformation — the GMO debate, discussed below. Conjoint with that debate a new round of critics, especially Vandana Shiva, brought the earlier criticisms of the Green Revolution to broader public attention. Though largely neglected by mainstream agricultural science, Shiva re-clothed the Green Revolution critique in the accouterments of feminist philosophy. Its failures were now caused by essentialism and reductionism [Shiva, 1993; 1995]. Shiva’s version of the Green Revolution controversy ignores the earlier debate altogether, but has been widely embraced by contemporary scholars. At the same time, Malthusian arguments continue to be raised by scientists such as Paul Erlich, as well as by environmental philosophers, such as Philip Cafaro [Cafaro, 1997].

Philosophy of Agricultural Technology

5

1265

BIOTECHNOLOGY AND GMOS

Recombinant techniques for inserting well characterized genetic constructs into plant and animal genomes began to be widely discussed and developed in the agricultural sciences in the 1980s. Along with adult cell mammalian cloning, which was first demonstrated in 1997, this cluster of techniques came to be referred to collectively as ‘biotechnology,’ though many agricultural scientists include older methods of plant breeding under that heading. (See Part VI: “Philosophy of Bitotechnology” for a more detailed discussion of the term.) The apparent novelty of biotechnology precipitated widespread angst. Social theorists have stressed the paradigm of risk acceptability in analyzing this controversy [Frewer et al., 1994; Wynne, 2001; Bauer and Gaskel, 2002]. Thousands of plant varieties and dozens of animal breeds have been developed through genetic engineering including types that confer enhanced nutrition, disease resistance and stress tolerance. Two plant examples — herbicide tolerance and various genes from the Bacillus thuringiensus (Bt) bacteria that produce toxins specific to butterflies and moths — represent the vast majority of commercial applications. According to the International Service for Acquisition of Agri-Biotech Applications (ISAAA), an industry funded non-profit organization that promotes biotechnology in the developing world, acreage planted in GMOs has increased steadily since 1996 [ISAAA, 2007]. However, GMOs have been the subject of enormous controversy and opposition, and were almost unavailable in Europe and Japan as of 2008. The GMO controversy has been approached philosophically in a number of different ways. Philosophers such as Michael Reiss and Roger Straughan, Mark Sagoff and Gregory Pence have interpreted opponents as taking the view that GMOs are “unnatural,” thus reprising the “Can foods be artifacts?” theme discussed above [Reiss and Straughan, 1996; Sagoff, 2001; Pence, 2002]. Bernard Rollin, Paul B. Thompson, and Hugh Lacey, have argued that concerns about the naturalness of GMOs must be set within a context of asking questions about the risks of these technologies [Rollin, 1986; 1995; Thompson, 1987; 1997; Lacey, 2005]. Jeffrey Burkhardt and Gary Comstock have seen all these debates as a special case of the long squabble over agrarian transition and the decline of family farms [Burkhardt, 1988; 1992; Comstock, 1989; 1990]. These alternative philosophical framings have been reinvented many times, especially since 2000, as new voices enter the fray. Whatever the overarching philosophical narrative, the GMO controversy is played out through a series of key episodes. Arpad Pusztai’s disturbing (but unreproduced) findings of toxicity in genetically engineered potatoes, John Losey’s study showing that pollen from Bt crops kills monarch butterflies, Ingo Potrykus’s thus far unrealized nutrition-enhanced “Golden Rice,” and Ignacio Chapella’s discovery of Bt transgenes growing in the center of diversity for maize led each to become embroiled in controversy, as their results were challenged and their motives questioned. Other episodes include the contamination of the US food supply with an unapproved “Starlink” version of the Bt gene, the refusal of GMO food

1266

Paul Thompson

aid by several African countries in 2002, and the long history of the Monsanto Corporation’s animal drug rBST, ultimately rejected by consumers even in the generally tolerant United States [Thompson, 2007]. So-called Terminator seed is an illustrative example of these episodes. “Terminator” is the anti-biotechnology community’s name for technology known within the industry as Genetic Use Restriction Technology (GURT). These are gene constructs that provide companies with a means to protect their investment in developing a GMO. Some accomplish this by “turning on” the gene only when a proprietary chemical is applied (a chemical that farmers would be required to purchase). But the first widely discussed GURT limited farmers’ unauthorized appropriation of the gene by rendering seeds from the crop infertile. Had these seeds been released, they would have posed a serious risk of crop failure, especially in developing countries where seed saving is still common and where seed dealers are often not reputable. In fact, Terminator-style GURTs have never been developed and released. Nevertheless, extensive campaigning by Shiva and other activists has led to the widespread perception that the biotechnology industry has deployed this technology. Some seem to think that Terminator genes can become established in plants, leading to successive reproductive failure and starvation, though this scenario is wholly contrary to the biological mechanism involved. Terminator is thus a poster child for complaints about the unethical nature of GMOs. The story is complex. Developing country farmers have planted seeds that they believed to be GMOs and experienced crop failures, though industry alleges that poor quality seeds sold by unauthorized sources were the cause. Shiva continues to campaign against GMOs citing Terminator as a key issue [YouTube, 2007]. As a result the question of whether Terminator crops even exist appears to depend on one’s political orientation to the GMO debate. The object lesson for philosophy of technology consists primarily in the extreme polarization of this debate, with neither side showing willingness to credit testimony or evidence produced by the other. Seemingly reasonable calls for more democratic and open-ended debate over new technologies are tested by a political environment in which all evidence will be regarded as tainted and unreliable by one side or another. Although the GMO debate provides a particularly apt case study for those who see technological risk as an acid that dissolves all forms of public trust (see [Thompson, 2007]), it is perhaps more fitting to see it as raw material for a wide array of philosophical and narrative framings. In particular, it is possible to see GMOs as just the most recent episode in a continuing political struggle over the promise of technology and its exploitation by dominant social groups. As such, GMOs will almost certainly continue to be of interest in the philosophy of technology for some time to come. 6

MODERN ORGANIC AGRICULTURE

Green revolution thinking is not unchallenged within the agricultural sciences, and assuredly not among a cadre of farmers committed to alternative approaches.

Philosophy of Agricultural Technology

1267

These alternatives are diverse in both philosophy and specific technology, but organic agriculture is illustrative. Organic methods are also relatively uniform on a global basis due to efforts at harmonizing organic standards through the International Federation of Organic Agriculture Movements (IFOAM). Based on a two-year consultative process, IFOAM defines organic agriculture as follows: Organic agriculture is a production system that sustains the health of soils, ecosystems and people. It relies on ecological processes, biodiversity and cycles adapted to local conditions, rather than the use of inputs with adverse effects. Organic agriculture combines tradition, innovation and science to benefit the shared environment and promote fair relationships and a good quality of life for all involved. [IFOAM, 2008] Although this definition was intentionally developed to include farmers whose ability to participate in costly certification schemes is limited, in Europe and North America organic farming is tied closely to explicitly articulated organic standards and to fee-based certification by third parties. Historically, organic standards were developed by associations of organic farmers, but the United States Department of Agriculture (USDA) implemented a uniform National Organic Program (NOP) for products marketed in the United States in 2002. “The NOP regulations prohibit the use of genetic engineering, ionizing radiation, and sewage sludge in organic production and handling. As a general rule, all natural (non-synthetic) substances are allowed in organic production and all synthetic substances are prohibited” [USDA, 2008]. Like any set of technical specifications intended to regularize and harmonize technical practices, organic standards stipulate specific tools and techniques and place significant limitations on who can participate in the technical practice [Busch, 2000]. Current organic standards thus represent a negotiated consensus that reflects principal areas of overlap among competing philosophical visions of organic agriculture. Some of those visions were overtly political. Organic movements that began to take shape in the 1960’s reflected anti-capitalist ideals and promoted social justice, as well as encouraging a “return to nature,” and a commitment to localism and community solidarity. However, as organic producers began to achieve economic success, the process of developing standards began to focus on technical standards specifying types of fertilizer and methods of pest control. More socially oriented goals, such as fair pay for agricultural workers or limitations on farm acreage, proved difficult to articulate in standards that could be enforced and agreed upon by members of voluntary organic agriculture associations [Guthman, 2004]. Thus, when USDA arrived at the specifications noted above, this reflected much of the consensus reached by organic agriculture associations around the world. It is worth noting that a century ago most of the world’s agricultural production would have qualified as organic given the IFOAM definition or USDA standards. Although chemical fertilizers and arsenical pesticides began to be utilized to some degree in the last half of the 19th century, the massive expansion of global capac-

1268

Paul Thompson

ity for nitrogen production needed to supply explosives for World War I enabled widespread use of synthetic fertilizers, while first generation chemical pesticides such as DDT were developed under the auspices of defense spending in World War II [Kroese, 2002]. As such, organic methods often reflect agricultural practices of traditional origin that have survived through being handed down from one generation to the next. In one sense, then, organic methods do not arise from any explicit scientific or philosophical principles. But two individuals working independently in the first half of the 20th century articulated rationales for organic agriculture that continue to guide the development of organic methods today. 7 PATRIARCHS OF ORGANIC FARMING Rudolf Steiner (1861-1925) was an Austrian philosopher with an extraordinarily diverse portfolio of achievements in the arts, literature and educational theory. He was a scholar of Johan Wolfgang Goethe (1749-1832), with a strong interest in the scientific writings where Goethe developed his theory of archetypes. Yet Steiner’s writings and teachings fall almost entirely outside the accepted boundaries of academic philosophy as understood both in his lifetime and in recent years [Hansson, 1991]. Steiner became associated with followers of the Russian psychic Madame Helena Blavatsky (1831-1891), modifying her teachings and leading a breakaway group called the Anthroposophical Society. In the year before his death, Steiner delivered a series of eight lectures on agricultural methods and supervised some limited agricultural trials. Steiner provides no details about the source of his approach to agriculture. Some of the specific techniques appear to have been handed down to him from farmers he had known during his youth in Eastern Austria and what is now Croatia. Yet the text of Steiner’s lectures implies that his agricultural methods could be derived from anthroposophical principles stressing cosmic forces that permeate the universe. Consistent with Steiner’s reputation as an esotericist, comments and questions from the audience suggest that he was simply able to intuit his agricultural principles in virtue of his general wisdom and spiritual excellence. Steiner’s agriculture involves methods for creating a series of preparations that are subsequently applied to crops or crop manures in homeopathic doses. These preparations are themselves formulated from organic materials such as dung or plant and insect parts, then composted according to detailed procedures that involve burying at specified depths and at times and places determined in accordance with lunar phases and planetary alignments [Steiner, 1958]. Steiner’s background, his interest in mystical intuitionism, his implicit alliance with vitalism, teleology and other discredited scientific approaches in biology and his opaque references to cosmic forces all tended to discredit his legitimacy as a scientifically informed theorist of agriculture. Yet Steiner’s preparations have attracted a worldwide following among organic farmers who refer to his methods as “biodynamic farming.” Biodynamic farming generally meets organic standards, but not all organic methods meet standards proposed by various national biodynamic agricultural as-

Philosophy of Agricultural Technology

1269

sociations. Biodynamic standards do not unilaterally stipulate the use of Steiner’s preparations. Instead, biodynamic associations emphasize the organic unity of the farm, encouraging farmers to derive their own fertilizer by keeping animals and specifying, for example, that farmers may not use more fertilizer on any given plot of land than would have been produced by growing fodder crops on that land and feeding it to animals. Biodynamic associations stress the uniqueness of each farm, implying that uniform standards may not be appropriate, but they also encourage farmers to live fully in the organic spirit, including participation in association activities intended to promote the biodynamic way of life. Sir Albert Howard (1973-1947) was another patriarch of organic farming who had a far more conventional scientific career. Howard served in a series of agricultural posts throughout the British Commonwealth, and was President of the 13th session of the Indian Science Congress in 1936. His work in India emphasized the improvement of composting techniques already in widespread use by Indian farmers. The Indore Method — Howard’s approach to composting — stressed the storage of animal manures in slatted bins to accelerate aerobic decomposition and to facilitate mixing with soils and vegetative waste. Howard believed that composted manure was superior to synthetic fertilizer for plant nutrition in virtue of its ability to foster the growth of soil microbes. He stressed the mycorrhizal association between soil microbes and plant rhizomes, small horizontal offshoots from the plant’s principal root structure, believing this to be an unappreciated source of plant nutrition. In contrast to Howard’s faith in composting, the agricultural science of the 1930’s emphasized the stimulation of plant growth through copious administration of ammonium sulfates. The observed increases in yields for most agricultural crops subjected to this treatment led plant physiologists and farmers alike to conclude that any nutritional benefits being sacrificed through their neglect of the mycorrhizal association must be negligible. Furthermore, as already noted, the capacity for ammonia production had grown significantly at just about this time. Thus did the “chemical” philosophy of agricultural technology pioneered by Liebig finally win out over “biological” philosophies stressing organic unity or soil health [Uekoetter, 2006]. Howard was not convinced. He began to argue that neglect of the mycorrhizal association accounted for an overall weakening in plant health, a decline that is passed on to animals (including human beings) who feed on these weakened plants. For evidence Howard cited on the one hand vulnerabilities to plant fungal and microbial disease or insect pests, while on the other hand he drew support from a dubious theory of disease being promoted by a British medical doctor named J.E.R. McDonagh (1881-1965). At the same time, Howard lambasted what he took to be ill-considered reductionism in the agricultural sciences, arguing that proper agricultural research needed to be conducted under realistic conditions, and preferably on a working commercial farm [Howard, 2006]. Howard’s views became the basis of organic approaches promoted by the Soil Association in the United Kingdom and by the Rodale Institute in the United States.

1270

Paul Thompson

Though more grounded in mainstream science than Steiner, Howard’s references to “soil health” were far too reminiscent of the ´elan vital that had been one of Liebig’s primary targets in the 19th century. What is more, his reliance on McDonagh did little to win friends for organic farming in agricultural universities and scientific societies. Writing in The Quarterly Review of Biology, a reviewer has this to say of McDonagh’s “unitary theory of disease”: The result is a vague farrago of opaque terminology whose meaning (if any) it is almost impossible to interpret. The biochemical, microbiological, and clinical material seem almost equally meaningless. Some purveyors of eccentric literature are able to make their claims superficially plausible, even if they reason very loosely from very dubious assumptions. Thus it is sometimes interesting to read works which attempt to prove that the earth is flat, or that all civilizations originated from the continent Mu which later sank into the Pacific Ocean. Other eccentrics are merely tiresome and dull. It is the opinion of the reviewer that this book falls into the latter class. [Tobie, 1949] Thus though Howard and Steiner developed specific techniques for improving soil fertility that continue to be used widely by organic farmers, both became associated with ideas that were being rejected by the scientific establishment. In fact, it is only recently that agricultural scientists have begun to take organic methods seriously. They have been spurred in part by the evident success of contemporary organic methods, by rising consumer demand and by the recognition of environmental problems associated with chemical methods. As historian Frank Uekoetter notes, “Promoting a plurality of opinions, and of approaches, is an important contribution to the quest for an agriculture that is both productive and sustainable. . . . And researchers should not be discouraged by the fact that there is no Liebig quotation to legitimate this endeavor” [Uekoetter, 2006, p. 336]. 8 CONCLUSION In sum, agricultural technology is a rich source for philosophical reflection and debate, too often neglected both by philosophers and by scientists or engineers oriented to mechanical, medical and electronic applications of technology. Not only can agriculture be seen to figure in the writings of philosophers since antiquity, proving that the philosophy of technology is not as novel as some have thought, but an agricultural focus also integrates philosophy and technology alike into some of the pressing issues of the present day. Controversies over the Green Revolution, GMOs and their presumed alternative, organic farming, show that agricultural technology is central to the sustainability of contemporary society. Emphasis on the role of agricultural productivity in neo-Malthusian debates shows the relevance of philosophy of technology to the philosophical literature on population puzzles that arose in connection with Derek Parfit’s Reasons and Persons [Parfit, 1984].

Philosophy of Agricultural Technology

1271

The debate over social consequences associated with these technologies brings philosophy of technology squarely into the growing field of ethics and development. The fact that agriculture cannot help but have dramatic environmental implications means that the philosophy of agricultural technology is also an important subfield in environmental philosophy. These possible extensions of themes covered in this chapter illustrate further opportunities for research in the philosophy of agricultural technology. BIBLIOGRAPHY [Allen, 1994] R. Allen. Agriculture in the Industrial Revolution, in R. Floud, (ed.) The Economic History of Great Britain since 1700. Cambridge: Cambridge U. Press, 96-122, 1944. [Bauer and Gaskel, 2002] M. Bauer and G. Gaskel, eds. Biotechnology: The Making of A Global Controversy. Cambridge, UK: Cambridge University Press, 2002. [Borgmann, 2006] A. Borgmann.Real American Ethics: Taking Responsibility for Our Country. Chicago: University of Chicago Press, 2006. [Borlaug, 1983] N. Borlaug. Contributions of conventional plant breeding to food production. Science 219: 689–693, 1983. [Burkhardt, 1988] J. Burkhardt. Biotechnology, Ethics and the Structure of Agriculture, Agriculture and Human Values 5(3):53-60, 1988. [Burkhardt, 1992] J. Burkhardt. Ethics and Technical Change: The Case of BST, Technology in Society 14:221-243, 1992. [Burkhardt, 2001] J. Burkhardt. Agricultural Biotechnology and the Future Benefits Argument, Journal of Agricultural and Environmental Ethics 14: 135-145, 2001. [Busch, 2000] L. Busch. The Moral Economy of Grades and Standards, Journal of Rural Studies 16: 273-283, 2000. [Cafaro, 1997] P. Cafaro. Enough Already, Conservation Biology 11:1258-1259, 1997. [Carson, 1962] R. Carson. Silent Spring. Boston: Houghton Mifflin, 1962. [Cochrane, 1983] W. Cochrane. . The Development of American Agriculture: A Historical Analysis. Minneapolis: University of Minnesota Press, 1983. [Comstock, 1989] G. Comstock. Genetically Engineered Herbicide Resistance, Part One, Journal of Agricultural Ethics 2:263-306, 1989. [Comstock, 1990] G. Comstock. Genetically Engineered Herbicide Resistance, Part Two, Journal of Agricultural Ethics 3: 114-146, 1990. [Dahlberg, 1979] K. Dahlberg. Beyond the Green Revolution: The Ecology and Politics of Global Agricultural Development. New York: Plenum, 1979. [Dalrymple, 1986] D. Dalrymple. Development and Spread of High Yielding Wheat Varieties in Developing Countries (Bureau for Science and Technology, Agency for International Development, Washington, DC), 1986. [Evenson and Golin, 2003] R. E. Evenson and D. Gollin. Assessing the Impact of the Green Revolution, 1960 to 2000. Science 300: 758–762, 2003. [Fitzgerald, 1993] D. Fitzgerald. Farmers Deskilled: Hybrid Corn and Farmer’s Work, Technology and Culture, 4, 324-343, 1993. [Frewer et al., 1994] L. F. Frewer, R. Shepherd, and P. Sparks. Biotechnology and food production: knowledge and perceived risk, British Food Journal, Vol. 96 (9):26-33. [Gaud, 1968] W. Gaud. The Green Revolution: Accomplishments and Apprehensions, 1968. http://www.agbioworld.org/biotech-info/topics/borlaug/borlaug-green.html Accessed February 18, 2008. [Goodman, 1988] M. M. Goodman. The history and evolution of maize, Critical Reviews in Plant Sciences 7:197-220, 1988. [Griffin, 1974] K. Griffin. The Political Economy of Agrarian Change: An Essay on the Green Revolution. Cambridge, MA: Harvard University Press, 1974. [Guthman, 2004] J. Guthman. Agrarian Dreams: The Paradox of Organic Farming in California. Berkeley, CA: U. California Press, 2004.

1272

Paul Thompson

[Hanson, 1995] V. D. Hanson. The Other Greeks: The Family Farm and the Agrarian Roots of Western Civilization. New York: The Free Press, 1995. [Hansson, 1991] S. O. Hansson. Is Anthroposophy Science? Conceptus, 25,37–49, 1991. [Hardin, 1968] G. Hardin. The tragedy of the commons. Science 162: 1243–1248, 1968. [Howard, 2006] A. Howard. The Soil and Health: A Study of Organic Agriculture, with a new Introduction by Wendell Berry. Lexington, KY: U. Kentucky Press, 2006. [IFOAM, 2008] IFOAM. IFOAM Definition of Organic Agriculture. http://www.ifoam.org/organic facts/doa/index.html Accessed August 14, 2008. [ISAAA, 2007] ISAAA. Global Status of Commercialized Biotech/GM Crops: 2007. ISAAA Brief 37-2007. Manilla: International Service for Acquisition of Agri-biotech Applications. [Kautsky, 1899] K. Kautsky. Die Agrarfrage; eine Uebersicht u ¨ber die Tendenzen des modernen Landwirthschaft und die Agrarpolitik der Sozialdemokratie. Stuttgart: J. H. W. Dietz Nachf, 1899. [Kroese, 2002] R. Kroese. Industrial Agriculture’s War Against Nature, in The Fatal Harvest Reader: The Tragedy of Industrial Agriculture, A. Kimbrell, Ed. Washington, DC: Island Press, pp. 92-105, 2002. [Kuhn, 1957] T. Kuhn. The Copernican Revolution. Cambridge, MA: Harvard University Press, 1957. [Lacey, 2005] H. Lacey. Values and Objectivity: The Current Controversy about Transgenic Crops. Lanham, MA: Lexington Books, 2005. [Lipton and Longhurst, 1988] M. Lipton and R. Longhurst. New Seeds and Poor People. Baltimore, MD: Johns Hopkins University Press, 1988. [Locke, 1952] J. Locke. The Second Treatise of Government. Thomas P. Reardon, ed. Indianapolis: Bobbs-Merrill Publishing Co, 1690/1952. [Mauss, 1967] M. Mauss. The Gift: Forms of Exchange in Archaic Societies. New York: Norton, 1954/1967. [Mazoyer and Roudart, 2006] M. Mazoyer and L. Roudart. A History of World Agriculture from the Neolithic to the Current Crisis. London: Earthscan, 2006. [Parfit, 1984] D. Parfit. Reasons and Persons. Oxford: Oxford University Press, 1984. [Pence, 2002] G. Pence. Designer Food: Mutant Harvest or Breadbasket of the World? Lanham, MD: Rowman and Littlefield, 2002. [Perelman, 1976] M. Perelman. The Green Revolution: American Agriculture in The Third World. New York: Harper & Row, 1976. [Reiss and Straughan, 1996] M. J. Reiss and R. Straughan. Improving Nature? The Science and Ethics of Genetic Engineering. Cambridge: Cambridge U. Press. [Rollin, 1986] B. Rollin. The Frankenstein Thing, in Genetic Engineering of Animals: An Agricultural Perspective, J.W. Evans and A. Hollaender, eds. New York, Plenum Press, 285-298, 1986. [Rollin, 1995] B. Rollin. The Frankenstein Syndrome: Ethical and Social Issues in the Genetic Engineering of Animals, New York, Cambridge University Press. [Sagoff, 2001] M. Sagoff. Biotechnology and the Natural, Philosophy and Public Policy Quarterly 21: 1-5, 2001. [Schmitt, 1987] R. Schmitt. Introduction to Marx and Engels: A Critical Reconstruction. Boulder, CO: Westview Press, 1987. [Shiva, 1993] V. Shiva. Monocultures of the Mind: Perspectives on Biodiversity and Biotechnology. London: Zed Books, 1993. [Shiva, 1995] V. Shiva. Epilogue: Beyond Reductionism, in Biopolitics: A Feminist and Ecological Reader on Biotechnology, Vandana Shiva and Ingunn Moser, eds. London: Zed Books, pp 267-284, 1995. [Shiva, 1997] V. Shiva. Biopiracy: The Plunder of Nature and Knowledge. Boston: South End Press, 1997. [Steiner, 1958] R. Steiner. Agriculture. George Adams, Tr. London: Bio-Dynamic Agricultural Association, 1958. [Tobie, 1949] W. C. Tobie. The Nature of Disease Institute First Annual Report by J. E. R. McDonagh, Mark Clement, The Quarterly Review of Biology, Vol. 24, No. 3, p. 261, 1949. [Thompson, 1972] E. P. Thompson. The Moral Economy of the English Crowd, Past and Present, 1972 [Thompson, 1987] P. B. Thompson. Agricultural Biotechnology and the Rhetoric of Risk: Some Conceptual Issues, The Environmental Professional, 9:316-326, 1987.

Philosophy of Agricultural Technology

1273

[Thompson, 1997] P. B. Thompson. Science Policy and Moral Purity: The Case of Animal Biotechnology, Agriculture and Human Values 14: 11-27, 1997. [Thompson, 2007] P. B. Thompson. Food Biotechnology in Ethical Perspective, 2nd ed. Dordrecht, NL: Springer, 2007. [Uekoetter, 2006] F. Uekoetter. Know Your Soil: Transitions in Farmers’ and Scientists’ Knowledge in Germany, in Soils and Societies: Perspectives from Environmental History, J. R. McNeill and V. Winiwarter, Eds. Isle of Harris, UK: White Horse Press, pp. 322-340, 2006. [USDA, 2008] USDA. National Organic Program Background Information. http://www.ams.usda.gov/AMSv1.0/getfile?dDocName=STELDEV3004443& acct=nopgeninfo. Accessed August 14, 2008. [Wynne, 2001] B. Wynne. Creating Public Alienation: Expert Cultures of Risk and Ethics on GMOs, Science as Culture 16: 445–481, 2001. [YouTube, 2007] YouTube. Vandana Shiva on Terminator Seed. http://www.youtube.com/ watch?v=wrwUecuK8WM. Accessed February 26, 2008.

This page intentionally left blank

PHILOSOPHY OF MEDICAL TECHNOLOGY

Sven Ove Hansson

1

INTRODUCTION

It has often been remarked that one of the foremost characteristics of modern medicine is its extensive use of technology. Medicine has always used technology, but since the late 19th century its reliance on technology has expanded dramatically. One of the many consequences of this is a change the location of the physician’s activities. The use of special equipment made it necessary to move consultations from home visits to hospitals and physician’s offices. As an example of this, the number of hospitals in the US increased from 200 to 4000 from 1873 to 1910 [Davis, 1981, p. 8]. Not surprisingly, various uses of technology figure prominently in discussions on medical ethics. However, few attempts have been made to give a comprehensive philosophical perspective on medical technology, and in the philosophy of technology medical applications are in fact marginal [Vos and Willems, 2000, p. 2]. Medicine and technology have much in common. Contrary to the natural sciences, neither of them is aimed at obtaining knowledge for its own sake. Both have an emphasis on techne rather than episteme, i.e. their goal is to find means of achieving practical results, to change the world rather than just to understand it [Hansson, 2007a; 2007b]. Medicine and technology also have a large and rapidly growing intersection, namely the use of technological methods to achieve the goals of healthcare. However, “[e]ven the most mechanical elements of medicine. . . are rarely, if ever, described as technology by its practitioners. Physicians are reluctant to see themselves as technicians or applied scientists” [Davis, 1981, p. 3]. The use of ever more complicated technology in hospitals has increased the role of engineers in healthcare. Engineers are responsible for the operation of essential diagnostic, therapeutic and palliative equipment. Due to the need for their expertise, some technological and engineering personnel are moving closer to the patient and assume more clinical roles in multidisciplinary healthcare teams [Deber and Leatt, 1986; Fielder, 1991; Wood, 2002]. Unfortunately, their role is often insufficiently understood by the public and by members of the more wellestablished healthcare professions. “Unlike other health professionals who have a firmly established role within the hospital system, clinical engineers often assume Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

1276

Sven Ove Hansson

new and greater responsibilities without the needed authority or institutional support” [Saha and Saha, 1997, p. 189]. By investigating philosophical issues in medical technology, we can obtain a better understanding of clinical and biomedical engineering that are important branches of modern technology. Such studies will also help us to achieve a better understanding of the nature of medicine itself. There are five major categories of medical or healthcare-related technology: • Diagnostic technology identifies diseases and other conditions for treatment or palliation. • Therapeutic technology is used in the treatment of diseases. • Enhancing technology improves human functioning beyond what is needed to cure diseases. • Enabling technology alleviates the impact of disease or a disability. This includes personalized equipment such as eyeglasses and artificial limbs but also universal technologies such as entrances that are accessible via wheelchair. • Preventive technology reduces the risk or severity of accidents, toxic exposures, and other social and environmental mechanisms that give rise to disease or injury. This includes a wide variety of technologies, from sewage treatment plants to airbags. Diagnostic, therapeutic, and enhancing technologies are integrated in healthcare. Enabling technology includes both technology that is part of healthcare, such as prosthetic technology, and technology that has little connection with healthcare. Preventive technology is usually not closely connected with healthcare, but in many cases, such as automobile safety, it makes extensive use of medical knowledge. This chapter contains sections on diagnostic technology (Section 2), therapeutic technology (Section 3), enhancing technology (Section 4), and enabling technology (Section 5). Preventive technology is not treated here, but some aspects of it are discussed in Risk and Safety in Technology in part V of this handbook. The final Section 6 of this chapter is devoted to some issues that concern medical technology in general, namely how technology shifts responsibilities, what effects it has on the quality of care and human contact, and whether it gives rise to unsound and perhaps unnatural dependence on artificial devices. 2

DIAGNOSTIC TECHNOLOGY

Up to the 19th century, diagnosis was primarily an oral and visual process, unaided by instruments (the main exception being uroscopy). Physical diagnosis, often including measurements, was developed to a high degree of precision in the early 19th century [Davis, 1981, p. 183]. Around 1840 clinical laboratories were

Philosophy of Medical Technology

1277

introduced, offering an increasingly sophisticated repertoire of biochemical tests [B¨ uttner, 2002]. In the 1880s and 1890s clinical photography rose to importance as a means of documentation. After R¨ ontgen’s discovery of X-rays in 1895 photography was overshadowed by X-ray diagnosis that had a deep impact on most clinical disciplines [Kr¨ oner, 2005]. Today, medical diagnosis is based on a combination of anamnesis (information obtained by interviewing the patient), physical examination of the patient, laboratory examination primarily of blood but also of other tissues and excretions, and imaging techniques including classical X-ray images, tomography and ultrasonography. In recent years some types of diagnostic technology that were previously in the hands of physicians have been made available to the patients themselves. Asthmatics can use a peak flow meter to regulate their medication, and insulin-dependent diabetics can measure their blood-sugar levels and adapt the dosage. In particular the latter practice has had large impacts on therapy. With frequent measurements of blood sugar, blood sugar regulation has been made tighter, i.e. lower values can be kept without risking hypoglycaemia. This reduces long-term risks of blindness, neuropathy and atherosclerosis. It also makes it possible for diabetics to lead a less regular life, since they can adjust dosage to food intake and physical activity [Willems, 2000; Mol, 2000]. Technologically mediated progress in medical diagnosis gives rise to several important philosophical questions: How does increased diagnostic precision influence our concepts of disease? Is diagnostic precision motivated even when it does not lead to better therapy, or can it even have negative side effects? Can excesses in medical diagnosis give rise to social discrimination? The recent introduction of genetic technology in the clinical laboratory gives rise to further issues, in particular: Should we avoid collecting genetic information that may tell the patient more about herself than what she wants to know?

2.1

An excess of diagnoses?

Diagnosis is essential for treatment. Some of the most important contributions of technology to medicine have been diagnostic procedures that made it possible to offer patients more specific therapies and to commence therapy at an earlier stage of the disease. In some cases, the recognition of previously unknown preclinical signs of disease have made it possible to begin therapy before the patient suffers from the disease. Important examples of this are the use of mass radiography to discover early stages of tuberculosis and the use of sphygmomanometry to diagnose hypertension. Not surprisingly, there are also cases when improved diagnosis has not been matched by corresponding developments in therapy so that, at least for a period of time, diagnosis has no effect on the patient’s health. It has often been questioned whether diagnosis can have any value when it does not lead to a therapeutic intervention. In this discussion it is important to distinguish between two cases. The first case is diagnostic information about a manifest disease. Consider for

1278

Sven Ove Hansson

instance a patient with a back pain, who is referred to an X-ray exam. A possible outcome of the exam is the discovery of physiological changes in the spine that are not accessible to specific treatment and do not change the advice that the physician had already given the patient. Is such a diagnosis useless or perhaps even of negative value? Experience from this particular diagnosis points in the opposite direction. Patients with back pain often want confirmation that their disease is real, and therefore appreciate knowledge about the physiological nature of the disease [Rhodes et al., 1999]. Generally speaking, patients often want to know what disease they have. Furthermore, an exact diagnosis is in most cases required before the physician knows that it is useless to search for other, perhaps treatable, causes of the symptoms. Therefore, although not all diagnoses of manifest disease lead to improved treatment, careful diagnostication is usually an unavoidable component of responsible medical management of the patient’s complaints. The other, more problematic, case is that of a diagnosis without a manifest disease. Alvan Feinstein introduced the term lanthanic disease for diseases that can be detected by technological means, but are not experienced in any way by the patient [Feinstein, 1967; Hofmann, 2003]. Since the 19th century, life insurance companies have been a driving force behind the development of such diagnoses. They need methods to prognosticate a prospective customer’s expected remaining length of life. Two technologies were shown in the early twentieth century to be efficient for this purpose, namely measurement of the person’s blood pressure and her vital capacity (the maximal volume of exhaled air after a maximal inhalation). Physical standards based on sphygmomanometry and spirometry were used as health indicators in insurance medicine in the early twentieth century, but these diagnoses were not then matched by therapies [Davis, 1981, p. 185]. A modern example of a possibly problematic lanthanic diagnosis is osteoporosis at an early stage (also called osteopenia), as diagnosed through low bone mineral density (BMD, bone mass). This is an X-ray diagnosis (dual energy X-ray absorptiometry, DXA); the patient has no symptoms other than a somewhat increased risk of fractures. A study of women who received this diagnosis revealed that for many of them the bone scan had influenced their social lives. They perceived their bodies as fragile and therefore chose not to participate in a number of social activities. It is a widespread misconception that a person who suffers from osteoporosis should avoid physical activity in order to avoid fractures. In actual fact, the contrary is the case: physical activity is an important means of preventing an aggravation of osteoporosis [Magnus et al., 1996; Dalsgaard Reventlow et al., 2006]. Hence in this case, information about a technology-mediated diagnosis can be counterproductive in terms of medical prognosis. However, it is important to observe that this is not a necessary consequence of the use of this technology. Its effect will be positive if the physician who informs the patient of the diagnosis also manages to encourage her to increase instead of decreasing her physical activity, and to take other measures that contribute to halting the development of the disease, such as to stop smoking and reduce the intake of soft drinks.

Philosophy of Medical Technology

2.2

1279

Diagnosis as a source of social discrimination

New diagnoses often have impact on our concepts of disease and health, and they can also influence the way in which we conceive our bodies [Vos and Willems, 2000]. Hence, the exact measurement of physiological functions has led to new concepts of normality and abnormality, such as the notion of hypertension [Davis, 1981, p. 5]. New diagnoses can also be used to classify persons in new ways. Such classifications can have negative social effects for the persons to whom they are applied; in particular they can be used to discriminate against the persons so classified. Discrimination means that certain persons receive a worse treatment, or less of some advantage, than others, without sufficient justification to select them for such inferior treatment. The most discussed types of discrimination are those that affect women, ethnic, religious, and sexual minorities, and people with certain handicaps and medical conditions. In some cases a diagnosis alone, i.e. a diagnosis without an accompanying actual condition, can have a discriminating effect [Hansson, 2005]. The clearest evidence of such discrimination can be found in the insurance sector. Insurance companies have a right to collect medical information about their customers. They also have economic incentives to use such information to the customers’ disadvantage. Hence, patients with hereditary hemochromatosis have found themselves excluded from insurance although they complied with therapeutic phlebotomy and therefore had no increased risk of disease or death. (Some relatives of patients with this diagnosis have avoided such discriminatory treatment by not having themselves tested but instead donating blood as often as phlebotomy is recommended for patients with the disease [Barash, 2000]). AfroAmericans who are carriers of the sickle-cell trait have been discriminated against by life insurers, although their condition does not give rise to an increased risk of death [Bowman, 2000]. It should be emphasized, however, that the extent to which insurance companies have incentives to discriminate customers with certain diagnoses depends on the politically chosen construction of the insurance system. Hence, the American insurance industry uses such information to reject applications for health insurance policies and to refuse payment for the treatment of illnesses [Alper and Beckwith, 1988; Anderlik and Rothstein, 2001]. The prevalence of this practice depends on the fragmentary nature of American health insurance [Wolf, 1995]. Most European countries have more developed health insurance programmes that cover everyone and have the same premium for all persons on the same income level. In such systems there is no incentive for health insurers to collect prognostic medical information about their customers. On the other hand, the system for life insurance seems to be more or less the same in all countries, and gives rise to such an incentive. Another situation where discrimination can be based on a diagnosis is the recruitment of personnel. Employers can require medical information about prospective employees. A well-known example concerns the sickle cell gene. The U.S. Air

1280

Sven Ove Hansson

Force barred Afro-Americans with the sickle-cell trait from becoming pilots due to an erroneous belief that they were prone to illness at high altitudes [Dolgin, 2001]. In later years worries have been expressed that genetic information can be used by employers to discover predispositions to certain diseases, recessive genes for inherited diseases, or (hypothetically) various psychological characteristics [Brady, 1995; Silvers and Stein, 2002; Persson and Hansson, 2003]. However, it should be emphasized that the use of diagnostic technology for such purposes is within social control. Several countries have passed laws that regulate what information an employer may acquire about a prospective employee. One of the best-known examples of maltreatment based on a mere diagnosis is the social discrimination of recessive carriers of the sickle-cell gene in the Greek village Orchemenos. Since the gene was unusually common in this village, all inhabitants were offered testing. The purpose was to make it possible for carriers of the gene to avoid marrying other carriers. However, this strategy failed, and instead testing led to stigmatization of the carriers. Non-carriers chose to only marry other non-carriers, and carriers were left to marrying each other [Moore, 2000]. Another example is the Ashkenazi Jews. This group has a long history of volunteering for genetic research, and therefore a disproportionate number of genetic alterations have been shown among them. This has given rise to a widespread though mistaken view that they are more prone to genetic disorders than others, and they have on occasions been discriminated for that reason [Dolgin, 2001].

2.3 Genetic diagnoses In recent debates about discrimination it has usually been taken for granted that genetic information is more sensitive than most non-genetic information. The use of genetic information is also much less accepted. While it seems to be fairly accepted that a person who has a manifest illness with a bad prognosis is denied a life insurance, rejections based on genetic tests have been vehemently protested against. The view that genetic information requires more protection to ensure privacy than most other forms of medical information has been called genetic exceptionalism [Green and Botkin, 2003]. Genetic exceptionalism is an example of a general tendency that is also seen in many social and ethical debates on biotechnology: The application of technology to a genetic material is conceived as particularly sensitive and is sometimes seen as ethically problematic in itself. More concretely, three major differences between genetic and non-genetic information have been invoked to defend genetic exceptionalism. First, genetic information is said to give more precise information about the likelihood of future disease than what is obtainable from non-genetic tests. Secondly, genetic tests provide information not only about the tested individual but to some extent also about her relatives. Thirdly, genetic information is said to reveal fundamental and immutable characteristics of the individual [Alper and Beckwith, 1988].

Philosophy of Medical Technology

1281

As one example of the first argument (the predictive power of genetic tests), Roche and Annas [2001] claim that DNA-sequence data differs from other types of medical data in providing information not only about a patient’s current health status but also about her future health risks. According to these authors, genetic information is in this sense analogous to a coded “future diary”. This, however, is a severely misleading statement. Although information about single-gene diseases may have a high predictive power, most health-related genetic information refers to diseases with a complex etiology involving several genes and several environmental factors. In such, more typical cases the predictive power of genetic tests is far from impressive. There are also several examples of non-genetic diagnostic technologies with a high degree of predictive power. Two practically important examples are sphygmomanometry and tests for fecal occult blood. They both have great value in detecting diseases (hypertension respectively colon cancer) in their early stages before the patient is aware of it. Concerning the second argument, it is certainly true that family members can be affected by results from genetic tests. However, the same applies to non-genetic tests for infectious diseases (not least sexual partners in the case of sexually transmitted diseases). An interesting comparison can be made been made between Huntington’s disease and HIV in this respect. Huntington’s disease is a rare genetic neurological disease that usually does not give rise to noticeable symptoms until the patient is in her thirties or forties. Having the abnormal Huntington gene is similar to being HIV-positive in at least two important respects: One may remain healthy for a number of years before the onset of the disease. Furthermore, both conditions are frequently transmitted to offspring [Gin, 1997]. Finally, concerning the third argument, genetic information is believed to reveal who the person “really is”. This view of personhood has been called “genetic essentialism” [Alper and Beckwith, 1988]. According to that view, genetic information is more intimately related to a person’s true nature than other sorts of information about the person. As Launis [2000] has argued convincingly, genetic essentialism is based on the highly controversial metaphysical presumption that there is such a thing as a person’s core nature, or essential identity. Furthermore, the available empirical evidence shows that we are constituted by a combination of genetics and environment, not by genetics alone. However, it is possible that the technological availability of genetic information will lead to more emphasis on genetic, inherited aspects not only of health but also of human personality. In this way, technologically mediated knowledge might have impact on how we view each other as persons: It might lead to a focus on inherited, unchangeable traits rather than on the social influence on personality. On the other hand, other technologies are also developing that may have an opposite effect. Proteomics, and information about the expression rather than the presence of a gene, may become more predictive than genetic sequencing. Biochemical tests can be developed that reveal environmental influences on the person. The development of future diagnostic technologies will in all probability provide us with tools that reveal both the genetic and the environmental influences

1282

Sven Ove Hansson

on our bodies and our personalities. It is not possible to predict in what way these developments will influence our views on human beings, but the philosophical impact may be substantial. 3 THERAPEUTIC TECHNOLOGY Therapy, the remediation or treatment of a health problem, is of course at the centre of medicine (although the prevention of disease or accidents is no less important). Therapy has always involved technological procedures; fairly advanced surgery such as trepanation was performed in Neolithic times.

3.1 Therapeutic knowledge and knowledge of side-effects Today it is taken for granted, at least in academic medicine, that therapy should be based on scientific knowledge. However, the connection between therapy and science is much more recent than that between therapy and technology. In Hippocratic medicine that dominated medicine for more then two millennia, the most common therapies were bloodletting, purging, and emetics, all of which were positively harmful to the patients. Although medicine has been taught in universities since the late thirteenth century, its practice was based on Hippocratic teachings. Important advances in understanding of human biology were made, such as Harvey’s discovery of the circulation of blood, but they led to no therapeutic advances [Wootton, 2006]. It was not until the nineteenth century that professors of medicine strove to make their discipline one of the sciences. Two major approaches were taken to achieve this. One was to make medical therapy essentially a branch of the natural sciences. By studies in the laboratory, diseased organs and tissues could be classified and causes of disease could be revealed. Claude Bernard was a leading proponent of this approach to the scientification of medicine. The other approach was treatment experiments, i.e. what we today call clinical trials. In the nineteenth century the first pioneers of clinical research began to evaluate the effectiveness of therapeutic methods through statistical comparisons of groups of patients who had received different treatments [Booth, 1993; Wilkinson, 1993]. Originally, the two approaches to scientific medicine were seen as competitors. Today it is generally recognized that laboratory research is as necessary to develop new therapies as is clinical research to validate, evaluate, and calibrate them. Hence, the crucial source of therapeutic knowledge is the clinical trial. In a clinical trial, groups of patients with the same disease receive different treatment, and statistical analysis is performed to determine both the therapeutic effects and the side effects in the different groups. In this way, the therapy with the best balance between therapeutic chances and (risks) of side effects can be identified. The ethical defensibility of clinical trials is far from self-evident. The consensus view is that a clinical trial is only acceptable if there is genuine uncertainty about which of the tested treatments is best, and informed consent has been obtained from all the subjects [Hansson, 2006].

Philosophy of Medical Technology

1283

Although clinical trials were proposed in the early nineteenth century, they were rare until after World War II. Today, a large part of the published medical research is reports from clinical trials. Since the 1990s, the use of information from clinical trials for clinical decision-making has been facilitated by the development of systematic procedures for evaluating clinical research (evidence-based medicine, EBM) [Evidence-Based Medicine Working Group, 1992]. The vast majority of clinical trials concerns pharmacological treatment. A major reason for this is that new drugs are not allowed unless they have been shown in clinical trials to be therapeutically useful in comparison to previously available therapy. Government control of medical devices is less extensive than for pharmaceutical products. In particular, there is no general system for premarketing testing similar to that for drugs [Altenstetter, 2003]. As a consequence of this, much less clinical research is performed on the therapeutic use of technical devices than on the therapeutic use of drugs. Not surprisingly, mechanical and other technological devices can fail in unforeseen ways, just like drugs. There is a long historical list of such failures. The majority of these did not give rise to severe injuries. But there have also been cases when technological failures had fatal outcomes. One of the best-known cases is the Bjork-Shiley heart valve, in which case regulators and industry seem to have been too slow in taking actions to prevent continued implantation of a defective product. The decision to withdraw the product came unnecessarily late according to critics. The decision was not made by the regulators but voluntarily by the company [Fielder, 1991]. It is important to relate the producer’s responsibility for the functioning of a device to the actual clinical settings in which it will be used. One critic complained that “most medical device designers appear to have envisioned the controlled, delicate, and precise choreography of a surgical team, not the frantic activity of the emergency room or a ’code-blue’ call. Consequently, many devices are not as rugged and easy to use as they could be” (Houston, quoted in [Saha et al., 1985]). However, this situation may change. One observer of the system described the current situation as follows: “The long-lasting honeymoon between the industry and European healthcare regulators seems to have ended. For healthcare payers and purchasers the case is clear: medical technology is a cost-driving force. Thus, medical devices and the medical device industry have come under increasing scrutiny and regulation” [Altenstetter, 2003]. A possible outcome of such increased scrutiny could be that more clinical trials are undertaken in order to determine the functionality of therapeutic technology.

3.2

Therapy vs letting die

Discussions on death have a central role in medical ethics, and they have often been connected to critique of technology. Some critics see the “modern” death in a technologically equipped hospital as “unnatural”, whereas they regard “natural” death without modern medical technology as more dignified. This is a highly ro-

1284

Sven Ove Hansson

manticized view. “Natural” death is often an extremely painful process, whereas modern technology can to some extent relieve the dying person of pain and distress [Barnard and Sandelowski, 2001]. Many critics also underestimate the quality of life that is obtainable with lifesustaining technology. Hence, it is often believed that a life with a ventilator could not be worth living. In actual fact, long-time use of a ventilator is perfectly compatible with a good quality of life [Bach and Barnett, 1994]. However, even after the exaggerations have been removed, difficult ethical problems remain in the use of medical technology on severely ill patients. Just as there are occasions when permanent use of a ventilator can help a patient to a meaningful life, there are also occasions when the use of a ventilator will keep alive the body of a person whose brain does not function any more. The issue of futility, and what technological means are justified in the treatment of a severely ill person, is mainly a medical issue. The crucial criteria are the patient’s condition and prognosis, in particular her level of consciousness, and her own preferences as far as they can be known. However, there are also some technological aspects to this question. One such issue is the distinction between act and omission, and correspondingly between causing someone’s death actively and causing it by refraining from doing something (e.g. refraining from a therapeutic action that is considered to be futile). This distinction has crucial role in the debate on euthanasia, but it is nevertheless far from clear [Hansson, 2008]. Hence, a physician who withdraws a respirator from a terminal patient with no hope of recovery is often seen as (passively) permitting death to occur through natural causes. In contrast, a well-meaning friend or relative who disconnects the respirator would run much greater risk of being accused of killing the patient. It seems as if the distinction between killing and letting die depends on social conventions and role norms [Winkler, 1988]. The withdrawal of nutrition from a terminally ill patient seems to be particularly problematic. It is an important part of medical and nursing tradition that patients should be given basic care and comfort even when the progress of the disease cannot be prevented or delayed. This includes the provision of food and fluid. Therefore, some maintain that the terminally ill should be provided with nutrition and water, even if this has to be done by technological means rather than by feeding them and giving them to drink. Others are unwilling to extend the requirement to provide nutrition and hydration to cases when this can only be done with a nasogastric tube or intravenously [Winkler, 1988, p. 165]. The continued use of new advanced devices on terminally ill persons has sometimes been questioned. This applies in particular to left ventricular assist devices (LVAD) and total artificial hearts (TAH). Although originally intended as bridging devices, LVADs have been used as destination therapy with good results. Total artificial hearts are at the time of writing still essentially an experimental therapy. Consider a case when an LVAD has been implanted as a bridging device, but circumstances have changed so that transplantation is no longer an option. It could then be claimed that since the device is no longer medically indicated,

Philosophy of Medical Technology

1285

it can be turned off or removed. However, both of these actions are expected to hasten the death of the patient [Bramstedt and Wanger, 2001]. Switching off the device under such circumstances would be contrary to generally accepted ethical principles. The same problem arises, perhaps in more drastic form, for total artificial hearts. Katrina Bramstedt has claimed that “the fact that a TAH (or any other implant or assist device) is functioning without flaw is of no relevance to the futility discourse. What is relevant to these discussions is whether the ’perfectly’ functioning device is serving the goals of medicine and the best interests of the patient. Just as with a ventilator, a TAH can be functioning ’perfectly’, yet be ethically inappropriate.” Furthermore, she says that “[a]s with implantable defibrillators, inactivation of a TAH is a simple procedure not involving surgery, and this inactivation should not be seen as ethically separate from the withdrawal of other life support measures such as dialysis or ventilation” [Bramstedt, 2003]. A contrary view was expressed by Robert Veatch [2003], who claims that Bramstedt “appears to be endorsing unilateral actions by physicians that will directly cause the death of their patients and do so against the will of the patient or surrogate. That should be called ’murder’.” According to Veatch, “[t]hrowing a switch that stops a TAH is more like injecting a drug that paralyzes the heart muscle or like excising the SA node. Either of these would be considered direct, active killing. How can it be that turning off the heart is any different?” Whereas other authors have emphasized the similarity between turning off an artificial heart and discontinuing other life-prolonging treatment [Miles et al., 1988], Veatch emphasizes the difference. It could be argued in favour of his view that a patient who has received an artificial heart will regard it as her own, and thus not as a device that somebody else has a right to stop. Future technological developments may provide us with other types of lifesustaining devices that give rise to essentially the same type of questions as the artificial heart. This would apply, for instance, to an artificial lung or kidney. A somewhat different type of end-of-life issue would arise from a brain implant that is not necessary for life but necessary to support consciousness. If the quality of the achieved consciousness deteriorates, arguments could be made in favour of turning off such an implant. This would, however, be a highly problematic standpoint for same reason that turning off a life-sustaining artificial organ is problematic.

3.3

Subcultures that resist therapy

Medical technology has effects not only on individuals but also on social groups and on society as a whole. Radical improvements in treatment will change the situation of disabled subcommunities in our societies. Perhaps surprisingly, therapeutic improvements are not always received positively in these subcommunities. The “fat is beautiful” movement denies that obesity is a disease requiring treatment and medical attention. Segments of the dwarf community have reacted against the introduction of therapies against their condition, seeing this as a threat to the future existence of their way of life and their organizations [Berreby, 1996].

1286

Sven Ove Hansson

By far the strongest such counter-reaction is the criticism from the Deaf World of cochlear implant surgery in prelingually deaf children [Crouch, 1997; Lane and Bahan, 1998]. The criticism of cochlear implantation is associated with a positive view of deafness. The Danish Deaf Association has stated that “deaf children are not sick or weak children, but normal Danish children, who just happen to use another language” (quoted in [Nunes, 2001]). Members of the Deaf World reject the idea that they have an impairment or disability. Instead, they view themselves as a minority culture with its own language, customs, attitudes, knowledge, and values. The use of cochlear implants will lead to a drastic decline in the population of this minority culture. Deaf activist have often referred to the ethical principle that minority cultures should be preserved. They claim that large-scale implantation of children conflicts with the right of the Deaf language and cultural minority to exist and flourish. The term “genocide” has sometimes been used to describe that prospect [Lane and Bahan, 1998]. This claim has given rise to an interesting discussion about the definition of a minority culture and whether cultures have intrinsic value [Levy, 2002]. Critics have pointed out the problematic nature of arguments that give precedence to the preservation of a culture over the interests of individual children. Some have noted that it is difficult to draw the line if cochlear implants are disallowed for this reason. If cochlear implants are unethical, then how should we judge the rubella vaccine [Balkany, 1996]? From the viewpoint of mainstream medical ethics the interests of a subculture that needs to recruit new members could hardly prevail over the physician’s responsibility towards the individual patient. Nevertheless, there are important lessons to be drawn from this debate. In particular, it shows that the ethical discussion on medical technology must take into account the social and cultural notions of disease. 4 ENHANCEMENT TECHNOLOGY Technological devices such as implants can be constructed not only to cure disease and restore human functioning to normal levels, but also to improve human functioning to levels above the normal. The philosophy of medical technology therefore has to deal with issues of normality and disease and with the admissibility of human enhancement. If it becomes possible to improve a healthy person’s physical strength or her memory to levels above her natural endowment, to what extent is it advisable to do so?

4.1 Enhancement and the limits of normality Much of the recent debate on enhancement has referred to genetic enhancement, which only few writers defend [Resnik, 2000]. In this area, the enhancement discussion is anticipatory since no enhancing genetic technology is currently available.

Philosophy of Medical Technology

1287

However, there are at least two branches of medicine that already deal with enhancement in everyday clinical decisions, namely cosmetic surgery and neuropharmacology. Many types of cosmetic surgery, including breast implants, have been criticized for not complying with the aims of medicine, since they do not treat a disease or malfunction [Jacobson, 1998; Miller et al., 2000]. Several drugs developed to treat diseases of the nervous system also have the ability to improve normal functioning. Hence, drugs developed for the treatment of narcolepsy are already in use in armed forces as wakefulness drugs. Drugs against depression are used for mood elevation by persons with no psychiatric diagnosis, and drugs against erectile dysfunction are used for pleasure [Wolpe, 2000]. Drugs developed to prevent cognitive deterioration in Alzheimer’s disease seem to be capable of improving cognitive functioning in the healthy. In addition to enhancement of capabilities that we already have, it is also possible to develop entirely new functions for the human body. Currently, microchip devices are implanted in animals for identification purposes. It is technically possible to implant similar devices into humans. One use of such chips would be to let airplane passengers travel without a ticket or identity document; instead they would be scanned. A more sophisticated read-write chip could carry a person’s medical history or her criminal record. An implanted radio transmitter can be used to track a person [Ramesh, 1997]. A related prospect is that of implanting a device in the body that continuously monitors levels of substances in the bloodstream, and adjusts drug release accordingly [Wood et al., 2003]. Some authors are against virtually all forms of enhancement since it transcends the traditional task of medicine that is to treat and prevent diseases, not to improve humanity generally. “[T]he goals of medicine concern not all human suffering, but only that suffering connected with a malady” [Miller et al., 2000]. There are at least two problems with this standpoint. First, the distinction between disease and health or normality is not as clear as it may first seem. Disease is not a biologically well-defined concept but one that depends largely on social values. Some conditions previously regarded as diseases are now regarded as normal states of the mind or body. Other conditions that were previously regarded as variations within normality are now regarded as diseases. Homosexuality is an example of the former, attention-deficit hyperactivity disorder (ADHD) an example of the latter. Secondly, it is easy to show with examples that our intuitions about whether treatment should be offered for a condition are strongly influenced by other factors than whether or not that condition is classifiable as a disease. One well-known example is the treatment of short stature. Both public and private insurers have chosen to pay for growth hormone treatment only if the child has some diagnosable growth hormone deficiency, not otherwise regardless of how short it is projected to be [Verweij and Kortmann, 1997]. As was noted by Norman Daniels [2000], this criterion for treatment is difficult to defend from an ethical point of view. If one person is short “just” because of her genotype and another due to some identified dysfunction, this does not mean that the first person suffers less or needs treatment less. Clearly, neither of them is short through a choice or fault of her own.

1288

Sven Ove Hansson

(In practice, however, we have been saved from ethical predicaments of growth hormone therapy by studies showing that this treatment does not affect the final, adult height of children who have a normal endogenous production of the hormone [Murray, 2002].) Presbyopia is a normal feature of aging, and should therefore not be regarded as a disease. Nevertheless, we do not hesitate to treat this condition (mostly with eyeglasses). Hopefully, no one would try to prevent ophthalmologists from treating this or other age-related conditions of the eye. Now suppose that a remedy becomes available for age-related cognitive decline. It is a good guess that — perhaps after some initial hesitation — our attitude to such a treatment would be the same as to presbyopia. (Or would anyone say: “Just let grandmother become confused. It is not a disease, so although there is a treatment she should not take it. Treatments are only for diseases.”) We already endorse improvements of the immune system (vaccinations). Other ways to improve the body’s resistance against disease would probably find acceptance relatively easily. There are also situations in which improved cognitive function would be seen by most of us as an advantage, such as improved driving ability and improved ability of surgeons to operate [Whitehouse et al., 1997]. It is also interesting to compare our views on improvements of the teeth and of the skin. In the middle of the 19th century it was normal for nearly all an adult’s teeth to display signs of decay. At that time, the type of dental work that is now routine would have been seen as remarkable and perhaps even as ethically doubtful. Today, it is about as difficult to provide old people with skin that looks youthful as it was then to make their teeth look youthful. How will we react if future developments make wrinkled skin as avoidable as discoloured tooth stubs are today? These examples show that the disease/normality limit does not tell us what treatments are acceptable. However, there may still be other arguments against enhancement, arguments that do not depend on the distinction between disease and normality. One obvious such argument is that enhancements may have serious side effects. Hence, we can expect genetic enhancement to have unknown negative effects [Goering, 2000]. In one experiment, mice that were genetically engineered to improve their performance on learning tasks turned out to have greater sensitivity to pain [Wei et al., 2001]. Perhaps a method to improve memory will have psychological side effects since it prevents us from forgetting things we cannot bear to think about. “Who needs to remember the hours waiting in the Department of Motor Vehicles staring at the ceiling tiles, or to recall the transient amnesia following a personal trauma” [Wolpe, 2000]? Other side effects may follow from other types of enhancement. However, although this type of argument can be used against many methods of enhancement, it is not a decisive argument against enhancement as such. At the bottom line, the enhancement issue concerns what kinds of human beings there should be. Should future people be stronger and more intelligent than we are? A common, often religiously motivated view is that human nature has been given

Philosophy of Medical Technology

1289

to us and should not be changed. Others see considerable scope for improvement of the human race. In one of the few scholarly papers devoted to the issue, James Hudson maintains that to the extent that we can influence the innate natures of future people, we should make them intelligent and probably without a sexual drive or “any drive. . . other than a drive to rational thought and action in general” [Hudson, 2000]. Needless to say, this is a controversial standpoint. The issue what kind(s) of persons there should be is among the most difficult ones to deal with rationally in moral philosophy. The very basis for the discussion is insecure. What criteria should we use? Should we judge future persons by our own criteria, or by the criteria that we predict (and partly determine) them to have? (Population ethics that deals with how many persons there should be has similar difficulties.) Possibly, the best way to tackle issues of enhancement is to deal with them incrementally, judging each individual case on the basis of our current values without even trying to take future values into account. However, such incrementalism needs to be informed by a discussion about possible long-term developments. The following words of warning are worth taking into account: Whereas one can make the case that future generations should have the right to decide by themselves about their fate, it should be prevented that we enter a slippery slope towards ever greater manipulation of the human body, without medical necessity, and do so without having fully considered the consequences. [Altmann, 2001]

4.2

Making man-machines

Microprobes implanted into nervous tissue can create interfaces for communication between a patient’s nervous system and devices that replace or supplement a malfunctioning organ. Currently the most important of these neural interface implants are cochlear implants (see above, Section 3.3). Brain implants are also used for bladder control and for blocking tremors for instance in Parkinson’s disease. There are several other promising applications, including the control of epileptic seizures [Pereira et al., 2007]. Experiments have been performed with chips implanted in the brain or a peripheral nerve in order to control a wheelchair or other compensatory technology, or a prosthetic device such as a prosthetic hand [Warwick et al., 2003; 2007]. Research is being conducted on prosthetic vision for the blind, based on essentially the same principles as cochlear implants, namely that stimuli from technological sensors are relayed to the nervous system via a nerveimplant interface. Two major alternatives are being investigated for the placement of this interface, namely retinal chips and chips implanted in the visual cortex of the brain. Prosthetic vision is currently primarily developed in animal models, but preliminary testing on human volunteers has taken place [Bertschinger et al., 2008; http://www.bostonretinalimplant.org]. If efficient implantable brain chips become available, then they can be used for various forms of enhancement. It has been speculated that military applications

1290

Sven Ove Hansson

can come first, with the purpose of producing soldiers with enhanced abilities [Maguire and McGee, 1999; Altmann, 2001]. Some computer visionaries dream of a future in which many or all humans have implantable computer chips that connect them to sensors, assist their memory, and provide them with a variety of capacities. The “cyborgs”, cybernetic organisms, of science fiction that are mixtures of man and machine would then become reality [Behling, 2005]. Some authors have hailed this as a positive development, since cyborgs can become better than men [Haraway, 1991]. It has also been argued that such neural implants could in the future be used to scan, upload and transfer (the contents) of a mind. Computer-brain connections will then allow electronic communications with other similarly connected individuals in a way that may require that we radically reassess the boundaries between self and society. However, this is even more speculative than the idea of a cyborg. We do not know whether or not complex sensory impressions, feelings and thoughts, can be communicated in either direction through an implant [White, 1999].

5

ENABLING TECHNOLOGY

The extent to which persons with impaired bodily functions are forced to live their lives differently than other people depends not only on therapeutic technology but also to a large part on a variety of other technologies, from wheelchairs to computer interfaces, from hearing aids to garage doors. Since the 1970s, handicap activists have urged us to see handicap less as a medical problem than as a consequence of social exclusion that is often mediated by technology. This standpoint was well expressed by Alison Davis: [I]f I lived in a society where being in a wheelchair was no more remarkable than wearing glasses and if the community was completely accepting and accessible, my disability would be an inconvenience and not much more than that. It is society which handicaps me, far more seriously and completely than the fact that I have spina bifida. (Quoted in [Newell, 1999, p. 172].) It is important to observe the difference between a medical condition (such as being blind) and a social condition that it contributes to (such as being unable to read the newspaper). This can be expressed with the distinction between disability and handicap. Disability is an impairment of a bodily or mental function. Handicap is the presence of obstacles that persons with disabilities are subject to in society. Hence disability is inherent in the person, whereas handicap is a relation between a person and her environment [Amundson, 1992]. Technology with capacity to reduce the negative impact of having a disease or disability can be called enabling technology [Hansson, 2007c]. Leaving aside therapeutic technology that we have already treated, enabling technology can be divided into three categories: compensatory, assistive, and universal technology.

Philosophy of Medical Technology

5.1

1291

Compensatory and assistive technology

Compensatory technology is technology that replaces (fully or in part) a lost biological function by a new function of a general nature. Hence, whereas therapeutic technology reduces handicap by reducing disability, compensatory technology reduces handicap by providing new abilities that compensate for the disability. Some examples of compensatory technology are eyeglasses, hearing aids, speech synthesis systems, walking sticks, crutches, wheelchairs, orthotic appliances, ventilators, and equipment for total parental nutrition. Rehabilitation medicine that aims at replacing lost functions by new compensating ones makes much use of compensatory technology. Assistive technology makes it possible for the individual to perform a task or activity despite an (uncompensated) disability or lack of function. Assistive technology provides abilities of a more specialized nature than what compensatory technology does. Typical examples are knives that require less strength than standard kitchen knives, plates and dishes that do not slide on the table, appliances for dressing, toileting, and bathing, remote controls for doors, windows, and light switches, textphones and videophones for the speech and hearing impaired, reading machines for the blind, etc. Adaptive interfaces of software products have become an increasingly important form of assistive technology, both for private life and on workplaces. However, the adaption of software has often lagged behind other technologies. As one example of this, many colleges and universities have ensured that handicapped persons have access to their buildings, but have failed to give them full access to their electronic information [Grodzinsky, 2000]. Household robots that assist disabled and elderly persons in a variety of daily activities are an important new development [Erlen, 2003]. Compensatory technology provides the person with general-purpose functions that can be used also in unforeseen situations, whereas assistive technology only provides solutions for more limited tasks. Therefore compensatory technology is more enabling than assistive technology. Hence, having a prosthesis that replaces a lost arm in a number of different tasks appears to be preferable to having a series of assistive appliances with which each of these tasks can be performed with only one arm.

5.2

Universal technology

Universal technology is technology that is intended for general use, not only for persons with a specific disease or disability. Without being restricted to persons with a disability, technology can be adjusted so that it includes them among its potential users. The difference between assistive and (adjusted) universal technology is often social rather than (in a restricted sense of the word) technological. Hence, a ramp that is used to enter a building both walking and in wheelchairs is universal technology; a wheelchair ramp at the back of the building intended only for those who cannot use the stairs at the front is assistive technology.

1292

Sven Ove Hansson

In the development of new technologies, accessibility for disabled persons is seldom treated more than at best as a side issue. Therefore, improvement or deterioration in terms of accessibility is often an unintended side effect of developments that have been driven by other aims. It is not easy to determine if the general trends in technological development are in general positive or negative for accessibility. There seem to be contradictory trends. One positive trend is mechanization that gradually decreases the need for physical strength in most occupations. Another positive trend is digitalization, that makes information more easily convertible to formats that are accessible to blind and deaf people [Cornes, 1993; Coombs 2003]. Mobile phones have also turned out to be more important for many handicapped people than for persons without a major handicap. A negative trend is increasing intellectual requirements, particularly on workplaces, that seem to be a consequence of many new technologies. This often makes life more difficult for mentally disabled persons. Hence, tentatively it seems as if ongoing technological developments make life easier for physically disabled persons but more difficult for those who are mentally disabled. Appropriately adapted universal technology has the advantage over compensatory and assistive technology that it makes it possible for disadvantaged people to interact with the technological environment in the same way as others. As one example of this, if a machine — such as an elevator — has both visual and auditory signals, then both blind and deaf people can use it in the same way as people who see and hear. Similarly, if a heavy door is operated from a panel that is accessible from a wheelchair, then both walking and wheelchair-bound persons can open it in the same way. Therefore, universal technology is, as a general principle, superior to compensatory or assistive technology. It is therefore a plausible ethical standpoint that if a problem cannot be solved with therapeutic technology, then it should if possible be solved with universal technology, even if alternative solutions with compensatory or assistive technology are available. However, contrary to therapeutic and compensatory technology, universal technology is usually not subject to decisions in the healthcare sector but rather in other sectors of society. This is in all probability a major reason why universal technology has often lagged behind therapeutic and compensatory technology.

6

GENERAL EFFECTS OF TECHNOLOGY IN MEDICINE

Technology has often been talked about very sweepingly in discussions on healthcare. In this chapter we instead focused on the impact on specific technologies and technological practices. However, there are some issues that do not relate to particular technologies but rather to the more general use of technology in healthcare. We will treat three major such issues: how technology shifts responsibilities, what effects it has on the quality of care and human contact, and whether it gives rise to unsound and perhaps unnatural dependence on artificial devices.

Philosophy of Medical Technology

6.1

1293

Shifting responsibilities

There are several ways in which current technological developments move responsibility for healthcare away from its traditional locus, i.e. physicians and nurses. The responsibility of companies that deliver medical equipment increases with the complexity of the equipment. In hospitals, bioengineers and clinical engineers take over some of the responsibilities of physicians, such as the calibration of advanced treatments. A quite different trend is the transfer of complex and sometimes lifecritical equipment from the hospital to the patient’s home, which confers more responsibility on patients and their relatives. Finally, as complex decisions are “delegated” to machines, some responsibilities become more diffuse, and bound to machines rather than to humans. Here, we will look more closely at the two lastmentioned of these trends, beginning with the shift of responsibility to patients and their relatives. More and more patients receive treatments in their homes such as ventilator therapy and artificial nutrition through infusion pumps. These are treatments that were previously only administered in hospitals [Arras, 1994]. The increase in homecare is partly a response to patients’ preferences, partly a response to economic pressures. “The combination of psychological benefits with cost containment makes home care seem an irresistible option” [Lantos and Kohrmann, 1992] (cf. [Kun, 2001]). Communications technology has an important role as facilitator of this development. Telemedicine allows for monitoring and diagnostics at home by the means of medical sensors connected to a personal computer. Temperature measurement, oximetry, electrocardiography, blood pressure measurement, and auscultation are among the diagnostic procedures that can be performed from a distance [Dansky et al., 1999; Stanberry, 2000; Elger and Burr, 2000]. The administration of advanced diagnostic and therapeutic technology in homes has many advantages. When things go well in homecare, patients receive “the best of both worlds” [Arras, 1994], advanced medical treatment in the privacy of their own homes. Telemedicine in home care can be a way to ensure that access to healthcare is not limited by geographical location and ability to travel [Bauer, 2000; Elger and Burr, 2000]. However, technological homecare is not without its problems. For an increasing number of families, it has erased the boundaries between hospital and home, between intensive care unit and living room [Arras, 1994]. Sometimes parents and other relatives take over tasks that nurses perform only after taking special courses [Kirk, 2001]. Advanced technological home care with life-sustaining machines can place excessive burdens on family members, typically women, who live with a constant fear of failure. One of the most important ethical issues in home care is what tasks and responsibilities can and should be taken over by laypersons. “As home healthcare broadens to include traditionally hospital-based therapies, it is unclear whether traditional hospital norms, which place ultimate responsibility for decisions on professionals, or traditional home care norms, which place responsibility on parents, should apply” [Lantos and Kohrmann, 1992].

1294

Sven Ove Hansson

This can have negative social consequences. Homecare can make familiar domestic settings alien, and may confuse family roles. In comparison, hospitals can often allow patients greater autonomy, and therefore better preserve family relationships. Sometimes patients have a greater sense of privacy in hospitals than in homecare [Ruddick, 1994]. A patient’s dependence on a spouse or a parent can be problematic for the relationship [Kohrmann, 1994]. Studies have shown stress and psychological problems in families who care for ventilator-dependent children at home [Lantos and Kohrmann, 1992; Arras and Neveloff Dubler, 1994; Kirk, 2001]. The other major shift in responsibilities emanates from a general tendency to automatize more and more advanced functions. Decisions are “taken over by machines” so that no human is directly responsible for them at the point in time at which they are made. Another way to express this is that decisions are predetermined in decision support systems. Healthcare is often seen as one of the most promising areas for the introduction of computerized decision support. It has been shown in several cases that decision support systems can help the clinician in important ways, for instance by decreasing the risk of kidney failure, and providing more rapid treatment of critical laboratory abnormalities [Bates, 1997]. If a decision support system is connected to electronic patient records, it can include mechanisms for following up and for automated learning. Like other applications of artificial intelligence, an advanced clinical decision support system will therefore have capabilities in addition to those explicitly programmed into it. We may very well be approaching systems in which computers perform what we usually see as the tasks of physicians: making diagnoses, performing therapies, and communicating with patients [Gell, 2002]. A system has already been tried out in which diabetes patients used a touch-tone telephone to obtain self-management instructions and dosage decision support from a computer. The result was encouraging; an improvement was shown in their diabetes management [Albisser, 2001]Nevertheless, important questions can be raised about the implications of such systems. If the advice was wrong, how important is it whether the patient communicated with a machine or with a human being? How can responsibilities be assigned when decisions are taken over by machines [Klieglis et al., 1986; Hucklenbroich 1986]? Furthermore, what will the effects be on the physician–patient and nurse–patient relationships if much of the therapeutic-technical support comes from a machine whereas the psychological part of the support presumably stays with the physicians and nurses?

6.2 Technology, care and human contact One of the most important effects of enabling technologies is to facilitate human communication. Hearing aids, textphones, appliances for reading and writing, speech reading programs, and various technologies for physical mobility are all examples of this. However, technology can also be used to replace human contacts or reduce the need for them. A phone call from a nurse can replace a personal

Philosophy of Medical Technology

1295

visit. A central electronic monitoring system can supersede assistant nurses at the bedside, and a nasogastric tube can be used instead of spoon-feeding. In public debates, medical technology has often been accused of causing the dehumanization and depersonalization of healthcare and the objectification of patients. However, there is no inbuilt conflict between care and technology; technology can be used both in ways that improve care and in ways that make it less humane [Haber, 1986; Barnard and Margarete Sandelowski, 2001; Widdershoven, 2002]. In a balanced discussion on technology in healthcare it has to be realized both (1) that technology is not in itself dehumanizing and (2) that technology cannot replace genuine human contact and care. For a practical example, we can consider the proposal to use virtual environments for training stroke patients. Virtual technology can be used to expose these patients to a wider range of sensory stimuli, over much longer periods, than what is otherwise possible in a hospital setting. This can yield benefits in terms of time and cost of therapy to stroke patients, who typically spend only 30-60 min per day in formal therapy. Thus, virtual reality “increases the possibility of stimulation and interaction with the world without increasing demands on staff time” [Wilson et al., 1997]. However, potentially this technology can also be used to reduce individual, staff-to-patient contact. This is then a negative effect of the way in which the technology is used, not of the technology itself. Recently, attempts to replace human contact with technology have in fact been made through the therapeutic use of companion robots. These products have been developed in Japan, where there is less resistance to robots with human features than in most other parts of the world. Elderly patients are invited to interact with robots such as the robot baby seal Paro that reacts when one speaks to it or pets its fur, and the “healing partner” Yumel that looks like a baby boy, has a vocabulary of 1200 phrases, and sings lullabies. Patients tend to appreciate these robots, cuddle with them and talk to them. Some patients with age-related dementia do not realize that they are interacting with a machine [Sullins, 2005]. Replacing human contact in this way is obviously problematic from an ethical point of view. It is questionable whether it is compatible with human dignity to provide demented patients with technological devices that they wrongly believe to be living beings. However, on the other hand, removing these robots without replacing them with true human contacts is not necessarily an improvement.

6.3

The technological imperative

Resistance to technological medicine has a long history. Around the year 1900 there was a “neohippocratic” movement among doctors who saw scientific medicine as a threat to the old art of medicine. One of the most prominent members of this movement was Ernst Schweninger, Bismarck’s personal physician [Koch, 1985]. A much stronger such movement developed in the 1960s and 1970s as a counterreaction to the rapidly growing use of mechanical and electronic equipment in healthcare. In 1968 economist Victor Fuchs introduced the term “technological

1296

Sven Ove Hansson

imperative”, by which he meant the tendency to give the best care that is technically possible, even if its costs are high [Fuchs, 1968; Barger-Lux and Robert P. Heaney, 1986]. Much of the criticism of medical technology was couched in the term “medicalization”. This term was used (and possibly invented) in 1961 by T. Szasz who originally used it to describe the incorporation into psychiatry of problems that should not be dealt with as psychiatric or otherwise medical [Nye, 2003]. The term was adapted by Ivan Illich (1926-2002), the foremost critic of technological medicine in this period. Illich, who has been incorrectly credited with inventing the term [Barnet, 2003, pp. 276 and 286], was an ardent critic of scientific medicine and in particular its use of technology [Illich, 1975]. In later years, the form of anti-medical movement that he represented has been significantly weakened. Critics of medical technology have done great service to society by pointing out various problems in the use of this technology. However, much of their criticism is weakened by an (explicit or implicit) technological determinism: a belief that medical technology of necessity must have certain negative traits, such as being dehumanizing and standing in the way of good care. On the other hand, blind belief in the progress of medical technology can be equally dangerous.

BIBLIOGRAPHY [Albisser et al., 2001] A. Albisser, S C. Michael, S.C. En Chao, I.D. Parson, and M. Sperlich. Information technology and home glucose clamping. Diabetes technology and therapeutics, 3, 377-386, 2001. [Alper and Beckwith, 1988] J. S. Alper and J. Beckwith. Distinguishing Genetic from Nongenetic Medical Tests: Some Implications for Antidiscrimination Legislation. Science and Engineering Ethics, 4, 141-150, 1988. [Altenstetter, 2003] C. Altenstetter. EU and member state medical devices regulation. International Journal of Technology Assessment in Health Care, 19, 228-248, 2003. [Altmann, 2001] J. Altmann. Military Uses of Microsystem Technologies. Dangers and Preventive Arms Control. M¨ unster: agenda Verlag 2001. [Amundson, 1992] R. Amundson. Disability, handicap, and the environment. Journal of Social Philosophy, 23, 105-119, 1992. [Anderlik and Rothstein, 2001] M. R. Anderlik and M.A. Rothstein. Privacy and confidentiality of genetic information: What rules for the new science? Annual Review of Genomics and Human Genetics, 2, 301-433, 2001. [Arras, 1994] J. D. Arras. The technological tether. Hastings Center Report, Supplement 24, S1-S3, 1994. [Arras and Neveloff Dubler, 1994] J. D. Arras and N. Neveloff Dubler. Bringing the Hospital Home. Ethical and Social Implications of High-Tech Home Care. Hastings Center Report, Supplement 24, S19-S28, 1994. [Bach and Barnett, 1994] J. R. Bach and V. Barnett. Ethical considerations in the management of individuals with severe neuromuscular disorders. American Journal of Physical Medicine and Rehabilitation, 73, 134-140, 1994. [Balkany et al., 1996] T. Balkany, A. V. Hodges, and K. W. Goodman. Ethics of cochlear implantation in young children. Otolaryngology and Head and Neck Surgery, 114, 748-755, 1996. [Barash, 2000] C. I. Barash. Genetic Discrimination and Screening for Hemochromatosis: Then and Now. Genetic Testing, 4(2), 213-218, 2000. [Barger-Lux and Heaney, 1986] M. J. Barger-Lux and R. P. Heaney. For better and worse: The technological imperative in health care. Social Science and Med icine, 22, 1313-1320, 1986.

Philosophy of Medical Technology

1297

[Barnard and Sandelowski, 2001] A. Barnard and M. Sandelowski. Technology and human nursing care: (ir)reconcilable or invented difference? Journal of Advanced Nursing, 34, 367-375, 2001. [Barnet, 2003] R. J. Barnet. Ivan Illich and the Nemesis of Medicine. Medicine, Health Care and Philosophy, 6, 273-286, 2003. [Bates, 1997] D. W. Bates. Commentary: Quality, Costs, Privacy and Electronic Medical Data. Journal of Law, Medicine and Ethics, 25, 111-112, 1997. [Bauer, 2000] K. A. Bauer. The ethical and social dimensions of home-based telemedicine. Critical Reviews in Biomedical Engineering 28, 541-544, 2000. [Behling, 2005] L. L. Behling. Replacing the Patient: The Fiction of Prosthetics in Medical Practice. Journal of Medical Humanities, 26, 53-66, 2005. [Berreby, 1996] D. Berreby. Up with people: dwarves meet identity politics. New Republic 214(18), 14-19, 1996. [Bertschinger et al., 2008] D. R. Bertschinger, E. Beknazar, M. Simonutti, A. B. Safran, J. A. Sahel, S. G. Rosolen, S. Picaud, and J. Salzmann. A review of in vivo animal studies in retinal prosthesis research. Graefe’s archive for clinical and experimental ophthalmology, 246(11): 1505-17, 2008. [Booth, 1993] C. C. Booth. Clinical Research. In W. F. Bynum and R. Porter, eds., Companion Encyclopedia of the History of Medicine, pp. 205–299. Routledge 1993. [Bowman, 2000] J. E. Bowman. Technical, Genetic, and Ethical Issues in Screening and Testing of African-Americans for Hemochromatosis. Genetic Testing, 4(2), 207-212, 2000. [Brady, 1995] T. Brady. The Ethical Implications of the Human Genome Project for the Workplace. International Journal of Applied Philosophy, 10, 47-56, 1995. [Bramstedt, 2003] K. A. Bramstedt. Contemplating total artificial heart inactivation in case of futility. Death Studies 27, 295-304, 2003. [Bramstedt and Wanger, 2001] K. A. Bramstedt and N. S. Wanger. When withdrawal of lifesustaining care does more than allow death to take its course: The dilemma of left ventricular assist devices. Journal of Heart and Lung Transplantation, 20, 544-548, 2001. [B¨ uttner, 2002] J. B¨ uttner. Naturwissenschaftliche Methoden im klinischen Laboratorium des 19. Jahrhundert und ihr Einfluß auf das klinische Denken. Berichte zur Wissenschaftsgeschichte, 25, 93-105, 2002. [Coombs, 2003] N. Coombs. Liberation Technology. Equal Access Via Computer Communication 2003. http://ittimes.ucdavis.edu/spring2003/stories/past fall92 liberation.htm. [Cornes, 1993] P. Cornes. Impairment, Disability, Handicap and New Technology. In M. Oliver, ed. Social work. Disabled People and Disabling Environments, pp. 98-114. Jessica Kingsley Publishers 1993. [Crouch, 1997] R. A. Crouch. Letting the deaf be Deaf. Reconsidering the use of cochlear implants in prelingually deaf children. Hastings Center Report, 27, 14-21, 1997. [Dalsgaard Reventlow et al., 2006] S. Dalsgaard Reventlow, L. Hvas, and K. Malterud. Making the invisible body visible. Bone scans, osteoporosis and women’s bodily experiences. Social Science and Medicine, 62, 2720-2731, 2006. [Daniels, 2000] N. Daniels. Normal Functioning and the Treatment-Enhancement Distinction. Cambridge Quarterly of Healthcare Ethics, 9, 309-322, 2000. [Dansky et al., 1999] K. H. Dansky, K.H. Bowles, and L. Palmer. How telehomecare affects patients. Caring Magazine, 18, 10-14, 1999. [Davis, 1981] A. B. Davis. Medicine and its Technology. An Introduction to the History of Medical Instrumentation. Greenwood Press 1981. [Deber and Leatt, 1986] R. B. Deber and P. Leatt. The multidisciplinary renal team: who makes the decisions?. Health Matrix 4, 3-9, 1986. [Dolgin, 2001] J. L. Dolgin. Ideologies of Discrimination: Personhood and the ‘Genetic Group’. Studies in History and Philosophy of Science Part C: Studies in History and Philosophy of Biological and Biomedical Sciences, 32(4), 705-721, 2001. [Elger and Burr, 2000] C. E. Elger and W. Burr. Advances in Telecommunications Concerning Epilepsy. Epilepsia 41, Suppl 5, S9-S12, 2000. [Erlen, 2003] J. A. Erlen. Technology. Possibilities and Pitfall. Orthopaedic Nursing, 22, 310-313, 2003. [Evidence-based Medicine Working Group, 1992] Evidence-Based Medicine Working Group. Evidence-based medicine. A new approach to teaching the practice of medicine. JAMA, 268, 2420-2425, 1992.

1298

Sven Ove Hansson

[Feinstein, 1967] A. R. Feinstein. Clinical Judgement. Krieger 1967. [Feinstein, 1996] A. R. Feinstein. Two Centuries of Conflict-Collaboration Between Medicine and Mathematics. Journal of Clinical Epidemiology, 49, 1339-1343, 1996. [Fielder, 1991] J. H. Fielder. Ethical issues in biomedical engineering: The Bjork-Shiley heart valve. IEEE Engineering in Medicine and Biology, 10, 76-78, 1991. [Fuchs, 1968] V. R. Fuchs. The growing demand for medical care. New England Journal of Medicine, 279(4), 190-195, 1968. [Gell, 2002] G. Gell. Safe, controllable technology? International Journal of Medical Informatics, 66, 69-73, 2002. [Gin, 1997] B. R. Gin. Genetic discrimination: Huntington’s disease and the Americans with Disabilities Act. Columbia Law Review, 97, 1406-1434, 1997. [Goering, 2000] S. Goering. Gene Therapies and the Pursuit of a Better Human. Cambridge Quarterly of Healthcare Ethics, 9, 330-341, 2000. [Green and Botkin, 2003] M. J. Green and R.J. Botkin. ‘Genetic exceptionalism’ in medicine: Clarifying the differences between genetic and nongenetic tests. Annals of Internal Medicine 138(7), 571-575, 2003. [Grodzinsky, 2000] F. S. Grodzinsky. Equity of Access: Adaptive Technology. Science and Engineering Ethics, 6, 221-234, 2000. [Haber, 1986] P. A. L. Haber. High Technology in Geriatric Care. Clinics in Geriatric Medicine 2, 491-500, 1986. [Hansson, 2005] S. O. Hansson. Privacy, Discrimination, and Inequality in the Workplace. In S. O. Hansson and E. Palm, eds., The Ethics of Workplace Privacy, pp. 119135. Peter Lang 2005. [Hansson, 2006] S. O. Hansson. Uncertainty and the Ethics of Clinical Trials. Theoretical Medicine and Bioethics, 27, 149-167, 2006. [Hansson, 2007a] S. O. Hansson. What is Technological Science? Studies in History and Philosophy of Science 38, 523-527, 2007. [Hansson, 2007b] S. O. Hansson. Praxis Relevance in Science. Foundations of Science, 12, 139154, 2007. [Hansson, 2007c] S. O. Hansson. The Ethics of Enabling Technology. Cambridge Quarterly of Healthcare Ethics, 16, 257-267, 2007c. [Hansson, 2008] S. O. Hansson. Three Bioethical Debates in Sweden. Cambridge Quarterly of Healthcare Ethics, 17, 261-269, 2008. [Haraway, 1991] D. Haraway. A Cyborg Manifesto: Science, Technology, and Socialist-Feminism in the Late Twentieth Century. In D. Haraway, ed., Simians, Cyborgs and Women: The Reinvention of Nature, pp 148-181. Routledge 1991. [Hofmann, 2003] B. Hofmann. Medicine as Techne – A Perspective from Antiquity. Journal of Medicine and Philosophy, 28, 403-425, 2003. [Hucklenbroich, 1986] P. Hucklenbroich. Automatisation and responsibility. Theoretical Medicine 7, 239-242, 1986. [Hudson, 2000] J. Hudson. What Kinds of People Should We Create? Journal of Applied Philosophy, 17, 131-143, 2000. [Illich, 1975] Illich, Ivan Medical nemesis: the expropriation of health. Calder and Boyars 1975. [Jacobson, 1998] N. Jacobson. The Socially Constructed Breast: Breast Implants and the Medical Construction of Need. American Journal of Public Health, 88, 1254-1261, 1998. [Kirk, 2001] S. Kirk. Negotiating lay and professional roles in the care of children with complex health care needs. Journal of Advanced Nursing, 34, 593-602, 2001. [Klieglis et al., 1986] U. Klieglis, A.C. Renirie, and J. Schaefer. Medicus Technologicus. Reflections on the conflict between the physician’s responsibility in decision-making and medicotechnical automation. Theoretical Medicine, 7, 233-238, 1986. [Koch and Lacquer, 1985] R. Koch and N. B. Laqueur. Schweninger’s Seminar, Journal of Contemporary History, 20, 757-779, 1985. [Kohrmann, 1994] A. F. Kohrmann. Chimeras and Odysseys. Toward Understanding the Technology-Dependent Child. Hastings Center Report, Supplement 24, S4-S6, 1994. ¨ [Kr¨ oner, 2005] H.-P. Kr¨ oner. Außere Form und Innere Krankheit: Zur klinischen Fotografie im sp¨ aten 19. Jahrhundert. Berichte zur Wissenschaftsgeschichte, 28, 123-134, 2005. [Kun, 2001] L. G. Kun. Telehealth and the global health network in the 21st century. From homecare to public health informatics. Computer Methods and Programs in Biomedicine, 64, 155-167, 2001.

Philosophy of Medical Technology

1299

[Lane and Bahan, 1998] H. Lane and B. Bahan. Ethics of cochlear implantation in young children: A review and reply from a Deaf-World perspective. Otolaryngology and Head and Neck Surgery, 119, 297-313, 1998. [Lantos and Kohrmann, 1992] J. D. Lantos and Arthur F. Kohrmann. Ethical aspects of pediatric home care. Pediatrics, 89, 920-924, 1992. [Launis, 2000] V. Launis. The Use of Genetic Test Information in Insurance: The Argument from Indistinguishability Reconsidered. Science and Engineering Ethics, 6, 299-310, 2000. [Levy, 2002] N. Levy. Reconsidering cochlear implants: The lessons of Martha’s Vineyard. Bioethics, 16, 134-153, 2002. [Magnus et al., 1996] J. H. Magnus, R. M. Joakimsen, G. K. Berntsen, A. Tollan and A. J. Søgaard. What do Norwegian women and men know about osteoporosis? Osteoporosis International, 6, 31-36, 1996. [Maguire and McGee, 1999] G. Q. Maguire and E.M. McGee. Implantable brain chips? Time for debate. Hastings Center Report, 29, 7-13, 1999. [Miles et al., 1988] S. H. Miles, M. Siegler, D. L. Schiedermayer , J.D. Lantos, and J. La Puma. The total artificial heart. An ethics perspective on current clinical research and deployment. Chest 94, 409-413, 1988. [Miller et al., 2000] F. G. Miller, H. Brody, and K.C. Chung. Cosmetic Surgery and the Internal Morality of Medicine. Cambridge Quarterly of Healthcare Ethics, 9, 353-364, 2000. [Mol, 2000] A. Mol. What Diagnostic Devices Do: The case of blood sugar measurement. Theoretical Medicine and Bioethics, 21, 9-22, 2000. [Moore, 2000] A. D. Moore. Owning genetic information and gene enhancement techniques. Bioethics, 14, 97-119, 2000. [Murray, 2002] T. H. Murray. Reflections on the ethics of genetic enhancement. Genetics in Medicine 4(6 Suppl), 27S-32S, 2002. [Newell, 1999] C. Newell. The social nature of disability, disease and genetics: a response to Gillam, Persson, Holtug, Draper and Chadwick. Journal of Medical Ethics, 25, 172-175, 1999. [Nunes, 2001] R. Nunes. Ethical dimension of paediatric cochlear implantation. Theoretical Medicine, 22, 337-349, 2001. [Nye, 2003] R. A. Nye. The evolution of the concept of medicalization in the late twentieth century. Journal of History of the Behavioral Sciences, 39(2), 115-129, 2003. [Pereira et al., 2007] E. A. Pereira, A. L. Green, D. Nandi, and T. Z. Aziz. Deep brain stimulation: indications and evidence. Expert Review of Medical Devices, 4(5), 591-603, 2007. [Persson and Hansson, 2003] A. P. Persson and S. O. Hansson. Privacy at Work – Ethical Criteria. Journal of Business Ethics, 42, 59-70, 2003. [Ramesh, 1997] E. M. Ramesh. Time Enough? Consequences of Human Microchip Implantation. Risk, 8, 373, 1997 [Resnik, 2000] D. B. Resnik. The Moral Significance of the Therapy-Enhancement Distinction in Human Genetics. Cambridge Quarterly of Healthcare Ethics, 9, 365-377, 2000. [Rhodes et al., 1999] L. A. Rhodes, C. A. McPhillips-Tangum, C. Markham, and R. Klenk. The power of the visible: the meaning of diagnostic tests in chronic back pain. Social Science and Medicine, 48,1189-1203, 1999. [Roche and Annas, 2001] P. A. Roche, P. A. and G. J. Annas. Protecting genetic privacy. Nature Reviews Genetics, 2(5), 392-396, 2001. [Ruddick, 1994] W. Ruddick. Transforming Homes and Hospitals. Hastings Center Report, Supplement 24, S11-S14, 1994. [Saha et al., 1985] S. Saha, S. Misra, and P. S. Saha. Bioengineers, health-care technology and bioethics. Journal of Medical Engineering and Technology, 9, 55-60, 1985. [Saha and Saha, 1997] S. Saha and P. S. Saha. Biomedical Ethics and the Biomedical Engineer: A Review. Critical Reviews in Biomedical Engineering, 25, 163-201, 1997. [Silvers and Stein, 2002] A. Silvers and M. A. Stein. An equality paradigm for preventing genetic discrimination. Vanderbilt Law Review 55, 1341-1499, 2002. [Stanberry, 2000] R. Stanberry. Telemedicine: barriers and opportunities in the 21st century. Journal of Internal Medicine, 247, 615-628, 2000. [Sullins, 2005] J. P. Sullins. Building the Perfect Companions: The Humane Design of Personal Robotics Technologies. Manuscript, 2005. [Veatch, 2003] R. M. Veatch. Inactivating a total artificial heart: special moral problems. Death Studies, 27, 305-315, 2003.

1300

Sven Ove Hansson

[Verweij and Kortmann, 1997] M. Verweij and F. Kortmann. Moral assessment of growth hormone therapy for children with idiopathic short stature. Journal of Medical Ethics, 23, 305309, 1997. [Vos and Willems, 2000] R. Vos and D. L. Willems. Technology in Medicine: Ontology, Epistemology, Ethics and Social Philosophy at the Crossroads. Theoretical Medicine and Bioethics, 21, 1-7, 2000. [Warwick et al., 2003] K. Warwick, M. Gasson, B. Hutt, I. Goodhew, P. Kyberd, B. Andrews, P. Teddy, and A. Shad. The Application of Implant Technology for Cybernetic System. Archives of Neurology, 60, 1369-1373, 2003. [Warwick et al., 2007] K. Warwick, M. N. Gasson, and A. J. Spiers. Therapeutic potential of computer to cerebral cortex implantable devices. Acta Neurochirurgica. Supplement, 97(2), 529-535, 2007. [Wei et al., 2001] F. Wei, G. D. Wang, G. A. Kerchner, S. J. Kim, H. M. Xu, Z. F. Chen, and M. Zhuo. Genetic enhancement of inflammatory pain by forebrain NR2B overexpression. Nature Neuroscience, 4(2), 164-169, 2001. [White, 1999] R. J. White. Brain chip: Postpone the debate. Hastings Center Report, 29, 4, 1999. [Whitehouse et al., 1997] P. J. Whitehouse, E. Juengst, M. Mehlman , and T. H. Murray. Enhancing cognition in the mentally intact. Hastings Center Report, 27(3), 14-22, 1997. [Widdershoven, 2002] G. Widdershoven. Technology and care from opposition to integration. In C. Gastmans, ed., Between Technology and Humanity. The Impact of Technology on Health Care Ethics, pp. 1262-1282. Leuven University Press 2002 [Wilkinson, 1993] L. Wilkinson. Epidemiology. In W. F. Bynum and R. Porter, Companion Encyclopedia of the History of Medicine, pp. 1262–1282. Routledge 1993. [Willems, 2000] D. Willems. Managing one’s body using self-management techniques: practicing autonomy. Theoretical Medicine and Bioethics 21, 23-38, 2000. [Wilson, 1977] P. N. Wilson. Nigel Foreman, and Dana¨e Stanton. Virtual reality, disability and rehabilitation. Disability and Rehabilitation, 19, 213-220, 1997. [Winkler, 1988] E. R. Winkler. Foregoing Treament: Killing vs. Letting Die and the Issue of Non-feeding. Pp. 155-171 in James En Thornton, Ethics and Aging: The Right to Live, The Right to Die, University of British Columbia Press 1988. [Wolf, 1995] S. M. Wolf. Beyond ‘Genetic Discrimination’: Toward the Broader Harm of Geneticism. Journal of Law, Medicine and Ethics, 23, 345-353, 1995. [Wolpe, 2000] P. R. Wolpe. Treatment, enhancement, and the ethics of neurotherapeutics. Brain and Cognition, 50, 387-395, 2000. [Wood, 2002] J. Wood. The role, duties and responsibilities of technologists in the clinical laboratory. Clinica Chimica Acta 319, 127-132, 2002. [Wood et al., 2006] S. Wood, R. Jones, and A. Geldart. The Social and Economic Challenges of Nanotechnology. Swindon: Economic and Social Research Council 2003. [Wooton, 2006] D. Wootton. Bad Medicine. Oxford University Press 2006.

PHILOSOPHY OF BIOTECHNOLOGY Henk van den Belt The great discoveries in molecular and cellular biology [. . . ] have provided us with a definitive understanding of the nature of life. Earlier in this century splendid discoveries in physics and chemistry provided us with a definitive understanding of the nature of matter. From that understanding has come the technology to reshape the inanimate world to human purpose. And many are less than pleased with the consequences. Now the description of life in molecular terms provides the beginning of a technology to reshape the living world to human purpose, to reconstruct our fellow life forms — each, as we are, the product of three billion years of evolution — into projections of the human will. And many are profoundly troubled by the prospect. [Sinsheimer, 1976] 1

1.1

INTRODUCTION: DEFINITIONS AND DISTINCTIONS

What is (modern) biotechnology?

Definitions are hardly ever innocuous. This holds for such contested concepts like ‘democracy’ and ‘freedom’, but also for ‘biotechnology’. For the meaning of the latter term we can turn to a standard lexicographic reference like the Concise Oxford Dictionary. The tenth edition from 1999 gives the following description: “the exploitation of biological processes for industrial and other purposes, especially the genetic manipulation of micro-organisms for the production of antibiotics, hormones, etc.” This definition might satisfy some but by no means all who hold a definite view on the meaning of ‘biotechnology’. The first part of the definition is so broad as to include agriculture, an endeavour in which humankind has been engaged since the Neolithic age. The second part centres on ‘genetic manipulation’ and may be thought unduly restrictive insofar as it limits this type of operation to micro-organisms and gives only an indication of the useful products that can be obtained. It is not only micro-organisms but also higher organisms like plants and animals that are being genetically ‘manipulated’ — by the way, due to its negative connotations the term ‘manipulation’ is nowadays usually avoided by the proponents of this technology and, for the same reason, eagerly embraced by its opponents. Proponents prefer to use the more neutral designation ‘genetic modification’ or ‘genetic engineering’. Not just the substantive definition but even the terminology to be used is an issue. Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

1302

Henk van den Belt

The Concise Oxford Dictionary entry does not deviate from common practice. In many other instances, the term ‘biotechnology’ is simultaneously used in a very broad and a more narrow sense. Thus in 1984 the US Office of Technology Assessment (OTA) offered the following broad definition of biotechnology: “Any technique using organisms and their components to make products, modify plants and animals to carry desired traits, or develop micro-organisms for specific uses”. In this sense, people have practised biotechnology from prehistoric times onwards simply by growing plants and raising livestock for food, by modifying them through breeding and selection, and also by baking bread, brewing beer, making wine, cheese, yoghurt and soy sauce — although they did not know they were in effect using micro-organisms to prepare those products. The OTA has also provided a more narrow definition of what it considered to be modern biotechnology: “‘New’ biotechnology is the industrial use of recombinant DNA, cell fusion and novel bioprocessing techniques.” [Office of Technology Assessment, 1984, p. 3]. Similarly, a European policy report also offers a definition covering both ‘modern’ and ‘traditional’ biotechnology: Modern biotechnology can be defined as the use of cellular, molecular and genetic processes in production of goods and services. Its beginnings date back to the early 1970s when recombinant DNA technology was first developed. Unlike traditional biotechnology which includes fermentation and plant and animal hybridisation — modern biotechnology involves a different set of technologies, including industrial use of recombinant DNA, cell fusion and tissue engineering amongst others. [Zika et al., 2007, p. 5]. It is not difficult to understand why recombinant DNA technology in particular is seen as a key component of modern biotechnology. This technology comprises a set of procedures by which segments of DNA from different organisms can be ‘cut’ and ‘pasted’ together and ‘inserted’ into a recipient organism. The cutting is done with a special category of enzymes called restriction enzymes; the resulting fragments can be pasted together with another type of enzymes called ligases; insertion can be achieved by using viruses, plasmids (circular pieces of DNA), plant-invading bacteria like Agrobacterium tumefaciens which can be used as Trojan horses, and techniques such as micropropulsion (gene gun) and microinjection. After the possibilities of restriction enzymes and ligases and the use of plasmids and viruses as insertion vectors first became known in the late 1960s and early 1970s, it did not take long before the potential of genetic engineering or genetic manipulation (‘gene splicing’) was widely recognized [Cohen, 1975]. James Watson, the doyen of molecular biology, sees recombinant DNA technology as similar to what a word processor can achieve: to cut, paste, and copy DNA [Watson, 2003, p. 84]. Before the development of PCR (Polymerase Chain Reaction) in the 1980s, the ‘copying’ of DNA (also called bacterial ‘cloning’) was done by the bacterial cell in which the piece of DNA has been inserted and which by its own self-replication also produces multiple identical copies of this particular piece of DNA. In Watson’s view, the

Philosophy of Biotechnology

1303

emergence of recombinant DNA technology represents “phase 2 of the molecular biology revolution” (“phase 1” being the descriptive analysis of the working of DNA in the cell). For once, Watson agrees with biotech’s long-term critic, Jeremy Rifkin, who compares the significance of this new technology with the importance of the discovery of fire: “This was no ordinary advance in lab techniques. Scientists were suddenly able to tailor DNA molecules, creating ones that had never before been seen in nature. We could “play God” with the molecular underpinning of all life. This was an unsettling idea to many people” [Watson, 2003, p. 84]. There are (or were) other conceptions of modern biotechnology, however, in which genetic engineering occupies a much less prominent place. In 1979, several years after the advent of recombinant DNA technology, a policy document prepared for the European Commission presented the following description of the subject of biotechnology: The meaning that is most widely accepted is that it is the industrial processing of materials by micro-organisms and other biological agents to provide desirable products and services. It incorporates fermentation and enzyme technology, water and waste treatment, and some aspects of food technology. Its scope and potential therefore are enormous . . . and with the advent of techniques of genetic engineering, the modification of such micro-organisms can be added to this list. Nor should ‘industrial processing’ exclude the application of biological science to agriculture and other non-industrial fields. [FAST/ACPM/79/14-3E, 1979, p. 3; quoted in Bud, 1994, p. 161] Here, genetic engineering or modification is mentioned almost as an afterthought, as an extra set of techniques to be added to the long list already held to make up biotechnology. This view of biotechnology was the outgrowth of a vision previously propounded by the German chemical industry which put special emphasis on the integration of microbiology and biochemistry with technical chemistry and process engineering [Buchholz, 2007]. Well into the 1980s, such alternative definitions continued to contest the hegemony of the view that the core of modern biotechnology consisted of recombinant DNA techniques, until the latter definition finally gained the upper hand in national and international forums. In this chapter I will follow current usage that sees genetic engineering as the central component of (modern) biotechnology, but I will also attend to the sometimes subtle shifts in meaning this designation has undergone in the course of time. The terms we use may be informed by our views on the predominance of continuous, incremental change or the occurrence of radical ruptures in the history of science and technology (think of the ‘Scientific Revolution’ or the ‘Industrial Revolution’). Alternatively, the vocabulary used may also tend to prejudge such issues.

1.2

The twentieth-century ‘prehistory’ of modern biotechnology

The English historian of science Robert Bud has traced the historical vicissitudes of the term ‘biotechnology’ (and of its foreign counterparts like Biotechnik and

1304

Henk van den Belt

Biotechnologie) in different countries throughout the 20th century and the various meanings that in the course of many decades have been attributed to this term. His monograph on the subject, The Uses of Life, bears the subtitle A History of Biotechnology [Bud, 1994]. However, if we consider genetic engineering a key component of (modern) biotechnology, it would be more adequate to view Bud’s work as a major contribution to the prehistory of biotechnology or, alternatively, to the history of the term ‘biotechnology’. What remains to be determined is how this prehistory, or the earlier projects and activities that have been dignified with the term ‘biotechnology’ or its equivalents, relates to biotechnology in the current sense. The word ‘biotechnology’ (or rather the German word ‘Biotechnologie’) was coined in 1919 by the Hungarian economist Karl Ereky to denote a new branch of knowledge based on the interaction of biology and engineering that aimed to deal with production processes in which raw materials are converted into end products with the aid of living organisms [Bud, 1994, pp. 32-37]. Animal husbandry was one of the production processes that Ereky had especially in mind when he invented the word: he described pigs in large-scale factory farms as ‘biotechnologische Arbeitsmachinen’ or ‘biotechnological labour machines’ [ibid., p. 34]. The new term was picked up by others and subsequently given new connotations and meanings. It thus embarked on an international career across countries like Hungary, Germany, Denmark, the United States, Japan, Sweden and the United Kingdom. Travelling from country to country, the word acquired new meanings with every passage. It became associated with fermentation technology and new combinations of microbiology and chemical engineering, with attempts to find new applications for agricultural raw materials such as energy and biofuels, but was also invested with visionary expectations about industrial renewal and the creation of a new harmony between man and nature. The prospect of replacing mechanical and chemical industries with a new type of economic activity, based on working with life itself, often induced utopian vistas. It would not be correct, however, to see the word ‘biotechnology’ as an empty container to be filled with any content that any actor might be interested to pour in. Bud presents it as a ‘boundary object’, which combines a central core meaning with enough semantic flexibility to be open to different interpretations by different parties [Bud, 1991, p. 419]. The central meaning is suggested by the word itself and refers to the frontier between biology and technology or between life and engineering. Throughout its history, this boundary has been mobile and contested [Bud, 1994, p. 220]. Modern biotechnology may have inherited some of the elements described in Bud’s prehistory. Industrial or ‘white’ biotechnology, in particular, which deals with the use of genetically modified micro-organisms and tailored enzymes in industrial processes, can be seen as a continuation of fermentation technology and earlier combinations of microbiology and chemical engineering. Another legacy concerns the social matrix of expectations in which modern biotechnology became embedded and which had been prepared by its historical precursors. Promises of an

Philosophy of Biotechnology

1305

environmentally benign and sustainable mode of production, nurtured by the expectation of a coming ‘Bio-society’, have been taken over by modern biotechnology. It is also not too difficult to see current scenarios for a future ‘Knowledge-Based Bio-Economy’ [OECD, 2001; European Union, 2007], with their radiant hopes for an ‘eco-efficient’ and sustainable production, as a continuation of earlier utopian projections.

1.3

Ancient and modern biotechnology

When the historical pedigree of (modern) biotechnology is discussed, it is mostly not its twentieth-century precursors to which attention is directed. Many biotech companies, industry organizations and research institutes present ‘timelines’ on their websites that afford a quick overview of the development of biotechnology. Almost invariably, the timeline starts about 8 to 10 millennia ago with the domestication of crops and livestock, and then moves on to the ancient Sumerians, Egyptians and Chinese who practised fermentation by using bacteria and yeast to make bread, wine, beer, cheese, yoghurt and soy products. The message is that there is nothing new under the sun. The Center for Integrated Biotechnology of Washington State University states it explicitly on its website: A common misconception among teachers is the thought that biotechnology includes only DNA and genetic engineering. To keep students abreast of current knowledge, teachers sometimes have emphasized the techniques of DNA science as the ‘end-and-all’ of biotechnology. This trend has also led to a misunderstanding in the general population. Biotechnology is NOT new. Man has been manipulating living things to solve problems and improve his way of life for millennia. Early agriculture concentrated on producing food. Plants and animals were selectively bred, and microorganisms were used to make food items such as beverages, cheese, and bread. [Center for Integrated Biotechnology, n.d.; my italics]. One advantage of the broad definition of biotechnology is that the most recent products of (modern) biotechnology can be presented in the reassuring light of familiarity. Thus Monsanto proclaimed in one of its newsletters: “Farmers Discovered Biotechnology 10,000 Years Ago: It’s Getting Better With Age”. Another advantage has to do with the legal regulation of the new technology. If biotechnology is just business as usual, then there is no need for special legislation to deal with its potential risks and hazards. This particular ‘framing’ of biotechnology triumphed in the United States, where no new laws were enacted, regulation focused on the final product and the process of genetic engineering was declared irrelevant from a regulatory point of view. For a variety of reasons, Europe would eventually opt for the diametrically opposed ‘process frame’, thus giving rise to regulatory polarization on a global scale and disputes within the World Trade Organization [Bernauer, 2003; Jasanoff, 2005a; Meijer and Stewart, 2004; Drezner 2007].

1306

Henk van den Belt

There are no definitions that are simply correct or incorrect, only definitions that are more or less pragmatically adequate in view of the aims one pursues. If one thinks it a little exaggerated to portray Stone Age farmers as the first biotechnologists, one might choose the narrow definition of ‘biotechnology’ that puts the emphasis on the use of techniques intervening at the molecular level of manipulation. Even so, one may feel compelled to add the seemingly qualifying epithet ‘modern’ so as not to be misunderstood. Dictionaries also follow this dual strategy. What is at issue in these definitional quarrels is the appraisal of genetic engineering as either an extension of age-old practices or a radically new departure in our dealings with the living world. Such a fundamental issue cannot be settled by definitional fiat.

2

THE ‘INFORMATISATION’ OF LIFE

2.1 From molecular biology via biotechnology to synthetic biology Ever since Watson and Crick elucidated the structure of DNA in 1953, molecular biology has continued to think about life in ‘informational’ terms, as is testified by the ubiquitous use of expressions such as ‘genetic code’, ‘translation’, ‘transcription’, ‘editing’, ‘expression’, ‘messenger RNA’ and the like [Yoxen, 1983, 44-45] (see also [Maynard Smith, 1999; Godfrey-Smith, 2006]).1 The famous Central Dogma of the discipline is also formulated in terms of information flows: sequence information is transferred from DNA to RNA to proteins, but never back from proteins to nucleic acids [Crick, 1970]. It is by no means unusual to compare an organism’s DNA with the software of a computer. From the 1950s onwards, the ‘informatisation’ and ‘digitisation’ of living organisms has gone on relentlessly, to the extent that the informational view of life that is implicit in molecular biology is increasingly being realized in technological forms. The advent of biotechnology made all life forms effectively re-programmable, because the cut-and-paste tricks of recombinant DNA technology provided a “molecular editing system” [Watson, 2004, p. 84]. Jeremy Rifkin, a well-known critic of biotechnology, speculated in the early 1980s that future generations would inhabit a world in which nature was “no longer something they are born into but rather something they program” [Rifkin, 1984, p. 23]. He expected the computer and information sciences to be1 In fact, the ‘informational’ view already dates from before the elucidation of the DNA structure. In 1943, the famous quantum physicist Erwin Schr¨ odinger unfolded his interesting ideas on ‘What is life?’ and speculated that the chromosomes “contain in some kind of code-script the entire pattern of the individual’s future development and its functioning in the mature state” [Schr¨ odinger, 1944, p. 20]. A substance in the form of an a-periodic solid would admit a nearly infinite number of arrangements, sufficient for a “highly complicated and specified plan of development” [ibid., p. 62] — Schr¨ odinger compared the endless number of possibilities with that of a simple Morse code. His elegant booklet influenced many pioneers of molecular biology. As James Watson wrote in retrospect, “Schr¨ odinger’s book was tremendously influential” [Watson, 2004, p. 35].

Philosophy of Biotechnology

1307

come “the means of communication humankind will use to reorder living material in the biotechnical age” [ibid., p. 21]. With the advent of biotechnology the gene pool of the entire microbial, plant and animal kingdom on Earth suddenly turned into a ‘Bestand ’ or standing-reserve (Heidegger), a collection of ‘genetic resources’ on which the genetic engineers could draw at will and which they could combine in all possible ways to create novel organisms.2 Under the biotechnological ‘gaze’ living organisms appear as the accidental (and in principle interchangeable) material repositories of valuable genetic ‘information’. This particular ‘gaze’ also manifests itself in the manner in which organisms are considered in patent law. Thus the European Directive on the Legal Protection of Biotechnological Inventions (Directive 98/44/EC) gives the following definition of the key object of protection: “‘biological material’ means any material containing genetic information and capable of reproducing itself or being reproduced in a biological system” [art. 2.1. sub a]. Not just isolated genes and cells but complete plants and animals can be brought under this definition, which reduces them to the status of raw material or carrier of genetic information. (There would be no reason to exclude human beings, were it not for the fact that by way of exception they have been declared unpatentable — human genes, however, insofar as they have been isolated and purified, can be patented!) The Convention on Biological Diversity provides another illustration of turning the infinite variety of life on Earth into a collection of resources. Article 2 of the Convention defines the key notion: “‘Genetic resources’ means genetic material of actual or potential value” (italics mine). Thanks to biotechnology, any genetic resource, each and every seemingly stupid plant in the tropical forest, has potential value in principle. Whether or not this potential value will actually be realized is another matter, of course. Nonetheless, the changed perception of the value of biological diversity, largely as a consequence of the rise of biotechnology, has led to attempts especially by developing countries to acquire more control over this potential source of wealth (see also [Raustiala and Victor, 2004]). It is not without reason that the Convention on Biological Diversity abandons the idea of the common heritage of humankind and proclaims the principle of national sovereignty over genetic resources — as if they could be controlled in the same way as mineral resources. However, the process of ‘informatisation’ and ‘digitisation’ did not stop with biotechnology. The Human Genome Programme and the parallel projects starting in the 1990s to sequence various non-human genomes gave a further impetus to this trend. These sequencing programmes would have been impossible without the new marriage between biology and information science that was forged in the new inter-disciplinary field of bioinformatics or computational biology. In a review article two representatives of this new discipline, Minoru Kanehisa and Peer Bork, write that its emergence and growth would have been unthinkable without the 2 See Søren Riis about Heidegger: “When ‘revealed’ through technology, nature is understood and visualized as a resource” [Riis, 2008, p. 290]. In the case under discussion, nature is ‘revealed’ by biotechnology as an almost inexhaustible collection of genetic resources.

1308

Henk van den Belt

infrastructure of the Internet [Kanehisa and Bork, 2003, p. 395]. They attribute a remarkable telos to bioinformatics: The ultimate integration of biological databases will be a computer representation of living cells and organisms, whereby any aspect of biology can be examined computationally. [Ibid., p. 309] The processing of information, or computation, is not only what computational biologists do; the activity of life itself is described in such terms. Kanehisa and Bork appear to take this view completely for granted. What was originally only a metaphor is now taken literally. In the end the object of knowledge and the subject of knowledge change places: When we have a complete computer representation of living cells and organisms and know the principles of how they compute, then, in the words of Sidney Brenner, ‘computational biology will become biological computation’. [Ibid.; my italics] The recent emergence of synthetic biology gives a further push to the ‘informatisation’ and ‘digitisation’ of life. Synthetic biologists attempt to build new life forms in a much more radical way than the first generation of genetic engineers succeeded in doing. For them it would betray lack of ambition just to move one or a few genes around and put them into an otherwise intact organism; they rather undertake to design entire genetic circuits from standardized building blocks or to insert cassettes with scores of synthetic genes into a host organism with a ‘minimal’ genome that can be used as a kind of chassis or technological platform (for a detailed sketch of the entire field of synthetic biology, see [De Vriend, 2007]). For some synthetic biologists, the ultimate dream is to create life ‘from scratch’. One anti-biotech NGO has portrayed synthetic biology as “extreme genetic engineering” [ETC Group 2007]. The development of synthetic biology is supported by technical advances in DNA synthesis. Worldwide some 70 firms are specialized in making DNA sequences with strings of nucleotides specified according to the client’s wishes. In conformity with a variant of Moore’s Law (well-known in the area of microelectronics), the cost-price of long DNA strands (more than 500 base pairs) is being halved every year [Endy, 2005, p. 451]. In fact, ever longer strands can be made (almost) faultlessly for ever lower costs. On the basis of this trend synthetic biologist Drew Endy foresees a decoupling between the design and the manufacturing of biological systems. In his view of the future a new category of creative artists, comparable to the graphic artists of today, will design new biological devices and systems from behind their laptops and then e-mail their designs to the synthesizing companies to have them manufactured [ETC Group, 2007, p. 34]. Thus the future biologist would be confined to the computer screen and would no longer need to dirty his hands in the ‘wet’ lab. The prominent ‘bio-entrepreneur’ Craig Venter also entered the new field of synthetic biology after his success in rivalling the public Human Genome Programme

Philosophy of Biotechnology

1309

with his own audacious shortcuts in genome sequencing. Venter has described the transition from genome sequencing to synthetic biology as a shift from reading to writing the genetic code. In a BBC lecture, he elaborated on what he called the ‘digitising’ of biology: For the past 15 years at ever faster rates we have been digitising biology. By that I mean going from the analog world of biology through DNA sequencing into the digital world of the computer. I also refer to this as reading the genetic code. The human genome is perhaps the best example of digitising biology. Our computer databases are growing faster per day than during the first 10 years of DNA sequencing. [. . . ] We and others have been working for the past several years on the ability to go from reading the genetic code to learning how to write it. It is now possible to design in the computer and then chemically make in the laboratory, very large molecules. [Venter, 2007] Viewed in this light the rise of synthetic biology can be seen as an attempt at closing the circle, so that in future researchers might easily pass from the ‘analogue world’ of life to the digital world of the computer and just as easily from the digital world back to the analogue world. The entire movement is predicated on the notion of biological ‘information’ and would, if fully completed, provide the ultimate technological realization of the ‘informatisation’ of life.

2.2

Postmodern glosses and philosophical scepticism

In the postmodern vision of Eugene Thacker transitions from the analogue to the digital world and from the digital to the analogue world (and also from one part of the digital world to another) are closely linked to the global networks of economic circuits in which contemporary biology is embedded: [. . . ] not only do we see biology being networked and distributed across the globe, but we are also seeing a hegemonic understanding of biological ‘life itself’ that ceases to make a hard distinction between the natural and the artificial, the biological and the informatic. [Thacker, 2005, pp. xvi-xvii] Presumably, Drew Endy’s expectations about biological design from behind the computer screen and Craig Venter’s views on ‘digitising’ biology exemplify this new “hegemonic understanding of biological ‘life itself’ ”. Thacker undertakes to study the ‘political economy’ (in a broad sense) of gene technology and the modern life sciences by systematically tracing the transitions from biological matter to digital information (encoding), from one form of information to another (recoding), and from digital information back to biological matter (decoding): in an economic sense these transformations correspond successively with production, distribution (circulation) and consumption. The entire process is not a self-enclosed circle but an expanding spiral. It seems that Thacker, despite his somewhat idiosyncratic

1310

Henk van den Belt

terminology and his all too fashionable postmodernism, has indeed put his finger on some salient aspects of the ongoing ‘informatisation’ of life. Given this continuing process of ‘digitising’ life, it may not be surprising that biotechnology and the modern life sciences are often perceived by non-biologists as branches of information and communication technology. Thus a leading business magazine claims that modern biology “is, itself, an information technology” [Economist, 1999, p. 97] (quoted in [Griffiths, 2001]). The well-known theorist of the ‘network society’, Manuel Castells, argues that genetic engineering actually is an information and communication technology: While genetic engineering is often considered as an independent process from the Information Technology revolution, it is not. First, because from an analytical perspective, these technologies are obviously information technologies, focused on the decoding and actual reprogramming of the DNA, the code of the living matter. And since biologists know that cells do not work in isolation, the real issue is to understand their networks of communication. Thus genetic engineering is both an information and communication technology [. . . ]. [Castells, 2005, p. 12] (quoted in [Engelbrecht, 2007, p. 42]) For the philosopher of biology Paul Griffiths, however, such an assertion would be on a par with the claim that the planets compute their orbits around the sun [Griffiths, 2001]. With Sahotra Sarkar and a few other philosophers of biology, Griffiths is highly critical about the idea that an organism’s DNA encodes ‘information’ about its developmental programme and sees no useful application of the concept of information in molecular genetics at all. His colleague Sarkar expresses this view most sharply: “ ‘Information’ is little more than a metaphor that masquerades as a technical concept and leads to a misleading picture of the conceptual structure of molecular biology” [Sarkar, 1996, p. 857]. The status and usefulness of the concept of information in modern biology is actually a central issue in a specialist debate within the philosophy of biology (see, among others, [Sarkar, 1996; Griffiths, 2001; Maynard Smith, 1999; Godfey-Smith, 2006]). It would not be appropriate to review that debate here, but suffice it to say that the belief in the usefulness and centrality of ‘information’ is a much embattled position, with John Maynard Smith remaining a staunch but lonely defender. The interesting point is that there is a big discrepancy in views on the scientific status of the concept of ‘information’ within modern biology between, on the one hand, an ‘esoteric’ circle of highly sceptical philosophers of biology and, on the other, a wider ‘exoteric’ circle of practising biological researchers and non-biologists for whom ‘information’ appears to be the first and last word of the scientific Gospel. If we are to believe sceptical philosophers of biology like Griffiths and Sarkar, the whole idea of the ‘informatisation’ of life or the ‘digitising’ of biology would be no more than a rhetorical construction or, insofar as it is actually believed by many involved, a collective delusion. Sure, bioinformatics plays such a prominent role in modern biological research, because there is a huge amount of data to

Philosophy of Biotechnology

1311

be analysed. But these vast amounts of information about genes should not, as Griffiths emphasizes, be assimilated with information encoded in genes [Griffiths 2001, p. 409]. The pervasive delusion is created precisely by such slippages and equivocations. Yet the distinction is not always easy to maintain. In recent years American companies and research institutes like Human Genome Sciences (HGS), the Institute of Genome Research (TIGR) and the J. Craig Venter Institute — all three, perhaps not coincidentally, associated with Craig Venter — have filed patent applications in which genome sequences in a ‘computer-readable medium’ (that is, digital versions of sequences) are being claimed. Rebecca Eisenberg, an internationally recognized expert on patent law who normally does not give in to the postmodernist fashion of erasing basic distinctions, shows herself quite baffled in the face of new artefacts like microarrays or ‘gene chips’: The distinction between computer-readable and molecular versions of DNA sequence is particularly difficult to maintain in the context of DNA array technology. DNA array technology involves immobilizing thousands of short oligonucleotide molecules on a substrate to detect the presence of particular sequences in a sample using specialized robotics and imaging equipment. [reference omitted]. In effect, this technology enables people to use computers to perceive information stored in DNA molecules in a sample. When contemporary technology blurs the distinction between computer-readable and molecular forms of DNA, what logic is there to drawing this distinction in determining the patent rights of DNA sequencers? [Eisenberg, 2002, p. 128; italics mine] The enchanting power of modern technology has induced Eisenberg to miss the ambiguity of “information stored in DNA molecules” — here, it is still information about a particular sequence (i.e. the precise order of nucleotides) and not information encoded in the sequence (i.e. the ‘instructions’ for protein synthesis or for the developmental programme). 3 EPISTEMOLOGICAL AND ONTOLOGICAL ISSUES

3.1

Biotechnology as technoscience

The term ‘technoscience’ is increasingly being used to refer to such contemporary disciplines as information and communication technology, nanotechnology, artificial intelligence and also to biotechnology. The popularity of the term is illustrated by the fact that even a group of social scientists reporting on the changing attitudes of European publics towards biotechnology invoke the term to mark the departure of this particular field from the norms and values once held to be essential to the ethos of science: “The evident commercialization and industrialization of biotechnology, with the pursuit of private knowledge, patents and profits, hardly

1312

Henk van den Belt

meets Merton’s criteria of universalism, communism, disinterestedness and skepticism, or the public’s expectations about the values, accountability and social responsibility of science. Biotechnology has become a technoscience, a commercial enterprise accountable to financial markets and to shareholders” [Gaskell et al., 2001, p. 117]. It is interesting to note that the general public, according to these social scientists, have not yet fully accommodated to the new commercial realities of ‘technoscience’ and still entertain expectations about science that modern biotechnology is unlikely going to meet. James Watson also describes the transition from molecular biology to biotechnology as a shift from the old academic ethos of science to a new ‘mind-set’: “Traditionally in academic biology, all that matters was precedence: who made the discovery first. One was rewarded in kudos, not in cash. [. . . ]. With the advent of biotechnology, all that changed. The 1980s would see changes in the relationship of science and commerce that were unimaginable a decade before. Biology was now a big-money game, and with the money came a whole new mind-set, and new complications” [Watson, 2003, p. 114].3 Elsewhere, he describes the transition from classical molecular biology to biotechnology as a shift from molecular ‘naturalism’ to a more proactive and manipulative approach: “Molecular biology had come a long way in the first twenty years after the discovery of the double helix. We understood the basic machinery of life, and we even had a grasp on how genes are regulated. But all we had been doing so far was observing; we were molecular naturalists for whom the rain forest was the cell — all we could do was describe what was there. The time had come to be proactive. Enough observation: we were beckoned by the prospect of intervention, of manipulating living things. The advent of recombinant DNA technologies, and with them the ability to tailor DNA molecules, would make all this possible” [ibid., 81-82]. Although Watson does not speak of ‘technoscience’, it is precisely the aspects of intervention and manipulation highlighted in his description that the term intends to capture. (One might suspect, however, that Watson makes a too strong contrast and portrays the molecular biology of the first twenty years after 1953 in too ‘naturalist’ colours — even in the old days, experiments had to be performed to find out about the machinery of life and the regulation of genes. It was never just simple observation.) The term ‘technoscience’ is thus not only a useful pointer to the highly commercialized setting in which modern biotechnology and many other contemporary undertakings are conducted, it also suggests that science and technology, which presumably were once distinct activities, have become so much intertwined as to be virtually indistinguishable nowadays. The present popularity of the term may thus reflect the historical process in which science and technology have become increasingly interwoven, but it may also partly reflect the dominant preoccupations and concerns of those who use the term. After all, postmodernist analysts 3 In the introduction to his novel Jurassic Park, Michael Crichton also discussed the rise of biotechnology as a “gold rush of astonishing proportions” and even called the commercialization of molecular biology “the most stunning ethical event in the history of science” [Crichton, 1991, p. x].

Philosophy of Biotechnology

1313

like Donna Haraway, Bruno Latour, Don Ihde and Andrew Pickering who are especially enamoured by this notion (they are all represented in the volume Chasing Technoscience, edited by Ihde and Selinger [2003]), are well-known for their tendency to cross conceptual boundaries and blur basic distinctions. It may thus be hard to provide convincing evidence for the historical thesis that we have entered the era of technoscience or that at present we have more technoscience than, say, one hundred or two hundred years ago. Is contemporary plant biotechnology more a ‘technoscience’ than traditional plant breeding or 20th -century plant breeding after the rediscovery of Mendelian genetics? Is contemporary nanotechnology more a ‘technoscience’ than electrical engineering? It is not even clear that we have the criteria to render such comparisons meaningful at all. However, historian of science John Pickstone holds that the concept of ‘technoscience’ can still be historically illuminating: “For our analysis ... we gain more by recognising the dense interweavings of universities and industry, and of science and technology, than we do by trying to separate them” [Pickstone, 2000, p. 163].The argument seems somewhat circular.4 Use of the term ‘technoscience’ is sometimes accompanied by rather unusual and problematic perspectives on epistemology and ontology. The notion was introduced by Latour [1987, p. 174] for the express purpose of providing a framework in which he could describe and analyze the construction of facts and artefacts (machines) in the same terms. For Latour, technoscientists are engaged in the ceaseless proliferation of ever new entities, ‘hybrids’ or ‘quasi-objects’ or ‘hairy objects’ or whatever these products may be called [Latour, 1993; 2004]. In the process, nature and society or nature and culture are said to be produced simultaneously — or ‘co-produced’. This description is also supposed to hold for the more traditional pursuits of natural science. When natural scientists succeed in getting their views and theories accepted among their colleagues, they ipso facto, according to Latour, enrich the world with the new kinds of entities that are postulated in those theories, such as atoms, molecules, microbes, genes, black holes, quarks, gravity waves, global warming, or the hole in the ozone layer. They are continuously proliferating new entities, just as their more technologically oriented colleagues are always busy to place new artifacts like plastics, electron tubes, transistors, vaccines or transgenic organisms into the world. Both the scientific and the technological branch of the technoscientific complex (if we can make such a distinction 4 In his big-picture history of science, technology and medicine (STM), Pickstone distinguishes four ideal-typical ‘ways of knowing’ that are held to be dominant in successive historical periods: (1) natural history, (2) analysis, (3) experimentalism and (4) technoscience. However, his fourth ideal-type is rather ambiguous, as it is defined both as a (more or less) timeless category and as a historically specific category. “Technoscience”, according to Pickstone, “refers to ways of making knowledge that are also ways of making commodities, or such quasi-commodities as stateproduced weapons” [Pickstone, 2000, pp. 13-14]. He thereby replaces the methodology-based criterion used for classifying the first three ideal-types with a new criterion of embeddedness within academic-industrial and military-industrial complexes. Indeed, in the second instance Pickstone distinguishes ‘natural-historical’, ‘analytical’ and ‘experimental’ forms of ‘technoscience’, equating the experimental form with the historically specific category of technoscience that emerged in the second part of the 19th century.

1314

Henk van den Belt

between two branches at all) are thus seen to be engaged in the same type of ‘world-making’ activity. Latour refuses to make a principled distinction between the different types of ‘entities’ or ‘hybrids’ that are being ‘constructed’. He would treat Robert Koch’s discovery of the cholera bacillus and Ananda Chakrabarty’s creation of a genetically modified oil-consuming bacterium on the same epistemological and ontological footing and is thus committed to the rather outlandish view that microbes entered the world only with Pasteur’s and Koch’s discoveries and did not exist in earlier ages [Sokal and Bricmont, 1997, p. 145 note 124] (see also my criticism of ‘radical constructivism’ in [Van den Belt, 2003]). We would do well, therefore, to keep some distance from the particular philosophical views in which analyses of contemporary technoscience are often packaged. This is not to deny that this body of literature raises important questions that need to be addressed and that are often neglected or ignored in traditional philosophy of science (cf. [Radder, 2004a]). It is indeed useful to stress that the production of knowledge is not only situated in particular historical and social contexts but also instrumentally mediated and thus embedded in larger material networks. It may also be fruitful to engage with the intimate relationship between science and technology and to focus on the instrumental practice of knowledge production rather than on the presumed correspondence between theory and reality, or in Ian Hacking’s terms, on intervening rather than representing. This would also warrant giving more attention to experimentation as the link between science and technology. For the philosophical reflection on biotechnology in particular it would be necessary to analyze the relationship between knowing and making, between nature and artifice, between animate and inanimate nature, and to clarify the ontological status of the products of this technology that are often considered to be ‘unsettling’ and problematic. It so happens that many of the above questions have also been raised in the context of patent law as it was confronted with the ascendance of biotechnology. Patent law can thus be seen as a practical laboratory or test site for the philosophy of biotechnology.

3.2 Knowing and making, nature and artifice Watson contrasted classical molecular biology aiming to understand the basic machinery of life with modern biotechnology seeking to manipulate living things or to create new life forms by tailoring DNA molecules. We already noticed that his contrast may be a little too stark. It may be hard to satisfy the will to know by pure observation alone. In the tradition of western philosophical thought it is by no means unusual to combine the ambition to understand with the ambition to create and to achieve useful effects. It is indeed often suggested that there is a very close connection between knowing and making, sometimes even to the point of reducing this connection to an intrinsic equivalence: whoever can make a particular object, thereby shows that he knows it, and only he who is able to make it may legitimately pretend to know the object [Kuypers, 1974]. This view was characteristic for the ‘mechanistic’ research programme in the natural sciences and can

Philosophy of Biotechnology

1315

be found in the Preface of Kant’s Critique of Pure Reason, where he remarks that prominent researchers like Galilei, Torricelli and Stahl knew very well that “reason only understands that which it produces after its own design”. Giambattista Vico’s verum factum principle from 1710 — “The criterion and the rule for the true is to have made it” — is an earlier formulation of the same idea. A late echo of Vico’s principle can be found in the epigram that the great physicist Richard Feynman wrote on the blackboard when he left his Caltech office in 1988: “What I cannot create I do not understand” [Morton, 2005]. In the second part of the 19th century, organic chemists in Germany followed the same principle. Guided by August Kekul´e’s structure theory, they learned to synthesize an ever increasing range of carbon compounds and to become adept in the art of molecular design [Travis, 1993]. Structural formulas were used by these ‘molecule smiths’ as construction drawings for the compounds that had to be made; conversely, a successful synthesis counted as proof that the structural formula gave the correct representation of the molecule in question. Knowing a particular chemical substance was thus equated with the ability to make it. Science imperceptibly merged into technology, as the close relations between academic chemistry and the synthetic dye industry in Germany also testified. It is no wonder, therefore, that this industry, which somewhat later also embarked on the production of pharmaceuticals, figures as a prominent example in Pickstone’s historical description of the rise of ‘technoscience’ as a new way of knowing [Pickstone, 2000, p. 176 ff]. Vico’s principle also provides a rationale for modern biotechnology and especially for its recent continuation, synthetic biology. Steven Benner, an organic chemist who is involved in synthetic biology, spells out the lessons of synthetic chemistry for the synthetic approach in biology: “Chemists know that if one truly understands a phenomenon, one should be able to synthesize another, different system that generates that phenomenon. [. . . ] Now that genetic engineering is available, biologists are benefiting. By attempting to create synthetic genetic systems, we will learn more about how natural genetic systems work; by attempting to create synthetic metabolisms, we learn about how natural metabolisms work; by attempting to create synthetic regulatory circuits, we learn about how natural regulatory circuits work” [Benner, 2008, p. 694]. Take for example the metabolic engineering performed by Jay Keasling and his Berkeley team on a strain of yeast to turn the latter into a biochemical factory for the production of the antimalarial drug precursor artemisinic acid [Ro et al., 2006]. This attempt to redirect the metabolic pathway of the yeast strain involved far more than inserting a single foreign gene into the host organism, as was done in the early days of biotechnology with the E. coli bacterium to produce human insulin. Keasling and his co-workers had to insert a whole series of genes from different sources (12 genes from the sweet wormwood plant Artemisia annua and 20 from other sources), ‘upregulate’ them and suppress the effects of other genes, in order to ensure the optimal execution of the successive reaction steps in the artificially installed metabolic pathway. The effort requires much more than throwing the necessary enzymes together in a cell.

1316

Henk van den Belt

Even while striving to achieve an eminently practical goal, along the way the team had to learn a lot about cell metabolism and its regulation. In this way, “synthesis drives discovery and learning” [Benner and Sismour, 2005, p. 538]. The huge complexity of life, however, erects obstacles against a straightforward implementation of the adage that to know something is to be able to make it. Natural evolution has not ‘optimized’ living organisms with regard to their being known by the human intellect. Evolved genomes are a mess with overlapping gene segments and large stretches of apparently meaningless junk. To understand the working of biological systems, the bio-engineer first has to introduce some rather drastic simplifications. MIT synthetic biologist Drew Endy bluntly explains his approach in this way: “Screw it! Let’s build new biological systems; systems that are easier to understand because we made them that way” [Silver, 2007; my italics]. If the bio-engineer succeeds in building such a deliberately simplified system, he may rightly claim to know it. The obvious objection, however, would be that in this way he has only obtained knowledge of his own artificial creation, not of Nature itself. To extrapolate from artificial systems to natural biological systems remains tricky. For the biotechnologists and synthetic biologists who follow Vico’s principle, it would hardly make sense to oppose the ‘natural’ and the ‘synthetic’. Yet, this seems to be a dichotomy that is deeply ingrained within our culture. The rise of synthetic dyes already caused a cultural backlash in the late nineteenth century, when the romantic artist William Morris blamed them for having destroyed the ancient art of textile dyeing. Later on, the advent of plastics led to a renewed passion for ‘natural’ fibres and other products. In cultural terms, what is ‘synthetic’ and ‘artificial’ is often considered ‘unnatural’ and ‘inauthentic’. This same polarized cultural repertoire would also play itself out in the public response to biotechnology [Sagoff, 2001]. In her book Designs on Nature [2005a] Sheila Jasanoff has explored the culturally ambivalent status of biotechnology and its products more deeply from the perspective of STS. In the next section I will turn to her views.

3.3 Proliferating monsters? Biotechnology can be seen as culturally ‘unsettling’ or ‘disruptive’. It is characteristic of this ‘technoscience’ that it continually places into the world new entities which from a cultural point of view may be described as ‘monsters’, that is to say, hybrids of nature and culture that have not yet found a recognized place within existing frames of reference and systems of classification.5 It is precisely because of its “zeal for hybridity”, according to Jasanoff, that biotechnology inevitably requires “ontological ordering” [Jasanoff, 2005b, p. 151]. Categories and classifications that are called into question by biotechnology “include the fundamental divisions between nature and culture, moral and immoral, safe and risky, 5 “Monsters” are defined as “entities that threaten disorder by crossing the settled boundaries of nature or society” [Jasanoff, 2005b, p. 151].

Philosophy of Biotechnology

1317

god-given and human-made” [Jasanoff, 2005a, p. 26].6 Jasanoff uses the following set of examples to illustrate the idea: We can import genes from spinach into pigs, from jellyfish into rabbits, and from fish into tomatoes; the technique of xenotransplantation allows cells from genetically altered pigs or chimpanzees to be inserted into biologically compatible humans. We can contemplate altering the human genome so as to produce enhanced human beings, with characteristic that today would be regarded as out of the ordinary, even superhuman. What, then, is nature and what is being human? [Jasanoff, 2005a, p. 26] It must be emphasized that the term ‘monster’ does not necessarily convey a negative connotation. Many postmodernist writers are so excited by crossing borders and blurring distinctions that they almost seem to fall in love with any ‘hybrid’, ‘monster’ or ‘cyborg’ that comes along. Within STS, Latour [1993; 2004] and Haraway [1992] are obvious examples. Other authors use the notion of ‘monsters’ in a more detached way as a conceptual instrument to study public responses to newly discovered or created phenomena. The locus classicus for this approach is Mary Douglas’ anthropological study on Purity and Danger (1966). The Dutch philosopher of technology, Martijntje Smits, adopted and elaborated her approach to illuminate public controversies on plastics and on the release of genetically modified organisms [Smits, 2002]. Jasanoff also treads in Douglas’s footsteps. There is a problem with the ‘monster creation’ thesis if it is taken as a specific claim about the culturally disruptive impact of modern biotechnology and the life sciences. After all, according to Latour’s Actor Network Theory (ANT), all new facts and artifacts produced by whatever ‘technoscience’ are to be considered as nature-culture hybrids or ‘monsters’. So then what, if anything, is so special about biotechnology? Interestingly, a similar criticism has been made by Noortje Marres with regard to Latour’s recent thesis that scientific and political institutions nowadays are faced predominantly with “hairy objects” (“the partly unknown entities that risk disturbing social life, from ‘BSE’ to ‘GM food”’) as against the relatively simple “smooth objects” of earlier days [Latour, 2004, p. 24]. This fact is thought to induce a major institutional crisis. Marres notes that Latour’s newly invented “hairy objects” take on many of the properties he earlier ascribed to all new entities leaving the laboratories of technoscience. So he too suggests a historical discontinuity that cannot be justified by the ANT approach [Marres, 2005, pp. 102-104]. On her part, Jasanoff acknowledges that questions about the ontological and moral status of new entities have arisen “in connection with 6 Compare the following passage: “Metaphysical disorder, or confusion about how to classify things, was an inevitable by-product of genetic modification, producing attendant confusion in the practices of governance. Biotechnology disrupted ancient classifications and transgressed boundaries that had for centuries been accepted as given in Western legal and political thought. Distinctions between nature and artifice, animate and inanimate, living and nonliving, body and property suddenly became problematic, and thus in principle political, in many areas of decision making” [Jasanoff, 2005a, pp. 280-81].

1318

Henk van den Belt

other technological developments”, but “perhaps never with quite the urgency generated at the fast-moving frontiers of biotechnology” [Jasanoff, 2005b, p. 151]. In other words, what may be valid to some extent for the hybrids created by other forms of technoscience, is even more strongly applicable to the products of modern biotechnology. Another and related criticism might be that the STS analyst, by subscribing to the monster creation thesis, illegitimately prejudges the outcomes of the very process of framing he sets out to explore. Robin Williams has expressed similar reservations about how “the activist wing of the STS community” takes up the study of the impacts of genomics and nanotechnology, new technologies which in his view are “conceived from the outset as being challenging in terms of risks and social values” [Williams, 2006, p. 327]. This goes against old constructivist tenets of agnosticism and impartiality: These commitments seem to conflict with the emphasis in most STS academic analysis on the need to deconstruct the objects of study, and in particular to be sceptical about claims regarding the character and implications of technology. [ibid.] By endorsing the monster creation thesis, Jasanoff effectively abandons her impartiality as an STS analyst and implicitly opposes the framing of biotechnology as product, or what could also be called the ‘business-as-usual’ frame. This frame transpires in the review of Jasanoff’s book in Nature, written by the European top-level civil servant, Mark Cantley: The perception — widespread in Europe — that biotechnology is something fundamentally new, like the discovery of electricity, or akin to black magic, is unfortunate. It has led to the assumption that there are technology-specific risks requiring ad hoc regulations and associated bureaucracies, and to consequent conflicts with sectoral regulations, as well as to international trade disputes. But not for the first time, perceptions, laws and the course of development may be driven by delusion. [Cantley, 2005] Back in the 1980s, Cantley attempted in vain to align European biotech policy with the ‘product’ frame adopted in the U.S. [Jasanoff, 2005a, p. 79 ff]. It is, of course, ironic that he does not recognize his own view as reflecting a particular framing but sees it as simply based on objective science; the other frames, by contrast, are dismissed as “delusion”. Here, however, we are concerned with the possible shortcomings of Jasanoff’s approach. What is problematic from a larger STS viewpoint is that she precludes the legitimacy of the ‘product’ or ‘businessas-usual’ frame by attributing a priori a particular character to biotechnology. This technology is seen as inherently disruptive because it inevitably challenges culturally entrenched categories and classifications, so anybody who merely sees it as business as usual must surely misjudge the issues.7 7 It is no coincidence that in the WTO dispute between the USA and the European Union, Jasanoff took the side of the latter. With two American and two British colleagues, she wrote

Philosophy of Biotechnology

1319

A possible remedy might be to change the monster-creation thesis from an a-priori into an a-posteriori judgment. In other words, it is not by any ‘inherent’ properties that biotechnology challenges existing categories and classifications. However, as a (contingent) matter of fact it turns out that many applications of biotechnology have indeed called into question many deep-rooted views and distinctions. The latter claim can hardly be disputed. This reformulation of the monster-creation thesis has an additional advantage. It makes clear that it depends not only on the properties of a particular technology whether or not that technology has a culturally disruptive or unsettling effect, but just as much on the prevailing categories and classifications that may be challenged or ‘offended’ by it. There are two variables in the equation.

3.4

Patent law as a laboratory for the philosophy of biotechnology

The emergence of modern biotechnology coincided in time with sustained efforts to strengthen the regime of intellectual property (IP) rights in general and of patent law in particular, by extending the domain of eligible subject matter for which IP rights can be sought and enhancing the geographical scope and degree of legal protection of such rights. The beginnings of this expansionary trend can be dated around 1980. In that year the US Supreme Court decreed in the Diamond v. Chakrabarty case that “anything under the sun that is made by man” is in principle eligible for patenting. The decision was a legal landmark that occasioned a huge influx of venture capital into the newly emerging biotechnology sector. Through a series of judicial decisions the domain of ‘patentable subject matter’ was further enlarged and extended in subsequent years to include computer software, methods of surgery, financial products and business methods, in addition to genetically modified organisms and gene sequences [Jaffe, 2000]. The European Union eventually followed the American example, albeit with hesitation and delay. It even took 10 years of negotiation and revision before the European Directive for the protection of biotechnological inventions, which codified the patentability of genes among other things, was finally passed in 1998. The campaign to strengthen the international IP regime, led by a coalition of American firms in the amusement, pharmaceutical and biotech sectors, met with success in 1995, when the TRIPs agreement was concluded as part of an overall WTO package. The TRIPs agreement (standing for Trade-Related aspects of Intellectual Property rights) sets worldwide minimum standards for the protection of IP rights that are binding for each WTO member state. Although legally it is by now an accomplished fact that genes, DNA sequences, cultivated cells and tissues and transgenic organisms are indeed considered patentable in principle, at least in the American and European jurisdictions, this by no means implies that the legal protection of biotechnological ‘inventions’ is an amicus brief for the WTO dispute panel. See [Winickoff et al., 2005].

1320

Henk van den Belt

fully accepted. Indeed, the issues remain highly controversial. Moreover, new court cases and policy decisions in this area continue to stir debate. 3.4.1

The patentability of organisms and genes

Let us briefly review the situation by focusing on the legal case for patenting transgenic organisms and genes and indicating the main objections and counter arguments. In the landmark case of Diamond v. Chakrabarty a 5-to-4 majority of the US Supreme Court held that anything new under the sun that is made by man, whether living or non-living, can in principle be patented. Referring to the ‘invention’ under consideration, Chief Justice Burger argued on behalf of the majority: “[T]he patentee has produced a new bacterium with markedly different characteristics from any found in nature and one having the potential for significant utility. His discovery is not nature’s handiwork, but his own; accordingly it is patentable subject matter under § 101” [U.S. Supreme Court, 1980; my italics]. Two objections can be made to the patentability of genetically modified or transgenic organisms. First, the mere fact that a researcher inserted a foreign gene in an existing organism does not mean that he made the entire organism. Claiming that the resulting organism is “not nature’s handiwork”, but the researcher’s, would grossly underestimate the relative contribution of nature and overestimate the relative contribution of the purported inventor. (Suppose my neighbour builds a car and I put a radio-set in it. Could I then claim the car as “my own handiwork” and not that of my neighbour?) The second objection to the patentability of organisms has to do with the character of living beings as self-reproducing entities. Surely this is a property that is given by nature and not created or invented by genetic engineers. Yet by claiming a patent on a transgenic organism, the inventor annexes said property to the substance of his invention, as the legal protection extends to all the offspring of the first specimen(s) of his modified organism. In other fields, an invention has to be ‘reproducible’, but in the case of genetic engineering the inventor has only to create the first ancestral individuals and leave the rest of the job to ‘Nature’. After his claims have been granted, a natural process of biological reproduction is turned into an act of illegal imitation or patent infringement. In this sense, it is no exaggeration to say that patents on transgenic organisms virtually represent “patents on life”. In the past decades the domain of patentable subject matter has also been extended to genes and other DNA sequences. However, the patentability of genes is hardly less controversial than the patentability of transgenic organisms. In practice, patent claims on isolated genes are often combined with patent claims on the transgenic organisms that can be ‘made’ by inserting those genes and on the proteins that they encode. Thus the patentability of genes is the crucial axis for the legal protection of biotechnological inventions. In patent law, the case for the patentability of genes is mainly argued by drawing on the analogy with natural substances that have been isolated and purified. According to American and European case law, such substances are indeed patentable. In a famous precedent-

Philosophy of Biotechnology

1321

setting case dating from 1911 (Parke-Davis & Co. v. H.K. Mulford & Co.), it was held that the researcher who first isolated adrenaline from the cortex of the adrenal gland and purified it from admixtures had for all practical purposes created a new substance, which would therefore count as a patentable invention. Along similar lines the legal case is nowadays made for the patentability of isolated genes and other DNA sequences [Doll, 1998; Goldstein and Golod, 2002]. Opponents object however that genes are not invented but discovered. They also point out that the customary patentability requirements of ‘inventive step’ (US: ‘non-obviousness’) and ‘industrial applicability’ (US: ‘utility’) are often not sufficiently satisfied. An interesting objection has been made by the British genome researcher and Nobel laureate, John Sulston, who contests the analogy with isolated natural substances. He asks what is so special about ‘isolating’ a human gene that would entitle this achievement to be rewarded with a patent. All we actually do is ‘copying’ the gene in bacteria or other unicellular organisms. Sulston formulates his argument in a most forceful way: “The European Patent Directive, approved by the European Parliament in 1998, accepts that a sequence or partial sequence is eligible for a ‘composition of matter’ patent once it has been replicated outside the human body — say, copied in bacteria as we do for sequencing. This argument has always seemed to me absurd. The essence of a gene is the information — the sequence — and copying it into another format makes no difference. It is as though I took a hardback book that you had written, published it in a paperback, and called it mine because the binding is different” [Sulston and Ferry, 2002, p. 299; italics mine]. In Sulston’s view patenting genes virtually amounts to a glaring infringement of Mother Nature’s original ‘copyright’ ! Interestingly, his position also puts the theoretical status of the gene concept in modern biology on the agenda of patent law. 3.4.2

Discovery versus invention

From the perspective of the philosophy of technology it is interesting that the debate on the validity of patents on genes and other biological materials partly turns on the distinction between discoveries and inventions. This distinction is firmly entrenched in European and American patent law. Patents are intended only for inventions or products of human ingenuity, not for discoveries or products of nature. (Article 52.2.a of the European Patent Convention states that “discoveries, scientific theories and mathematical methods” shall not be regarded as inventions within the meaning of paragraph 1.) However, modern commentators are often sceptical about the possibility to settle the dispute between the proponents and opponents of gene patents by appealing to the discovery/invention distinction. David Resnik, for example, holds that this distinction does not mark a purely factual or descriptive boundary, but is inevitably linked up with various values and interests. Authoritative decisions on the problematic hard cases in the intermediary grey zone, such as cloned or modified genes, he argues, can only be arrived at through a broader process of public deliberation in which those values

1322

Henk van den Belt

and interests can be brought in [Resnik, 2002]. Hans Radder, on his part, points out that different philosophical positions such as constructivism and realism may lead to different views on what is an invention and what is a discovery [Radder 2004b, p. 279]. While some constructivists might hold that all purported discoveries are really inventions and are therefore in principle eligible for patents, some realists (like Steve Luper) might be inclined to assimilate all inventions to discoveries and therefore argue against their patentability. This leaves us with a paradox. The discovery/invention distinction is firmly enshrined in modern patent law and considered to be of fundamental importance, yet philosophers of science and technology seem to hold that it cannot do any useful work in helping to decide concrete cases. We may gain some advantage of historical distance when we try to look at the issue through the eyes of Josef Kohler (1849-1919), a leading German jurist and legal philosopher at the turn of the 20th century. In his handbook on German patent law in comparative perspective [Kohler, 1900], he presented a systematic account of the legal and philosophical foundations of patent law. Kohler had no doubt that the distinction between discovery and invention was constitutive for this domain of law and that it could be effectively used to decide questions of patent-eligible subject matter. He stipulated that an invention is a human creation and not a creation of nature and that as such it must “wring a new aspect from nature” (die Natur eine neue Seite abgewinnen; see [Kohler, 1900, p. 85]). A discovery, by contrast, is defined as a reflection (Widerspiegelung) of the outside world within the human mind. It is not limited to the perceptible aspects of the world but also extends to underlying mechanisms, regularities and laws. Kohler also stressed that special human-made arrangements (menschliche Veranstaltungen) and experimental interventions may be required to make discoveries. So the invention/discovery distinction did not hinge on the contrast between active intervention and passive contemplation. Kohler drew some interesting corollaries for the patentability of findings in the field of chemistry and biology. He held that the above definitions entailed the inadmissibility of product patents on chemical substances (what in American patent law would be called ‘composition of matter patents’). As a matter of fact, the German Patent Law of 1877 (Section 1.2) excluded chemical substances from protection, allowing patents only for the particular processes (Verfahren) used to produce them. However, Kohler argued that the exclusion of Stoffpatente (‘composition-ofmatter’ patents or chemical product patents) is not merely the result of a legal fiat (Machtspruch des Gesetzes): “The chemical products either are already present in nature; then it is self-evident that the substance cannot be patented as a substance; or at least they might be available in nature, as there are many substances in nature that are still unknown, so that they can only be discovered but not invented” [ibid., p. 85]. It would seem that this reasoning ignores the possibility that new synthetic substances can be created that would never be found in nature — Kohler denied their status of being inventions on the basis of the theoretical possibility that one day these synthetically produced substances might also be

Philosophy of Biotechnology

1323

found in nature. He further argued that the tendency of chemical substances to combine with each other reflects an inherent, natural disposition, so that man’s contribution consists not so much in the creation of a new compound as in the removal of the obstacles that block its formation. (The latter achievement may be rewarded with a process patent.) He finally presented more pragmatic grounds for the exclusion of patents on chemical substances, arguing that if the first inventor who invented a new process to produce a new substance would be granted a product patent on the substance as such, this would discourage further efforts to invent more efficient processes for manufacturing the same substance. (Actually, this was the official rationale behind the exclusion of chemical substances in the German Patent Law of 1877). Nowadays, we would probably judge Kohler’s attempt to argue the exclusion of chemical product patents by sole appeal to the discovery/invention distinction as not entirely convincing. This is not to say that this exclusion could not be defended on other grounds, such as the more pragmatic reasons that were also adduced by Kohler.8 In any event, the legal exclusion of chemical substances from patentability, which was customary in Germany and some other jurisdictions well into the 20th century, provides a notable contrast with the doctrine of the patentability of isolated and purified natural substances that was introduced in the USA in the wake of the Parke-Davis & Co. v. H.K. Mulford & Co. case of 1911 and later also adopted in other parts of the world. The latter doctrine, allowing special exceptions to the general exclusion of products of nature from patentability, is not beyond dispute, however, if only because the exceptions are granted in a highly selective way to natural substances with a ‘biotherapeutic’ value but not to other natural substances. It is not transparent, for example, why purified adrenaline should be considered a patentable invention and purified tungsten should not. A closer look at the verdict on purified adrenaline shows moreover that the decision suffered from several incoherencies [Gipstein, 2003]. Thus the legal precedent on which the case for the patentability of DNA sequences as isolated natural substances has been built, is itself not immune to criticism. Indeed, some critics hold that the practice of issuing patents on isolated and purified, or otherwise slightly modified, natural substances makes a mockery of the requirement that inventions must be products of human ingenuity and not products of nature [Demaine and Fellmeth, 2003]. Needless to say, the latter criticism still invokes the discovery/invention distinction, as it asserts that the mere act of isolating and purifying a pre-existing natural substance would not qualify as an invention of a ‘novel’ compound. Hans Radder has argued the case for excluding product patents in general (not just for isolated natural substances) without invoking the discovery/invention distinction. His central argument is that a product patent pre-empts all possible 8 In a recent thesis, Reinier Bakels also rejects product patents on chemical substances as incompatible with the principles of patent law and only allows patents for two distinct types of inventions, namely concerning the preparation or manufacturing of a substance and concerning the use or application of a substance [Bakels, 2008].

1324

Henk van den Belt

processes by which the product can be produced and utilized, while the patent description discloses only one or a limited number of processes for obtaining the product. Thus the customary written description and enablement requirements of patent law may be satisfied for this particular process or these particular processes (reproducibility in the narrow sense), but not for the full range of actual and possible processes that are covered by the patent (reproducibility in the broad sense or replicability). A claim that covers the full range of possibilities is actually a theoretical claim. However, it is a standard tenet of patent law that theories are unpatentable. Hence, product patents should not be allowed [Radder, 2004b]. They grant inventors more than they have made available. 3.4.3

Living inventions

Interestingly, Kohler also drew some corollaries on the patentability of new biological creations. He speculated that in future living beings might be created by direct human combination “without relying on a naturally created cell” [Kohler, 1900, p. 88] — by the way, this is still far beyond the most advanced achievements of contemporary synthetic biology. Such artificial creatures would in Kohler’s opinion be eligible for product patents. However, product patents would be impossible for existing species produced in a new way or for new varieties of plant and animal species (in Kohler’s time, there was no established system of plant variety protection or breeders’ rights). Kohler justified this exclusion by pointing out that nature itself is able to generate an endless range of new varieties and that the appropriation of a small subset of these by human breeders would be unfounded. He added a very interesting consideration to this argument: “By the way, a patent would indeed prevent that someone created the same variety by means of human art, but it would not prevent that such organisms multiplied themselves naturally and that the same varieties resulted from the seeds and shoots. For patent law can govern only human action, it cannot constrain nature in those cases in which nature causes everything or at least the main part” [ibid., p. 88]. It would seem that this consideration is still topical in the age of biotechnology. In Diamond v. Chakrabarty, the US Supreme Court ruled that anything under the sun, whether living or inanimate, can in principle be patented. Yet living inventions may create special problems for patent holders, their users and others. As Kohler already anticipated, they may start to wander about and multiply themselves on their own accord. This problem was at issue in the notorious legal conflict between biotech company Monsanto and Percy Schmeiser, a canola farmer from the Canadian province of Saskatchewan. Monsanto has inserted gene constructs into various crops to make them resistant to its herbicide glyphosate or ‘Roundup’. Farmers who want to use the ‘Roundup Ready’ seeds with their built-in herbicide resistance are compelled to sign rather restrictive ‘technology use agreements’, which subject them to an inspection regime that severely encroaches upon their sphere of privacy. Saving and replanting from one’s own harvest — an age-old farming practice to

Philosophy of Biotechnology

1325

save part of the seed and reuse it in the next season, which in the legal system of plant variety protection or breeders’ rights is still recognized as the so-called ‘farmers privilege’ — has been labelled an act of infringement. A complication, however, is that plants containing patented genes may inadvertently shoot up on a farmer’s field through loss of transgenic seed from passing trucks, contamination in purchased seed, or cross-fertilization with pollen from neighbouring or more remote fields that has been dispersed by wind or insects. But even in those cases the farmer can be sued for patent infringement, as this particular offence does not presume deliberate intent. A report written by a critical NGO uses the appropriate expression ‘guilt by contamination’ to refer to such instances [Center for Food Safety, 2005, p. 14]. This factor proved indeed of paramount importance in the legal proceedings which Monsanto initiated in 1998 against Schmeiser and which the Canadian Supreme Court on May 21, 2004 decided in favour of the biotech company, although Schmeiser was eventually not condemned to pay full damages to Monsanto (for the text of the verdict, see [S.C.R. 1, 2004]). During the proceedings Schmeiser’s lawyer formulated the legal absurdity in the following terms: Had [Monsanto] maintained control over its invention, it may have maintained its exclusive rights. However, inventions do not usually spread themselves around. They do not normally replicate and invade the property and land of others . . . . [Monsanto] cannot on the one hand unleash self-propagating matter uncontrolled into the environment and then claim exclusivity wherever it invades. (Defense brief, § 170, quoted in [Lezaun, 2004, p. 151].) As the non-intentional spread of modified genes among crops grown in the United States and Canada continues unimpeded, a legal imbroglio is beginning to show up: Over time, virtually every farmer of a given crop may have his or her crop polluted with the genetically engineered variety of that crop, leading to a legally chaotic scenario where virtually every farmer in the United States is an infringer of the plant protections for genetically engineered crops. [Center for Food Safety, 2005, p. 50] This would seem to be the definitive reductio ad absurdum of a system for the legal protection of biotechnological inventions that defines the natural spread of genes as acts of ‘infringement’. Meanwhile Mother Nature helps quite a lot to tighten the hold that intellectual property owners can exercise over the real property of their baffled fellow citizens. Thus the new regime of intellectual property that has been created for the sake of modern biotechnology appears to seriously erode the age-old property basis of farming (see also [DeBeer, 2005] and [Ziff, 2005]). These difficulties ultimately spring from the decision to extend the legal protection of genetically engineered organisms also to their offspring and to extend the

1326

Henk van den Belt

protection of patented genes to every organism in which they may be found. Proponents of the existing regime usually argue that without such an arrangement biotechnological inventors would not have much opportunity to actually benefit from their inventions. That may be true, but should not distract us from the fact that the arrangement involves a considerable departure from the normal rules of patent law and as such constitutes a major concession to the holders of biotech patents. As Graham Dutfield remarks: [O]ne clear point of divergence from conventional patent norms is that since living things have a tendency to reproduce themselves, or at least with willing partners, in granting patents on life forms we are being very generous to the owners when we allow them to claim ancestors and progeny. [Dutfield, 2004; italics mine] The biotech industry and other supporters of this legal arrangement often invoke the TRIPs agreement to argue that the application of patent law should be ‘technology-neutral’ and not discriminate against biotechnology. They forget that the alleged neutrality has already been violated by the special concessions that are granted to biotech inventors. 3.4.4

Molecules versus information

In American patent law, genes have been construed as chemical substances to fit them into one of the available boxes of statutory subject matter, to wit ‘compositions of matter’. A similar policy has been followed by legislatures, patent offices and courts in other jurisdictions. Many biological researchers and legal experts, however, consider this view of genes as fundamentally flawed or at least as partial and one-sided. It is also widely expected that the dominant legal view on the nature of genes and other DNA sequences will increasingly clash with new developments in the life sciences. One prominent critic of gene patents is the British genome researcher John Sulston, who stresses the information content of genes and genomes: “It makes more sense to think of genes as software rather than chemical entities. The information can just as well be held in a computer or written in a book; composition of matter is irrelevant, because the conversion of the information from one form to another is unsurprising” [Sulston 2006, p. 412]. Sulston’s view is not exceptional among genome researchers. Indeed, it has been characteristic of molecular biology from the very beginning to conceive of an organism’s DNA as containing the ‘programme’ for its life. This conception has been technologically reinforced by the still ongoing trend of what Craig Venter refers to as the ‘digitization’ of biology [Venter, 2007]. Due to the character of DNA as an information-carrying molecule [Rai, 1999, p. 836], or in other words its dual nature as both a chemical compound and a template for protein synthesis, product patents for gene sequences tend to effectively capture the genetic code and thus to encroach upon the sphere of laws of nature. After

Philosophy of Biotechnology

1327

all, the genetic code specifies a fixed relationship between DNA nucleotide triplets and amino acids (the building blocks of proteins) and can thus be seen as a law of nature. A product patent on a gene ‘pre-empts’ this law of nature because it gives the patent holder leverage over the corresponding proteins [Kane, 2004; 2006]. According to jurisprudence of the U.S. Supreme Court, laws of nature are an inalienable part of the public domain and should not be patentable. Until now, however, America’s highest legal court has not pronounced on this issue in the context of gene patenting. (This argument against product patents on genes can be seen as a specific version of Hans Radder’s argument against product patents in general.) Viewing gene sequences as encoded information may not solve all problems of patentability, however, as legal systems everywhere still seem to have enormous difficulty in dealing with that other major category of the modern information society, computer programmes. Software patents were initially not allowed (article 52 of the European Patent Treaty excludes patents on computer programmes ‘as such’, whereas U.S. patent law traditionally recognizes the abstract idea exception and the mathematical algorithm exception), but under pressure from parts of the growing software industry both patent offices and legal courts have become increasingly accommodating and have granted and validated patents on so-called ‘computer-implemented inventions’. Controversy persists, however, and the official legal doctrine on the patentability of software inventions appears to remain in a state of utter confusion. With the further advance of the ‘digitization’ of biology in the wake of highthroughput sequencing and the rise of bioinformatics, some companies abandon the pretence of laying claim to tangible molecules only and also stake out claims on the informational content of gene sequences or even entire genome sequences. Examples are the patent application filed by Human Genome Sciences (HGS) on the sequence of the Haemofilus Influenzae genome, the application of The Institute of Genome Research (TIGR) on the genome of Mycoplasma genitalium and of the J. Craig Venter Institute on the genome of the ‘synthetic’ organism Mycoplasma laboratorium. In all three cases the original applications included claims on the ‘computer-readable medium’ on which the genome sequence was stored, or in other words claims on the digital version of the sequence. It seems that until now the US Patent and Trademark Office is not yet ready to grant such claims (in the first two cases, where patents have already been issued, the reference to the computerreadable media only occurs in the description parts), but the pressure may be mounting for this agency to allow such claims [O’Malley et al., 2005]. If such claims were eventually admitted, it would mean a radical departure from existing patent law and lead to rather absurd consequences. Protection of gene or genomic sequences in computer-readable form would entail that downloading and printing out the patent document from the Patent Office’s website, or even this website publication itself, constitutes an act of infringement [Eisenberg, 2002]. This would undermine the traditional patent bargain in which the inventor obtains protection for his invention in exchange for full disclosure.

1328

3.4.5

Henk van den Belt

The future of the international patent system

The preceding decades have seen the growth and consolidation of a more and more unified international patent regime. In the words of historian of technology Eda Kranakis, the general tendency has been “in the direction of enhancing the rights of patent holders by broadening the definition of patentability, widening the territorial reach of patent protection, and curtailing national restrictions on the use of patents” [Kranakis, 2007, p. 726]. Choices that were once available to countries that are now economically advanced at an earlier stage of their development are denied to the developing countries of today. The decision of nineteenth-century Switzerland and the Netherlands to have no patent law at all (which seems to have served their national economic interests quite well) would be out of the question for developing countries today. Even less radical restrictions like a ban on product patents for chemical and pharmaceutical inventions, which in the past has been practised by Germany and several other countries, would now be contrary to the TRIPs agreement. It looks as if the developed countries attempt to kick away the ladder to prevent others from climbing it too. indent It would not be justified, however, to extrapolate from current integrating and globalizing tendencies and predict the prevalence of an even more tightly organized international IP regime in the near and mid-term future. As always, there are also countervailing tendencies. Even a key institution like the European Patent Office in Munich is aware that the future may hold several surprises in store, as is testified by the interesting outcomes of an exploratory forecasting exercise which it commissioned in 2007. The scenarios show that in the not-so-distant future the international patent system may be thoroughly challenged by, inter alia, shifting geopolitical relationships indicated by the rise of new economic powers (e.g. China, India, Brazil), more incisive critiques from global civil society questioning the legitimacy of gene patents and the strategic control over agriculture and food, or the rise of complex technologies which call the one-size-fits-all approach of the current system into question [European Patent Office, 2007]. At present, serious tensions and fault lines are already visible in the international IP system. Kranakis refers to the lingering conflict between the American biotech company Monsanto and the Argentine government over the intellectual property and control of GM herbicide-resistant (“Roundup Ready”) soybeans that are cultivated in Argentina as an example illustrating the global reach and effects of biotech patents. Argentina was one of the first countries in the mid-1990s to allow the introduction of Monsanto’s GM soybeans containing a gene construct that provided resistance against the herbicide glyphosate (‘Roundup’). The country exports large quantities of soy in the form of soybean meal to Europe to serve as cattle feed. The Argentine Seed Law does not allow the patenting of plant varieties, however, so Monsanto first sold its Roundup Ready soybean seeds with a slight price mark-up and later concluded detailed contracts with farmers in an attempt to protect its intellectual property [Gallo and Kesan, 2006]. In the end the

Philosophy of Biotechnology

1329

company proved unable to retain control over GM soybeans in Argentina: most farmers procure their seed from the black market or replant from their own harvest (which is perfectly legal). Monsanto subsequently attempted to collect royalties on export shipments and exerted pressure on the Argentine government to compensate its foregone technology fees, but all to no avail. In 2005 the company started to attack Argentine soybean meal as it entered several European ports, asking customs officials to detain freighters and instituting lawsuits against imports of products said to infringe their European patent on the gene construct conferring resistance to glyphosate (the European patent had been granted in 1996 and revalidated in 2005, after opposition from Greenpeace). Even before any legal verdict was reached, these actions did much to disrupt the soybean trade between Argentina and Europe: “Monsanto’s European patent thus proved to be a powerful agent for disciplining not only European importers and traders, but also Argentine exporters, farmers, and the Argentine government itself” [Kranakis, 2007, p. 725]. However, it may be too early to declare victory for the North American company. On 10 October 2007, Monsanto suffered legal defeat in a verdict rendered by the United Kingdom Patents Court (lawsuits in other European countries are still going on). The British judge held that Monsanto’s patent claim on its gene construct did not extend to any derivative product in which this particular gene construct could be found; after all, the Roudup Ready gene works in the soybean plant to confer resistance to glyphosate but does not fulfil this function in soybean meal [Cohen and Morgan, 2008]. If this verdict is to set a legal precedent, it will clearly limit the scope of product claims on DNA sequences in Europe. If not, traders in derivative products (in fact, soybean derivatives are used in a endless number of consumer products) must be worried that any product containing an intact DNA molecule derived from any GMO might be construed as an infringement of any patent right on the sequences used to construct that GMO (ibid.). In the latter case, product claims on DNA sequences could have an enormous viral effect on the whole chain of economic transactions. Opponents of biotech patents often point out that such patents are concentrated in the hands of a few giant multinationals, giving them considerable power over the world food supply. The example of GM soybean cultivation in Argentina, however, shows that the control of multinational corporations over their (alleged) intellectual property may be far from perfect. What holds for Argentina, also holds for other parts of the developing world. In India, Monsanto and its Indian business partner Mahyco have been unable to retain intellectual property control over the so-called Cry1Ac gene (derived from Bacillus thuringiensis or Bt), which they inserted in their cotton variety “Bollgard” to confer insect resistance, especially against the cotton bollworm. Farmers in Gujarat and other Indian states have somehow been able to appropriate a few transgenic seeds (possibly from fields testing the new Bt cotton for regulatory purposes), multiplied them and crossed them out with indigenous varieties, in the process creating a huge and flourishing market in ‘stealth seeds’; state and federal authorities are unwilling or unable to suppress the illicit production of and trade in Bt cotton [Herring, 2007b]. Something similar holds for

1330

Henk van den Belt

Brazilian soybean farmers who initially obtained Roundup Ready soybeans from Argentina: “Some transgenes are spread so widely underground that they resemble open-access or open-source technology more than monopoly, more Linux than Microsoft” [Herring, 2007a, p. 17]. The reality of ‘stealth seeds’ on the ground indicates that the activities of farmers may be beyond the control of companies seeking to protect what they perceive as their intellectual property and of supervisory agencies charged with biosafety regulation. Some authors are worried about the circumvention of biosafety regulations implied by these illicit activities [Gupta and Chandak, 2005]. The genie is definitely out of the bottle.

4

ETHICS, REGULATION AND GOVERNANCE

Modern biotechnology in agriculture and food has proven to be one of the most contested technologies in contemporary society, comparable perhaps to nuclear energy. Applications of biotechnology in the production of medicines (‘red biotechnology’) and in industrial fermentation (‘white biotechnology’) have been far less controversial, however. Nobody seems to object to the use of GM bacteria for the production of human insulin (one of the first applications of recombinant-DNA technology); nor is there much protest against the use GM microbes for the production of re-engineered enzymes to catalyze industrial fermentation processes. Opposition to agrifood biotechnology may be based on the inherent characteristics of the technology, on its expected consequences for the natural environment, for animal welfare or for human health, or on its likely or presumed socioeconomic impacts on consumer autonomy, the position of small farmers and the relations between North and South. It will be simply impossible to review the debates on all these divergent issues in any detail here. Instead, I will focus on a few key principles and central questions that have surfaced in several of these debates. Discussions on social and ethical issues related to agricultural biotech do not occur in a political vacuum. What is fundamentally at stake in many debates is how society should set up the regulation and governance of this branch of technology. In broad outline there have been two major policy answers to this question. One response is to treat agricultural biotech as business as usual, to regulate its final products in the same manner as those of any other technology, and to declare the process (i.e. genetic engineering) by which the products are generated as simply irrelevant from a regulatory point of view. The second response is to recognize the process of genetic modification itself as a relevant factor for regulation and to consider agricultural biotechnology as something more than business as usual. The first response, the so-called product frame is characteristic of US biotech policy; the second response, the so-called process frame, defines the dominant biotech policy of the European Union, Japan and South Korea [Jasanoff, 2005a]. The confrontation of these two policy frames in the international arena has given rise to “regulatory polarization” [Bernauer, 2003] on a global scale and even led to serious disputes before the World Trade Organization.

Philosophy of Biotechnology

4.1

1331

Crossing species boundaries and playing God

For some, biotechnology is scary and unnatural, because it involves breaching the boundaries between biological species that have been created by God or naturally evolved during life’s long history on earth. The British prince Charles has made himself into a mouthpiece for articulating the public discontent on agricultural biotechnology. In 1998, at the height of the GM food scare in the United Kindom, the Prince of Wales personally poured oil on the fire by launching the term ‘Frankenfoods’, a phrase alluding to the title of Mary Shelley’s gothic novel. He also declared that “genetic modification takes mankind into realms that belong to God, and to God alone” [Prince of Wales, 1998]. Ten years later, in August 2008, Charles renewed his attack on biotech foods at a time when many leading UK politicians were thinking about giving GM crops a second chance in view of rising food prices on the world market. He now called genetic modification “the biggest disaster, environmentally, of all time” and accused unnamed “gigantic corporations” of “conducting a gigantic experiment with nature, and the whole of humanity, which has seriously gone wrong” [Booth, 2008]. Such statements annoy many British scientists, but find support among segments of the public and especially among the adherents of organic agriculture. Critics with some biological literacy will be quick to point out that species barriers in nature are not so rigidly fixed as to block all exchange of genetic material. Horizontal gene transfer is common among bacteria and not altogether rare among eukaryotes. Indeed, Agrobacterium tumefaciens, the bacterium that causes galls at wound sites of various plants, was already engaged in genetic engineering, so to speak, before modern plant biotechnology turned it into a Trojan horse or ‘vector’ for transferring recombinant genes. Triticale resulted from crossing species (wheat and rye) long before the rise of recombinant-DNA technology. These examples do not justify to naturalize and trivialize modern genetic engineering as nothing new under the sun (indeed, the range for making new re-combinations between the most remotely related species has been enormously enlarged), but they certainly militate against ascribing an absolute significance to the crossing of species barriers. Moreover, if crossing species boundaries through genetic modification is what supposedly takes mankind into realms that belong to God, then the same objection would apply to medical applications of biotechnology. Yet even Prince Charles made exception for “certain highly beneficial and specific medical applications” [Prince of Wales, 1998]. It is quite remarkable that the charge of ‘playing God’ is regularly levelled by secular environmentalist organizations against biotechnology and synthetic biology. Indeed, as the Christian philosopher Gordon Graham notes, “. . . if, as the secular world believes, there is no God, how could there be any danger of human beings illegitimately abrogating to themselves His function?” [Graham, 2002, p. 145]. A possible answer to the paradox of “playing God without God” may be found in Lutheran theologian Ted Peters’s hint that the God of “playing God” is not necessarily the God of the Bible, but rather “divinized nature” [Peters,

1332

Henk van den Belt

2005]. It is the presumed sacredness of nature that the modern life sciences risk to profane. Also to modern secular minds, there appear to be boundaries that may not be breached, even in the absence of a god who has the authority to institute such boundaries. This would agree with the view of the American molecular biologist Lee Silver, who explains the negative attitude of many post-Christian and left-leaning Europeans towards modern biotechnology from a “deep-seated sense of spirituality, one that encompasses all of Mother Nature” [Silver, 2004]. By way of illustration, Silver refers to an article in which three Dutch bioethicists condemn the potential use of biotechnology to create “quasi-chickens — genetically engineered humps of living chicken flesh that do nothing but lay eggs” [Bovenkerk et al., 2002]. What could possibly be wrong with that, asks Silver, assuming that those ‘quasi-chickens’ are incapable of suffering or feeling pain? The Dutch bioethicists argue that the creation of such entities would violate chicken integrity. Silver is stunned: “What can possibly be violated when no animals are harmed or killed? It can only be the imagined spirit of the chicken species” [Silver, 2004]. Overstepping the boundaries supposedly instituted by God or Mother Nature may also be seen as inviting unknown and unprecedented risks. To accuse biotechnologists of playing God may thus be just another way of warning the wider public for the recklessness of their pursuits in the relentless quest for profit and glory. Nonbelievers may also recognize the dangers of hubris and the wisdom of the Proverb “Pride comes before disaster, and arrogance before a fall” [Proverbs 16:18]. This takes us to the next subsection.

4.2 The precautionary principle Opponents of agricultural biotech often invoke the so-called Precautionary Principle (henceforth to be abbreviated as PP) as a justification for banning the cultivation of GM crops. This principle is already enshrined in international agreements like the Convention on Biological Diversity and the Cartagena Protocol on Biosafety, but existing definitions of it are at best partial and incomplete. In the context of dealing with environmental hazards, the Rio Declaration of 1992 presented the following formulation of what the PP would entail: “Where there are threats of serious or irreversible damage, lack of full scientific certainty shall not be used as a reason for postponing cost-effective measures to prevent environmental degradation”. Existing formulations of the PP do not clearly specify the ‘triggering’ condition under which preventative action would be demanded or indeed the type of action that should be taken. A crucial question is whether we need a minimal threshold of scientific plausibility before initiating preventative action. In practice, suggestions for the type of preventative action to be taken range from implementing a ban, imposing a moratorium while further research is conducted, allowing the potentially harmful activity to proceed while closely monitoring its effects, to just conducting more research.

Philosophy of Biotechnology

1333

Environmentalist organizations like Greenpeace solve the definitional ambiguities of the PP by adopting a very strong and absolute version of it. On their view, even the mere possibility that a particular product or activity might do harm to human health or the natural environment is sufficient to demand an outright ban on the incriminated product or activity. If the prospective harm is of catastrophic size, even an infinitesimally low probability of occurrence would not annul our supreme obligation to avoid this harm. Greenpeace looks at the cultivation of GM crops in this apocalyptic light: agricultural biotech contains the potential seeds of ecological destruction. Once the genie is out of the bottle, it cannot be recalled. The strong and absolute version of the PP embraced by Greenpeace and other environmentalist organizations exhibits a remarkable resemblance to another example of a ‘zero-infinity dilemma’ that is well-known in the history of philosophy, namely Pascal’s famous ‘wager’. When it comes to wagering on the existence of God, the seventeenth-century French philosopher argued incisively in his Pens´ees that it is better to be safe than sorry. Given an unknown but nonzero probability of God’s existence and the infinity of the reward of an eternal life, the rational choice would be to conduct one’s earthly life as if God exists. Pascal’s reasoning, however, is vulnerable to the “many gods” objection. Consider the possible existence of another deity than God, say Odin. If Odin is jealous, he will resent our worship of God, and we will have to pay an infinite price for our mistake. Never mind that Odin’s existence may not seem likely or plausible to us. It is sufficient that we cannot exclude the possibility that he exists with absolute certainty. Therefore, the very same logic of Pascal’s wager would lead us to adopt the opposite conclusion not to worship God. Pascal’s argument, then, cannot be valid. For the same reasons, the strong version of the PP is logically untenable (for a more extensive discussion of the PP, see [Van den Belt, 2003]). Just as Pascal’s wager argument leads to contradictory advice (worshipping and not worshipping God), the strong version of the PP can be construed as leading to the following contradictory recommendations: (a) We must not develop GM crops, and (b) We must develop GM crops. The first alternative is argued vehemently by Greenpeace and other anti-biotech groups. Philosopher Gary Comstock provides backing for the second alternative. He conjures up a dramatic scenario in which people are forced to seize upon the remaining reserves of nature on earth in a desperate effort to overcome food shortages resulting from global warming. He then argues, in the style of the anti-biotech groups, that “lack of full scientific certainty that GM crops will prevent environmental degradation shall not be used as a reason for postponing this potentially cost-effective measure” [Comstock, 2000].

4.3

Consumer autonomy and labelling

Consumer freedom of choice or autonomy is an important principle underpinning democratic market economies. One would think that this freedom includes the possibility to choose either GM or non-GM foods for consumption, irrespective of what nutritionists or other experts think about the merits of the two categories

1334

Henk van den Belt

(for consumers, several other values may be at stake than just health). To exercise this choice, consumers need to be able to tell GM from non-GM foods. Hence the need for mandatory labelling of GM foods. Although the ethical case for labelling looks simple and straightforward, the ‘product-based’ system for the assessment of biotechnology that was adopted in the mid-1980s as the basis of the US regulatory regime makes it virtually impossible to respect the American consumer’s “right to know”. In 1992 the Food and Drug Administration (FDA) indeed concluded that there should be no mandatory labelling for GM foods, a decision that has remained basically unaltered until this day. The US Biotechnology Industry Organization (BIO) summarizes the policy position of the FDA as follows: The FDA’s 1992 policy states that there is no reason to conclude that bioengineered foods differ from other foods in any meaningful or uniform way or that they present a greater safety concern than foods developed using traditional methods. Its position is that foods should be labeled according to their characteristics, not their method of production. Therefore, FDA does not require special labeling of biotechnology foods or the products of animals fed these foods. [Biotechnology Industry Organization, n.d.] For many consumers, however, there is an eminently “meaningful” difference between GM foods and foods made with traditional methods, even if there would be no difference at all in the physical characteristics of the end product (although in many cases the difference in the method of production also results in more or less significant physical differences of the end product). This difference in method of production could be “meaningful” and important to consumers/citizens for a great variety of underlying beliefs and values, yet the FDA and the US biotech industry wilfully deprive them of the crucial information that would enable them to incorporate their values in their purchasing decisions on the market. The irony is that in focus group discussions commissioned by the FDA itself, “virtually all participants said that bioengineered foods should be labelled as such” [Food and Drug Administration, 2000]. Participants were also presented with factual information about the extent to which GM foods had entered American food stores and supermarkets: [M]ost participants expressed great surprise that food biotechnology has become so pervasive in the U.S. food supply. [. . . ]. The typical reaction of participants was not one of great concern about the immediate health and safety effects of unknowingly eating bioengineered foods, but rather outrage that such a change in the food supply could happen without them knowing about it. [Food and Drug Administration, 2000] This sense of “outrage” is quite understandable given that the autonomy of consumers has been violated by companies and regulatory authorities implementing

Philosophy of Biotechnology

1335

and allowing massive changes in the food supply without their knowledge. In this connection an interesting and consistent finding from the series of public opinion polls that the Pew Initiative on Food and Biotechnology has published over the years is that a considerable percentage of American consumers actually believe that they have never eaten GM foods (in the November 2006 report this was 60%), while an estimated 75% of processed foods in grocery stores are GM foods or contain GM ingredients. Most, if not all, Americans must have eaten GM foods in one form or another, but only 26% say they believe to have eaten them [Pew Initiative, 2006, p. 2]. A large number of polls done by various groups (industry, news organisations, academics, etcetera) all point to the overwhelming public support for mandatory labelling among the American population [Streiffer and Rubel, 2004]. In a democratic society, such a strong public opinion should be a weighty prima facie reason to provide labelling. Yet there is no chance that Congress and the US government will act upon this strong ethical case for labelling. American biotech companies continue to oppose it, largely on the grounds that they fear that GM labels would induce the public to buy GM-free foods. In their view, GM labels are misleading, because they suggest that GM foods are inherently different from non-GM foods. The ethical case for labelling is largely based on deontological considerations. A utilitarian case against mandatory labelling may be built on the huge administrative and economic burdens that the practical implementation of such a labelling system is likely to entail. In the European Union labelling is part of a wider policy that also includes segregation and traceability (GM products must be traced along the food chain) and legal measures to ensure the possibility of “coexistence” between GM and conventional agriculture. There is no doubt that this particular set of policies imposes high costs and does not offer much encouragement to the cultivation and processing of GM crops in European countries, if only because of the liability rules vis-`a-vis conventional producers. In the end the freedom of the European consumer may de facto become limited to consuming non-GM foods.

4.4

The question of corporate control

Some critics oppose agricultural biotech not so much because of any inherent characteristics of this technology or its likely consequences for the natural environment or human health as on the grounds that its development and deployment occur at the behest of and to the benefit of a few powerful multinational corporations. This view has been forcefully expressed by the British publicist George Monbiot. For him the question is simply whether we want a few corporations to monopolise the global food supply. If not, we should reject agricultural biotech: GM technology permits companies to ensure that everything we eat is owned by them. They can patent the seeds and the processes that give rise to them. They can make sure that crops can’t be grown without their patented chemicals. They can prevent seeds from reproducing themselves. By buying up competing seed companies and closing them

1336

Henk van den Belt

down, they can capture the food market, the biggest and most diverse market of all. [Monbiot, 2004] For Monbiot, discussions about the possible contributions of agricultural biotechnology to feeding the hungry in the world and making Third World agriculture both more productive and more ecologically sustainable are just a diversion leading away from the main issue of corporate control. He even concedes that such contributions are feasible in principle and in practice: Now in some places governments or unselfish private researchers are producing GM crops that are free from patents and not dependent on the application of proprietary pesticides, and these could well be of benefit to small farmers in the developing world. [ibid.] If it is not so much the tool that is wrong as the purposes to which it is put by the powerful actors that wield it, one might think that giving up the tool itself is not the most sensible strategy. An alternative approach might be to wrest control over the tool from the incumbent possessors and deploy it for more benign purposes. We should also bear in mind that Monsanto and its likes (Syngenta, DuPont, Bayer . . . ) may indeed be immensely powerful, but they are not all-powerful. As we have seen above in Section 3.4, farmers in India, Argentina and Brazil already effectively challenge Monsanto’s alleged intellectual property rights. Greater involvement in plant breeding and biotechnology by government agencies could also help to tilt the balance of power more towards the public interest. The American molecular biologist Richard Jefferson, working at CAMBIA in Canberra, Australia, attributes public hostility towards GM crops also largely to the inordinate amount of power and control that big multinationals exercise over agricultural biotech, but he tries hard to wrest at least some part of the control from them. Jefferson considers the international patent system as a major obstacle for developing innovations in agricultural biotechnology that will benefit the poorest part of the world’s population. Not just various DNA sequences, but also the strategic techniques for inserting transgenes into plants (like the socalled ‘gene gun’ and the use of various Agrobacterium strains as Trojan horses) are tied up in patents. The huge transaction costs that must be made to obtain licenses have the unfortunate consequence that only the most spectacular potential ‘blockbusters’ are selected as targets for transgenic plant development and that less prominent crops are ignored. (“But blockbusters alone don’t make for good agriculture, good environmental management or good public health” [Jefferson, 2006, p. 21]). Jefferson holds that the early promises of agricultural biotechnology have actually been betrayed by the patenting frenzy: In 1985 the [biotech] sector was viewed as exhilarating, entrepreneurial and vibrant, with almost unlimited possibility for doing good in world agriculture; within a decade or so it had all but stalled into a corporate oligopoly, with vertical integration, ossified and oppressive business models, and massive patent portfolios tying up almost every key technology and platform used in the field. [Jefferson, 2006, p. 21]

Philosophy of Biotechnology

1337

Inspired by the Open Source Software Movement, Jefferson has set up his BiOS initiative (Biological Innovation for Open Society or Biological Open Source) at CAMBIA with support from the Rockefeller Foundation. The aim is to make and to keep the basic techniques of agricultural biotechnology, the ‘tools’ and the ‘technology platforms’, accessible to everybody. Firms are then free to acquire IP rights on the biotech ‘applications’ and ‘products’ which they might develop using these tools and platforms. Freeing the tools from the stranglehold of patents would make the development of numerous potential applications benefiting the poor and needy of the world economically viable. As a first practical contribution, Jefferson and his group have themselves already found a new bacterial vector for transferring genes into plants and made it available as a patent-free alternative to the patented Agrobacterium technique. It remains to be seen what difference initiatives like Jefferson’s will make for the further development of agricultural biotechnology in the developing world. The future may be more open-ended and surprising than the bleak vision of increasing corporate control suggests.

BIBLIOGRAPHY [Bakels, 2008] R. Bakels. Techniek, de Vierde Dimensie van het Octrooirecht. Maastricht (Thesis), 2008. [Benner, 2008] S. Benner. Biology from the bottom up. Nature, 452, 692-694, 2008. [Benner and Sismour, 2005] S. Benner and A. M. Sismour. Synthetic Biology. Nature Reviews Genetics, 6, 533-544, 2005. [Booth, 2008] R. Booth. Charles warns GM farming will end in ecological disaster. The Guardian, 13 August 2008. [Bernauer, 2003] T. Bernauer. Genes, Trade, and Regulation. Princeton University Press, 2003. [Biotechnology Industry organisation, 2001] Biotechnology Industry Organization. Labelling of Biotechnology Food Products (www.bio.org/foodag/background/foodlabeling.asp?p=yes). [Bovenkerk et al., 2002] B. Bovenkerk, F. W. A. Brom, and B. J. Van den Bergh, B.J. Brave New Birds: The Use of ‘Animal Integrity’ in Animal Ethics. Hastings Center Report, 32, 16-22, 2002. [Buchholz, 2007] K. Buchholz. Science — or not? The status and dynamics of biotechnology. Biotechnology Journal, 2, 1154-1168, 2007. [Bud, 1991] R. Bud. Biotechnology in the Twentieth Century. Social Studies of Science, 21, 415-457, 1991. [Bud, 1994] R. Bud. The Uses of Life: A History of Biotechnology. Cambridge University Press, 1994. [Calvert, 2007] J. Calvert. Patenting Genomic Objects: Genes, Genomes, Function and Information. Science as Culture, 16.2, 207-223, 2007. [Cantley, 2005] M. Cantley. In our own hands. Nature, 437, 193-94, 2005. [Castells, 2005] M. Castells, editor. The Network Society: A Cross-Cultural Perspective. Edward Elgar, 2005. [Center for Integrated Biotechnology, no date] Center for Integrated Biotechnology. Definition and Scope: What is Biotechnology? (no date) (www.biotechnology.wsu.edu/definition scope/). [Center for Food Safety, 2005] Center for Food Safety. Monsanto vs. U.S. Farmers. Washington, D.C., 2005. [Cohen, 1975] S. N. Cohen. The manipulation of genes. Scientific American, July 1975, 80-90. [Cohen and Morgan, 2008] S. Cohen and G. Morgan. Monsanto Technology LLC v. Cargill: a matter of construction. Nature Biotechnology, 26.3, 289-291, 2008.

1338

Henk van den Belt

[Comstock, 2000] G. Comstock. Are the policy implications of the precautionary principle coherent? Talk at Harvard University, 23 September 2000 (www.cid.harvard.edu/cidbiotech/comments/comments72.htm). [Crichton, 1991] M. Crichton. Jurassic Park. Arrow Books, 1991. [Crick, 1970] F. Crick. Central Dogma of Molecular Biology. Nature, 227, 561-563, 1970. [DeBeer, 2005] J. DeBeer. Reconciling Property Rights in Plants. The Journal of World Intellectual Property, 8.1, 5-31, 2005. [Demain and Fellmeth, 2003] L. J. Demaine and A. X. Fellmeth. Natural Substances and Patentable Inventions. Science, 300, 1375-1376, 2003. [De Vriend, 2007] H. De Vriend. Constructing Life: Early Social Reflections on the Emerging Field of Synthetic Biology. Rathenau Institute, 2007. [Doll, 1998] J. Doll. The Patenting of DNA. Science, 280, 689-690, 1998. [Douglas, 1966] M. Douglas. Purity and Danger. Routledge, 1966 [Drezner, 2007] D. Drezner. All Politics is Global. Princeton University Press, 2007. [Dutfield, 2004] G. Dutfield. From Mousetraps to (Onco)mice: The Law and Politics of Biotech Patenting. Draft, EASST/4S Conference, Paris, 24 August 2004. [Eisenberg, 2002] R. S. Eisenberg. How can you patent genes? In Who Owns Life?, D. Magnus, A.-L. Caplan and G. McGee, eds., pp. 117-134, Prometheus Books, 2002. [Endy, 2005] D. Endy. Foundations for Engineering Biology. Nature, 238, 449-453, 2005. [Engelbrecht, 2007] D. Engelbrecht. Patenting and the Informational Gene Concept: A Philosophical Exploration. Vrije Universiteit Amsterdam (Student Thesis), 2007. [ETC Group, 2007] ETC Group. Extreme Genetic Engineering: An Introduction to Synthetic Biology. Ottawa: ETC Group, 2007 (www.etcgroup.org/upload/publication/602/01/synbioreportweb.pdf). [European Patent Office, 2007] European Patent Office. Scenarios for the Future. EPO, 2007. [European Union, 2007] European Union. En Route to the Knowledge-Based Bio-Economy. Cologne Paper, 2007. [Food and Drug Administration, 2000] Food and Drug Administration. Report on Consumer Focus Groups on Biotechnology, October 20, 2000 (www.cfsan.fda.gov/ comm/biorpt.html). [Gallo and Kesan, 2006] A. A. Gallo and J. P. Kesan. Property Rights Legislation in Agricultural Biotechnology: United States and Argentina. Minnesota Law Journal of Science and Technology, 7.2, 565-598, 2006. [Gaskell et al., 2001] G. Gaskell, H. Torgerson, N. Allum, and M. W. Bauer. Biotechnology, technoscience and the public sphere. In Biotechnology 1996-2000: the years of controversy, G. Gaskell and M. Bauer, eds., pp. 116-127, Science Museum Press, 2001. [Gipstein, 2003] R. S. Gipstein. The isolation and purification exception to the general unpatentability of products of nature. Columbia Science and Technology Law Review, 4, 1-44, 2003. [Godfrey-Smith, 2006] P. Godfrey-Smith. Information in Biology. In The Cambridge Companion to the Philosophy of Biology, D. Hull & M. Ruse, eds., pp. 103-119, Cambridge University Press, 2006. [Goldstein and Golod, 2002] J. A. Goldstein and E. Golod. Human Gene Patents. Academic Medicine, 77, 1315-1328, 2002. [Graham, 2002] G. Graham. Genes: A Philosophical Inquiry. Routledge, 2002. [Griffiths, 2001] P. E. Griffiths. Genetic Information: A Metaphor in Search of a Theory. Philosophy of Science, 68, 394-412, 2001. [Gupta and Chandak, 2005] A. Gupta and V. Chandak. Agricultural biotechnology in India: ethics, business and politics. International Journal of Biotechnology, 7.1/3, 212-227, 2005. [Haraway, 1992] D. Haraway. The Promises of Monsters: A Regenerative Politics for Inappropriate/d Others. In Cultural Studies, L. Grossberg, C. Nelson, P.A. Teichler, eds., pp. 295-337, Routledge, 1992. [Herring, 2007a] R. J. Herring. The Genomics Revolution and Development Studies: Science, Poverty and Politics. Journal of Development Studies, 43.1, 1-30, 2007. [Herring, 2007b] R. J. Herring. Stealth Seeds: Bioproperty, Biosafety, Biopolitics. Journal of Development Studies, 43.1, 130-157, 2007. [Ihde and Selinger, 2003] D. Ihde and E. Selinger, eds. Chasing Technoscience. Matrix for Materiality. Indiana University Press, 2003. [Jaffe, 2000] A. B. Jaffe. The U.S. patent system in transition: policy innovation and the innovation process. Research Policy, 29, 531-557, 2000.

Philosophy of Biotechnology

1339

[Jasanoff, 2005a] S. Jasanoff. Designs on Nature: Science and Democracy in Europe and the United States. Princeton University Press, 2005. [Jasanoff, 2005b] S. Jasanoff. In the democracies of DNA: ontological uncertainty and political order in three states. New Genetics and Society, 24.2 , 139-155, 2005. [Jefferson, 2006] R. Jefferson. Science as Social Enterprise: The CAMBIA BiOS Initiative. Innovations, 1(4), 13-44, 2006. [Kane, 2004] E. M. Kane. Splitting the gene: DNA patents and the genetic code. Tennessee Law Review, 71, 707-767, 2004. [Kane, 2006] E. M. Kane. Patent Ineligibility: Maintaining a Scientific Public Domain. St. John’s Law Review, 80, 519, 2006. [Kanehisa and Bork, 2003] M. Kanehisa and P. Bork. Bioinformatics in the post-sequence era. Nature Genetics, 33, 305-310, 2003. [Kohler, 1900] J. Kohler. Handbuch des Deutschen Patentrechts in rechtsvergleichender Darstellung. J. Bensheimer, 1900. [Kranakis, 2007] E. Kranakis. Patents and Power: European Patent-System Integration in the Context of Globalization. Technology & Culture, 48, 689-728, 2007. [Kuypers, 1974] K. Kuypers. The Relation Between Knowing and Making as an Epistemological Principle. Philosophy and Phenomenological Research, 35.1, 60-78. 1974. [Latour, 1987] B. Latour. Science in Action. Open University Press, 1987. [Latour, 1993] B. Latour. We Have Never Been Modern. Harvard University Press, 1993. [Latour, 2004] B. Latour. The Politics of Nature: How to Bring the Sciences into Democracy. Harvard University Press, 2004. [Lezaun, 2004] J. Lezaun. Pollution and the Use of Patents: A Reading of Monsanto v. Schmeiser. In Biotechnology: Between Commerce and Civil Society, N. Stehr, ed., pp. 135-158, Transaction Publishers, 2004. [Marres, 2005] N. Marres. No Issue, No Public: Democratic Deficits after the Displacement of Politics. University of Amsterdam (Thesis), 2005. [Maynard Smith, 1999] J. Maynard Smith. The Idea of Information in Biology. The Quarterly Review of Biology, 74.4, 395-400, 1999. [Meijer and Stewart, 2004] E. Meijer and R. Stewart. The GM Cold War: How Developing Countries Can Go from Being Dominos to Being Players. RECIEL, 13.3, 247-262, 2004. [OECD, 2001] OECD. The Application of Biotechnology to Industrial Sustainability — A Primer, without place and year [2001]. [Monbiot, 2004] G. Monbiot. Starved of the truth. The Guardian, 9 March 2004. [Morton, 2005] O. Morton. Life, Reinvented. Wired Magazine, Issue 13.01, January 2005, (www.wired.com/wired/archive/13.01/mit pr.html). [Office of Technology Assessment, 1984] Office of Technology Assessment. Commercial Biotechnology: An International Analysis. U.S. Government Printing Office. 1984. [O’Malley et al., 2005] M. O’Malley, A. Bostanci, and J. Calvert. Whole-genome patenting. Nature Reviews Genetics, 6, 502-507, 2005. [Peters, 2005] T. Peters. ITE Brief on Science in Pastoral Ministry, 12 October 2005 (www.plts.edu/docs/ite science.pdf). [Pew Initiative, 2006] Pew Initiative on Food and Biotechnology. Public Sentiment about Genetically Modified Food (November 2006 report) (http://pewagbiotech.org/research/2006update/2006summary.pdf). [Pickstone, 2000] J. V. Pickstone. Ways of Knowing: A New History of Science, Technology and Medicine. The University of Chicago Press, 2000. [Prince of Wales, 1998] Prince of Wales. The Seeds of Disaster. The Daily Telegraph, 8 June 1998. [Radder, 2004a] H. Radder. Review of Don Ihde and Evan Selinger (eds.), Chasing Technoscience. Matrix for Materiality. Philosophy of Science, 71.4, 614-619, 2004. [Radder, 2004b] H. Radder. Exploiting abstract possibilities: a critique of the concept and practice of product patenting. Journal of Agricultural and Environmental Ethics, 17, 275-291, 2004. [Rai, 1999] A. K. Rai. Intellectual Property Rights in Biotechnology: Addressing New Technology. Wake Forest Law Review, 34, 827-847, 1999. [Raustiala and Victor, 2004] K. Raustiala and D. G. Victor. The Regime Complex for Plant Genetic Resources. International Organization, 58, part 2, 277-310, 2004.

1340

Henk van den Belt

[Resnik, 2002] D. B. Resnik. Discoveries, Inventions, and Gene Patents. In Who Owns Life?, D. Magnus, A.-L. Caplan and G. McGee, eds., pp. 135-150, Prometheus Books, 2002. [Rifkin, 1984] J. Rifkin. Algeny: A New Word — A New World. Penguin Books, 1984. [Riis, 2008] S. Riis. The Symmetry Between Bruno Latour and Martin Heidegger. Social Studies of Science, 38.2, 285-301, 2008. [Ro et al., 2006] D.-K. Ro, E. M. Paradise, and M. Ouellet, et al. Production of the antimalarial drug precursor artemisinic acid in engineered yeast. Nature, 440, 940-943, 2006. [Sagoff, 2001] M. Sagoff. Genetic Engineering and the Concept of the Natural. Philosophy & Public Policy Quarterly, 21 (2/3), 2-10, 2001. [Sarkar, 1996] S. Sarkar. Decoding “coding” — information and DNA. BioScience, 46.11, 857864, 1996. [Schr¨ odinger, 1944] E. Schr¨ odinger. What is Life? Cambridge University Press, 1944. [SCR, 2004] S.C.R. 1. Monsanto Canada Inc. v. Schmeiser (2004) (www.lexum.umontreal.ca/cscscc/en/pub/2004/vol1/html/2004scr1 0902.html). [Sinsheimer, 1976] R. L. Sinsheimer. Recombinant DNA — On Our Own. BioScience, 26, 599, 1976. [Silver, 2004] L. Silver. The God Effect. Newsweek, 5 April 2004. [Silver, 2007] L. Silver. Scientists Push the Boundaries of Human Life. Newsweek International, June 4, 2007. [Smits, 2002] M. Smits. Monsterbezwering: de Culturele Domesticatie van Nieuwe Technologie. Boom, 2002. [Sokal and Bricmont, 1998] A. Sokal and J. Bricmont. Impostures Intellectuelles. Odile Jacob, 1998. [Streiffer and Rubel, 2004] R. Streiffer and A, Rubel. Democratic Principles and Mandatory Labeling of Genetically Engineered Food. Public Affairs Quarterly, 18, 223-248, 2004. [Sulston, 2006] J. Sulston. Staking claims in the biotechnology Klondike. Bulletin of the World Health Organization, 84.5, 412-414, 2006. [Sulston and Ferry, 2003] J. Sulston and G. Ferry. The Common Thread: Science, politics, ethics and the Human Genome. Corgi Books, 2003. [US Supreme Court, 1980] U.S. Supreme Court Diamond v. Chakrabarty, 447 U.S. 303. (1980) (http://caselaw.lp.findlaw.com/scripts/getcase.pl?court=us& vol=447& invol=303). [Thacker, 2005] E. Thacker. The Global Genome: Biotechnology, Politics, and Culture. The MIT Press, 2005. [Travis, 1993] A. S. Travis. The Rainbow Makers: The Origins of the Synthetic Dyestuffs Industry in Western Europe. Lehigh University Press, 1993. [Van den Belt, 2003a] H. Van den Belt. How to engage with experimental practices? Moderate versus radical constructivism. Journal for General Philosophy of Science, 34, 201-219, 2003. [Van den Belt, 2003b] H. Van den Belt. Debating the Precautionary Principle: “Guilty until Proven Innocent” or “Innocent until Proven Guilty”? Plant Physiology, 132, 1122-1126, 2003. [Venter, 2007] J. C. Venter. A DNA-driven world. The 32nd Richard Dimbleby Lecture. BBC One, December 4, 2007 (www.bbc.co.uk/pressoffice/pressreleases/stories/2007/12 december/05/dimbleby.shtml). [Watson, 2004] J. Watson. DNA: The Secret of Life. Arrow Books, 2004. [Williams, 2006] R. Williams. Compressed Foresight and Narrative Bias: Pitfalls in Assessing High Technology Futures. Science as Culture, 15.4, 327-348, 2006. [Winickoff et al., 2005] D. Winickoff, S. Jasanoff, L. Busch, R. Grove-White, and B. Wynne. Adjudicating the GM Food Wars: Science, Risk, and Democracy in World Trade Law. The Yale Journal of International Law, 30, 81-123, 2005. [Yoxen, 1983] E. Yoxen. The Gene Business: Who Should Control Biotechnology? Pan Books, 1983. [Ziff, 2005] B. Ziff. Travels with my plant: Monsanto v. Schmeiser revisited. University of Ottawa Law and Technology Journal, 2.2, 493-509, 2005. [Zika et al., 2007] E. Zika, I. Papetryfon, O. Wolf, M. Gomez-Barbero, A. J. Stein, and A.K. Bock. Consequences, Opportunities and Challenges of Modern Biotechnology for Europe. Office for Official Publications of the European Communities, 2007.

PHILOSOPHY OF COMPUTING AND INFORMATION TECHNOLOGY Philip Brey and Johnny Hartz Søraker Philosophy has been described as having taken a ‘computational turn’, referring to the ways in which computers and information technology throw new light upon traditional philosophical issues, provide new tools and concepts for philosophical reasoning, and pose theoretical and practical questions that cannot readily be approached within traditional philosophical frameworks. As such, computer technology is arguably the technology that has had the most profound impact on philosophy. Philosophers have studied computer technology and its philosophical implications extensively, and this chapter gives an overview of the field. We start with definitions and historical overviews of the field and its various subfields. We then consider studies of the fundamental nature and basic principles of computing and computational systems, before moving on to philosophy of computer science, which investigates the nature, scope and methods of computer science. Under this heading, we will also address such topics as data modeling, ontology in computer science, programming languages, software engineering as an engineering discipline, management of information systems, the use of computers for simulation, and human-computer interaction. Subsequently, we will address the issue in computing that has received the most attention from philosophers, artificial intelligence (AI). The purpose of this section is to give an overview of the philosophical issues raised by the notion of creating intelligent machines. We consider philosophical critiques of different approaches within AI and pay special attention to philosophical studies of applications of AI. We then turn to a section on philosophical issues pertaining to new media and the Internet, including the convergence between media and digital computers. The theoretical and ethical issues raised by this relatively recent phenomenon are diverse. We will focus on philosophical theories of the ‘information society’, epistemological and ontological issues in relation to Internet information and virtuality, the philosophical study of social life online and cyberpolitics, and issues raised by the disappearing borders between body and artifact in cyborgs and virtual selves. The final section in this chapter is devoted to the many ethical questions raised by computers and information technology, as studied in computer ethics. 1

INTRODUCTION

Philosophers have discovered computers and information technology (IT) as research topics, and a wealth of research is taking place on philosophical issues in Handbook of the Philosophy of Science. Volume 9: Philosophy of Technology and Engineering Sciences. Volume editor: Anthonie Meijers. General editors: Dov M. Gabbay, Paul Thagard and John Woods. c 2009 Elsevier BV. All rights reserved. 

1342

Philip Brey and Johnny Hartz Søraker

relation to these technologies. The research agenda is broad and diverse. Issues that are studied include the nature of computational systems, the ontological status of virtual worlds, the limitations of artificial intelligence, philosophical aspects of data modeling, the political regulation of cyberspace, the epistemology of Internet information, ethical aspects of information privacy and security, and many more. There are specialized journals, conference series, and academic associations devoted to philosophical aspects of computing and IT as well as a number of anthologies and introductions to the field [Floridi, 1999, 2004; Moor and Bynum, 2002], and the number of publications is increasing every year. Philosophers have not agreed, however, on a name for the field that would encompass all this research. There is, to be fair, not a single field, but a set of loosely related fields — such as the philosophy of artificial intelligence, computer ethics and the philosophy of computing — which are showing some signs of convergence and integration. Various names have been considered for such a field, including “philosophy of computing”, “philosophy of computer science,” “cyberphilosophy” and “philosophy of information technology”. We find none of these names sufficiently broad in scope. Without aiming to settle the issue for good, we propose philosophy of computing and information technology. We define philosophy of computing and IT as the study of philosophical issues in relation to computer and information systems, their study and design in the fields of computer science and information systems, and their use and application in society. We propose that on the basis of both conceptual and historical grounds this area can be divided up into five subfields, which we will survey in the following five sections. They are the philosophy of computing (section 2), the philosophy of computer science (section 3), the philosophy of artificial intelligence (AI) (section 4), the philosophy of new media and the Internet (section 5), and computer and information ethics (section 6). Conceptually, these areas have distinct subject matters and involve distinct philosophical questions, as we will try to show in these sections. Historically, the philosophy of AI is the oldest area within philosophy of computing and IT; taking shape in the 1960s, and maturing through the 1970s and 1980s. Philosophy of AI is the philosophical study of machine intelligence and its relation to human intelligence. It is an area of philosophy that emerged in close interaction with development in the field of artificial intelligence. The philosophy of AI studies whether computational systems are capable of intelligent behavior and human-like mental states, whether human and computer intelligence rest on the same basic principles, and studies conceptual and methodological issues within various approaches in AI. In the philosophy of computing significant work is being done since at least the 1970s. The philosophy of computing studies fundamental concepts and assumptions in the theory of computing, including the notions of a computational system, computation, algorithm, computability, provability, computational complexity, data, information, and representation. As such, it is the philosophical cousin of theoretical computer science. This area, which is more loosely defined and contains much less research than the philoso-

Philosophy of Computing and Information Technology

1343

phy of AI, is the product of three historical developments. First, the philosophy of AI necessitated an understanding of the nature of computational systems, and some philosophers of AI consequently devoted part of their research to this issue. Second, philosophically minded computer scientists working in theoretical computer science occasionally started contributing to this area. Lastly, philosophers working in philosophical logic and philosophy of mathematics started considering fundamental issues in computing that seemed to be an extension of the issues they were studying, such as issues in computability and provability of algorithms. By the late 1980s, the landscape of philosophical research on computers and IT consisted almost entirely of studies on AI and theoretical issues in computing. But grounds were shifting. With the emergence of powerful personal computers and the proliferation of usable software, computers were becoming more than an object of study for philosophers, they were becoming devices for teaching and aids for philosophical research. In addition, philosophers were becoming increasingly concerned with the social impact of computers and with ethical issues. Playing into this development, some philosophers started advancing the claim that philosophy was gearing up for a “computational turn”, an expression first introduced by Burkholder [1992] and further advanced by Bynum and Moor [1998]. The computational turn in philosophy is a perceived development in which an orientation towards computing would transform the field in much the same way that an orientation towards language restructured the field in the so-called linguistic turn in twentieth-century Anglo-American philosophy. At the heart of the argument for the computational turn was that computing did not just constitute an interesting subject matter for philosophy, but that it also provided new models and methods for approaching philosophical problems [Moor and Bynum, 2002]. The application of computational tools to philosophy has been called computational philosophy. Computational philosophy regards the computer as “a medium in which to model philosophical theories and positions” [Bynum and Moor, 1998, p. 6] that can serve as a useful addition to thought experiments and other traditional philosophical methods. In particular, the exploration of philosophical ideas by means of computers allows us to create vastly more complex and nuanced thought experiments that must be made “in the form of fully explicit models, so detailed and complete that they can be programmed” [Grim, Mar and St. Denis, 1998, p. 10]. In addition it is usually possible to make (real-time) changes to the model, and thereby “explore consequences of epistemological, biological, or social theories in slightly different environments” [Grim, 2004, p. 338]. Thus, computer modeling has successfully been applied to philosophy of biology (see also Section 4.2), economics, philosophy of language, physics and logic. Thagard has also pioneered a computational approach to philosophy of science, arguing that computational models can “illustrate the processes by which scientific theories are constructed and used [and] offers ideas and techniques for representing and using knowledge that surpass ones usually employed by philosophers” [1988, p. 2]. Another area in which computer modeling has been employed is ethics. For instance, Danielson [1998] argues that computational modeling of ethical scenarios can help us keep

1344

Philip Brey and Johnny Hartz Søraker

our theories open to counter-intuitive ideas and serve as checks on consistency. Closely related, computer models have also been used to explore topics in social philosophy, such as prejudice reduction [Grim et al., 2005]. Despite the significant advantages, computational philosophy also has limitations. Importantly, it is limited to those kinds of philosophical problems that lend themselves to computational modeling. Additionally, addressing a problem by means of a computer leads to a very specific way of asking questions and placing focus, which might not be equally helpful in all cases. For instance, theories of social dynamics can most easily be computationally modeled by means of rational choice theory, due to its formal nature, which in itself contains particular assumptions that could influence the results (such as methodological individualism). Another problem is that computational modeling can in some cases run counter to fundamental philosophical ideals, because computational models are often built upon earlier computational models or libraries of pre-programmed constructs and, as such, a number of unexamined assumptions can go into a computational model (cf. [Grim, 2004, pp. 339-340]). There are hence reasons for caution in the performance of a computational turn in philosophy. As a matter of fact, the impact of computational modeling on philosophy is as of yet quite limited. Nevertheless, the notion of a computational turn is referred to explicitly in the mission statement of the International Association of Computing and Philosophy (IACAP). IACAP, a leading academic organization in the field, was founded in 2004. It was preceded by a conference series in computing and philosophy that started in 1986. In its mission statement, IACAP refers to a field of “computing and philosophy” that encompasses any interaction between philosophy and computing, including both the philosophy of computing and IT, as defined earlier, and computational philosophy, along with other uses of computers for the benefit of philosophy. In spite of this significant philosophical interest in computer systems, artificial intelligence, and computational modeling, it was not until the late 1990s that philosophers started to pay serious attention to computer science itself, and to develop a true philosophy of computer science. The philosophy of computer science can be defined, in analogy with the philosophy of physics or the philosophy of biology, as the philosophical study of the aims, methods and assumptions of computer science. Defined in this way, it is a branch of the philosophy of science. Work in the philosophy of computing did not, or hardly, address questions about the nature of computer science, and the philosophy of AI limited itself to the nature and methods of only one field of computer science, AI. The relative neglect of computer science by philosophers can perhaps be explained in part by the fact that the philosophy of science has tended to ignore applied science and engineering. The philosophy of science has consistently focused on sciences that aim to represent reality, not on fields that model and design artifacts. With its aim to investigate the nature of intelligence, AI was the only field in computer science with a pretense to represent reality, which may account for much of the attention it received. Other fields of computer science were more oriented

Philosophy of Computing and Information Technology

1345

towards engineering. In addition, computer science did not have many developed methodologies that could be studied. Yet, since the late 1999s, there has been a trickle of studies that do explicitly address issues in computer science [Longo, 1999; Colburn, 2000; Rapaport, 2005; Turner and Eden, 2007a, b], and even an entry in the Stanford Encyclopedia of Philosophy [Turner and Eden, forthcoming b]. The philosophy of computer science is shaping up as a field that includes issues in the philosophy of computing, but that also addresses philosophical questions regarding the aims, concepts, methods and practices of computer science. In Section 3, we use the limited amount of literature in this area to lay out a set of issues and problems for the field. The rise of the personal computer and multimedia technology in the 1980s and the Internet and World Wide Web in the 1990s ushered in a new era in which the computer became part of everyday life. This has brought along major changes in society, including changes in the way people work, learn, recreate and interact with each other, and in the functioning of organizations and social and political institutions. It has even been claimed that these technologies are fundamentally changing human cognition and experience. These social and cultural changes have prompted philosophers to reflect on different aspects of the new constellation, ranging from the epistemology of hyperlinks to the ontology of virtual environments and the value of computer-mediated friendships. We tie these different investigations together under the rubric philosophy of the Internet and new media. Whereas most work in other areas discussed here has been in the analytic tradition in philosophy, a large part of the research in this area is taking place in the Continental tradition, and includes phenomenological, poststructuralist and postmodernist approaches. Additionally, philosophical work in this area is often affiliated with work in social theory and cultural studies. Computer ethics, a fifth area to be surveyed, started out in the late 1970s and gained traction in the mid-1990s, quickly establishing itself as a field with its own journals and conference series. Computer ethics developed largely separately from other areas in the philosophy of computing and IT. Its emergence was driven by concerns of both computer scientists and philosophers about social and ethical issues relating to computers and to address issues of professional responsibility for computer professionals. While its initial emphasis was on professional ethics, it has since broadened to include ethical issues in the use and regulation of information technology in society. 2

PHILOSOPHY OF COMPUTING

Philosophy of computing is the investigation of the basic nature and principles of computers and the process of computation. Although the term is often used to denote any philosophical issue related to computers, we have chosen to narrow this section to issues focusing specifically on the nature, possibilities and limits of computation. In this section, we will begin by giving an outline of what a computer is, focusing primarily on the abstract notion of computation developed by Turing.

1346

Philip Brey and Johnny Hartz Søraker

We will then consider what it means for something to be computable, outline some of the problems that cannot be computed, and discuss forms of computation that go beyond Turing. Having considered which kinds of problems are Turing non-computable in principle, we then consider problems that are so complex that they cannot be solved in practice. Finally, computing is always computing of something; hence we will conclude this section with a brief outline of central notions like data, representation and information. Since these issues constitute the basics of computing, the many philosophical issues are raised in different contexts and surface in one way or another in most of the following sections. We have chosen to primarily address these issues in the contexts in which they are most commonly raised. In particular, computer science is addressed in Section 3, the limits of computation are further addressed in Section 4 on artificial intelligence, and many of the issues regarding computers as (networked) information technologies are discussed in Section 5.

2.1 Computation, computational systems, and Turing machines At the most fundamental level, philosophy of computing investigates the nature of computing itself. In spite of the profound influence computational systems have had in most areas of life, it is notoriously difficult to define terms like ‘computer’ and ‘computation’. At its most basic level, a computer is a machine that can process information in accordance with lists of instructions. However, among many other variations, the information can be analogue or digital, the processing can be done sequentially or in parallel, and the instructions (or, the program) can be more or less sensitive to non-deterministic variables such as user input (see also 2.2 and 4.3). Furthermore, questions regarding computation are sometimes framed in normative terms, e.g. whether it should be defined so as to include the human brain (see Section 4) or the universe at large (see e.g. [Fredkin, 2003]). At the same time, claims to the effect that computers have had a profound influence on modern society presuppose that there is a distinctive class of artifacts that are computers proper. Indeed, the work of Alan Turing pioneered this development and his notion of a Turing Machine is often invoked in order to explain what computation entails. Turing’s way of defining computation, in effect, was to give an abstract description of the simplest possible device that could perform any computation that could be performed by a human computer, which has come to be known as a Turing machine [Turing, 1937]. A Turing machine is characterized as “a finite-state machine associated with an external storage or memory medium” [Minsky, 1967, p. 117]. It has a read/write head that can move left and right along an (infinite) tape that is divided into cells, each capable of bearing a symbol (typically, some representation of ‘0’ and ‘1’). Furthermore, the machine has a finite number of transition functions that determines whether the read/write head erases or writes a ‘0’ or a ‘1’ to the cell, and whether the head moves to the left or right along the tape. In addition to these operations, the machine can change its internal state, which allows it to remember some of the symbols it has seen previously. The in-

Philosophy of Computing and Information Technology

1347

structions, then, are of the form, “if the machine is in state a and reads a ‘0’ then it stays in state a and writes a ‘1’ and moves one square to the right”. Turing then defined and proved the existence of one such machine that can be made to do the work of all: a Universal Turing Machine (UTM). Von Neumann subsequently proposed his architecture for a computer that can implement such a machine — an architecture that underlies computers to this day. The purely abstract definition of ‘computation’ raises a number of controversial philosophical and mathematical problems regarding the in-principle possibility of solving problems by computational means (2.2) and the in-practice possibility of computing highly complex algorithms (2.3). However, it is still debatable whether UTMs really can perform any task that any computer, including humans, can do (see Sections 2.2 and 4). Sloman [2002] and others have argued that computation, understood in the abstract syntactic terms of a Turing machine or lambda calculus, are simply too far removed from the embodied, interactive, physically implemented and semantic forms of computation at work in both real-world computers and minds [Scheutz, 2002, p. x]. That is, although computation understood in terms of a Turing machine can yield insights about logic and mathematics, it is entirely irrelevant to the way computers are used today — especially in AI research.

2.2

Computability and the Church–Turing thesis

Computability refers to the possibility of solving a mathematical problem by means of a computer, which can either be a technological device or a human being. The discussion surrounding computability in mathematics had partly been fuelled by the challenge put forward by mathematician David Hilbert to find a procedure by which one can decide in a finite number of operations whether a given first-order logical expression is generally valid or satisfiable [Hilbert and Ackermann, 1928, pp. 73-74]. This challenge, known as the Entscheidungsproblem, led to extensive research and discussion. However, in the 1930’s, Church and Turing independently proved that the Entscheidungsproblem is unsolvable; Church in terms of lambda calculus and Turing in terms of computable functions on a Turing machine (which were also shown to be equivalent). In part due to the seminal work of Church and Turing, effectiveness has become a condition for computability. A method is judged to be effective if it is made up of a finite number of exact instructions that require no insight or ingenuity on the part of the computer and can be carried out by a human being with only paper and pencil as tools. In addition, when such a method is carried out it should lead to the desired result in a finite number of steps. The Church–Turing thesis holds that a Universal Turing Machine (UTM) is able to perform any calculation that any human computer can carry out (but see Shagrir [2002] for a distinction between the human, the machine and the physical version of the thesis). An equivalent way of stating the thesis is that any effectively computable function can be carried out by the UTM. On the basis of the Church Turing thesis it became possible to establish whether an effective method existed for a certain mathematical task by

1348

Philip Brey and Johnny Hartz Søraker

showing that a Turing Machine Program could or could not be written for such a task. The thesis backed by ample evidence soon became a standard for discussing effective methods The development of the concept of the Universal Turing Machine and the Church Turing thesis made it possible to identify problems that cannot be solved by Turing Machines. One famous example, and one of Turing’s answers to the Entscheidungsproblem, is known as the halting problem. This involves deciding whether any arbitrarily chosen Turing machine will at some point halt, given a description of the program and its input. Sometimes the machine’s table of instructions might provide insight, but this is often not the case. In these cases one might propose to watch the machine run to determine whether it stops at some point. However what conclusion can we draw when the machine is running for a day, a week or even a month? There is no certainty that it will not stop in the future. Similar to the halting problem is the printing problem where the challenge is to determine whether a machine will at some point print ’0’. Turing argued that if a Turing machine would be able to tell for any statement whether it is provable through first-order predicate calculus, then it would also be able to tell whether an arbitrarily chosen Turing machine ever prints ’0’. By showing that first-order predicate calculus is equivalent to the printing problem, Turing was able to transfer the undecidability result for the latter to the former [Galton, 2005, p. 94]. Additionally, Turing argued that numbers could be considered computable if they could be written by a Turing machine. However since there are only countably many different Turing-machine programs, there are also only countably many computable numbers. Since there are uncountably many real numbers, not all real numbers are computable simply because there are not enough Turing machines to compute them [Barker-Plummer, 2007]. Rice’s theorem [Rice, 1953] goes even further and states that there is no algorithm that can decide any non-trivial property of computations [Harel, 2000, p. 54]. Thus, it is important to recognize that the undecidability problems outlined above, and many more, are not of mere theoretical interest. Undecidability is not an exception, it is the rule when it comes to algorithmic reasoning about computer programs (cf. [Harel and Feldman, 2004; Harel, 2000]). As these examples show, the Turing machine and the Church–Turing thesis are powerful constructs and can provide deep insights into the nature of computation as well as notions well beyond philosophy of computing. Indeed, Copeland [2004] has argued that some have taken it too far, pointing out many misunderstandings and unsupported claims surrounding the thesis. In particular, many have committed the “Church–Turing fallacy” by claiming that any mechanical model, including the human brain, must necessarily be Turing-equivalent and therefore in-principle possible to simulate on a Turing machine [Copeland, 2004, p. 13]. This claim, sometimes distinguished as the strong Church–Turing thesis, presupposes that anything that can be calculated by any machine is Turing computable, which is a much stronger claim than the thesis that any effective method (one that could in-principle be carried out by an unaided human) is Turing computable.

Philosophy of Computing and Information Technology

1349

Although Turing proved that problems like the halting problem are unsolvable on any Turing machine, alternative forms of computation have been proposed that could go beyond the limits of Turing-computability — so-called hypercomputation (for an overview of the history of hypercomputation and its challenges, see [Copeland, 2002a]). Questions surrounding hypercomputation are primarily of theoretical importance, however, since there is still substantial disagreement on whether a genuine hypercomputer can actually be realized in the physical world (cf. [Shagrir and Pitowsky, 2003; Hagar, 2007; Copeland and Shagrir, 2007]). The question is also closely related to pancomputationalism and the question whether the universe itself is (hyper-) computational in nature (see e.g. [Lloyd, 2006; Dodig-Crnkovic, 2006]). MacLennan [2003] has also argued that although Turingcomputability is relevant to determining effective computability in logic and mathematics, it is irrelevant when it comes to real-time, continuous computation — such as the kind of natural computation found in nature. He further outlines theoretical work that has shown that certain analogue computers can produce non-Turing computable solutions and solve problems like the halting problem.

2.3

Computational complexity

Even in cases where it is in-principle possible to compute a given function, there still remains a question whether it is possible in practice. Theories of computational complexity are concerned with the actual resources a computer requires to solve certain problems, the most central resources being time (or the number of operations required in the computation) and space (the amount of memory used in the computation). Urquhart [2004] argues that complexity is important to philosophy in general, because many philosophical thought experiments do depend on computational resources for their feasibility. If we do take complexity into account, it becomes possible to differentiate between constructs that only exist in a purely mathematical sense and ones that can actually be physically constructed — which in turn can determine the validity of the thought experiment. Computational complexity theory has shown that the set of problems that are solvable fall into different complexity classes. Most fundamentally, a problem can be considered efficiently solvable if it requires no more than a polynomial number of steps, even in worst-case scenarios. This class is known as P. To see the difference between efficiently solvable and provably hard problems, consider the difference between an algorithm that requires a polynomial (e.g. n2 ) and one that requires an exponential (e.g. 2n ) number of operations. If n=100, the former amounts to 10.000 steps whereas the latter amounts to a number higher than the number of microseconds elapsed since the Big Bang. Again, the provably hard problems are not exceptions; problems like Chess and complex route planning can only be achieved by simplified shortcuts that often miss the optimal solution (cf. [Harel, 2000, pp. 59-89]). Some problems are easily tractable and some have been proven to require resources way beyond the time and space available. Sometimes, however, it remains

1350

Philip Brey and Johnny Hartz Søraker

a mystery whether there is a tractable solution or not. The class of NP refers to problems where the answer can be verified for correctness in polynomial time — or, in more formal terms, the set of decision problems solvable in polynomial time by a non-deterministic Turing machine. A non-deterministic Turing machine differs from a normal/deterministic Turing machine in that it has several possible actions it might choose when it is in a certain state receiving certain input. As a result, the time it would take a non-deterministic Turing machine to compute an NP problem would be the number of steps needed in the sequence that leads to the correct answer. That is, the sequences that turn out to be false do not count towards the number of steps needed to solve the problem, as they do in a normal, deterministic machine. Another way of putting it is to say that the answer to an NP problem can be verified for correctness in polynomial time, but the answer itself cannot necessarily be computed in polynomial time (on a deterministic machine). The question, then, becomes: If a given NP problem can be solved in polynomial time on such a machine, is it possible to solve it in polynomial time on a deterministic machine as well? This is of particular importance when it comes to so-called NP-complete problems. A problem is NP-complete when it is in NP and all other NP problems can be reduced to it by a transformation computable in polynomial time. Consequently, if it can be shown that any of the NP-complete problems can be solved in polynomial time, then all NP problems can; P=NP. Such proof would have vast implications, but in spite of tremendous effort and the large class of such problems, no such solution has been found. As a result, many believe that P =NP, and many important problems are thus seen as being intractable. On the positive side, this feature forms the basis of many encryption techniques (cf. [Harel, 2000, pp. 157ff]). Traditionally, the bulk of complexity theory has gone into the complexity of sequential computation, but parallel computation is getting more and more attention in both theory and practice. Parallel computing faces several additional issues such as the question of the amount of parallel processors required to solve a problem in parallel, as well as questions relating to which steps can be done in parallel and which need to be done sequentially.

2.4 Data, information and representation s Although ‘data’ and ‘information’ are among the most basic concepts in computing, there is little agreement on what these concepts refer to, making the investigation of the conceptual nature and basic principles of these terms one of the most fundamental issues in philosophy of computing. In particular, philosophy of information has become an interdisciplinary field of study on its own, often seen as going hand in hand with philosophy of computing. The literature on ‘information’ and related concepts, both historically and contemporary, is vast and cannot be done justice to within this scope (Volume 8 [Adriaans and van Benthem, 2009] in this series is dedicated to philosophy of information. See also [Bar-Hillel, 1964; Dretske, 1981; Floridi, 2004a; 2004b; 2007]). In short, the fundamental question

Philosophy of Computing and Information Technology

1351

in this field is “what is the nature of information?” This question is not only itself illuminated by the nature of computation, but the ‘open problems’ (cf. [Floridi, 2004a]) in philosophy of information often involve the most fundamental problems in computing, many of which are addressed in other sections (see especially 2.2, 3.3, 3.5, 4.4 and 5.2). It should also be pointed out that this is an area in which philosophy of computing not only extends far beyond computational issues, but also closely intersects with communication studies, engineering, biology, physics, mathematics and cognitive science. Although it is generally agreed that there can be no information without data, the exact relation between the two remains a challenging question. If we restrict ourselves to computation, it can be added that the data that constitute information must somehow be physically implemented. In practice, data is implemented (or encoded) in computers in binary form, i.e. as some representation of 1 or 0 (on or off), referred to as a bit. This satisfies the most fundamental definition of a datum, being “a lack of uniformity between two signs” [Floridi, 2004b, p. 43]. Furthermore, a string of these bits can represent, or correspond to, specific instructions or information. For instance, a computer can be given the instruction ‘1011000001100001’ corresponding to a particular operation, and a computer program can interpret the string ‘01100001’ as corresponding to the letter ‘a’. This underlines, however, that when dealing with questions regarding data, information and representation, it is important to emphasize that there are different levels of abstraction. For instance, a physical object can be represented by a word or an image, which in turn can be represented by a string of binary digits, which in turn can be represented by a series of on/off switches. Programming the computer and entering data can be done at different abstraction levels, but the instructions and data have to be converted into machine-readable code (see Section 3.2). The level at which we are operating will determine the appropriate notion of ‘representation’, what it entails to be well-formed and meaningful and whether or not the information must be meaningful to someone. As can be seen in the remarks above, there are at least three requirements for something to be information, which is known as the General Definition of Information (GDI); It must consist of data, be well-formed, and (potentially) meaningful. It is, however, controversial whether data constituting semantic information can be meaningful “independent of an informee” [Floridi, 2004b, p. 45]. This gives rise to one of the issues concerning the nature of information that has been given extraordinary amount of attention from philosophers: the symbol grounding problem [Harnad, 1990]. In short, the problem concerns how meaningless symbols can acquire meaning, and the problem stems from the fact that for humans, the “words in our heads” have original intentionality or meaning (they are about something) independently of other observers, whereas words on a page do not have meaning without being observed — their intentionality is derived. However, if it is the case that the human brain is a computational system (or Turing-equivalent), especially when seen as instantiating a “language of thought” [Fodor, 1975, cf. Section 4.2], and if the human brain can produce original intentionality, then computers must

1352

Philip Brey and Johnny Hartz Søraker

be able to achieve the same, at least in principle. This problem, perhaps best illustrated by Searle’s Chinese room argument [see Section 4.2], is not only one of the central issues in the philosophy of AI, it also constitutes one of the challenges involved in making semantically blind computers perform reliable operations. This is the subject of ‘computational semantics’, where the aim is to accurately and reliably formalize the meaning of natural language. The main challenges are to define data structures that can deal with the ambiguity and context-sensitivity inherent in natural language and to train or program the computer to make reliable inferences based on such formalizations (cf. [Blackburn and Bos, 2005]).

3 PHILOSOPHY OF COMPUTER SCIENCE As argued in the introduction, although philosophers have reflected quite extensively on the nature of computers and computing, they have hardly reflected on the nature of computer science. It is the aim of this section to summarize the scarce philosophical literature that does focus on issues concerning the nature of computer science, and to speculate on what a philosophy of computer science might look like. We hypothesize that a philosophy of computer science would, in analogy to the philosophy of science in general, philosophically reflect on the concepts, aims, structure and methodologies of computer science and its various fields. It would engage in at least the following research activities: 1. Analysis, interpretation and clarification of central concepts in computer science and the relation between them. What, for example, is a program? What is data? What is a database? What is a computer model? What is a computer network? What is human-computer interaction? What is the relation between software engineering and computer programming? What is the difference between a programming language and a natural language? These questions would be answered with the tools and methods of philosophy, and would aim at a philosophical rather than a technical understanding of these concepts. The result would be a deeper, more reflective understanding of these concepts, and possibly an analysis of vaguenesses, ambiguities and inconsistencies in the way that these concepts are used in computer science, and suggestions for improvement. 2. Analysis, clarification and evaluation of aims and key assumptions of computer science and its various subfields and the relations between them. What, for example, is the aim of software engineering? What is the aim of operating systems design? How do the aims of different subfields relate to each other? Also, how should these aims be evaluated in terms of their feasibility, desirability, or contribution to the overall aims of computer science? On what key assumptions do various subfields of computer science rest, and are these assumptions defensible?

Philosophy of Computing and Information Technology

1353

3. Analysis, clarification and evaluation of the methods and methodologies of computer science and its various subfields. What, for example, are the main methodologies used in software engineering or human-computer interaction design? How can these methodologies be evaluated in terms of the aims of these various subfields? What are their strengths and weaknesses? What better methodologies might be possible? 4. Analysis of the scientific status of computer science and its relation to other academic fields. Is computer science a mature science or is it still in a preparadigmatic stage? Is computer science a science at all? Is it an engineering discipline? In addition, how do the methodologies of computer science compare to the methods used in natural science, or other scientific fields? Where do the aims of computer science overlap with the aims of other fields, and how does and should computer science either make use of or contribute to other fields? What, for example, is the proper relation between computer science and mathematics, or computer science and logic? 5. Analysis of the role and meaning of computer science for society as a whole, as well as for particular human aims and enterprises. How do the aims of computer science contribute to overall human aims? How are the enterprises and projects of computer science believed to make life or society better, and to what extent do they succeed? To what extent is a reorientation of the aims of computer science necessary? In this section, we will begin with a discussion of attempts to give a general account of the nature, aims and methods of computer science, its status as a science, and its relation to other academic fields. We will then move to five important subfields of computer science, and discuss their nature, aims, methods, and relation to other subfields, as well as any specific philosophical issues that they raise. The subfields that will be discussed are computer programming and software engineering, data modeling and ontology, information systems, computer simulation, and human-computer interaction. Some areas, such as the nature of programming languages, will naturally be dispersed across many of these sub fields. Another subfield of computer science, artificial intelligence, will be discussed in a separate section because it has generated a very large amount of philosophical literature.

3.1

Computer science: its nature, scope and methods

One of the fundamental questions for a philosophy of computer science concerns the nature and scientific status of computer science. We will discuss four prominent accounts of computer science as an academic field and discuss some of their limitations. The first account that is sometimes given may be called the deflationary account. It holds that computer science is such a diverse field that no unified definition can be given that would underscore its status as a science or even as a coherent academic field. Paul Graham [2004], for example, has claimed that

1354

Philip Brey and Johnny Hartz Søraker

“computer science is a grab bag of tenuously related areas thrown together by an accident of history”, and Paul Abrahams has claimed that “computer science is that which is taught in computer science departments” [Abrahams, 1987, p. 1]. An objection to deflationary accounts is that they do not explain how computer science is capable of functioning as a recognized academic field, nor do they address its scientific or academic credentials. Rejecting a deflationary account, others have attempted to characterize computer science as either a science, a form of engineering, or a branch of mathematics [Wegner, 1976; Eden, 2007]. On the mathematical conception of computer science, computer science is a branch of mathematics, its methods are aprioristic and deductive, and its aims are to develop useful algorithms and to realize these in computer programs. Theoretical computer science is defended as the core of the field of computer science. A mathematical conception has been defended, amongst others, by Knuth [1974a], who claims that computer science centrally consists of the writing and evaluation of programs, and that computer programs are mere representations of algorithms that can be realized in computers. Knuth defines an algorithm as a “preciselydefined sequence of rules telling how to produce specified output information from given input information in a finite number of steps” [Knuth, 1974a, p. 2]. Since algorithms are mathematical expressions, Knuth argues, it follows that computer science is a branch of applied mathematics. A similar position is taken by Hoare [1986]. The scientific conception of computer science holds that the apriorism of the mathematical conception is incorrect, and that computer science is an ordinary empirical science. The aim of computer science is to explain, model, understand and predict the behavior of computer programs, and its methods include deduction and empirical validation. This conception has been defended by Allen Newell and Herbert Simon, who have defined computer science as “the study of the phenomena surrounding computers” and who have claimed that it is a branch of natural (empirical) sciences, on a par with “astronomy, economics, and geology” [1976, pp. 113-114]. A computer is both software and hardware, both algorithm and machinery. Indeed, it is inherently difficult to make a distinction between the two [Suber, 1988]. The workings of computers are therefore complex causal-physical processes that can be studied experimentally like ordinary physical phenomena. Eden claims that the scientific conception seems to make a good fit with various branches of computer science that involve scientific experiments, including “artificial intelligence, machine learning, evolutionary programming, artificial neural networks, artificial life, robotics, and modern formal methods” [2007, p. 138]. An objection to the scientific conception has been raised by Mahoney [2002], who argues that computers and programs cannot be the subject of scientific phenomena because they are not natural phenomena. They are human-made artifacts, and science does not study artifacts but natural phenomena. Newell and Simon have anticipated this objection in their 1976 paper, where they acknowledge that programs are indeed contingent artefacts. However, they maintain that they are nonetheless appropriate subjects for scientific experiments, albeit of a novel sort.

Philosophy of Computing and Information Technology

1355

They argue that computers, although artificial, are part of the physical world and can be experimentally studied just like natural parts of the world (see also [Simon, 1996]). Eden [2007] adds that analytical methods fall short in the study of many programs, and that the properties of such programs can only be properly understood using experimental methods. The engineering conception of computer science, finally, conceives of computer science as a branch of engineering concerned with the development of computer systems and software that meet relevant design specifications (see e.g. [Loui, 1987]). The methodology of computer science is an engineering methodology for the design and testing of computer systems. Theoretical computer science does not constitute the core of the field and has only limited applicability. The engineering conception is supported by the fact that most computer scientists do not conduct experiments but are rather involved in the design and testing of computer systems and software. The testing that is involved is usually not aimed at validating scientific hypotheses, but rather at establishing the reliability of the systems that is being developed and in making further improvements in its design. Eden [2007] has argued that the engineering conception of computer science seems to have won out in recent decades, both in theory and in practice. The mathematical conception has difficulties accounting for complex software systems, and the scientific conception does not make a good fit with the contemporary emphasis on design. A worrisome aspect of this development, Eden argues, is that the field seems to have developed an anti-theoretical and even anti-scientific attitude. Theoretical computer science is regarded to be of little value, and students are not taught basic science and the development of a scientific attitude. The danger is that computer science students are only taught to build short-lived technologies for short-term commercial gain. Eden argues that computer science has gone too far in jettisoning theoretical computer science and scientific approaches, and that the standards of the field have suffered, resulting in the development of poorly designed and unreliable computer systems and software. For computer science (and especially software engineering) to mature as a field, Eden argues, it should embrace again theoretical computer science and scientific methods and incorporate them into methods for design and testing.

3.2

Computer programming and software engineering

Two central fields of computer science are software engineering and programming languages. Software engineering is the “application of a systematic, disciplined, quantifiable approach to the development, operation, and maintenance of software” [Abran et al., 2004]. Theories of programming languages study the properties of formal languages for expressing algorithms and methods of compiling and interpreting computer programs. Computer programming, an important phase in the software development process, is the process of writing, testing, debugging and maintaining the source code of computer programs.

1356

Philip Brey and Johnny Hartz Søraker

In theorizing about the nature of software engineering, Parnas has argued that it ought to be radically differentiated from both computer science and programming, and that it should be more closely modeled on traditional forms of engineering. That is, an engineer is traditionally regarded as one “who is held responsible for producing products that are fit for use” [Parnas, 1998, p. 3], which means that software engineering involves a lot more than computer programming and the creation of software. Thus, a software engineer should be able to determine the requirements that must be satisfied by the software, participate in the overall design specification of the product, verify and validate that the software meets the requirements, and take responsibility for the product’s usability, safety and reliability [Parnas, 1998, p. 3-5]. Software engineering differs, however, from more traditional forms of engineering because software engineers are often unable to avail themselves of pre-fabricated components and because there is a lack of quantitative techniques for measuring the properties of software. For instance, the cost of a project often correlates with its complexity, which is notoriously difficult to measure when it comes to software [Brookshear, 2007, pp. 326ff]. The importance of software engineering is due to the staggering complexity of many software products, as well as the intricate and often incompatible demands of shareholders, workers, clients and society at large. This complexity also involves the specification of requirements, developing a design overview, as well as verifying and validating that the software satisfies internal and external requirements. The verification and validation of software is a critical part of software engineering. A product can work flawlessly but fail to meet the requirements set out initially, in which case it fails validation (“The right product was not built”). Or, it can generally meet the requirements set out initially, but malfunction in important ways, in which case it fails verification (“The product was not built right”). The methods employed in verification often reflect the overall perspective on what computer science and computer programs are. Eden [2007] outlines three paradigms of computer science (cf. Section 3.1), in which software is verified by means of a priori deductive reasoning (rationalist), by means of a posteriori, empirical testing (technocratic) or by means of a combination (scientific). The rationalist paradigm is most closely related with the question of ‘formal program verification’. This long-lasting debate is concerned with the question whether software reliability can (in some cases) be ensured by utilizing deductive logic and pure mathematics [Fetzer, 1988; 1991; 1998]. This research stems from dissatisfaction with “technocratic” means of verification, including manual testing and prototyping, which are subjective and usually cannot guarantee that the software is reliable (cf. Section 2.2). On the other hand, as mentioned in 3.1, the complexity involved in modern software engineering has left a purely formal approach unfeasible in practice. Although software engineering encompasses a range of techniques and procedures throughout the software development process, computer programming is one of the most important elements. Due to the complexity of modern software, hardly anyone programs computers in machine code anymore. Instead, programming languages (PL) at a higher abstraction level are used, usually being closer

Philosophy of Computing and Information Technology

1357

to natural language constructs. These ‘source codes’ are then compiled into instructions that can be executed by the computer. Indeed, programming languages can be seen as ‘virtual machines’, i.e. abstract machines that do not exist, but must be capable of being translated into the operations of an existing machine [McLaughlin, 2004, p. 139]. The investigation of appropriate mechanisms for abstraction is one of the primary concerns in the design and creation of both computer programs and the programming languages themselves [Turner and Eden, forthcoming b]. Once the choice of PL has been made, the PL more or less forces the programmer to solve problems in a particular way — within a given conceptual framework (cf. Turner and Eden [forthcoming a, p. 148]). This underlying conceptual framework can be referred to as the programming paradigm. The initial choice of programming language (or paradigm) depends on a number of factors, primarily its suitability for the problem at hand. However, the notion of programming paradigm carries some of the more irrational connotations of Kuhn’s [1970] concept, meaning that the use of a particular PL is often determined by social, commercial and ad hoc considerations, and sometimes lead to polarization and lack of communication within the field of software engineering [Floyd, 1978]. These ontological commitments do concern questions considered with regard to data modeling (see Section 3.3), but are more closely related to the control structures that operate on the data. For instance, abstraction necessarily entails some form of ‘information hiding’. This is, however, a different kind of abstraction than that found in formal sciences. In many sciences, certain kinds of information are deemed irrelevant, such as the color of triangle in mathematics, and therefore neglected. In PL abstraction, as Colburn and Shute [2007] has pointed out, information that is “hidden” at one level of abstraction (in particular, the actual machine code needed to perform the operations) cannot be ignored at a lower level. Finally, PLs differ immensely with regard to the structure and flow of the control structures. For instance, Edsger Dijkstra’s seminal paper “Go To Statement Considered Harmful” [Dijkstra, 1968], which has spurred dozens of other “x considered harmful” papers, criticized the then common use of unstructured jumps (goto’s) in programming, advocating a structured approach instead. These discussions on ‘good’ and ‘bad’ programming differ enormously depending on the underlying justification, whether it is ease of learning, reliability, ease of debugging, ease of cooperation or a notion of aesthetic beauty (see e.g. [Knuth, 1974b]).

3.3

Data Modeling and Ontology

One of the most common uses of computer technology, and a central concern in computer programming and software engineering, is to store vast amounts of data in a database so as to make the storage, retrieval and manipulation as efficient and reliable as possible. This requires a specification beforehand of how the database should be organized. Such a specification is known as a data model theory. Al-

1358

Philip Brey and Johnny Hartz Søraker

though the term is used in many different senses (cf. Marcos [2001]), a data model typically consists of 1) the structural part, i.e. a specification of how to represent the entities or objects to be modeled by the database; 2) the integrity part, i.e. rules that place constraints in order to ensure integrity; and 3) the manipulation part, i.e. a specification of the operations that can be performed on the data structures. The purpose of these parts is to ensure that the data are stored in a consistent manner, that queries and manipulations are reliable and that the database preserves its integrity. Data integrity refers to the accuracy, correctness and validity of the data, which in lack of a comprehensive data model theory, might be compromised when new data is entered, when databases are merged or when operations are carried out. We can distinguish between 2-dimensional databases, which can be visualized as a familiar spreadsheet of rows and columns, and ndimensional databases, where numerous databases are related to each other, for instance by means of shared ‘keys’. Floridi makes a distinction between an ‘aesthetic’ and a ‘constructionist’ view concerning the nature and utility of databases ([Floridi, 1999]; see [Marcos and Marcos, 2001] for a similar distinction between ‘model-as-copy’ and ‘model-asoriginal’). First, the “aesthetic” approach sees databases as a collection of data, information or knowledge that conceptualizes a particular reality, typically modeled on na¨ıve realism. This approach can in particular be seen in ‘knowledge engineering’, where human knowledge is collected and organized in a ‘knowledge base’, usually forming the basis of an ‘expert system’ (see Section 4.4). In a similar vein, Gruber [1995] defines the use of ontology in computer science as “a specification of a representational vocabulary for a shared domain of discourse [including] definitions of classes, relations, functions, and other objects” [Gruber, 1995, p. 908]. Although this is the most common use of data modeling, one of the philosophical problems with such “specification of conceptualization” is that these conceptualizations might not directly correspond to entities that exist in the real word but to human-constructed concepts instead. This is particularly a problem when it comes to science-based ontologies, where non-existent entities ought to be avoided [Smith, 2004]. According to Floridi, a second approach to data modeling can be termed ‘constructionist’, where databases are seen as a strategic resource whose “overall purpose is to generate new information out of old data” and the implemented data model “is an essential element that contributes to the proper modification and improvement of the conceptualized reality in question” [Floridi, 1999, p. 110). The distinction between ‘aesthetic’ and ‘constructionist’ also gives rise to an epistemological distinction between those sciences where the database is intended to represent actual entities, such as biology and physics, and those sciences where databases can provide requirements that the implementation in the real world must satisfy, including computer science itself [Floridi, 1999, p. 111]. Although data model theories are application- and hardware-independent, they are usually task-specific and implementation-oriented. This has raised the need for domain- and application-independent ontologies,the purpose being to establish a high-level conceptualization that can be shared by different data models — in

Philosophy of Computing and Information Technology

1359

different domains. Since such ontologies often aim to be task-independent, they typically describe a hierarchy of concepts, properties and their relations, rather than the entities themselves. This is known as a formal, as opposed to a descriptive ontology and is influenced by philosophical attempts to develop ontological categories in a systematic and coherent manner. The impetus of much of this research stems from a common problem in computer science, sometimes referred to as the tower of Babel problem. Especially with the advent of networked computers, the many different kinds of terminals, operating systems and database models — as well as the many different domains that can be represented in a database — posed a problem for successful exchange of data. Rather than dealing with these problems on an ad hoc basis, formal ontologies can provide a common controlled vocabulary [Smith et al., 2007, p. 1251] that ensures compatibility across different systems and different types of information. Such compatibility does not only save man hours, but opens up new possibilities for cross-correlating and finding “hidden” information in and between databases (so-called ‘data mining’). The importance of such ontologies has been recognized in fields as diverse as Artificial Intelligence and knowledge engineering (cf. Section 4), information systems (cf. 3.4), natural language translation, mechanical engineering, electronic commerce, geographic information systems, legal information systems and, with particular success, biomedicine (cf. [Guarino, 1998; Smith et al., 2007]). Paradoxically, however, the very success of this approach has led to a proliferation of different ontologies that sometimes stand in the way of successful integration [Smith et al., 2007]. Closely related, these ontologies cannot always cope with specific domain-dependent requirements. This could be one reason why, despite the philosophical interest and heritage, the importance of philosophical scrutiny has often been “obscured by the temptation to seek immediate solutions to apparently localized problems” [Fielding et al., 2004].

3.4

Information systems

‘Information’ and ‘system’ are both highly generic terms, which means that the term ‘information system’ is used in many different ways. In light of this, Cope et al. [1997] conducted a survey of different uses of the term and identified four major conceptions that form a hierarchy. The most general conception of IS simply refers to a database where information can be retrieved through an interface. At the other end, the more specific conception considers IS to encompass the total information flow of a system, typically a large organization — including “people, data, processes, and information technology that interact to collect, process, store, and provide as output the information needed to support an organization” [Whitten, 2004, p. 12]. As such, IS is not the same as information technology but a system in which information technology plays an important role. Du Plooy also argues that the social aspect of information systems is of such importance that it should be seen as the core of the discipline [du Plooy, 2003] and we will focus on this notion in this sub section, given that many of the non-social issues are discussed elsewhere.

1360

Philip Brey and Johnny Hartz Søraker

Although IS includes many factors in addition to the technology, the focus in IS research has typically been on the role of the technology, for instance how the technology can be optimized to improve the information flow in an organization. Among the many philosophical issues raised by such systems, one of the most important ones is the relation between scientifically-based, rationalist theories of information systems design and the actual practice of people involved in management. Introna [1997] argues that the (then) reigning techno-functionalist paradigm in the information systems discipline fails to take actual practices into account. Based on insights from hermeneutics, he stresses instead the importance of the involved manager and the changing demands of being-there as a part of the information system. In a similar manner, Butler and Murphy [2007] argue that computerization of organizations means that we rationalize what is easy to rationalize, and therefore place too much emphasis on decontextualized information processes rather than the reality of the human actors. As can be seen in these examples, theories of information systems often address the (power) relationship between humans and technology — especially the over-emphasis on technology at the expense of humans — which means that hermeneutics and theorists like Giddens, Heidegger, Habermas, Foucault and Latour often lend themselves to such analysis. It should also be pointed out that IS research often involves assessment of actual information systems and as such pre-supposes certain methodologies and assessment criteria. Dobson points out that this raises a number of epistemological questions regarding the IS researcher’s theoretical lens, skill and political biases, as well as a number of ontological questions regarding which entities to include as part of the information system and their relation to each other [Dobson, 2002]. Given the complexity of large information systems, the ontological and epistemological questions become particularly challenging, because large organizations often involve fundamentally different kinds of entities. For instance, an Information system can include specifications of such things as databases of physical entities, algorithms for efficient scheduling, information flow relations, and interfaces for retrieving information. Dobson argues that the adopted methodologies have been dominated by the kind of social theorists mentioned above and suggests that IS studies should pay more attention to philosophical approaches to epistemology and ontology, in particular Bhaskar’s critical realism. Many of the perennial issues in computer ethics also revolve around information systems, including problems surrounding surveillance in the workplace, automation of manual labor and the problems with assigning responsibility (see also Sections 4.6 and 6). These are issues that usually fall under professional ethics because they deal with computer science and information systems professionals — management in particular. Other issues in philosophy of information systems overlap with philosophy of computer science in general (3.1), software engineering (3.2), and often address issues concerned with data modeling and ontology (3.3) as well as human-computer interaction (3.6).

Philosophy of Computing and Information Technology

3.5

1361

Computer simulation

In addition to the kind of data modeling outlined above, which is primarily occupied with commercial and management systems, computers are frequently employed for scientific modeling and simulation. Computers are particularly useful when simulating micro- or macroscopic phenomena, where traditional forms of experimentation is not feasible. Computer simulations differ from data models in that they often employ visualization of the data (especially simulations at microor macro level) or real-time input from users (e.g. flight simulation). One of the main issues in the establishment of philosophy of computer science as distinct from traditional philosophy of science revolves around the latter’s ability to adequately account for computer simulations. Clearly, computers allow us to simulate events that we have not been able to simulate before, primarily because of the immense processing power and visualization possibilities, and therefore becomes a valuable tool for many sciences. The question remains, however, whether computer simulation raises any novel philosophical issues. Humphreys [forthcoming] argues that computer simulation raises the need for a new philosophy of computer science in a number of ways. Most importantly, computer simulations are epistemically opaque in the sense that it is often impossible for a researcher to know all of the epistemically relevant elements of the process. This is not only a result of the complexity involved, but also the fact that scientists need to delegate substantial amounts of authority to the programmers and software engineers. Furthermore, the question of what we can know in philosophy of science has become in part a question of computational possibilities and limitations. Similar concerns have been raised by Brian Cantwell Smith, who argues that there are inherent limitations to what can be proven about computers and computer programs; the ‘correctness’ of a computer simulation is vulnerable to the combination of computational complexity, unpredictable human-computer interaction, the many levels at which a computer can fail and the lack of precision regarding what ‘correctness’ entails [Smith, 1996]. Morgan and Morrison [1999] have argued that it is the lack of material similarity that makes computer simulations unique; that computer simulations thereby have a reduced potential to make strong inference back to the world. Parker [forthcoming] and others have argued against this view by pointing out that noncomputer simulations often have less validity because of the complexity involved — for instance with respect to weather forecasting. Winsberg [forthcoming], although disagreeing with Morgan and Morrison, makes a related but ultimately more plausible argument. He argues that it is not the degree of materiality that determines the validity of the simulation in question, but the justification of this validity itself. Non-computer simulations typically justify the similarity between the object and target systems in terms of material or structural similarities (e.g. a miniature airplane in a wind tunnel being similar to its full-scale counterpart in some material respects), whereas the validity of computer simulations are typically justified in terms of strictly formal and mathematical similarities. In other words,

1362

Philip Brey and Johnny Hartz Søraker

computer simulations require a justification that is entirely different from, and raise philosophical questions typically not raised by, non-computer simulations — such as the relation between formalized algorithms and the physical world, the justifiability of heuristic algorithms, the use of pseudo-randomizations and truncated numbers, approximations of physical laws and so forth. Frigg and Reiss [forthcoming] argue that despite these characteristics, computer simulations pose no new philosophical problems. But, as Humphreys [forthcoming] points out, this conclusion seems to rest on a misinterpretation of the Church–Turing thesis (cf. Section 2.2). That is, although computer simulations in-principle can be carried out on a Turing machine and ipso facto by means of pencil and paper, the staggering complexity of e.g. weather forecasting forbids this in practice. Computer simulations also offer the possibility to visualize the results of the simulation and even to let the researchers intervene on the basis of visual feedback — thereby resembling experiments (“in silico”) more than simulations. In particular, these visualizations can be run at different speeds, in reverse and with different foci. Thus, computer simulation has opened up simulation possibilities that are otherwise prohibited by physical laws.

3.6 Human-Computer interaction Human-Computer Interaction (HCI) is a subfield within computer science concerned with the study of the interaction between people (users) and computers and the design, evaluation and implementation of user interfaces for computer systems that are receptive to the user’s needs and habits. It is a multidisciplinary field, which incorporates computer science, behavioral sciences, and design. A central objective of HCI is to make computer systems more user-friendly and more usable. Users interact with computer systems through a user interface, which consists of hard- and software that provides means of input, allowing users to manipulate the system, and output, allowing the system to provide information to the user. The design, implementation and evaluation of interfaces is therefore a central focus of HCI. It is recognized in HCI that good interface design presupposes a good theory or model of human-computer interaction, and that such a theory should be based in large part on a theory of human cognition to model the cognitive processes of users interacting with computer systems [Peschl and Stary, 1998]. Such theories of human cognition are usually derived from cognitive psychology or the multidisciplinary field of cognitive science. Whereas philosophers have rarely studied human-computer interaction specifically, they have contributed significantly to theorizing about cognition, including the relation between cognition and the external environment, and this is where philosophy relates to HCI. Research in HCI has initially relied extensively on classical conceptions of cognition as developed in cognitive psychology and cognitive science. Classical conceptions, alternatively called cognitivism or the information-processing approach, hold that cognition is an internal mental process that can be analyzed largely in-

Philosophy of Computing and Information Technology

1363

dependently of the body of the environment, and which involves the manipulation of discrete, internal states (representations or symbols) that are manipulated according to rules or algorithms [Haugeland, 1978]. These internal representations are intended to correspond to structures in the external world, which is conceived of as an objective reality fully independent of the mind. Cognitivism has been influenced by the rationalist tradition in philosophy, from Descartes to Jerry Fodor, which construes the mind as an entity separate from both the body and the world, and cognition as an abstract rational, process. Critics have assailed cognitivism for these assumptions, and have argued that cognitivism cannot explain cognition as it actually takes place in real-life settings. In its place, they have developed embodied and situated approaches to cognition that conceive of cognition as a process that cannot be understood without intimate reference to the human body and to the interactions of humans with their physical and social environment [Anderson, 2003]. Many approaches in HCI now embrace an embodied and/or situated perspective on cognition. Embodied and situated approaches share many assumptions, and often no distinction is made between them. Embodied cognition approaches hold that cognition is a process that cannot be understood without reference to the perceptual and motor capacities of the body and the body’s internal milieu, and that many cognitive processes arise out of real-time goal-directed interactions of our bodies with the environment. Situated cognition approaches hold that cognitive processes are co-determined by the local situations in which agents find themselves. Knowledge is constructed out of direct interaction with the environment rather than derived from prior rules and representations in the mind. Cognition and knowledge are therefore radically context-dependent and can only be understood by considering the environment in which cognition takes place and the agent’s interactions with this environment. Embodied and situated approaches have been strongly influenced by phenomenology, especially Heidegger, Merleau-Ponty and the contemporary work of Hubert Dreyfus (e.g., [Winograd and Flores, 1987; Dourish, 2001; Suchman, 1987]). Philosophers Andy Clark and David Chalmers have developed an influential embodied/situated theory of cognition, active externalism, according to which cognition is not a property of individual agents but of agent-environment pairings. They argue that external objects play a significant role in aiding cognitive processes, and that therefore cognitive processes extend to both mind and environment. This implies, they argue, that mind and environment together constitute a cognitive system, and the mind can be conceived of as extending beyond the skull [Clark and Chalmers, 1998; Clark, 1997]. Clark uses the terms “wideware” and “cognitive technology” to denote structures in the environment that are used to extend cognitive processes, and he argues that because we have always extended our minds using cognitive technologies, we have always been cyborgs [Clark, 2003]. Active externalism has been inspired by, and inspires, distributed cognition approaches to cognition [Hutchins, 1995], according to which cognitive processes may be distributed over agents and external environmental structures, as well as over the

1364

Philip Brey and Johnny Hartz Søraker

members of social groups. Distributed cognition approaches have been applied to HCI [Hollan, Hutchins and Kirsh, 2000], and have been especially influential in the area of Computer Supported Cooperative Work (CSCW). Brey [2005] has invoked cognitive externalist and distributed cognition approaches to analyze how computer systems extend human cognition in humancomputer interaction. He claims that humans have always used dedicated artifacts to support cognition, artifacts like calendars and calculators, which HCI researcher Donald Norman [1993] has called cognitive artifacts. Computer systems are extremely versatile and powerful cognitive artifacts that can support almost any cognitive task. They are capable of engaging in a unique symbiotic relationship with humans to create hybrid cognitive systems in which a human and an artificial processor process information in tandem. However, Brey argues, not all uses of computer systems are cognitive. With the emergence of graphical user interfaces, multimedia and virtual environments, the computer is now often used to simulate environments to support communication, play, creative expression, and social interaction. Brey argues that while such activities may involve distributed cognition, they are not primarily cognitive themselves. Interface design has to take into account whether the primary aim of applications is cognitive or simulational, and different design criteria exist for both. 4

PHILOSOPHY OF ARTIFICIAL INTELLIGENCE

Artificial Intelligence (AI) is commonly referred to as the science and engineering of intelligent machines — ‘intelligent’ commonly seen as relative to human intelligence. Given its close ties with numerous sub disciplines of philosophy, philosophy of mind and philosophy of language in particular, it has received tremendous attention from philosophers. The field is inherently interdisciplinary and has arguably had a more profound impact on philosophical discourse than any other technology. This section discusses issues and approaches in the philosophy of artificial intelligence, including its emergence and scope, the philosophy of major approaches in AI (symbolic AI, connectionist AI, artificial life, dynamical systems), the philosophy of AI applications (expert systems, knowledge engineering, robotics, and artificial agents) and concludes with a review of some ethical issues in AI.

4.1 Artificial intelligence and philosophy Artificial intelligence, or AI, is a field of computer science that became established in the 1950s. It was described at the time as a new science which would systematically study the phenomenon of ‘intelligence’. This goal was to be pursued by using computers to simulate intelligent processes. The central assumption of AI was that the logical operations of computers could be structured to imitate human thought processes. Because the workings of a computer are understood while those of the human mind are not, AI researchers hoped in this way to reach a scientific understanding of the phenomenon of ‘intelligence’.

Philosophy of Computing and Information Technology

1365

Intelligence is conceived of in AI as a general mental ability that encompasses several more specific abilities, such as the ability to reason, plan, solve problems, comprehend ideas, use language, and learn. AI research commonly focuses on a specific ability and attempts to develop programs that are capable of performing limited tasks involving that ability. The highest goal of AI was to construct a computer system with the intelligence and reasoning ability of an adult human being. Many early AI researchers claimed that this goal would be reached within only a few decades, thanks to the invention of the digital computer and to key breakthroughs in the fields of information theory and formal logic. In 1965, the noted AI researcher Herbert Simon predicted that computers would be able to execute any task that human beings could by 1985 [Simon, 1965]. How might it be demonstrated that a computer is as intelligent as a human being? Alan Turing [1950] proposed the Turing Test, in which a computer and a human being are placed behind a screen, and a test person is to ask questions to both in order to find out which of the two is human. If the test person cannot make such a judgment after a reasonable amount of time, the computer has supposedly been demonstrated to be in possession of general intelligence. The Turing Test is still often invoked, but has not remained without criticism as a test for general intelligence [Moor, 2003]. AI researchers agreed that AI studied intelligent processes and aimed to create intelligent computer programs, but they soon developed different viewpoints on the extent to which AI should be directed at the study of human intelligence. Some researchers, like Allen Newell and Herbert Simon, believed that intelligent computer programs could be used to model thought processes of humans, and made it their goal to do this. This is sometimes called the cognitive simulation approach in AI, or strong AI [Searle, 1980]. Strong AI holds that suitably programmed computers literally have cognitive states that resemble the cognitive states found in human minds, and are therefore capable of explaining human cognition. Some proponents of strong AI even go further and hold that a suitably programmed computer is capable of consciousness. Underlying these claims of strong AI is a belief in computationalism: the doctrine that mental states are computational states, and that cognition equals computation [Pylyshyn, 1984; Shapiro, 1995]. In the mid-1970s, computationalism became a widely held view within AI, linguistics, philosophy, and psychology, and researchers from these fields joined to create the field of cognitive science, a new field that engages in interdisciplinary studies of the mind and intelligence [Boden, 2006]. Whereas many researchers in AI embraced the cognitive simulation approach, many others merely wanted to develop computer programs that were capable of performing intelligent tasks. For all they were concerned, the underlying mechanism by which computers were capable of intelligent behavior might be completely different from the working of human minds. This approach has been called weak AI. Many proponents of this more cautious approach nevertheless believed that research in AI could contribute to an understanding of the phenomenon of intelli-

1366

Philip Brey and Johnny Hartz Søraker

gence, by uncovering general properties of intelligent processes, and that AI could therefore still meaningfully contribute to cognitive science. In recent decades, the view of AI as a science that studies the phenomenon of intelligence has been partially superseded by a view of AI as an engineering discipline in which researchers focus on developing useful programs and tools that perform in domains that normally require intelligence. AI has therefore in large part become an applied science, often merging with other fields of computer science. The philosophy of AI [Copeland, 1993; Haugeland, 1981; Boden, 1990; Fetzer, 2004] emerged in the 1960s and became an established field in the 1980s. For the most part, it focuses on assumptions and approaches within the scientific approach to AI, and its relation to cognitive science. Much less attention has been paid to developments in the engineering approach to AI The philosophy of AI considers the questions whether machines (and specifically computer systems) are capable of general intelligence, whether they are capable of having mental states and consciousness, and whether human intelligence and machine intelligence are essentially the same and the mind therefore is a computational system. Philosophers have also explored the relation between philosophical logic and AI [Thomason, 2003] and ethical issues in AI (Section 4.6).

4.2 Symbolic AI From the beginnings of AI research in the 1950s up to the early 1980s different approaches in AI research had so much in common that they constituted a research paradigm, called “symbolic AI” (or, alternatively, “classical AI” or GOFAI, which stands for Good Old Fashioned AI), that is still influential today. The central claim of symbolic AI is that intelligence, in both humans and machines, is a matter of manipulating symbols according to fixed and formal rules. This claim rests on several basic assumptions, made precise by Newell and Simon [1976], who introduced the notion of a physical symbol system. A physical symbol system was defined by them as a system that manipulates and produces physically realized symbol structures. Symbol structures, or expressions, are physical combinations of instances of symbols, which are unique physically realized patterns. The system contains continually changing sets of symbol structures as well as a set of processes for their creation, modification, reproduction and destruction. Symbol structures are capable of designating objects in the world and are also capable of designating processes which can be carried out (“interpreted”) by the system. Clearly, computer systems qualify as physical symbol systems, but the above definition leaves open the possibility that other entities, such as human brains, are also physical symbol systems. Based on this definition, Newell and Simon state the physical symbol system hypothesis, which is that a physical symbol system has the necessary and sufficient means to display general intelligence. Because this hypothesis implies that only physical symbol systems can display general intelligence, it also implies that the human mind implements a physical symbol system, and that minds are information-

Philosophy of Computing and Information Technology

1367

processing systems very similar to digital computers. This view is called computationalism, or the computational theory of mind. Newell and Simon’s hypothesis is therefore a version of strong (symbolic) AI. A weaker version, equivalent to weak (symbolic) AI is that being a physical symbol system is sufficient but not necessary for intelligence, which implies that computer systems are capable of general intelligence but that their architecture may not resemble that of human minds [Copeland, 1993]. Strong symbolic AI, and the corresponding computational theory of mind, have been both defended and criticized by philosophers. Philosopher Jerry Fodor has famously defended computationalism through a defense of his Language of Thought Hypothesis, which states that human cognition consists of syntactic operations over physically realized representations in the mind that have a combinatorial syntax and semantics [Fodor, 1975]. The most famous (or infamous) critic of strong symbolic AI, John Searle, has argued that when computers process symbols they do not have access to their content or meaning, whereas humans do, and that therefore computationalism and strong symbolic AI are false. He makes his case using a thought experiment, called the Chinese Room Argument, in which a human in a room is asked to follow English instructions for manipulating Chinese symbols [Searle, 1980]. The human receives questions in Chinese through a slot in the wall, and is capable of answering in Chinese, thus appearing to understand Chinese. Yet, Searle claims, the human does not understand Chinese. The example shows, he argues, that manipulating symbols on the basis of syntax alone, which is what the human does, does not imply understanding. Computer cognition and human cognition are therefore different, because humans normally do have understanding of the information they process. Searle’s argument against strong symbolic AI has met with numerous responses and attempted rebuttals from the AI community and fellow philosophers [Preston and Bishop, 2002]. Hubert Dreyfus [1972; 1992; 1996], who has been critiquing AI since the mid1960s, has argued, like Searle, that human cognition normally does not involve the application of rules or a use of internal representations. Dreyfus instead proposes a phenomenological view of human intelligence as situated — codetermined by the situation in which humans find themselves — and embodied — emergent out of real-time goal-directed sensorimotor interactions of the human body with the environment. (See also Section 3.6.) Computation differs from cognition, he argues, as it is disembodied and not situated but detached and abstracted from the world in which computers find themselves. The assumptions of symbolic AI about the nature of intelligence are so fundamentally mistaken, Dreyfus argues, that weak symbolic AI is false as well. Symbolic AI is therefore in-principle incapable of yielding general intelligence. The problem, Dreyfus argues, is that symbolic AI stands in the tradition of Cartesian rationalism, and inherits all of its false assumptions: that intelligence involves the disembodied application of formal rules, that the world we know has a formal, objective structure, and that all knowledge can be formalized. Dreyfus is

1368

Philip Brey and Johnny Hartz Søraker

particularly critical of the third assumption, which he calls the epistemological assumption. This assumption implies that everything that is known or understood by humans can be expressed in context-independent, formal rules or definitions that can be processed by machines. Dreyfus argues against this assumption that, while formal rules may be one way of describing human knowledge, they do not provide the basis for reproduction of such knowledge by an intelligent system. The problem is that formal rules do not contain their own criteria for application, and that additional contextual or background information is needed. They do not possess the elaborate system of background information possessed by humans in virtue of which they can interpret items effortlessly in the context in which they occur, and by which they know which interpretations are meaningful and which ones absurd or meaningless. Dreyfus calls this problem for symbolic AI the commonsense knowledge problem, and claims that it is unsolvable within a symbolic approach [Dreyfus and Dreyfus, 1986]. Many “hard” problems in symbolic AI, such as the well-known frame problem [Pylyshyn, 1987], can be analyzed as specific instances of the commonsense knowledge problem.

4.3 Connectionist AI, artificial life and dynamical systems Since the 1980s, a rival paradigm to symbolic AI has arisen, called neural networks or connectionism [Bechtel and Abrahamson, 1990; Clark, 1991]. Connectionist AI rejects the idea that intelligent behavior springs from the manipulation of symbols according to formal rules. The neural network approach derives its inspiration for the modeling of intelligent processes from the structure and operation of the human brain rather than from digital computers. Connectionist models consist of a large number of simple processors, or units, with relatively simple input/output functions that resemble those of nerve cells. These units are connected to each other and some also to input or output structures, via a number of connections. These connections have different “weight”. The weights in combination with the input signals determine the activation level of a unit. Units may be activated to different degrees and when the activation reaches a certain threshold they give off signals to connected units. A complete connectionist network consists of an input layer of input units, an output layer, and one or more “hidden” layers of units in between. Information processing in a connectionist system is then a process of excitation and inhibition of units. It is a massively parallel process, in which large numbers of simple computational units perform simple computations and influence each other, ultimately leading to an set of output signals. Representations in connectionism can be defined as patterns of activation across a unit layer. Neural networks turn out to be astoundingly good at carrying out certain types of intelligent tasks, like pattern recognition, categorization, and the coordination of behavior. They have been less successful, so far, in modeling “higher” cognitive

Philosophy of Computing and Information Technology

1369

tasks, like abstract reasoning, formal tasks, and problem solving, precisely the kinds of tasks that symbolic AI is best able to model. Attempts have been made to physically build such connectionist models, but in practice, most connectionist models are simulated on ordinary digital computers. Connectionist and symbolic AI share the assumption that cognition is a matter of information processing, and that such information processing is computational, meaning that it can be represented algorithmically and mathematically. There are four major differences between the two approaches. First, information processing in symbolic AI involves the application of explicit, formal rules, whereas no such rules are operative in connectionist networks. Second, information processing in symbolic AI is executive-driven, involving a central overviewer (processing unit) which controls processes, whereas information processing in networks is the result of many independently operating structures. Third, information processing in symbol systems is typically serial, whereas in networks it is massively parallel. Fourth, learning in symbolical AI is a deductive process consisting of hypothesis testing, whereas in connectionism, it is associationist, i.e. a process of strengthening or weakening (or growing or losing) connections between nodes. Connectionism was embraced enthusiastically by many philosophers in the 1980s and 1990s as a superior alternative to symbolic AI. Amongst its strongest proponents were Andy Clark [1991; 1993], who related it to many issues in the philosophy of mind and language, and Paul Churchland [1992], who also employed it as a new foundation for epistemology and philosophy of science. Yet, many proponents of symbolic AI were unconvinced. Jerry Fodor and Zenon Pylyshyn argued that connectionism had serious limitations because its representations lacked the systematicity, productivity and inferential coherence needed for human language use and reasoning. Only representations with a syntactic and semantic structure could provide these properties, they argued [Fodor and Pylyshyn, 1988]. Proponents of connectionism either denied that cognition had the properties described by Fodor and Pylyshyn or that syntactic and semantic structure were necessary to produce them, or argued that connectionist networks could approximate or instantiate symbol systems for the modeling of language use and reasoning [Smolensky, 1988; Clark, 1989]. Having previously rejected symbolic AI, Hubert Dreyfus praised connectionism for rejecting the rationalist conception of cognition in symbolic AI, and held that its basic assumptions were compatible with his own vision of intelligence [Dreyfus, 1992; Dreyfus and Dreyfus, 1988]. Yet, Dreyfus is ultimately pessimistic about its prospects for AI, because of the incredible complexity of human intelligence. The commonsense knowledge problem applies just as much to connectionism as it does to symbolic AI. In connectionist networks, the ability to deal intelligently with new situations depends on the ability to generalize intelligently from past experiences to new ones. This ability, Dreyfus argues, requires significant amounts of background knowledge. A neural network with such background knowledge would have to consist of millions or billions of processors, not the tens or hundreds found in most current networks. Acquisition of such knowledge would moreover require

1370

Philip Brey and Johnny Hartz Søraker

extended embodied interaction with an environment, whereas neural networks are still essentially disembodied. ‘Artificial Life’ shares the biological underpinning of connectionism, but rather than taking the brain’s processing power as inspiration, it takes the evolutionary processes that have created the brain as its source of inspiration. Thus, whereas AI in practice involves computer simulations of human-like intelligence, Artificial Life involves computer simulations of life and life-like processes [Bedau, 2003]. One of the key differences is that the mechanisms of life are better known and easier to conceptualize than intelligence. Two further distinguishing features of ALife research is that it tends to focus on the essential features of living systems and to understand such systems by artificially synthesizing extremely simple forms of them (cf. Bedau [2004]). Keeley [1998] claims that one can identify a strong and a weak version of ALife, in which the weak version holds that computers and/or robots can be designed in such a way that they can be effectively used as tools in the formulation and testing of biological theories. The strong version goes further by arguing that such systems could actually be considered as being biological and alive. Bedau argues that one of the most important differences between symbolic AI and ALife is that the former is top-down and the latter is bottom-up. That is, symbolic AI models involve a central controller that monitors the system’s global state and makes decisions that affect any aspect of the system. ALife, however, tends to take a bottom-up approach, utilizing simple “stupid” agents that interact with each other and only together determine the state and behavior of the system. In this regard, ALife shares many of the characteristic features of connectionism and Brooks’ anti-representationalism (see Section 4.5). Several other aspects of ALife are interesting for philosophical research: the conceptualization/definition of life, the possibility and ethics of creating life in a digital computer, the relationship between cognition and life, and its relevance to other forms of emergent phenomena in economics and social philosophy (cf. [Boden, 1996]). Dennett [1994] has argued that ALife also ought to be regarded as a method for doing philosophy; ALife can impose requirements on thought experiments that could never be imposed by reasoning alone, thereby yielding new insights about their feasibility and possible implications. ALife also involves the use of notions from biology and evolution to improve computing. The latter is referred to as evolutionary computing, which is a collective term for a range of different approaches that inherit many of the characteristics of natural evolution, in particular the ability to provide good, although usually not optimal, solutions to a wide range of (often unforeseeable) problems (cf. [Eiben and Smith, 2003]) on the basis of trial-and-error and some means of reinforcing the “fittest” solution. Finally, a recent trend in AI research and cognitive science has been increased focus on dynamical systems. Dynamical systems theory focuses on how all aspects of a system can be seen as changing from one total state to another [Port and van Gelder, 1995, p. 15]. In other words, DST stresses this kind of holism as an

Philosophy of Computing and Information Technology

1371

alternative to the modularity, representation and (especially sequential) computation found in traditional approaches to AI. It is primarily a theory of cognitive development and the theory itself need not take an explicit stand on the possibility of its realization in a computer [van Gelder, 2000, p. 9]. However, the approach does put new items on the agenda for AI researchers by further opposing representationalism and emphasizing the importance of the brain, body and environment as a dynamical, holistic system.

4.4

Knowledge engineering and expert systems

Knowledge engineering is the task of transferring human knowledge into a database (cf. Section 3.3) that can serve as a basis for enabling AI applications to perform human-like reasoning. A rough distinction can be made between the engineering of common sense knowledge and of expert knowledge, the former being inspired by Hayes’ seminal paper on “The na¨ıve physics manifesto” [Hayes, 1990]. One of the most famous examples of a common sense knowledge base is Lenat’s ambitious CYC-project (cf. [Lenat and Guha, 1990]). The CYC project aims to eventually have a suitable representation for the full range of human expression, so that expert knowledge bases can be created with CYC as its basis. Despite some success (cf. www.cyc.com) the endeavor has been heavily criticized. For instance, Drew McDermott, who for a long time was an avid supporter of Hayes’ program, recanted and argued that the approach commits the “logicist” fallacy of assuming that all human reasoning is necessarily deductive [McDermott, 1990]. A somewhat different approach to knowledge engineering is to focus on a limited domain of knowledge and try to understand and represent expert knowledge and reasoning. Expert systems, the first of which were developed in the middle of the 1970s, are computer systems which are intended to take over tasks from human experts in a particular specialized domain, for instance in medicine, law, mathematics and financial planning. Such expert systems typically include a knowledgebase of expert knowledge and advanced artificial intelligence to ensure that the system returns reliable and accurate answers in response to non-experts’ queries. Expert systems are mainly built according to the assumptions of symbolic AI. Their designers try to provide these systems with the required knowledge by interviewing experts with the goal of making their often tacit knowledge explicit and arrive at formal rules that experts are thought to follow. The quality of expert systems is usually assessed by comparing its performance with that of a human expert. Despite his criticism of symbolic AI, Dreyfus was relatively optimistic in his early work about the prospects of expert systems because of the formal nature of much expert reasoning. Later, Dreyfus famously reconsidered his view, concluding that humans do employ rules in early stages of learning, but that real experts replace this with an intuitive and holistic manner of problem solving [Dreyfus and Dreyfus, 1984; 1986]. In more philosophical terms, Dreyfus refuses both the epistemological and ontological assumptions behind expert systems, arguing that

1372

Philip Brey and Johnny Hartz Søraker

neither human knowledge nor physical reality has a formal structure that can be fully described in terms of rules. He thereby echoes a similar claim made by Weizenbaum, who already in 1976 attacked the tendency to reduce human problems to calculable, logical problems and emphasized the importance of intuitive human judgment even in specialized domains [Weizenbaum, 1976]. Manning [1987] argues that it is not only formalization of knowledge that poses a problem. Many problems arise because of the interaction between the expert system and the users. For instance, expert systems cannot match human experts when it comes to asking appropriate follow-up questions (e.g. if the user query does not contain enough information), to make context-sensitive distinctions between relevant and irrelevant information (cf. Section 4.2) and to separate between information that needs explanation and information that can provide explanation (e.g. in medicine, to separate between symptoms and causes of an illness). Another problem stems from the fact that expert systems must employ some kind of probability measure, since in most cases the available knowledge and user information is only sufficient to make one result more probable than the others. The question, then, becomes whether to utilize subjective measures of probability assigned by the experts or more objective measures of probability, for instance as a function of statistical data [Gillies, 2004]. These challenges give rise to a number of limitations in the range of application of (symbolic) expert systems. If expert systems cannot make decisions or form judgments at the level of an expert, they cannot be entrusted with tasks that require expertise. However, even Dreyfus admits that expert systems can often attain a certain degree of competence, which is a higher performance level than a human novice or advanced beginner. Expert systems therefore might indeed prove useful in applications that do not call for performance at the expert level. The decision when this is the case is related to pragmatic concerns regarding the availability of human experts and, more importantly, ethical/legal notions of risk and responsibility (see Section 4.6).

4.5 Robots and artificial agents The notion of (artificial) agency is often used in computer science to refer to a computer program that is able to act on and interact with its environment. Different sets of requirements have been proposed for what it means to be an agent, which has resulted in a complex and often inconsistent set of terms for different kinds of agency. Sycara argues that there are four properties that characterize an artificial agent and distinguishes it from object-oriented systems or expert systems. These are 1) situatedness, which means that the agent receives some form of input from its environment and can perform actions that change the environment in some way; 2) autonomy, which means that the agent can act without human intervention and control its own actions and states; 3) adaptivity, meaning that it is capable of taking initiative based on its (usually pre-programmed) goals, learn from its experiences and have a flexible repertoire of possible actions; and 4) so-

Philosophy of Computing and Information Technology

1373

ciability, referring to the ability to interact in a peer-to-peer manner with other agents or humans [Sycara, 1998, p. 11]. Insofar as these requirements are too restrictive, a taxonomy could be constructed on the basis of the satisfied conditions, respectively, ‘situated agents’, ‘autonomous agents’, ‘adaptive agents’, and ‘social agents’. Furthermore, it is important to note that ‘environment’ should be interpreted in such a way that it includes non-physical environments such as cyberspace; arguably the most successful artificial agents are those who perform one or more of the criteria above in cyberspace — usually referred to as bots. The notion of autonomy is also used in different senses. At a minimum, ‘autonomous’ carries some of its philosophical meaning in the sense that an autonomous agent should be able to make informed decisions (based on its knowledgebase, rules and sensory input) and act accordingly. However, Haselager [2007] has argued that the notion of autonomy in robotics and philosophy is radically different, referring respectively to independent performance of tasks and capacity to choose goals for oneself. He suggests that this gap can be bridged by interpreting ‘autonomy’ in terms of what it is that makes one’s goals genuinely one’s own, further emphasizing the importance of embodiment to genuine autonomy. Insofar as artificial agents need to act in a complex and dynamic world, many of the epistemological questions discussed through the history of philosophy become relevant. For instance, they should ideally be able to revise their “beliefs” in light of their experiences, which requires functions for preserving consistency between beliefs, differentiating between beliefs that require justification and those that do not and so forth. Importantly, such revision will differ depending on what kind of substantive theory of truth is used. For instance, an intelligent agent will operate differently depending on whether it will revise its beliefs according to a coherentist or correspondence theory of truth. Thus, there is a close connection between epistemological problems in philosophy and robotics (cf. [Lacey and Lee, 2003]). Colloquially, people tend to use the word ‘robot’ for artificial agents that have a human or animal appearance and these robots tend to produce more natural man-machine interaction. At least, this is the guiding principle behind the MIT Cog project. Cog is a humanoid robot designed to gain experience from natural interactions with humans. The guiding principles behind this project are that human-level intelligence requires social interactions akin to those of a human infant and that a humanoid robot is more likely to elicit natural interactions [Brooks, 2002]. As such, Cog and other advanced intelligent agents can be seen as a means of empirically testing the more abstract theories in philosophy of artificial intelligence. In particular, Cog illustrates a recurring theme in philosophy of AI. Many philosophers have claimed that “the nonformalizable form of information processing . . . is possible only for embodied beings” [Dreyfus, 1992, p. 237] and that robotics stands a better chance of producing human-like intelligence. This claim, originally made by Turing more than 50 years ago, has been taken to heart by Rodney Brooks, who claims that the visionary ideas of Turing can now be taken seriously; AI should focus on robotic architectures that are inspired by biology and interact with the actual world rather than simply reason according to a set of

1374

Philip Brey and Johnny Hartz Søraker

formal rules and knowledge bases that are far removed from the complexity that human-like intelligence must be able to handle (cf. [Brooks, 2002]).

4.6 AI and ethics If we regard the ultimate goal of AI to be machines capable of doing things that have traditionally required human intelligence, many ethical questions become obvious. For instance, can computers be trusted with tasks that involve considerable risk to others, and what are the social and existential implications of substituting man for machine to an even higher degree than we already have? One major problem with the development of autonomous systems is that it tends to erode the notion of ‘responsibility’, arguably the most important concept in both ethics and law. In general, it can be difficult to assign responsibility if computer malfunction results in loss of lives (see also Section 6.1). Are the designers, the users, or perhaps even the artificial agents themselves responsible? Indeed, Sullins [2006] argues that if the only way we can make sense of an artificial agent’s behavior is to ascribe to it an understanding of responsibility — then it should be treated as having both rights and responsibilities. Floridi and Sanders have argued that although artificial agents cannot be blamed or praised for their actions, they ought — when seen at a given level of abstraction — to be regarded as moral agents in the sense of being the sources of good or evil [Floridi and Sanders, 2004]. They further claim that regarding artificial agents as moral agents does not reduce the responsibility of the designers. On the contrary, seeing them as sources of immorality should prompt us to pay extra attention to the kinds of agency these entities have. Johnson, however, sees a danger in assigning moral agency to artificial agents, “because it disconnects computer behavior from human behavior, the human behavior that creates and deploys the computer systems” [Johnson, 2006, p. 204]. That is, contrary to Floridi and Sanders, Johnson argues that the design of artificial agents is more likely to be subject to moral scrutiny if we focus on computer systems as human-made rather than as independent moral agents. A clear distinction between humans and computers also underlies Moor’s conclusion that computers should not be allowed to make decisions about our basic goals, values, and priorities between them [Moor, 1979]. As these viewpoints on artificial moral agency show, the source of many ethical problems in AI stems from the fact that these systems tend to be opaque. That is, they make choices and decisions according to criteria of which the users have little or no understanding. These operations can even be opaque to the designers themselves, especially when built upon a connectionist or evolutionary architecture (cf. Sections 3.2 and 4.2). With the complexity required for a machine to act intelligently, it could even be argued that it is impossible to safeguard against their malfunction and that the creation of war robots (cf. [Asaro, forthcoming]) and other AI systems capable of massive destruction is inherently unethical. Indeed, this line of reasoning famously led David Parnas to take a public stand against the so-called Star Wars program during the cold war, arguing that it would be

Philosophy of Computing and Information Technology

1375

impossible to create artificial intelligence that can reliably be trusted to prevent nuclear attacks (cf. [Parnas, 1985]). Some more futuristic ethical problems have also been raised, for instance surrounding the discussion whether artificial agents can acquire moral status and rights comparable to humans, or at least simpler life forms. These kinds of questions have long been the subject of literature and movies. Many of the most fundamental problems are illustrated by Asimov’s famous laws of robotics, which state that 1) a robot may not injure a human being, or, through inaction, allow a human being to come to harm; 2) a robot must obey orders given it by human beings, except where such orders would conflict with the First Law; 3) a robot must protect its own existence as long as such protection does not conflict with the First or Second Law [Asimov, 1968]. The ethical problems that arise in Asimov’s works typically arise as a result of irresolvable conflicts between these laws, illustrating the difficulty with which ethical guidelines can be formalized. The laws also presuppose that robots have mere instrumental value and consequently should be regarded as means and not as ends in themselves. This raises the question of what it would do to our sense of humanity if machines were to become better at reasoning than humans, rationality having traditionally been seen as the very essence of humanity (see e.g. [Mazlish, 1993]). Furthermore, Asimov’s laws illustrate two of the biggest problems in creating the kinds of robots that would need such laws. First, such a robot must be able to make nuanced and reliable distinctions between moral patients, non-moral patients — and its own being. Second, it must be able to understand the consequences of its actions — and inactions. Allen, Smit and Wallach suggest that the latter can be done either in a top-down approach, which involves turning moral theories into algorithms, or bottom-up, which involves attempts to train artificial agents in such a way that their behavior emulates morally praiseworthy human behavior (see [Allen et al., 02005]; see also [Clarke, 1994]). However, regardless of how successful we are in trying to create artificial morality, the mere attempt can be advantageous for many of the same reasons that AI in general could lead to new insights without necessarily leading to success. As Knuth puts it, in speaking of the impact of computers on mathematics, the mere attempt to “formalize things as algorithms leads to a much deeper understanding than if we simply try to understand things in the traditional way” [Knuth, 1974a, p. 327]. Finally, if computers become reliable to such a degree that we willingly leave our deliberations and decisions to the computer, does this entail that our autonomy is reduced? Perhaps the first to raise these kinds of issues was Joseph Weizenbaum [1976], who himself had created the famous artificial therapist ELIZA. Weizenbaum became increasingly worried about the effects of ELIZA, especially due to the fact that people confided in it despite knowing that it was a computer program and that psychologists considered developing it further and putting it into real practice. Weizenbaum argued that the question is not what intelligent machines can do, but whether we should allow them to do it. Ethics is not a question

1376

Philip Brey and Johnny Hartz Søraker

that should be raised by AI, but it should be the very foundation of AI; the justification for building a system in the first place should be an ethical question.

5 PHILOSOPHY OF THE INTERNET AND NEW MEDIA This section discusses issues and approaches in the newly emerged field of philosophy of the Internet and new media, sometimes called cyberphilosophy, which has emerged together with the multidisciplinary field of new media studies. The first section will give a broad outline of new media and discuss theories on how society has increasingly become an information society. In the two subsequent sections, we consider epistemological and ontological issues relating to the Internet and other new media. Section 5.4 considers new media as a platform for communication and the establishment of virtual communities, followed by a related section on the Internet as a political venue. The chapter concludes with a section on how our identity is affected by the disappearing barriers between body and technology and between real and virtual selves. Ethical issues will be considered occasionally throughout, but will also be discussed separately in Section 6.

5.1 Theories of new media and the information society The emergence of multimedia computers in the 1980s and the Internet as a mass medium in the early 1990s created a new role for computer technology. This development moved computers beyond scientific, administrative and organizational applications, and made them into a social and mass medium — a general-purpose tool and environment for games, social networking, collaborative work, creative expression, art, film and photography, etc. ‘New media’ generally refers to the recent forms of media that rely on digital computers, including both the development of unique forms of digital media, such as virtual worlds, and the transformation of traditional media, such as movies streamed on demand on the Internet [Flew, 2002]. The development of new media is also closely related to the development of increasingly mobile, ubiquitous and interconnected devices, which enable access to new media at any time and any place. Another important feature of new media is its facilitation of user contributions; users can for instance generate and share original content, find and retrieve content on demand, or publicly rate, comment and recommend content. New media often involves many-to-many communication. This cannot be achieved by simply transferring (or pushing) information simultaneously from many to many. Instead, a venue must be created in which many can leave information and many can retrieve (or pull) information; anything else would amount to chaos and information overload. Thus, traditional forms of media are sometimes described as channels of information, but a more apt analogy for new media is a place for information — which is reflected in terms like cyberspace, infosphere, virtual worlds and virtual environments. This form of interactivity entails that

Philosophy of Computing and Information Technology

1377

users are not left with a choice between ’on’ or ’off’, but also what, when and how. Thus, new media has increasingly relied upon a community of users and often places emphasis on sharing and collaboration — what has also been referred to as a bazaar rather than a cathedral model [Raymond, 2000]. There is now considerable agreement among social theorists, economists and historians that contemporary society can be characterized as an information society, which is in important respects different from the industrial society that preceded it until the 1970s [Webster, 1995]. The information society is a society in which the production and use of information has a dominant role in economic and social processes. According to social theorist Manuel Castells [1996] the information society is the result of a transformation of the capitalist economy through information technology, which has made capitalism more flexible, more global and more self-sustained. In this new model, the basic unit of economic organization is no longer the factory but the network, made up of subjects and organizations and continually modified in adaptation to its (market) environment, with great ramifications for culture and society. The transition to an information society is also theorized by Van Dijk [2006], who argues that the information revolution in the 1970s was preceded by a crisis of control in organizations, which were held back by uncontrolled bureaucracy, limitations in transportation systems, and inadequacies of mass communication in an individualizing and diversifying society. He argues that new media technologies enabled a revolution in information and communication that helped solve these problems and enbabled the transition from a Fordist to a postfordist mode of production in which organizations become more streamlined and flexible and better able to operate on a global scale. These characterizations of the transition of an industrial to a postindustrial information society are accepted by postmodern thinkers like David Harvey, Frederick Jameson, Jean Baudrillard, Mark Poster, and Jean-Fran¸cois Lyotard, who add that these technological and economic changes are accompanied by distinct social, cultural and epistemological changes which designate a shift from a modern to a postmodern culture and society. In general, postmodern authors characterize the information society as a society in which modern life has become saturated by information, signals and media, in which there is a decline of epistemic and political authorities, and which is characterized by consumerism, commodification, simulation, a blurring of the distinction between representation and reality, and the fragmentation of experience and personal identity. Harvey [1989] has argued that the new economic system has led to a new dynamism in which work and consumption are sped up and made more competitive, and a new, postmodern culture which rejects the faith in reason and objective reality and accepts heterogeneity and commodification. Paul Virilio [1994] holds that the marriage of capitalism and new (media) technologies have created a culture of speed, which ultimately leads to a feeling of incarceration and confinement in the world. Jean Baudrillard [1995] theorizes a shift from an economy of goods to an economy of signs and spaces and characterizes the new era as an era of simulation, which leads to a disappearance of the distinction between representation and

1378

Philip Brey and Johnny Hartz Søraker

reality, past and future, catching people in a disorienting postmodern hyperspace. Baudrillard claims, along with Poster [1990] and Lyotard [1984] that contemporary life is ruled by a new ‘mode of information’, in which life is quintessentially about symbolization, about exchanging and receiving messages about ourselves and others. The emergence of the Internet as a mass medium in the 1990s has further strengthened the arguments of theorists of the Information society. However, it has also left some of the older theories dated, due to its explosive development. Floridi is one of few who have theorized the significance of the Internet. He argues that those born in the 21st century are probably the last generation to experience a clear difference between offline and online. Its dissolution is due to three fundamental trends. First, there has been a steady increase in the kinds of information that can be represented digitally, already encompassing all other forms of media. The same holds for the amount of information produced. Second, technologies like Radio Frequency Identifiers (RFID) will increasingly allow us to be continuously online and to communicate with non-living entities like cookware and clothing (and for them to communicate with each other). Third, the information society is becoming a collection of “connected information organisms” or ‘Inforgs’ and we are about to become the “only biological species capable of creating a synthetic environment to which it then must adapt” [Floridi, 2006]. Floridi’s vision emphasizes the emergence of new and pervasive forms of connectedness, between humans and humans, humans and machines, and machines and machines, all mediated by the flow of information. How the new form of sociality this entails for humans differs from — and whether it is less valuable than — traditional forms gives rise to many of the issues discussed in the following sections.

5.2 Internet epistemology The Internet is a global tool for the production, storage, dissemination, and consumption of information and knowledge on which a large percentage of the world population relies. Given this dominant role an investigation of the epistemic properties of Internet technology and practices of information production, management and utilization on the Internet is warranted. The term “Internet epistemology” was first introduced by Paul Thagard [2001] to refer to the epistemology of scientific information practices on the Internet, but is now gaining a wider usage to include everyday information practices as well. Issues in Internet epistemology include the epistemic quality of Internet information, the normative implications of the Internet for information production and consumption and the epistemology of Internet-related information practices, including information utilization, management and production. In addition it could propose improved epistemic practices and technologies. The quality of information on the Internet, and the epistemic value of the Internet as a source of information, has been questioned by Alvin Goldman [1999, 161-189]. Goldman argues that while the Internet strongly increases the amount of

Philosophy of Computing and Information Technology

1379

information available to us, it need not be true that we know more as a result. As he notes, the quality of the Internet as a source of information depends on both its support for intelligent information retrieval that gives users access to information relevant to their interests, and the reliability of the information that is thus made accessible. The first issue will be referred to as the problem of relevance, while the second will be called the problem of reliability. It is generally agreed that the problem of relevance has not yet been solved adequately. The Internet is often criticized for offering users vast amounts of information without sufficient means for identifying and retrieving the information that is most relevant to their interests. This has been argued to contribute to information overload, or information glut, which is a condition in which recipients are being provided with more information than they can meaningfully process [Shenk, 1998; Himma, 2007b]. Information overload leads to information fatigue and to recipients becoming paralyzed in their decision-making capabilities or to them remaining uninformed about topics. Proper information management, both by information providers and recipients, can provide a solution to information overload. Tagging and categorization of web documents and sites, search engines, hierarchical directories, filtering techniques, hyperlinking, personalized information retrieval profiles, and the development of semantic web technologies can further facilitate information retrieval. Many of these techniques depend on automated procedures. Dreyfus [2001, p. 8-26] argues that these will ultimately fail because computers are insufficiently capable of discerning relevant from irrelevant information (see also Section 4.2). Levy [2008] argues that good information management is not enough to avoid information overload, and that the creation of space and time for thinking, reflection and extensive reading is necessary as well. The problem of reliability also looms large for the Internet. Issues include the fact that anyone can place whatever information on the Internet that they please, that the source of information on the Internet is often unclear, and that it is often unclear whether information on the Internet is still current. Furthermore, websites usually lack information on criteria used in the selection of the information provided or referenced on them. The problem of reliability of Internet information has been addressed by Anton Vedder [2008; Vedder and Wachbroit, 2003]. Vedder argues that generally, persons evaluate the reliability of information presented to them by means of two types of criteria: content criteria and pedigree criteria. Content criteria are criteria of reliability inferred from the information itself. They include criteria of consistency, coherence, accuracy, accordance with observations, and interpretability, accessibility, and applicability relative to the user’s capacities and interests. Vedder argues that people often cannot determine the reliability of information on content criteria alone. Reliability is often evaluated through pedigree criteria: epistemic criteria to assess the authority, trustworthiness and credibility of the persons or organizations behind the information. Pedigree information is often not available in Internet information, so that according to Vedder recipients are dependent on content criteria alone. Two such criteria have become dominant: ac-

1380

Philip Brey and Johnny Hartz Søraker

cessibility and usability. Many users choose to rely on Internet information solely based on it being easily available and applicable for their purposes. Vedder argues that this undesirable state-of-affairs can be remedied through two strategies: developing critical attitudes in recipients and making pedigree criteria visible on the Internet by creating institutionally licensed seals and procedures for verifying the credibility of websites. Fallis [2004] offers a somewhat similar argument based on the epistemology of testimony. In addition to the quality of Internet information, philosophers have also analyzed and evaluated the wider implications of the Internet for information production and consumption. Floridi [1999, 81-87] outlines eleven implications of the Internet for organized knowledge, including the decentralization and possible fragmentation of knowledge, a potential blurring of the distinction between information and knowledge, the emergence of the computerized scholar, and the loss of information on paper. He argues that we are “giving the body of organized knowledge a new electronic life” and argues that we must do so carefully and wisely. Hubert Dreyfus [1999, 10-11] argues that we are moving from a library culture, built on classification, careful selection and permanent collections, to a hyperlinked culture, involving diversification, access to everything, and dynamic collections. The old library culture presupposes a modern subject with a fixed identity seeking a more complete and reliable model of the world, whereas the new hyperlinked culture assumes a postmodern subject not interested in collecting and selecting but in connecting to whatever information is out there. Dreyfus claims that the old information culture is superior to the new one; poststructuralists and postmodernists would argue the opposite. Similar conservative positions are taken by Albert Borgmann [1999], and Phil Mullins [1996], who worries that in a culture which relies on electronic documents and hypertext, books become fluid and decentered, and canons dissolve. Authors who study the organization of information on the Internet agree that the dominant mode of organization is hypertextual. Hypertext is text which contains dynamic links, called hyperlinks, to other texts. When using a text, users can retrieve related texts, by clicking the relevant hyperlink. Dreyfus [1999, 826] argues that hyperlinks link pieces of information to each other based on some subjective perceived relationship, removing hierarchy and authority from the organization of information, and with it, meaningfulness. Hyperlinks, he argues, are not designed to support the retrieval of meaningful and useful information, but rather to support lifestyles oriented at surprise and wonder. Floridi [1999, pp. 116-131) presents an opposite point of view, arguing that hypertext significantly contributes to the meaningfulness of information on the Internet by providing semantic structure between its separate texts. Hypertext, he argues, allows for multi-linear narratives and a more open and flexible space of knowledge which he does not characterize as postmodern, but rather as marking a return of the Renaissance mind. The Internet has become a medium for collaborative knowledge creation, enabling the utilization of both synchronous (i.e., real-time) media such as real-time

Philosophy of Computing and Information Technology

1381

groupware, instant messaging, and multi-user editors, and asynchronous media such as email, version control systems, open source software development tools, wiki’s and blogs. What are the epistemological properties of both these media and the associated collaborative practices, and how can we compare the epistemic quality of alternative collaborative practices? These questions fall within the scope of social epistemology [Goldman, 1999], which is the study of the social dimensions of knowledge or information. The most innovative development in knowledge production on the Internet is the emergence of nonhierarchical forms of mass collaboration, including the creation of wiki’s, open source software design, and collaborative blogging [Tapscott and Williams, 2008; Sunstein, 2006]. These practices are community-based, voluntary, egalitarian, and based on self-organization rather than top-down control. The relative success of these practices seems to show that such systems of knowledge creation can be as successful as more traditional systems. Philosophers have already considered the reliability of Wikipedia [Fallis, 2008], of news blogging [Goldman, 2008] and the quality of scientific knowledge production and assimilation through the Internet [Thagard, 2001]. Turning to knowledge utilization, one development that merits attention is distance learning, or distance education. Hubert Dreyfus [1999] has argued that education centrally involves the transmission of skills and the fostering of commitment by educators in students to develop strong identities. According to Dreyfus such aspects of education cannot adequately be transferred in distance education since they require bodily presence and localized interactions between students and teachers. Prosser and Ward [2000] add that the transfer of “practical wisdom” in education requires communities with interpersonal connectivity among its members, something virtual communities in distance education cannot provide because of their relative anonymity, the lack of mutual commitments, and the risk of an overload of trivial information. Nissenbaum and Walker [1998] provide a more nuanced view, arguing that the implications of information technology for learning depend on the actions and attitudes of instructors and policy makers.

5.3

The ontology of cyberspace and virtual reality

The software constructs with which computer users interact, such as files, folders, and web pages, exist virtually rather than physically. Although they are realized and sustained by means of physical systems, they do not exist solely or primarily as physical entities. The existence of virtual objects, and correlated virtual spaces, actions and events raises questions regarding their ontological status: what is their mode of existence, and what is their place in philosophical ontology? Let us call nonphysical software-generated objects and spaces with which users interact virtual entities. Virtual entities are represented as part of the user interface of computer programs. They manifest themselves to the user through symbolic or graphical representations, and they interactively respond to actions of the user. Contemporary user interfaces are in most cases graphical, representing virtual ob-

1382

Philip Brey and Johnny Hartz Søraker

jects as ordinary, manipulable physical objects. Virtual reality(VR) is special kind of graphical user interface which presents a computer-generated immersive, threedimensional, interactive environment that is accessed and manipulated using, for instance, stereo headphones, head-mounted stereo television goggles, and datagloves. VR allows for a representation of virtual entities with a great degree of realism. The Internet and other computer networks define collective, multi-user virtual entities in a collective user environment that is often called cyberspace [Benedikt, 1991]. Although cyberspace is accessed using graphical user interfaces, most virtual entities it contains are informational objects like web pages, text and image files, and video documents. Located conceptually in between cyberspace and VR one finds virtual environments or virtual worlds. Although the term virtual environment is sometimes used synonymously with virtual reality, it more often is used to denote any interactive computer simulation of an environment, whether represented textually or graphically, and whether immersive or nonimmersive, which can be navigated by users. (For a discussion of these and related conceptual distinctions, see [Brey, 2008; 1998]). Brey [1998] argues that the analogy between virtual and physical spaces goes deep because both are topological spaces as defined by mathematical topology, a branch of mathematics. Topological spaces are mathematical structures that define abstract relations of closeness and connectedness between objects in terms of relationships between sets rather than geometrical properties. Directories and hyperlinked websites are topological in this sense. This characterization of virtual space does not yet answer our question regarding the ontological status of virtual entities. A common conception is that “virtual” contrasts with “real” and that therefore virtual entities are not real in an ontologically robust sense. They are hence constructions of the mind, or mere representations. Borgmann [1999] argues that virtual reality is therefore always only a make-believe reality [Borgmann, 1999]. Philip Zhai [1998] takes a radically opposing point of view, arguing that something is real when it is meaningful to us, and that consequently there is no principled ontological distinction between virtual and physical reality. Steering between idealist and realist conceptions of virtual entities, Brey [2003] has argued that virtual is not the opposite of real, and that some virtual entities are virtual and real at the same time. Brey argues that a distinction can be made between two types of virtual entities: simulations and ontological reproductions. Simulations are virtual versions of real-world entities that have a perceptual or functional similarity to them, but do not have the pragmatic worth or effects of the corresponding real-world equivalent. Ontological reproductions are computer simulations of real-world entities that have (nearly) the same value or pragmatic effects as their real world counterparts. He argues that two classes of physical objects and processes can be ontologically reproduced on computers. A first class consists of physical entities that are defined in terms of visual, auditory or computational properties that can be fully realized on multimedia computers, such as images, movies, musical pieces, stereo systems and calculators.

Philosophy of Computing and Information Technology

1383

A second class consists of what John Searle [1995] has called institutional entities, which are entities that are defined by a status or function that has been assigned to them within a social institution or practice. Examples of institutional entities are activities like buying, voting, owning, chatting, playing chess, trespassing and joining a club, and requisite objects like contracts, money, chat rooms, letters and chess pieces. Most institutional entities are not dependent on a physical medium and can thus exist in virtual or electronic form, because they are only dependent on the collective assignment of a status or function. Institutional virtual entities are the focus of David Koepsell [2000], who critiques the existing legal ontology of entities in cyberspace. Because virtual entities may have different ontological statuses, users of cyberspace may encounter ontological confusion. Although Brey hence blurs the distinction between reality and virtuality, he does maintain a distinction between reality and fiction. Some authors have argued that this distinction is also being erased with the emergence of computer-generated realities. Jean Baudrillard [1995] has claimed that information technology, media, and cybernetics have ushered in an era of simulation, in which models, signs and codes mediate access to reality and define it to the extent that people cannot meaningfully distinguish between simulations and reality anymore. Albert Borgmann [1999] has argued that virtual reality and cyberspace have incorrectly led many people to confuse them for alternative realities that have the same actuality as the real world, whereas he holds that VR and cyberspace are merely forms of information that should be treated as such. Next to these ontological and epistemological questions regarding the distinction between the virtual and the real, there is the moral question of the goodness of virtuality [Brey, 2008]. First of all, are virtual things better or worse, more or less valuable, than their physical counterparts? Some authors have argued that they are in some ways better: they tend to be more beautiful, shiny and clean, and more controllable, predictable, and timeless. They attain, as Michael Heim [1993] has argued, a supervivid hyper-reality, like the ideal forms of Platonism, more perfect and permanent than the everyday physical world. Critics have argued that these shiny objects are mere surrogates that lack authenticity [Borgmann, 1999] and that presence in VR and cyberspace gives a disembodied and therefore false experience of reality and present one with impoverished experiences [Dreyfus, 2001]. More optimistically, Mooradian [2006] claims that virtual environments and entities are good at creating hedonistic value as well as certain types of perfectionist value, notably intellectual and aesthetic value, though not value located in excellent physical activities.

5.4

Computer-mediated communication and virtual communities

The Internet has become a medium for communication and social interaction, and an increasing part of social life is now taking place online. The study of online social life is being undertaken in a number of new and overlapping interdisciplinary fields that include new media studies, Internet studies, cyberculture

1384

Philip Brey and Johnny Hartz Søraker

studies, cyberpsychology and computer-mediated communication. In philosophy, online social life has been studied within philosophy of computing and computer ethics, but the relevant parent discipline is social philosophy, which is the philosophical study of behavior, social structure and social institutions. Let us first consider philosophical issues in computer-mediated communication. Computer-mediated communication (CMC) is interactive communication across two or more networked computers. It includes both synchronous and asynchronous communication, and such formats as e-mail, instant messaging, chatrooms, bulletin boards, listservs, MUDs, blogs, video conferencing, shared virtual reality and software-supported social networking. The field of CMC studies these different forms of communication as well as their social effects. Philosophical studies of CMC are quite diverse, including issues like online identity, virtual public spheres, Internet pornography, and power relations in cyberspace, and thus seem to consider many issues in the philosophical study of social life online if not the Internet at large [Ess, 1996; 2004]. A key philosophical issue for CMC is how different types of CMC formats can be normatively evaluated. Is CMC epistemically and socially inferior to offline communication, including face-to-face communication, or is it in some ways superior? Shank and Cunningham [1996] argue that many forms of CMC involve multiloguing, which is unscripted, simultaneous, non-hierarchical conversation involving multiple participants. They argue that multiloguing supports diversity in perspectives, integration of knowledge, equality in participation, and access to archived dialogue, and therefore presents a superior mode of communication. Dreyfus [2001] takes a more negative view, arguing that important qualities are lost in CMC, including the movements and expressions of the body, a sense of context, and genuine commitment and risk-taking. Another major philosophical topic in CMC is communication across different cultures and worldviews [Ess, 2002; Ess and Sudweeks, 2005]. Ess [2002] has argued that cross-cultural studies of CMC can help resolve long-standing questions about the nature of culture, knowledge, politics, and the self. Another important development is online social networking. The Internet is being used to build up networks of acquaintances, and to forge and maintain friendships and even love relationships. A question for philosophy is how this development and the resulting new types of relationships should be evaluated. One central issue is whether online social relationships can include mutual trust [Weckert, 2005; Nissenbaum, 2001]. Pettit [2004] argues that genuine trust (as opposed to mere reliance) is not possible in exclusively online relationships because the Internet does not sufficiently support the justification of beliefs in loyalty and the communication of trust. De Laat [2005] presents an opposing view, arguing that enough social and institutional cues can be used online to develop trust. Cocking [2008] argues that fully computer-mediated personal relationships cannot be as rich and genuine as offline relationships because people have too much control over our self-presentation online. Briggle [2008] takes issue with this position, arguing that the Internet is well-suited for fostering close friendships based

Philosophy of Computing and Information Technology

1385

on mutual self-exploration because it creates distance and supports deliberate behavior. Ben-Ze’ev [2004] argues that the Internet enhances love relationships because it allows for meaningful online relationships in which people can express themselves very directly and in which they can live out interactive fantasies. Briggle [2008b] presents a general framework for the interpretation and evaluation of different types of online love relationships. Brey [1998], finally, criticizes the increasing substitution of social interaction by interaction with software agents and the proliferation of social relationships with virtual characters and pets. People do not only form individual social relationships online, they also form online communities, or virtual communities, that have their existence in cyberspace. Virtual communities can be closed or open and may be intended to maintain existing relationships, or explore various kins of shared interests. It has been questioned whether virtual communities constitute genuine communities and whether they are inferior to traditional, local communities, whether they effectuate social integration or fragmentation, and whether they cause a harmful erosion of traditional, local communities [Feenberg and Barney, 2004]. Many authors have defended virtual communities, arguing that they can embody all the qualities of traditional communities, including mutual trust, care and a sense of belonging [Rheingold, 1993]. Virtual communities have been assessed mostly positively by postmodern philosophers like Lyotard and Bolter because of the non-Cartesian, decentered, fragmented, and hypertextual nature of the identities portrayed by their users (cf. Section 5.6). Others have argued that virtual communities are inferior to traditional ones. Borgmann makes a distinction between instrumental, commodified and final communities and argues that virtual communities can at best be instrumental or commodified, because they do not contain “the fullness of reality, the bodily presence of persons and the commanding presence of things” found in final communities [Borgmann, 2004, p. 63]. In a similar fashion Barney [2004] sees virtual communities as inferior due to their lack of physical practices, and Dreyfus is critical of what he describes as the nihilist, irresponsible and often uninformed nature of virtual communities [Dreyfus, 2004]. Winner, finally, has criticized the fact that any kind of online network is called a community, since this broad definition ignores the importance of “obligations, responsibilities, constraints, and mounds of sheer work that real communities involve” [Winner, 1997, p. 17]. Interestingly, both Rheingold and Bolter have recently also adopted more conservative positions on virtual communities. Does the proliferation of virtual communities and online social networks support social integration or does it lead to social fragmentation? Many years before the Internet, Marshall McLuhan [1962] already claimed that electronic mass media were bringing about a global village in which people are globally interconnected by electronic communications. It has subsequently been claimed that the Internet, more than other electronic media, has instantiated a global village. This view has met with serious criticism. The notion of a global village suggests civic engagement and a unified public sphere. Instead, Robert Putnam [2001] has argued, the

1386

Philip Brey and Johnny Hartz Søraker

ubiquitous creation of interest-based online communities has brought about a cyberbalkanization of online social life: a process in which cyberspace is divided into narrowly focused groups of individuals with shared views and experiences, that cut themselves off from alternative views and critique. A similar view is presented by Sunstein [2001], who emphasizes that this process is not only caused by the creation of interest-based virtual communities, but also by the increasing ability to individually filter information on the Internet in accordance with one’s previously formed beliefs.

5.5 The Internet and politics Political philosophy studies what kinds of political institutions we should have. It analyzes, criticizes and defends major political ideologies like liberalism, socialism, and conservatism, and tries to give content to central concepts in political theory like power, liberty, democracy and the state. The Internet is becoming an object of study for political philosophy for two reasons. First, the Internet is becoming an important means by which politics is pursued. It is being used for political dialogue between politicians and citizens, for political organization and activism, for electronic voting, for political reporting, and even for terrorist attacks. The use of the Internet for political activity has been termed cyberpolitics. A political philosophy of the modern state that does not take the existence of cyberpolitics into account runs the risk of using an outdated conception of the political process. In addition, cyberpolitics itself is a worthy object of study, since legitimate questions can be raised concerning the way cyberpolitics ought to be conducted. The second reason that the Internet ought to be studied by political philosophy is because of the emergence of virtual communities and social networks in cyberspace (Section 5.4). These social structures have emerged in a medium, cyberspace, that is not subjected to the political authority of any nation of conglomerate of nations, and therefore constitutes a “stateless society”. Nevertheless, cyberspace has a politics; it has processes by which individuals and groups negotiate conflicts of interests and attempt to exercise power and authority. A question for political philosophy is what the politics of cyberspace ought to be, that is, what political institutions and regulations ought to be in place in cyberspace. Although the politics of cyberspace and cyberpolitics are conceptually distinct, their relation should also be considered. The way in which cyberspace is organized politically may have serious consequences for the extent to which it can be used as a means for politics in the “real” world by different groups. Conversely, agents that use the Internet for “real-world” politics may adopt a presence in cyberspace and establish interactions with other agents in it, thereby becoming part of the social fabric of cyberspace and hence of its politics. The politics of cyberspace have been an issue long before the emergence of cyberpolitics. Early pioneers of the Internet considered it a free realm, a new “electronic frontier” not subjected to laws. As Langdon Winner [1997] has argued, the dominant political ideology of Internet users and authors in the 1980s and 1990s was

Philosophy of Computing and Information Technology

1387

cyberlibertarianism, which he defines as an ideology that embraces radical individualism, including strong conceptions of individual rights along with an embrace of free-market capitalism, a rejection of regulative structures and a great optimism about technology as a means for liberation. Winner criticizes cyberlibertarianism for its neglect of issues of justice and equality, and its conceptions of citizenship and community. In its place, he proposes cybercommunitarianism, in which political structures are put in place that support communities rather than individuals. Another issue concerns the democratic nature of cyberspace. Some have argued that the Internet is inherently a democratic technology, as it is designed as consisting of a network of equal nodes with equal opportunities for sending and receiving information. This design obliterates hierarchies, it has been claimed, and supports direct, participatory democratic processes. Deborah Johnson [1997] cautions that although the Internet can indeed empower people, the filtering of information by authorities and the insulation from diverse perspectives for which the Internet allows can counter its democratic tendencies (cf. [Sunstein, 2008]). Søraker [2008], however, has argued that the increasing use of frameworks within which Internet users can contribute nontextual information constitutes a serious obstacle to government attempts to censor and monitor Internet traffic. Whether or not the Internet has an inherent tendency to support democratic processes, it has often been argued that cyberspace ought to be organized to support direct democracy and citizenship by functioning as a public sphere, an area in social life in which people get together to discuss issues of mutual interest, and to develop shared opinions and judgments, and take political action when appropriate [Gimmler, 2001; Bohman, 2008]. Both the idea that cyberspace should function as a public sphere and that it is capable of doing so have been criticized [Dean, 2002]. The functioning of cyberspace as a public space has, in any case, come under pressure since the mid-1990s with the emergence of e-commerce and the concomitant processes of commodification and privatization. E-commerce has brought along increased governmental regulation of cyberspace, and enhanced attention to issues of intellectual property rights, security and cybercrime, as well as to commercial free speech consumer privacy, and ethical issues of pornography. Let us now turn to philosophical issues in cyberpolitics. An overarching question here is how the Internet can and should be used to support democratic political processes in the “real” world [Chadwick, 2006; Shane, 2004]. What should governments, providers and others do to strengthen or utilize the possibilities of the Internet for supporting “real-life” democratic politics, including better means for political communication, deliberation, participation and activism? Should all government information be made available online? Should electronic voting be introduced [Pieters, 2006]? What role can and should the Internet have in international and global political processes? A specific political issue is raised by the digital divide, the existence of a gap of effective access to information technology because of preexisting imbalances in resources and skills [Norris, 2001; Van Dijk, 2005]. The digital divide has been argued to exacerbate inequalities in society, since effective access to information

1388

Philip Brey and Johnny Hartz Søraker

technology has become an important condition for social and economic success. Van den Hoven and Rooksby [2008] argue that in contemporary societies information qualifies as a Rawlsian primary good, a necessary means to realizing one’s life plan, and derive a set of criteria for the just distribution of access to information. A final issue concerns the politics of national security and cyberterrorism. The Internet has become a critical infrastructure that is vulnerable to cyberattacks, and also functions as a platform where conventional terrorist attacks may be prepared and discussed. This raises the question of how much control the government should exercise over the Internet in the interest of national security [Nissenbaum, 2005] and where the boundaries should be drawn between cyberterrorism and other subversive online activities, such as cyberactivism, cybercrime and hacking [Manion and Goodrum, 2000].

5.6 Cyborgs and virtual subjects Information technology has become so much part of everyday life that it is affecting human identity (understood as character). Two developments have been claimed to have a particularly great impact. The first of these is that information technologies are starting to become part of our bodies and function as prosthetic technologies that take over or augment biological functions, turning humans into cyborgs, and thereby altering human nature. A second development is the emergence of virtual identities, which are identities that people assume online and in virtual worlds. This development has raised questions about the nature of identity and the self, and their realization in the future. Philosophical studies of cyborgs have considered three principal questions: the conceptual question of what a cyborg is, the interpretive and empirical question of whether humans are or are becoming cyborgs, and the normative questions of whether it would be good or desirable for humans to become cyborgs. The term “cyborg” has been used in three increasingly broad senses. The traditional definition of a cyborg, is that of a being composed of both organic and artificial systems, between which there is feedback-control, with the artificial systems closely mimicing the behavior of organic systems. On a broader conception, a cyborg is any individual with artificial parts, even if these parts are simple structures like artificial teeth and breast implants. On a still broader conception, a cyborg is any individual who relies extensively on technological devices and artifacts to function. On this conception, everyone is a cyborg, since everyone relies extensively on technology. Cyborgs have become a major research topic in cultural studies, which has brought forth the area of cyborg theory, which is the multidisciplinary study of cyborgs and their representation in popular culture [Gray, 1996]. In this field the notion of the cyborg is often used as a metaphor to understand aspects of contemporary — late modern or postmodern — society’s relationship to technology, as well as to the human body and the self. The advance of cyborg theory has been credited to Donna Haraway, in particular her essay “Manifesto for Cyborgs”

Philosophy of Computing and Information Technology

1389

[Haraway, 1985]. Haraway claims that the binary ways of thinking of modernity (organism-technology, man-woman, physical-nonphysical and fact-fiction ) traps beings into supposedly fixed identities and oppresses those beings (animals, women, blacks, etc.) who are on the wrong, inferior side of binary oppositions. She believes that the hybridization of humans and human societies, through the notion of the cyborg, can free those who are oppressed by blurring boundaries and constructing hybrid identities that are less vulnerable to the trappings of modernistic thinking (see also [Mazlish, 1993]). Haraway believes, along with many other authors in cyborg theory (cf. [Gray, 2004; Hayles, 1999]) that this hybridization is already occurring on a large scale. Many of our most basic concepts, such as those of human nature, the body, consciousness and reality, are shifting and taking on new, hybrid, informationalized meanings. Coming from the philosophy of cognitive science Andy Clark [2003] develops the argument that technologies have always extended and co-constituted human nature (cf. [Brey, 2000]), and specifically human cognition. He concludes that humans are “natural-born cyborgs” (see also the discussion of Clark in Section 3.6). Philosophers Nick Bostrom and David Pearce have founded a recent school of thought, known as transhumanism that shares the positive outlook on the technological transformation of human nature held by many cyborg theorists [Bostrom, 2005; Young, 2005]. Transhumanists want to move beyond humanism, which they commend for many of its values but which they fault for its belief in a fixed human nature. They aim at increasing human autonomy and happiness and eliminate suffering and pain (and possibly death) through human enhancement. Thus achieving a trans- or posthuman state in which bodily and cognitive abilities are augmented by modern technology. Critics of transhumanism and human enhancement, like Francis Fukuyama, Leon Kass, George Annas, Jeremy Rifkin and J¨ urgen Habermas, oppose tinkering with human nature for the purpose of enhancement. Their position that human nature should not be altered through technology has been called bioconservatism. Human enhancement has been opposed for a variety of reasons, including claims that it is unnatural, undermines human dignity, erodes human equality, and can do bodily and psychological harm [DeGrazia, 2005]. Currently, there is an increasing focus on ethical analyses of specific enhancements and prosthetic technologies that are in development, including ones that involve information technology [Gillett, 2006; Lucivero and Tamburrini, 2008]. James Moor [2004] has cautioned that there are limitations to such ethical studies. Since ethics is determined by one’s nature, he argues, a decision to change one’s nature cannot be settled by ethics itself. Questions concerning human nature and identity are also being asked anew because of the coming into existence of virtual identities [Maun and Corruncker, 2008]. Such virtual identities, or online identities, are social identities assumed or presented by persons in computer-mediated communication and virtual communities. They usually include textual descriptions of oneself and avatars, which are

1390

Philip Brey and Johnny Hartz Søraker

graphically realized characters over which users assume control. Salient features of virtual identities are that they can be different from the corresponding real-world identities, that persons can assume multiple virtual identities in different contexts and settings, that virtual identities can be used by persons to emphasize or hide different aspects of their personality and character, and that they usually do not depend on or make reference to the user’s embodiment or situatedness in real life. In a by now classical (though also controversial) study of virtual identity, psychologist Sherry Turkle [1995] argues that the dynamics of virtual identities appear to validate poststructuralist and postmodern theories of the subject. These hold that the self is constructed, multiple, situated, and dynamical. The next step to take is to claim that behind these different virtual identities, there is no stable self, but rather that these identities, along with other projected identities in real life, collectively constitute the subject. The dynamics of virtual identities have been studied extensively in fields like cultural studies and new media studies. It has been mostly assessed positively that people can freely construct their virtual identities, that they can assume multiple identities in different contexts and can explore different social identities to overcome oppositions and stereotypes, that virtual identities stimulate playfulness and exploration, and that traditional social identities based on categories like gender and race play a lesser role in cyberspace [Turkle, 1995; Bell, 2001]. Critics like Dreyfus [2001] and Borgmann [1999], however, argue that virtual identities promote inauthenticity and the hiding of one’s true identity, and lead to a loss of embodied presence, a lack of commitment and a shallow existence. Taking a more neutral stance, Brennan and Pettit [2008] analyze the importance of esteem on the Internet, and argue that people care about their virtual reputations even if they have multiple virtual identities. Matthews [2008], finally, considers the relation between virtual identities and cyborgs, both of which are often supported and denounced for quite similar reasons, namely their subversion of the concept of a fixed human identity.

6

COMPUTER AND INFORMATION ETHICS

This section surveys the field of computer and information ethics. The first section will define the field and will consider its aims and scope, its history, and major approaches and orientations. In the section thereafter, major topics in computer ethics will be surveyed, including privacy, security, free expression and content control, equity issues, intellectual property, and issues of moral responsibility. The final section will focus on the approaches of values in design and value-sensitive design, which aim to analyze embedded values in computer software and systems, and to devise methodologies for incorporating values into the design process.

Philosophy of Computing and Information Technology

6.1

1391

Approaches in computer and information ethics

Computer ethics is a field of applied ethics that addresses ethical issues in the use, design and management of information technology and in the formulation of ethical policies for its regulation in society. For contemporary overviews of the field, see [Tavani, 2007; Weckert, 2007; Spinello and Tavani, 2004; Himma and Tavani, 2008]. Early work in computer ethics or cyberethics had already started in the 1940s, soon after the invention of the computer. MIT Professor Norbert Wiener was a precursor of the field, already identifying many issues of computer ethics in his book The Human Use of Human Beings [Wiener, 1950]. The term “computer ethics” was first introduced in the mid-1970s by Walter Maner, who also promoted the idea of teaching computer ethics in computer science curricula [Maner, 1980]. The watershed year of 1985 saw the appearance of seminal publications by Jim Moor [1985] and Deborah Johnson [1985] that helped define the field. Since then, it has become a recognized field of applied ethics, with its own journals and conference series. In recent years, the field is sometimes also related to a more general field of information ethics, which includes computer ethics, media ethics, library ethics, and bioinformation ethics. Why would there be a need for computer ethics, while there is no need for a separate field of ethics for many other technologies, like automobiles and appliances? Jim Moor [1985] has argued that the computer has had an impact like no other recent technology. The computer seems to impact every sector of society, and seems to require us to rethink many of our policies, laws and behaviors. According to Moor, this great impact is due to the fact that computers have logical malleability, meaning that their structure allows them to perform any activity that can be specified as a logical relation between inputs and outputs. As a result computers can perform an incredible amount of functions, from word processor to communication device to gaming platform to financial manager. The versatility of computers is an important reason for the occurrence of a computer revolution, or information revolution, which is now transforming many human activities and social institutions. Many important things that humans do, including many that raise moral questions like stealing from someone, defaming someone, or invading someone’s privacy now also exist in electronic form. In addition, the computer also makes substantially new types of activities possible that are morally controversial, such as the creation of virtual child pornography for which no real children were abused. Because many of the actions made possible by computers are different and new, we often lack policies and laws to guide them. They generate what Moor has called policy vacuums, being the lack of clear policies or rules of conduct. The task of computer ethics, then, is to propose and develop new ethical policies, ranging from explicit laws to informal guidelines, to guide new types of actions that involve computers. Computer ethics has taken off since its birth in the mid-80s, and has established itself as a mature field with its own scientific journals, conferences and organizations. The field initially attracted most of its interests from computer scientists

1392

Philip Brey and Johnny Hartz Søraker

and philosophers. However, given the wide implications for human action sketched by Moor, computer ethics is also of interest to other fields that focus on human behavior and social institutions, such as law, communication studies, education, political science and management. Moreover, computer ethics is also an important topic of debate in the public arena, and computer ethicists regularly contribute to public discussions regarding the use and regulating of computer technology. Computer ethics is sometimes defined as a branch of professional ethics similar to other branches like engineering ethics and journalism ethics. On this view, the aim of computer ethics is to define and analyze the moral and professional responsibilities of computer professionals, for instance in the design, development and maintenance of computer hardware and software. Computer professionals are individuals employed in the information technology branch, for example as hardware or software engineer, web designer, network or database administrator, computer science instructor or computer-repair technician. Within this approach to computer ethics, most attention goes to the discussion of ethical dilemmas that various sorts of computer professionals may face in their work and possible ways of approaching them. Such dilemmas may include, for example, the question how one should act as a web designer when one’s employer asks one to install spyware into a site built for a client, or the question to what extent software engineers should be held accountable for harm incurred by software malfunction. Next to the discussion of specific ethical dilemmas, there is also general discussion of the responsibilities of computer professionals towards various other parties, such as clients, employers, colleagues, and the general public, and of the nature and importance of ethical codes in the profession. A recent topic of interest has been the development of methods for value-sensitive design, which is the design of software and systems in such a way that they conform to a desired set of (moral) values [Friedman, Kahn and Borning, 2006]. While the professional ethics view of computer ethics is influential, many in the field employ a broader conception that places the focus on general ethical issues in the use and regulation of information technology. This approach may be called the philosophical ethics approach to computer ethics. This conception holds, following Moor [1985], that computer ethics studies moral issues that are of broad societal importance, like information privacy and security, computer crime, issues of access and equity, and the regulation of commerce and speech on the Internet, and develops ethical policies to address them. It asks what ethical principles should guide our thinking about these issues, and what specific policies (laws, social and corporate policies, social norms) should regulate conduct with respect to them. Within this approach, some researchers focus on the development of ethical guidelines for users of computer technology. Others place more emphasis on policy issues, and try to formulate ethical policies for organizations, government agencies or lawmakers. Still others focus on computer technologies themselves, and try to identify and evaluate morally relevant features in their design. Some also focus on theoretical and metaethical issues.

Philosophy of Computing and Information Technology

6.2

1393

Topics in computer and information ethics

Introductions to computer ethics show considerable agreement on what the central issues for computer ethics are. They include ethical issues of privacy, security, computer crime, intellectual property, free expression, and equity and access, and issues of responsibility and professional ethics. Privacy Privacy is a topic that has received much attention in computer ethics from early on. Information technology is often used to record, store and transmit personal information, and it may happen that this information is accessed or used by third parties without the consent of the corresponding persons, thus violating their privacy. Privacy is the right of persons to control access to their personal affairs, such as their body, thoughts, private places, private conduct, and personal information about themselves. The most attention in computer ethics has gone to information privacy, which is the right to control the disclosure of personal data. Privacy issues come into play on the Internet, where cookies, spyware, browser tracking and access to the records of internet providers may be used to study the Internet behavior of individuals or to get access to their PCs. They also come into play in the construction of databases with personal information by corporations and government organizations, and the merging of such databases to create complex records about persons or to find matches across databases. Other topics of major concern include the privacy implications of video surveillance and biometric technologies, and the ethics of medical privacy and privacy in the workplace. It has also been studied whether people have a legitimate expectation to privacy in public areas, whether they can be freely recorded, screened and tracked whenever they appear in public and how the notion of “public” itself has changed in light of information technology. Security and crime Security has become a major issue in computer ethics, because of rampant computer crime and fraud, the spread of computer viruses, malware and spam, and national security concerns about the status of computer networks as breeding grounds for terrorist activity and as vulnerable targets for terrorist attacks. Computer security is the protection of computer systems against the unauthorized disclosure, manipulation, or deletion of information and against denial of service attacks. Breaches of computer security may cause harms and rights violations, including economic losses, personal injury and death, which may occur in so-called safety-critical systems, and violations of privacy and intellectual property rights. Much attention goes to the moral and social evaluation of computer crime and other forms of disruptive behavior, including hacking (non-malicious break-ins into systems and networks), cracking (malicious break-ins), cybervandalism (disrupting the operations of computer networks or corrupting data), software piracy (the

1394

Philip Brey and Johnny Hartz Søraker

illegal reproduction or dissemination of proprietary software), and computer fraud (the deception for personal gain in online business transactions by assuming a false online identity or by altering or misrepresenting data). Another recently important security-related issue is how state interests in monitoring and controlling information infrastructures to better protect against terrorist attacks should be balanced against the right to privacy and other civil rights [Nissenbaum, 2005]. Free expression and content control The Internet has become a very important medium for the expression of information and ideas. This has raised questions about whether there should be content control or censorship of Internet information, for example by governments or service providers. Censorship could thwart the right to free expression, which is held to be a basic right in many nations. Free expression includes both freedom of speech (the freedom to express oneself through publication and dissemination) and freedom of access to information. Several types of speech have been proposed as candidates for censorship. These include pornography and other obscene forms of speech, hate speech such as websites of fascist and racist organizations, speech that can cause harm or undermine the state, such as information as to how to build bombs, speech that violates privacy or confidentiality, and libelous and defamatory speech. Studies in computer ethics focus on the permissibility of these types of speech, and on the ethical aspects of different censorship methods, such as legal prohibitions and software filters (see also Section 5.5). Equity and access The information revolution has been claimed to exacerbate inequalities in society, such as racial, class and gender inequalities, and to create a new, digital divide, in which those that have the skills and opportunities to use information technology effectively reap the benefits while others are left behind. In computer ethics, it is studied how both the design of information technologies and their embedding in society could increase inequalities, and how ethical policies may be developed that result in a fairer and more just distribution of their benefits and disadvantages. This research includes ethical analyses of the accessibility of computer systems and services for various social groups, studies of social biases in software and systems design, normative studies of education in the use of computers, and ethical studies of the digital gap between industrialized and developing countries. Intellectual property Intellectual property is the name for information, ideas, works of art and other creations of the mind for which the creator has an established proprietary right of use. Intellectual property laws exist to protect creative works by ensuring that only the creators benefit from marketing them or making them available, be they

Philosophy of Computing and Information Technology

1395

individuals or corporations. Intellectual property rights for software and digital information have generated much controversy. There are those who want to ensure strict control of creators over their digital products, whereas others emphasize the importance of maintaining a strong public domain in cyberspace, and argue for unrestricted access to electronic information and for the permissibility of copying proprietary software. In computer ethics, the ethical and philosophical aspects of these disputes are analyzed, and policy proposals are made for the regulation of digital intellectual property in its different forms. Moral responsibility Society strongly relies on computers. It relies on them for correct information, for collaboration and social interaction, for aid in decision-making, and for the monitoring and execution of tasks. When computer systems malfunction or make mistakes, harm can be done, in terms of loss of time, money, property, opportunities, or even life and limb. Who is responsible for such harms? Computer professionals, end-users, employers, policy makers and others could all be held responsible for particular harms. It has even been argued that intelligent computer systems can bear moral responsibility themselves [Dodig-Crnkovic and Persson, 2008]. In computer ethics, it is studied how the moral responsibility of different actors can be defined, and what kinds of decisions should be delegated to computers to begin with. It is studied how a proper assignment of responsibility can minimize harm and allows for attributions of accountability and liability. Foundational issues in computer ethics Foundational, metaethical and methodological issues have received considerable attention in computer ethics. Many of these issues have been discussed in the context of the so-called foundationalist debate in computer ethics [Floridi and Sanders, 2002; Himma, 2007a]. This is an ongoing metatheoretical debate on the nature and justification of computer ethics and its relation to metaethical theories. Three central questions are: “Is computer ethics a legitimate field of applied ethics?”, “Does computer ethics raise any ethical issues that are new or unique?” and “Does computer ethics require substantially new ethical theories, concepts or methodologies different from those used elsewhere in applied ethics?”. The first question, whether computer ethics is a legitimate field of applied ethics, has often been discussed in the context of the other two questions, with discussants arguing that the legitimacy of computer ethics depends on the existence of unique ethical issues or questions in relation to computer technology. The debate on whether such issues exist has been called the uniqueness debate [Tavani, 2002]. Maner [1996] has argued that unique features of computer systems, like logical malleability, superhuman complexity and the ability to make exact copies, raise unique ethical issues to which no non-computer analogues exist. Others remain unconvinced that any computer ethics issue is genuinely unique. Johnson [2003] has proposed that issues in computer ethics are familiar in that they involve traditional

1396

Philip Brey and Johnny Hartz Søraker

ethical concepts and principles like privacy, responsibility, harm and ownership. The application of these concepts and principles is however not straightforward because of special properties of computer technology, which require a rethinking and retooling of ethical notions and new ways of applying them. Floridi and Sanders [2002; Floridi, 2003] do not propose the existence of unique ethical issues but rather argue for the need of new ethical theory. They argue that computer ethics needs a metaethical and macrotheoretical foundation that differs from the standard macroethical theories like utilitarianism and Kantianism. They propose a macroethical theory they call Information Ethics, which assigns intrinsic value to information. The theory covers not just digital or analogue information, but in fact analyzes all of reality as having an informational ontology, being built out of informational objects. Since informational objects are postulated to have intrinsic value, moral consideration should be given to them, including the informational objects produced by computers. In contrast to these various authors, Himma [2007a] has argued that computer technology does not need to raise new ethical issues or require new ethical theories to be a legitimate field of applied ethics. He argues that issues in computer ethics may not be unique and may be approached with traditional ethical theories. Nevertheless, computer ethics is a legitimate field because computer technology has given rise to an identifiable cluster of moral issues in much the same way like medical ethics and other fields of applied ethics. Largely separately from the foundationalist debate, several authors have discussed the issue of proper methodology in computer ethics, discussing standard methods of applied ethics and their limitations for computer ethics [Van den Hoven, 2008; Brey, 2000]. An important recent development that has methodological and perhaps also metaethical ramifications is the increased focus on cultural issues. In intercultural information ethics [Ess and Hongladarom, 2007; Brey, 2007], ethicists attempt to compare and come to grips with the vastly different moral attitudes and behaviors that exist towards information and information technology in different cultures. In line with this development, Gorniak-Kocykowska [1995] and others have argued that the global character of cyberspace requires a global ethics which transcends cultural differences in value systems. Other topics There are many other social and ethical issues that are studied in computer ethics next to these central ones. Some of these include the implications of IT for community, identity, the quality of work, and the quality of life, the relation between information technology and democracy, the ethics of Internet governance and electronic commerce, and the ethics of trust online. Recently, much attention has been devoted to ethical aspects of social networking sites like Facebook, MySpace and Youtube, to ubiquitous computing and ambient intelligence, and to robotics and artificial agents. The constant addition of new products and services in information technology and the emergence of new uses and correlated social and cultural consequences ensures that the field keeps meeting new challenges.

Philosophy of Computing and Information Technology

6.3

1397

Values and computer systems design

Although most ethical commentary in the philosophical approach is directed at the use of computers by individuals and organizations, attention has also started to be paid to systems and software themselves. It has come to be recognized that the systems themselves are not morally neutral but contain values and biases in their design. Approaches of this sort have been called values in design approaches [Nissenbaum, 1998; Flanagan, Howe and Nissenbaum, 2007]. Values in design approaches hold that computer software and systems can be morally evaluated partially or wholly independently of actual uses of them. They can be said to embody values in the sense that they have a tendency to promote or sustain particular values when used. This may sound like technological determinism, but proponents usually do not subscribe to the strong determinist thesis that embodied values necessitate certain effects in whatever way the system is used. Yet, they do hold a weak determinism according to which systems may embody values that systematically engender certain effects across a wide range of uses, at least including typical or “normal” ways of using the system. This observation has led proponents to argue that more attention should be paid to ethical aspects in the design of computer systems rather than just their use. Friedman and Nissenbaum [1996] have studied how values may enter into computer systems, with a focus on justice and bias. They argue that bias can enter into computer systems in three ways. Preexisting bias emerges from the practices and attitudes of designers and the social institutions in which they function. Technical bias arises from technical constraints. Emergent bias arises after the design of the system, when a context of use emerges that is different from the one anticipated. These three origins of bias may be generalized to apply to other values as well. Brey [2000] has proposed a particular values in design approach termed disclosive computer ethics. He claims that a significant part of the effort of computer ethics should be directed at deciphering and subsequently evaluating embedded moral values in computer software and systems. The focus should be on widely held public and moral values, such as privacy, autonomy, justice, and democracy. Research, Brey argues, should take place at three levels: the disclosure level, at which morally charged features of computer systems are detected and disclosed, the theoretical level, at which relevant moral theory is developed, and the application level, at which ethical theory is used in the evaluation of the disclosed morally charged features. He claims that such research should be interdisciplinary, involving ethicists, computer scientists and social scientists. The approach of value-sensitive design [Friedman, Kahn and Borning, 2006; Friedman and Kahn, 2003] is not so much concerned with the identification and evaluation of values in computer systems, but rather with the development of methods for incorporating values into the design process. It is an approach to software engineering and systems development that builds on values in design approaches and studies how accepted moral values can be operationalized and incorporated

1398

Philip Brey and Johnny Hartz Søraker

into software and systems. Its proposed methodology integrates conceptual investigations into values, empirical investigations into the practices, beliefs and intentions of users and designers, and technical investigations into the way in which technological properties and mechanisms support or hinder the realization of values. It also seeks procedures to incorporate and balance the values of different stakeholders in the design process. ACKNOWLEDGMENTS The authors would like to thank the following individuals for their helpful comments and recommendations: Adam Briggle, Terrell Ward Bynum, Andy Clark, Timothy Colburn, Gordana Dodig-Crnkovic, Juan Manuel Duran, Amnon H. Eden, Charles Ess, James H. Fetzer, Sven Ove Hansson, David Harel, Luciano Floridi, Patrick Grim, David Koepsell, Maurice Liebregt, Anthonie Meijers, Jim Moor, William J. Rapaport, Oron Shagrir, Herman Tavani, Raymond Turner and Alasdair Urquhart. We have also benefited from feedback by participants at the E-CAP ’08 conference. BIBLIOGRAPHY [Abrahams, 1987] P. Abrahams. What Is Computer Science? Communications of the ACM, 30(6), 472-473, 1987. [Abran et al., 2004] A. Abran, J.W. Moore, P. Bourque and R. Dupuis. SWEBOK — Guide to the Software Engineering Body of Knowledge. IEEE Computer Society Press, 2004. [Adriaans and van Benthem, 2009] P. Adriaans and J. van Benthem, eds. Handbook on the Philosophy of Information. Elsevier, 2009. [Allen et al., 2005] C. Allen, I. Smith and W. Wallach. Artificial Morality: Top-down, Bottomup, and Hybrid Approaches, Ethics and Information Technology, 7, 149-155, 2005. [Anderson, 2003] M.L. Anderson. Embodied Cognition: A field guide. Artificial Intelligence, 149(1), 91-130, 2003. [Asaro, forthcoming] P. M. Asaro. How Just Could a Robot War Be? In Current Issues in Computing and Philosophy, K. Waelbers, A. Briggle and P. Brey, eds. IOS Press, forthcoming. [Asimov, 1968] I. Asimov. I, Robot. Grafton Books, 1968. [Bar-Hillel, 1964] Y. Bar-Hillel. Language and Information: Selected Essays on Their Theory and Application. Addison-Wesley, 1964. [Barker-Plummer, 2007] D. Barker-Plummer. Turing Machines. In The Stanford Encyclopedia of Philosophy (Winter 2007 Edition), E.N. Zalta, ed., 2007. [Barney, 2004] D. Barney. The Vanishing Table, or Community in a World That is No World. In Community in the Digital Age: Philosophy and Practice, A. Feenberg and D. Barney, eds., pp. 31-52. Rowman and Littlefield, 2004. [Baudrillard, 1995] J. Baudrillard. Simulacra and Simulation. University of Michigan Press, 1995. [Bechtel and Abrahamson, 1990] W. Bechtel and A. Abrahamson. Connectionism and the Mind: An Introduction to Parallel Processing in Networks. Blackwell, 1990. [Bedau, 2003] M. A. Bedau. Artificial Life: Organization, Adaptation and Complexity from the Bottom Up, Trends in Cognitive Sciences, 7(11), 505-512, 2003. [Bedau, 2004] M. A. Bedau. Artificial Life. In The Blackwell Guide to Philosophy of Computing and Information, L. Floridi, ed., pp. 197-212, Blackwell, 2004. [Bell, 2001] D. Bell. An Introduction to Cybercultures. Routledge, 2001. [Benedikt, 1991] M. Benedikt, ed. Cyberspace: First Steps. MIT Press, 1991.

Philosophy of Computing and Information Technology

1399

[Ben-Ze’ev, 2004] A. Ben-Ze’ev. Love Online. Emotions on the Internet. Cambridge University Press, 2004. [Blackburn and Bos, 2005] P. Blackburn and J. Bos. Representation and Inference for Natural Language. Chicago University Press, 2005. [Boden, 1990] M. Boden, ed. The Philosophy of Artificial Intelligence. Oxford University Press, 1990. [Boden, 1996] M. Boden, ed. The Philosophy of Artificial Life. Oxford University Press, 1996. [Boden, 2006] M. Boden. Mind As Machine: A History of Cognitive Science. Oxford University Press, 2006. [Bohman, 2008] J. Bohman. The Transformation of the Public Sphere: Political Authority, Communicative Freedom, and Internet Publics. In Information Technology and Moral Philosophy, J. van den Hoven and J. Weckert, eds., pp. 66-92. Cambridge University Press, 2008. [Borgmann, 1999] A. Borgmann. Holding On to Reality: The Nature of Information at the Turn of the Millennium. University Of Chicago Press, 1999. [Borgmann, 2004] A. Borgmann. Is the Internet the Solution to the Problem of Community? In Community in the Digital Age: Philosophy and Practice, A. Feenberg and D. Barney, eds., pp. 53-68. Rowman and Littlefield, 2004. [Brennan and Pettit, 2008] G. Brennan and P. Pettit. Esteem, Identifiability, and the Internet. In Information Technology and Moral Philosophy, J. van den Hoven and J. Weckert, eds., pp. 175-194. Cambridge University Press, 2008. [Brey, 1998a] P. Brey. New Media and the Quality of Life. Techn´ e: Journal of the Society for Philosophy and Technology, 3(1), 1-23, 1998. [Brey, 1998b] P. Brey. Space-Shaping Technologies and the Geographical Disembedding of Place, In Philosophy & Geography vol. III: Philosophies of Place, A. Light and B. & J. B. Smith, pp. 239-263. Rowman & Littlefield, 1998. [Brey, 2000a] P. Brey. Disclosive Computer Ethics: The Exposure and Evaluation of Embedded Normativity in Computer Technology, Computers and Society, 30(4), 10-16, 2000. [Brey, 2000b] P. Brey. Technology as Extension of Human Faculties, In Metaphysics, Epistemology, and Technology. Research in Philosophy and Technology, vol 19, C. Mitcham, ed. Elsevier/JAI Press, 2000. [Brey, 2003] P. Brey. The Social Ontology of Virtual Environments. In John Searle’s Ideas About Social Reality: Extensions, Criticisms and Reconstructions, D. Koepsell and L. Moss, eds., pp. 269-282. Blackwell 2003. [Brey, 2005] P. Brey. The Epistemology and Ontology of Human-Computer Interaction. Minds and Machines, 15(3-4), 383-398, 2005. [Brey, 2007] P. Brey. Is Information Ethics Culture-Relative? International Journal of Technology and Human Interaction. Special Issue Information Ethics: Global Issues, Local Perspectives, 3(3), 12-24, 2007. [Brey, 2008] P. Brey. Virtual Reality and Computer Simulation. In The Handbook of Information and Computer Ethics, K. Himma and H. Tavani, eds. John Wiley and Sons: 2008. [Brooks, 2002] R. A. Brooks. Flesh and Machines, How Robots Will Change Us. Pantheon Books, 2002. [Brookshear, 2007] J. G. Brookshear. Computer Science: An Overview, 10th edn. AddisonWesley, 2007. [Burkholder, 1992] L. Burkholder, ed. Philosophy and the Computer. Westview Press, 1992. [Butler and Murphy, 2007] T. Butler and C. Murphy. Understanding the design of information technologies for knowledge management in organizations: a pragmatic perspective. Information Systems Journal, 17(2), 143-163, 2007. [Bynum and Moor, 1998] T. W. Bynum and J. Moor. How Computers are Changing Philosophy. In The Digital Phoenix: How Computers are Changing Philosophy. T. Bynum and J. Moor, eds., pp. 1-16. Blackwell, 1998. [Castells, 1996] M. Castells. The Rise of the Network Society. Information Age vol. 1. Blackwell, 1996. [Chadwick, 2006] A. Chadwick. Internet Politics: States, Citizens, and New Communication Technologies. Oxford University Press, 2006. [Churchland, 1992] P. M. Churchland. A Neurocomputational Perspective: The Nature of Mind and the Structure of Science. MIT Press, 1992. [Clark, 1991] A. Clark. Microcognition: Philosophy, Cognitive Science, and Parallel Distributed Processing. MIT Press, 1991.

1400

Philip Brey and Johnny Hartz Søraker

[Clark, 1993] A. Clark. Associative Engines: Connectionism, Concepts, and Representational Change. MIT Press, 1993. [Clark, 1997] A. Clark. Being There: Putting Brain, Body, and World Together Again. MIT Press, 1997. [Clark, 2003] A. Clark. Natural-born Cyborgs: Minds, Technologies, and the Future of Human Intelligence. Oxford University Press, 2003. [Clark and Chalmers, 1998] A. Clark and D. Chalmers. The Extended Mind. Analysis 58 (1): 7-19, 1998. [Clarke, 1994] R. Clarke. Asimov’s laws of robotics: Implications for information technology — Part 2. Computer, 27(1), 57-66, 1994. [Cocking, 2008] D. Cocking. Plural Selves and Relational Identity: Intimacy and Privacy Online. In Information Technology and Moral Philosophy, J. van den Hoven and J. Weckert, eds., pp. 123-141. Cambridge University Press, 2008. [Colburn, 1999] T. Colburn. Philosophy and Computer Science. Sharpe, 1999. [Colburn, 2004] T. Colburn. Methodology of Computer Science. In The Blackwell Guide to Philosophy of Computing and Information, L. Floridi, ed., pp. 318–326. Blackwell, 2004. [Colburn and Shute, 2007] T. Colburn and G. Shute. Abstraction in Computer Science. Minds and Machines, 17(2), 169-184, 2007. [Cope et al., 1997] C. Cope, P. Horan and M. Garner. Conceptions of an Information System and Their Use in Teaching about IS. Informing science 1(1), 9-22, 1997. [Copeland, 1993] B.J. Copeland. Artificial Intelligence: A Philosophical Introduction. WileyBlackwell, 1993. [Copeland, 2002a] B.J. Copeland. Hypercomputation, Minds and Machines, 12(4), 461-502, 2002. [Copeland, 2002b] B.J. Copeland. The Church–Turing Thesis. In The Stanford Encyclopedia of Philosophy (Fall 2002 Edition), E.N. Zalta, ed., 2002. [Copeland, 2004] B.J. Copeland. Computation. In The Blackwell Guide to Philosophy of Computing and Information, L. Floridi, ed., pp. 3–17. Blackwell, 2004. [Copeland and Shagrir, 2007] B.J. Copeland and O. Shagrir. Physical Computation: How General are Gandy’s Principles for Mechanisms? Minds and Machines, 17(2), 217-231, 2007. [Cordeschi, 2004] R. Cordeschi. Cybernetics. In The Blackwell Guide to Philosophy of Computing and Information, L. Floridi, ed., pp. 186-196, Blackwell, 2004. [Danielson, 1998] P. Danielson. How Computers Extend Artificial Morality. In The Digital Phoenix: How Computers are Changing Philosophy. T. Bynum and J. Moor, eds., pp. 292-307. Blackwell, 1998. [Dean, 2003] J. Dean. Why the Net is not a public sphere. Constellations, 10(1), 95-112, 2003. [DeGrazia, 2005] D. DeGrazia. Enhancement Technologies and Human Identity. Journal of Medicine and Philosophy, 30, 261-283, 2005. [Dennett, 1994] D. Dennett. Artificial Life as Philosophy. Artificial Life, 1(3), 291-292, 1994. [van Dijk, 2005] J. van Dijk. The Deepening Divide: Inequality in the Information Society. Sage, 2005. [van Dijk, 2006] J. van Dijk. The Network Society. Sage, 2006. [Dijkstra, 1968] E. Dijkstra. Go To Statement Considered Harmful. Communications of the ACM, 11(3), 147–148, 1968. [Dobson, 2002] P. J. Dobson Critical realism and information systems research: why bother with philosophy? Information Research, 7(2) , Rec. No, 124, 2002. [Dodig-Crnkovic, 2006] G. Dodig-Crnkovic. Investigations into Information Semantics and Ethics of Computing. M¨ alardalen University Press, 2006. [Dodig-Crnkovic and Persson, 2008] G. Dodig-Crnkovic and D. Persson. Sharing Moral Responsibility with Robots: A Pragmatic Approach. In Tenth Scandinavian Conference on Artificial Intelligence SCAI 2008. Volume 173 of Frontiers in Artificial Intelligence and Applications, A. Holst, P. Kreuger and P. Funk, eds., pp. 165-168. IOS Press, 2008. [Dourish, 2001] P. Dourish. Where the Action is. MIT Press, 2001. [Dretske, 1981] F. Dretske. Knowledge and the Flow of Information. MIT Press, 1981. [Dreyfus, 1972] H. L. Dreyfus. What Computers Can’t Do: A Critique of Artificial Reason. Harper and Row, 1972. [Dreyfus, 1992] H. L. Dreyfus. What Computers Still Can’t Do: A Critique of Artificial Reason. MIT Press, 1992.

Philosophy of Computing and Information Technology

1401

[Dreyfus, 1999] H. L. Dreyfus. Anonymity versus commitment: the dangers of education on the internet. Ethics and information technology, 1, 15-21, 1999. [Dreyfus, 2001] H. L. Dreyfus. On the Internet. Routledge, 2001. [Dreyfus, 2004] H. L. Dreyfus. Nihilism on the Information Highway: Anonymity versus Commitment in the Present Age. In Community in the Digital Age: Philosophy and Practice, A. Feenberg and D. Barney, eds., pp. 69-82. Rowman and Littlefield, 2004. [Dreyfus and Dreyfus, 1984] H. L. Dreyfus and S. E. Dreyfus. From Socrates to Expert Systems: The Limits and Dangers of Calculative Rationality. Technology in Society, 6(3), 217-233, 1984. [Dreyfus and Dreyfus, 1986] H. L. Dreyfus and S. E. Dreyfus. Mind over Machine: The Power of Human Intuition and Expertise in the Era of the Computer. Free Press, 1986. [Dreyfus and Dreyfus, 1988] H. L. Dreyfus and S. E. Dreyfus. Making a Mind Versus Modeling the Brain: Artificial Intelligence Back at a Branchpoint. Daedalus, 117, 15-43, 1988. [Eden, 2007] A. Eden. Three Paradigms of Computer Science. Minds & Machines, 17,135–167, 2007. [Eiben and Smith, 2003] A. E. Eiben and J. E. Smith. Introduction to Evolutionary Computing. Springer, 2003. [Elman, 1998] J. L. Elman. Connectionism, artificial life, and dynamical systems: New approaches to old questions. In A Companion to Cognitive Science, W. Bechtel and G. Graham, eds., pp. 488-505. Basil Blackwood, 1998. [Ess, 1996] C. Ess, ed. Philosophical Perspectives on Computer-Mediated Communication. SUNY Press, 1996. [Ess, 2002] C. Ess. Cultures in collision: Philosophical lessons from computer-mediated communication. Metaphilosophy, 33 (1/2), 229-253, 2002. [Ess, 2004] C. Ess. Computer-Mediated Communication and Human-Computer Interaction. In The Blackwell Guide to Philosophy of Computing and Information, L. Floridi, ed., pp. 76-91. Blackwell, 2004. [Ess and Hongladarom, 2007] C. Ess and S. Hongladarom, Information Technology Ethics: Cultural Perspectives. IGI Global, 2007. [Ess and Sudweeks, 2005] C. Ess and F. Sudweeks, eds. Special Theme: Culture and ComputerMediated Communication. Journal of Computer-Mediated Communication, 11(1), 179-394, 2005. [Fallis, 2004] D. Fallis. On Verifying the Accuracy of Information: Philosophical Perspectives. Library Trends, 52(3), 463-487, 2004. [Feenberg and Barney, 2004] A. Feenberg and D. Barney. Community in the Digital Age. Philosophy and Practice. Rowman & Littlefield, 2004. [Fetzer, 1988] J.H. Fetzer. Program Verification: The Very Idea. Communications of the ACM, 31(9), 1048-1063, 1988. [Fetzer, 1991] J.H. Fetzer. Philosophical aspects of program verification. Minds and Machines, 1(2), 197-216, 1991. [Fetzer, 1998] J.H. Fetzer. Philosophy and Computer Science: Reflections on the Program Verification Debate. In The Digital Phoenix: How Computers are Changing Philosophy, T. W. Bynum and J. H. Moor, eds., pp. 253-273. Blackwell, 1998. [Fetzer, 2004] J.H. Fetzer. The Philosophy of AI and Its Critique. In The Blackwell Guide to Philosophy of Computing and Information, L. Floridi, ed., pp. 119-134. Blackwell, 2004. [Fielding et al., 2004] J. Fielding, J. Simon, W. Ceusters and B. Smith. Ontological Theory for Ontological Engineering: Biomedical Systems Information Integration. Proceedings of the Ninth International Conference on the Principles of Knowledge Representation and Reasoning, 2004. [Flanagan et al., 2008] M. Flanagan, D. Howe and H. Nissenbaum. Embodying Values in Technology: Theory and Practice. In Information Technology and Moral Philosophy, J. Van den Hoven and J. Weckert, eds., pp. 322-353. Cambridge University Press, 2008. [Flew, 2002] T. Flew. New Media: An Introduction, Oxford University Press, 2002. [Floridi, 1999a] L. Floridi. Philosophy and Computing. Routledge 1999. [Floridi, 1999b] L. Floridi. Information Ethics: On the Theoretical Foundations of Computer Ethics. Ethics and Information Technology, 1(1), 37-56, 1999. [Floridi, 2002] L. Floridi. What is the Philosophy of Information? Metaphilosophy, 33(1/2), 123-145, 2002. [Floridi, 2003] L. Floridi. On the Intrinsic Value of Information Objects and the Infosphere. Ethics and Information Technology, 4(4), 287-304, 2003.

1402

Philip Brey and Johnny Hartz Søraker

[Floridi, 2004a] L. Floridi. Open Problems in the Philosophy of Information. Metaphilosophy, 35(4), 554-582, 2004. [Floridi, 2004b] L. Floridi. Information. In The Blackwell Guide to Philosophy of Computing and Information, L. Floridi, ed., pp. 40-62. Blackwell, 2004. [Floridi, 2004c] L. Floridi, ed. The Blackwell Guide to Philosophy of Computing and Information. Blackwell, 2004. [Flroidi, 2007] L. Floridi. Semantic Conceptions of Information. In The Stanford Encyclopedia of Philosophy (Spring 2007 Edition), E.N. Zalta, ed., 2007. [Floridi, 2006] L. Floridi. Peering into the Future of the Infosphere. TidBITS, Sep 25, 2006 [Retrieved on May 13, 2008, from http://db.tidbits.com/article/8686], 2006. [Floridi and Sanders, 2002] L. Floridi and J. W. Sanders. Mapping the Foundationalist Debate in Computer Ethics. Ethics and Information Technology, 4(1), 1-9, 2002. [Floridi and Sanders, 2004] L. Floridi and J. W. Sanders. On the Morality of Artificial Agents. Minds and Machines, 14(3), 349-379, 2004. [Floyd, 1979] R. Floyd. The paradigms of programming. Communications of the ACM, 22(8), 455-160, 1979. [Fodor, 1975] J. Fodor. The Language of Thought. Harvard University Press, 1975. [Fodor and Pylyshyn, 1988] J. Fodor and Z. Pylyshyn. Connectionism and Cognitive Architecture: a Critical Analysis, Cognition, 28, 3-71, 1988. [Fredkin, 2003] E. Fredkin. An Introduction to Digital Philosophy. International Journal of Theoretical Physics, 42, 189-247, 2003. [Friedman and Kahn, 2003] B. Friedman and P. H. Kahn. Human values, ethics, and design. In The Human-Computer Interaction Handbook, J. A. Jacko and A. Sears, eds., pp. 1177–1201. Lawrence Erlbaum Associates, 2003. [Friedman and Nissenbaum, 1996] B. Friedman and H. Nissenbaum. Bias in Computer Systems, ACM Transactions on Information Systems, 14 (3), 330–347, 1996. [Friedman et al., 2006] B. Friedman, P. H. Kahn and A. Borning. Value Sensitive Design and Information Systems. In Human-Computer Interaction in Management Information Systems: Foundations, P. Zhang and D. Galletta, eds., pp. 348-372. M.E. Sharpe, 2006. [Frigg and Reiss, forthcoming] R. Frigg and J. Reiss: The Philosophy of Simulation: Hot New Issues or Same Old Stew? Synthese, forthcoming. [Galton, 2006] A. Galton. The Church–Turing thesis: Still valid after all these years? Applied Mathematics and Computation, 178, 93–102, 2006. [van Gelder, 2000] T. van Gelder. The dynamical hypothesis in cognitive science. Behavioral and Brain Sciences, 21, 615-628, 2000. [Gillett, 2006] G. Gillett. Cyborgs and moral identity. Journal of Medical Ethics, 32, 79-83, 2006. [Gillies, 2004] D. Gillies. Probability in Artificial Intelligence. In The Blackwell Guide to Philosophy of Computing and Information, L. Floridi, ed., pp. 276-288. Blackwell, 2004. [Gimmler, 2001] A. Gimmler. Deliberative democracy, the public sphere and the internet. Philosophy and Social Criticism, 27(4), 21-39, 2001. [Goldman, 1999] A. Goldman. Knowledge in a Social World. Oxford University Press, 1999. [Goldman, 2008] A. Goldman. The Social Epistemology of Blogging. In Information Technology and Moral Philosophy, J. van den Hoven and J. Weckert, eds., pp. 111-122. Cambridge University Press, 2008. [Gornisk-Kocikowska, 1995] K. Gorniak-Kocikowska. The Computer Revolution and the Problem of Global Ethics. Science and Engineering Ethics, 2(2), 177-190, 1995. [Graham, 2004] P. Graham. Hackers and Painters: Big Ideas from the Computer Age. O’Reilly and Associates, 2004. [Gray, 1996] C. Gray, ed. The Cyborg Handbook. Routledge, 1996. [Grim, 2004] P. Grim. Computational Modeling as a Philosophical Methodology. In The Blackwell Guide to Philosophy of Computing and Information, L. Floridi, ed., pp. 337–347. Blackwell, 2004. [Grim et al., 2005] P. Grim, E. Selinger, W. Braynen, R. Rosenberger, R. Au, N. Louie, and J. Connolly. Modeling Prejudice Reduction. Public Affairs Quarterly, 19(2), 95-125, 2005. [Gruber, 1995] T.R. Gruber. Toward Principles for the Design of Ontologies Used for Knowledge Sharing. International Journal of Human-Computer Studies, 43(5-6), pp. 907-928, 1995. [Guarino, 1998] N. Guarino, ed. Formal Ontology in Information Systems. IOS Press, 1998.

Philosophy of Computing and Information Technology

1403

[Hagar, 2007] A. Hagar. Quantum Algorithms: Philosophical Lessons. Minds and Machines, 17(2), 233–247, 2007. [Haraway, 1985] D. Haraway. A Manifesto for Cyborgs: Science, Technology, and Socialist Feminism in the 1980’s, Socialist Review, 80, 56-107, 1985. [Harel, 2000] D. Harel. Computers Ltd. What They Rreally Can’t Do. Oxford University Press, 2000. [Harel and Feldman, 2004] D. Harel and Y. A. Feldman. Algorithmics: The Spirit of Computing, 3rd edn. Pearson, 2004. [Harnad, 1990] S. Harnad. The Symbol Grounding Problem. Physica D, 42, 335-346, 1990. [Harnad, 1994] S. Harnad. Artificial life: Synthetic vs. virtual. In Artificial Life III, C. Langton, ed., pp. 539-552. Addison-Wesley, 1994. [Harvey, 1989] D. Harvey. The Condition of Postmodernity: An Enquiry into the Origins of Cultural Change. Blackwell, 1989. [Haselager, 2007] W. F. G. Haselager. Robotics, philosophy and the problems of autonomy. In Cognitive Technologies and the Pragmatics of Cognition, I. Dror, ed, pp. 61-77. John Benjamins, 2007. [Haugeland, 1978] J. Haugeland. The Nature and Plausibility of Cognitivism. Behavioral and Brain Sciences, 2, 215-260, 1978. [Haugeland, 1981] J. Haugeland, ed. Mind Design: Philosophy, Psychology, and Artificial Intelligence. MIT Press, 1981. [Hayes, 1990] P. Hayes. The Na¨ıve Physics Manifesto. In The Philosophy of Artificial Intelligence. M. Boden, ed., pp. 171-205. Oxford University Press, 1990. [Hayles, 1999] K. Hayles How We Became Posthuman: Virtual Bodies in Cybernetics, Literature and Informatics, University of Chicago Press, 1999. [Heim, 1993] M. Heim. The Metaphysics of Virtual Reality. Oxford University Press, 1993. [Hilbert and Ackermann, 1928] D. Hilbert and W. Ackermann. Grundz¨ uge der theoretischen Logik. Springer, 1928. [Himma, 2007a] K. E. Himma. Foundational Issues in Information Ethics. Library Hi Tech, 25(1), 79-94, 2007. [Himma, 2007b] K. E. Himma. The concept of information overload: A preliminary step in understanding the nature of a harmful information-related condition. Ethics and Information Technology, 9(4), 259-272, 2007. [Himma and Tavani, 2008] K. Himma and H. Tavani, eds. Information and Computer Ethics. John Wiley and Sons, 2008. [Hoare, 1986] C. Hoare. Mathematics of programming, Byte, August 1986, 115-49, 1986. [Hollan et al., 2000] J. Hollan, E. Hutchins and D. Kirsh. Distributed Cognition: Toward a New Foundation for Human-Computer Interaction Research’, ACM Transactions on ComputerHuman Interaction, 7(2), 174–196, 2000. [van den Hoven, 2008] J. van den Hoven. Moral Methodology and Information Technology. In The Handbook of Information and Computer Ethics. K. E. Himma and H. T. Tavani, eds., pp. 49-68. John Wiley and Sons, 2008. [van den Hoven and Rooksby, 2008] J. van den Hoven and E. Rooksby. Distributive Justice and the Value of Information: A (Broadly) Rawlsian Approach. In Information Technology and Moral Philosophy, J. van den Hoven and J. Weckert, eds., pp. 376-398. Cambridge University Press, 2008. [Humphreys, forthcoming] P. Humphreys: The Philosophical Novelty of Computer Simulation Methods. Synthese, forthcoming. [Hutchins, 1995] E. Hutchins. Cognition in the Wild, MIT Press, 1995. [Immerman, 2006] N. Immerman. Computability and Complexity. In The Stanford Encyclopedia of Philosophy (Fall 2006 Edition), E.N. Zalta, ed., 2006. [Introns, 1997] L. Introna. Management, Information and Power: A Narrative of the Involved Manager. Macmillan, 1997. [Johnson, 1985] D. Johnson. Computer Ethics [1st ed]. Prentice Hall, 1985. [Johnson, 1997] D. Johnson. Is the global information infrastructure a democratic technology? Computers and Society, 27(3), 20–26, 1997. [Johnson, 2003] D. Johnson. Computer Ethics. In Blackwell Companion to Applied Ethics, R.G. Frey and C.H. Wellman, eds., pp. 608-619. Blackwell, 2003. [Johnson, 2006] D. Johnson. Computer systems: Moral entities but not moral agents. Ethics and Information Technology, 8, 195-204, 2006.

1404

Philip Brey and Johnny Hartz Søraker

[Keeley, 1998] B. L. Keeley. Artificial life for philosophers. Philosophical Psychology, 11(2), 251– 260, 1998. [Kendal and Creen, 2006] S. Kendal and M. Creen. An Introduction to Knowledge Engineering. Springer, 2006. [Knuth, 1974a] D. Knuth, Computer Science and Its Relation to Mathematics. American Mathematical Monthly, 81(4), 323-343, 1974. [Knuth, 1974b] D. Knuth. Computer Programming as an Art. Communications of the ACM, 17(12), 667–673, 1974. [Koepsell, 2000] D. R. Koepsell. The Ontology of Cyberspace: Law, Philosophy, and the Future of Intellectual Property. Open Court, 2000. [Kuhn, 1970] T. S. Kuhn. The Structure of Scientific Revolutions, 2nd edn.. University of Chicago Press, 1970. [de Laat, 2005] P. de Laat. Trusting Virtual Trust. Ethics and Information Technology, 7(3), 167–180, 2005. [Lacey and Lee, 2003] N. J. Lacey and M. H. Lee. The Epistemological Foundations of Artificial Agents. Minds and Machines, 13, 339–365, 2003. [Langton, 1995] C. Langton. Artificial Life: An Overview. MIT Press, 1995. [Lenat and Guha, 1990] D. Lenat and R. V. Guha. Building Large Knowledge-Based Systems: Representation and Inference in the Cyc Project. Addison-Wesley, 1990. [Levy, 2008] D. Levy. Information Overload. In The Handbook of Information and Computer Ethics, K. Himma and H. Tavani, eds. John Wiley and Sons: 2008. [Lloyd, 2006] S. Lloyd. Programming the Universe: A Quantum Computer Scientist Takes on the Cosmos. Alfred A. Knopf, 2006. [Longe, 1999] G. Longo, ed. Philosophy of Computer Science, special issue of The Monist, 82(1), 1999. [Loui, 1987] M.C. Loui. Computer Science Is an Engineering Discipline. Engineering Education, 78(3), 175-178, 1987. [Lucivero and Tamburrini, 2008] F. Lucivero and G. Tamburrini. Ethical monitoring of brainmachine interfaces. AI & Society, 22(3), 449–460, 2008. [Lyotard, 1984] J. Lyotard. The Postmodern Condition. Manchester University Press, 1984. [MacLennan, 2003] B.J. MacLennan. Transcending Turing Computability. Minds and Machines, 13(1), 3-22, 2003 [Mahoney, 2002] M.S. Mahoney. Software as Science—Science as Software. In Proceedings of the international Conference on History of Computing: Software Issues, U. Hashagen, R. Keil-Slawik, and A. Norberg, eds., pp. 25-48. Springer, 2002. [Mainon and Goodrum, 2000] D. Mainon and A. Goodrum. Terrorism or Civil Disobedience: Toward a Hacktivist Ethic. Computers and Society, 30(2), 14-19, 2000. [Maner, 1980] W. Maner. Starter Kit in Computer Ethics. Helvetia Press, 1980. [Maner, 1996] W. Maner. Unique Ethical Problems in Information Technology. Science and Engineering Ethics, 2(2), 137-154, 1996. [Manning, 1987] R. Manning. Why Sherlock Holmes can’t be replaced by an expert system. Philosophical Studies, (511), 19–28, 1987. [Marcos and Marcos, 2001] E. Marcos and A. Marcos. A Philosophical Approach to the Concept of Data Model: Is a Data Model, in Fact, a Model? Information Systems Frontiers, 3(2), 267–274, 2001. [Matthews, 2008] S. Matthews. Identity and Information Technology. In Information Technology and Moral Philosophy, J. van den Hoven and J. Weckert, eds., pp. 142-160. Cambridge University Press, 2008. [Maun and Corrunker, 2008] C. Maun and L. Corrunker, eds. Virtual Identities: The Construction of Selves in Cyberspace. Eastern Washington University Press, 2008. [Mazlish, 1993] B. Mazlish. The Fourth Discontinuity: The Co-evolution of Humans and Machines. Yale University Press, 1993. [McDermott, 1990] D. McDermott. A Critique of Pure Reason. In The Philosophy of Artificial Intelligence. M. Boden, ed., pp. 206-230. Oxford University Press, 1990. [McLaughlin, 2004] B. P. McLaughlin. Computationalism, Connectionism, and the Philosophy of Mind. In The Blackwell Guide to Philosophy of Computing and Information, L. Floridi, ed., pp. 135-152. Blackwell, 2004. [McLuhan, 1962] M. McLuhan. The Gutenberg Galaxy: The Making of Typographic Man. University of Toronto Press, 1962.

Philosophy of Computing and Information Technology

1405

[Minsky, 1967] M. Minsky. Computation: Finite and Infinite Machines. Prentice-Hall, 1967. [Moor, 1979] J. H. Moor. Are There Decisions Computers Should Never Make? Nature and System, 1, 217-229, 1979. [Moor, 1985] J. H. Moor. What is Computer Ethics? Metaphilosophy, 16, 266-275, 1985. [Moor, 2003] J. H. Moor, ed. The Turing Test: The Elusive Standard of Artificial Intelligence. Kluwer, 2003. [Moor, 2005] J. H. Moor. Should we let computers get under our skin?” In The Impact of the Internet on Our Moral Lives, R. Cavalier, ed., pp. 121-137. SUNY press, 2005. [Moor and Bynum, 2002] J. H. Moor and T.W. Bynum, eds. Cyberphilosophy. Blackwell, 2002. [Mooradian, 2006] N. Mooradian. Virtual Reality, Ontology, and Value. Metaphilosophy, 37(5), 673–690, 2006. [Morgan and Morrison, 1999] M. Morgan and M. Morrison. Models as Mediators: Perspectives on Natural and Social Science. Cambridge University Press, 1999. [Mullins, 1996] P. Mullins. Sacred Text in the Sea of Texts: The Bible in North American Electronic Culture. In Philosophical Perspectives on Computer-Mediated Communication, C. Ess (ed.). SUNY Press, 1996. [Newell and Simon, 1976] A. Newell and, H.A. Simon. Computer Science as Empirical Inquiry: Symbols and Search. Communications of the ACM, 19(3), 113–126, 1976. [Nissenbaum, 1998] H. Nissenbaum. Values in the Design of Computer Systems. Computers and Society, March 1998, 38-39, 1998. [Nissenbaum, 2001] H. Nissenbaum. Securing Trust Online: Wisdom or Oxymoron. Boston University Law Review, 81(3), 635-664, 2001. [Nissenbaum, 2005] H. Nissenbaum. Where Computer Security Meets National Security. Ethics and Information Technology, 7, 61-73, 2005. [Nissenbaum and Walker, 1998] H. Nissenbaum and D. Walker. Will computers dehumanize education: A grounded approach to values at risk. Technology in Society, 20, 237-273, 1998. [Norman, 1993] D.A. Norman. Things That Make Us Smart: Defending Human Attributes in the Age of the Machine. Addison-Wesley, 1993. [Norris, 2001] P. Norris. Digital Divide: Civic Engagement, Information Poverty, and the Internet Worldwide. Cambridge University Press, 2001. [Parker, forthcoming] W. Parker: Does Matter Really Matter? Computer Simulations, Experiments, and Materiality. Synthese, forthcoming. [Parnas, 1985] D. L. Parnas. Software Aspects of Strategic Defense Systems. Communications of the ACM, 28(12), 1326-1335, 1985. [Parnas, 1999] D. L. Parnas. Software Engineering Programs Are Not Computer Science Programs. IEEE Software 16(6), 19-30, 1999. [Peschl and Stary, 1998] M. Peschl and C. Stary. The Role of Cognitive Modeling for User Interface Design Representations: An Epistemological Analysis of Knowledge Engineering in the Context of Human-Computer Interaction. Minds and Machines, 8(2), 203–236, 1998. [Pettit, 2004] P. Pettit. Trust, Reliance and the Internet, Analyse & Kritik, 26, 108-121, 2004. [Pieters, 2006] W. Pieters. Internet Voting: A Conceptual Challenge to Democracy. In IFIP International Federation for Information Processing, 208, 89-103, 2006. [du Plooy, 2003] N. F. du Plooy. Information systems as social systems. In Critical Reflections on information Systems: A Systemic Approach, J. J. Cano, ed., pp. 105-121. IGI Publishing, 2003. [Port and van Gelder, 1995] R. F. Port and T. van Gelder. Mind as Motion: Explorations in the Dynamics of Cognition. MIT Press, 1995. [Poster, 1990] M. Poster. The Mode of Information: Poststructuralism and Social Context. Polity, 1990. [Preston and Bishop, 2002] J. Preston and M. Bishop, eds. Views into the Chinese Room: New Essays on Searle and Artificial Intelligence. Oxford University Press, 2002. [Prosser and Ward, 2000] B. T. Prosser and A. Ward. Kierkegaard and the Internet: Existential reflections on education and community. Ethics and information technology, 2, 167-180, 2000. [Putnam, 2001] R. Putnam. Bowling Alone: The Collapse and Revival of American Community. Simon & Schuster, 2001. [Pylyshyn, 1984] Z. W. Pylyshyn. Computation and Cognition. MIT Press, 1984. [Pylyshyn, 1987] Z. Pylyshyn, ed. The Robot’s Dilemma: The Frame Problem in Artificial Intelligence. Ablex, 1987. [Quine, 1961] W. V.O. Quine. From a Logical Point of View. Harvard University Press, 1961.

1406

Philip Brey and Johnny Hartz Søraker

[Rapaport, 2005] W. Rapaport. Philosophy of Computer Science: An Introductory Course. Teaching Philosophy, 28(4), 319-41, 2005. [Raymond, 2000] E. Raymond. The Cathedral and the Bazaar: Musings on Linus and Open Source by an Accidenal Revolutionary, O’Reilly Media, Inc, 2000. http://www.catb.org/ esr/writings/cathedral-bazaar/ [Rheingold, 1993] H. Rheingold. The Virtual Community: Homesteading on the Electronic Frontier. Perseus, 1993. [Rice, 1953] H. G. Rice. Classes of Recursively Enumerable Sets and Their Decision Problems. Transactions of the American Mathematical Society, 74(2), 358-366, 1953. [Rowe, 2000] D. Rowe. The Hilbert Challenge. Oxford University Press, 2000. [Scheutz, 2002] M. Scheutz (ed.). Computationalism: New Directions. MIT Press, 2002.. [Searle, 1980] J. Searle. Minds, Brains and Programs. Behavioural and Brain Sciences, 3(3), 417-424, 1980. [Searle, 1995] J. Searle. The Construction of Social Reality. MIT Press,1995. [Shagrir, 2002] O. Shagrir. Effective Computation by Humans and Machines. Minds and Machines, 12(2), 221-240, 2002. [Shagrir and Pitowski, 2003] O. Shagrir and I. Pitowski. Physical hypercomputation and the Church Turing thesis. Minds and Machines, 13(1), 87-101, 2003. [Shane, 2004] P. M. Shane. Democracy Online: The Prospects for Political Renewal Through the Internet. Routledge, 2004. [Shannon, 1993] C. Shannon. Collected Papers. IEEE Press, 1993. [Shapiro, 1995] S. C. Shapiro. Computationalism. Minds and Machines, 5 (4), 467-87, 1995. [Shenk, 1998] D. Shenk. Data Smog: Surviving the Information Glut [revised ed.]. HarperOne, 1998 . [Simon, 1965] H. A. Simon. The Shape of Automation for Men and Management. Harper & Row, 1965. [Simon, 1996] H. A. Simon. The Sciences of the Artificial, 3rd edn. MIT Press, 1996. [Sloman, 2002] A. Sloman. The irrelevance of Turing machines to Artificial Intelligence. In Computationalism: New directions, M. Scheutz, ed., pp. 87-128. MIT Press, 2002. [Smith, 1996] B. C. Smith. Limits of Correctness in Computers. In Computerization and Controversy, R. Kling, ed., pp. 810-825. Academic Press, 1996. [Smith, 2004] B. Smith. Ontology. In The Blackwell Guide to Philosophy of Computing and Information, L. Floridi, ed., pp. 155-166. Blackwell, 2004. [Smith et al., 2007] B. Smith, M. Ashburner, C. Rosse, J. Bard, W. Bug, W. Ceusters, L. J. Goldberg, K. Eilbeck, A. Ireland, C. J. Mungall, The OBI Consortium, N. Leontis, P. RoccaSerra, A. Ruttenberg, S.Sansone, R. H. Scheuermann, N. Shah, P. L. Whetzel and S. Lewis. The OBO Foundry: Coordinated Evolution of Ontologies to Support Biomedical Data Integration. Nature Biotechnology, 25(11), 1251-1255, 2007. [Smolensky, 1988] P. Smolensky. On the Proper Treatment of Connectionism. Behavioral and Brain Sciences, 11, 1-74, 1988 [Spinello and Tavani, 2004] R. Spinello and H. Tavani. Readings in Cyberethics, 2nd ed. Tandem, 2004. [Suber, 1988] P. Suber. What is Software? Journal of Speculative Philosophy, 2(2), 89-119, 1988. [Suchman, 1987] L. A. Suchman. Plans and Situated Actions: The Problem of Human-machine Communication. Cambridge University Press, 1987. [Sullins, 2006] J. P. Sullins. When is a Robot a Moral Agent? International Review of Information Ethics, 6, 24-30, 2006. [Sunstein, 2001] C. Sunstein. Republic.com. Princeton University Press, 2001. [Sunstein, 2006] C. Sunstein. Infotopia: How Many Minds Produce Knowledge. Oxford University Press, 2006. [Sunstein, 2008] C. Sunstein. Democracy and the Internet. In Information Technology and Moral Philosophy, J. van den Hoven and J. Weckert, eds., pp. 93-102. Cambridge University Press, 2008. [Sycara, 1998] K. P. Sycara. The Many Faces of Agents. AI Magazine, Summer 1998, pp. 11-13, 1998. [Søraker, 2008] J. H. Søraker. Global Freedom of Expression within Non-Textual Frameworks. The Information Society, 24(1), 40-46, 2008.

Philosophy of Computing and Information Technology

1407

[Tapscott and Williams, 2008] D. Tapscott and A. D. Williams. Wikinomics: How Mass Collaboration Changes Everything. Portfolio, 2008. [Tavani, 2002] H. T. Tavani. The Uniqueness Debate in Computer Ethics: What Exactly Is at Issue, and Why Does it Matter? Ethics and Information Technology, 4(1), 37-54, 2002. [Tavani, 2007] H. T. Tavani. Ethics and Technology: Ethical Issues in an Age of Information and Communication Technology, 2nd edn. John Wiley and Sons, 2007. [Thagard, 1988] P. Thagard. Computational Philosophy of Science. MIT Press, 1988. [Thagard, 2001] P. Thagard. Internet epistemology: Contributions of new information technologies to scientific research. In Designing for Science: Implications from Everyday, Classroom, and Professional Settings, K. Crowley, C. Schunn and T. Okada, eds., pp. 465–485. Lawrence Erlbaum Associates, 2001. [Thomason, 2003] R. Thomason. Logic and Artificial Intelligence. In The Stanford Encyclopedia of Philosophy (Fall 2003 Edition), E.N. Zalta, ed., 2003. [Turing, 1937] A. Turing. On computable numbers, with an application to the Entscheidungsproblem. Proceedings of the London Mathematical Society, Series 2, 42, 230-265, 1937. [Turing, 1950] A. Turing. Computing Machinery and Intelligence. Mind, LIX, 433-460, 1950. [Turkle, 1995] S. Turkle. Life on the Screen. Identity in the Age of the Internet. Simon & Schuster, 1995. [Turner and Eden, 2007a] R. Turner and A. Eden. eds. The Philosophy of Computer Science, special issue of Minds & Machines, 17(2), 2007. [Turner and Eden, 2007b] R. Turner and A. Eden. The Philosophy of Computer Science. Minds & Machines, 17(2), 129-133, 2007. [Turner and Eden, forthcoming a] R. Turner and A. Eden. Towards a programming language ontology. In Computing, Philosophy, and Cognitive Science, G. Dodig-Crnkovic, ed. Cambridge Scholars Press, forthcoming. [Turner and Eden, forthcoming b] R. Turner and A. Eden. Philosophy of Computer Science. In The Stanford Encyclopedia of Philosophy, E.N. Zalta, ed., forthcoming. [Urquhart, 2004] A. Urquhart. Complexity. In The Blackwell Guide to Philosophy of Computing and Information, L. Floridi, ed., pp. 18-27. Blackwell, 2004. [Vedder and Wachbroit, 2003] A. Vedder and R. Wachbroit. Reliability of Information on the Internet: Some Distinctions. Ethics and Information Technology, 5 (4), 211-215, 2003. [Vedder, 2008] A. Vedder. Responsibilities for Information on the Internet. In The Handbook of Information and Computer Ethics, K. Himma and H. Tavani, eds., pp. 339-359. John Wiley and Sons: 2008. [Virilio, 1994] P. Virilio. The Vision Machine. Indiana University Press, 1994. [Webster, 1995] F. Webster. Theories of the Information Society. Routledge, 1995. [Weckert, 2005] J. Weckert. Trust in Cyberspace. In The Impact of the Internet on Our Moral Lives, R. Cavalier, ed., pp. 95-117. SUNY press, 2005. [Weckert, 2007] J. Weckert. ed. Computer Ethics. International Library of Essays in Public and Professional Ethics Series. Ashgate, 2007. [Wegner, 1976] P. Wegner. Research paradigms in computer science. In Proceedings of the 2nd international Conference on Software Engineering, pp. 322-330. IEEE Computer Society Press, 1976. [Weizenbaum, 1976] J. Weizenbaum. Computer Power and Human Reason: From Judgment to Calculation. Freeman, 1976. [Whitten et al., 2003] J. Whitten, L. Bentley and K. Dittman. Systems Analysis and Design Methods [6th ed.]. McGraw-Hill, 2003. [Wiener, 1950] N. Wiener. The Human Use of Human Beings: Cybernetics and Society. Houghton Mifflin, 1950. [Winner, 1997] L. Winner. Cyberlibertarian myths and the prospects for community. SIGCAS Computers and Society, 27(3), 14-19, 1997. [Winograd and Flores, 1987] T. Winograd and C.F. Flores. Understanding Computers and Cognition: a New Foundation for Design. Addison-Wesley, 1987. [Winsberg, forthcoming] E. Winsberg: A Tale of Two Methods. Synthese, forthcoming. [Young, 2005] S. Young. Designer Evolution: A Transhumanist Manifesto. Prometheus Books, 2005. [Zhai, 1998] P. Zhai. Get Real. A Philosophical Adventure in Virtual Reality. Rowman and Littlefield, 1998.

This page intentionally left blank

INDEX Compiled by Marcel Scheele and Andreas Spahn

Aalto, A., 1202, 1204, 1213, 1233, 1235, 1247-1251 Abrahamsen, S., 239 abstract, 121, 158-159, 164, 168, 198, 255, 279-280, 289-290, 353, 425, 428, 460, 465, 543, 545547, 548-555, 561-562, 655, 677, 701, 731, 745, 751-752, 1061, 1351, 1356-1357 abstract entity, 280, 724 abstract machine, 1357 accuracy, 339, 681-682, 858, 1073, 1358, 1379 Achinstein, P, 650-651, 659, 866 Ackrill, J. L., 383 action, 15, 45, 52, 58, 70, 73, 79, 85, 158-160, 166, 219, 226, 228, 339, 384, 387, 391, 393, 398, 426, 434-435, 446-447, 527, 530-531, 570, 574-577, 579, 582, 642, 913, 915, 1021, 1072, 1075, 1092, 1157-1158 actualism, 1091 adaptive agents, 1373 adaptive systems, 538, 613-614 Addison, J. W., 657 Adorno, T., 1024-1025, 1157 aesthetic computing, 1062 aesthetic judgement/judgment, 1043, 1056, 1230-1231, 1236 aesthetic quality, 1035, 1043-44, 1063 aesthetic standards, 1057, 1231 aesthetic value, 7, 12, 14, 884, 94648, 984, 1004, 1031-1067, 1229,

1383 aesthetic virtue, 1049-1050 aesthetics, 1031-1065, 1229-1238 conception of, 1230-1232 affordance, 544, 701, 751 agent/agency, adaptive, 1373 artificial, 1197, 1364, 1372-1375 causal, 1014 moral, 957, 1374 agriculture, 1019, 1197, 1257-1271, 1301, 1303, 1305, 1330, 1335-1336 agricultural chemistry, 74-75, 77 Akkerman, A., 1223 Akrich, M., 107, 887, 909-912 Alberti, L. B., 123, 639-640, 643, 653, 1200, 1206-1207, 1213, 1231, 1235, 1239 Alexanders dictum, 192 Alexander, C., 440-446, 449, 1040, 1203, 1249 algorithm, 146, 425, 472, 591, 604, 742, 745, 1135, 1140, 1184, 1343, 1347-1349, 1354-1355, 1362-1363, 1369, 1375 Allen, C., 1375 Allen, R., 1266 Amdahl, E., 99, 100 analogy, 646, 648-650, 877 analytic and synthetic methods, 405407 analytic ontology, 1 Anders, G., 1105, 1160 Anderson, J., 133-134, 140, 1024-1025

1410

Marcel Scheele and Andreas Spahn

Annas, G., 1272, 1389 Anscombe, G. E. M., 380, 1163 anthropocentric, 1166, 1203 application conditions, 160, 197-198, 208 applied ethics, 1103, 1127, 1154, 1163, 1169, 1197, 1391, 1395-1396 applied mechanics, 1016 applied ontology, 7-8, 11, 274, 276 applied science, 1, 29, 40, 48, 54, 66, 69-74, 77, 127, 150, 162, 188, 309-312, 340, 344-345, 347348, 413, 418, 516, 518-519, 802, 863, 1082, 1344, 1366 applied science debate/thesis, 197, 309311, 340, 344-345, 347-348 appropriation, 93, 96, 101-112, 1266, 1324 Archer, B., 430, 432-435, 437, 439, 442, 445-448, 450 Archimedes, 119-121, 124-125, 179, 1013 architecture, 411, 638-639, 643, 985, 1003, 1033-1044, 1057, 10641065, 1199-1257 Baroque, 1208, 1218, 1231, 1237 Bauhaus, 416-417, 426, 1224, 12351236, 1240 classicism, 1038, 1208-1209, 1242 culture, 1204, 1209-1210, 12131214, 1216, 1230-1231, 1238 ethics, 1201, 1204-1205, 1218-1229 function/functionalism, 1212, 1214, 1216, 1220-1221, 1224, 1230, 1235-1237, 1240 Gothic, 1040, 1200, 1202, 12081209, 1215, 1218, 1237 international style, 1036, 1038, 1040 postmodernism, 1204, 1216 Renaissance, 1199-1200, 1202-1204, 1206, 1208 safety, 1200, 1219, 1221 symbolism, 1218, 1225, 1246

truth, 1209, 1216, 1217, 1229 Aris, R., 1052 Aristotle/aristotelian, 29, 33-35, 44, 48, 120-122, 378-380, 396, 410, 416, 422, 433, 572, 635, 870, 976, 978, 1011, 1014, 1029, 1031, 1149-1150 Armando, H. D. van, 1247 Arrows theorem, 591 Arrow, K. A., 590-592 Art Nouveau, 1210 artefact/artifact cognitive, 756, 1364 dual nature of, 9, 13, 216, 295, 323, 535, 545 malfunctioning, 297, 305, 494, 936 social 275, 291-292, 301-303, 1164 artefact/artifact epistemology, 67, 189 artefact/artifact function, 188-189, 214232, 314 322, 341, 376, 497, 548, 870, 890, 892, 913, 935, 981 artefact/artifact kind, 198-210, 930931, 936, 938-939, 944 artefact/artifact type, 14, 227, 229, 290, 300-301, 304 artefact/artifact purpose, 220 artefactual/artifactual, 195, 224, 293, 295, 302, 594, 694, 887-890, 892, 896-897, 903, 911, 924, 939, 944 artefactual/artifactual normativity, 919 artefactual/artifactual system, 889, 892 artificial heart, 1186, 1284-1285 artificial intelligence, 146, 424, 426, 527, 648, 1197, 1294, 1354, 1364-1376 classical, 1366 artificial life, 1368-1371 as-if theories, 314-315 Asaro, P. M., 1374 assistive technology, 1291-1292 Audi, R., 343, 378-379, 381, 387-389 automated learning, 1294

Index

autonomy/autonomous, 106, 311, 314, 326, 332, 344, 1333, 13721374 B¨ohme, G., 72, 75, 77, 78, 82, 128 background, 355, 362, 367, 399-400 background knowledge, 597, 717, 1369 Bacon, F., 30, 36, 44, 124-126, 128, 414, 639, 647, 653, 943, 11511152, 1159 Bacon, R., 122, 1150 Bailer-Jones, D. M., 696-697, 705 Baird, D., 9-10, 24, 80, 1197 Baker, L. R., 194-195, 290-291, 293, 304-305 Ball, P., 996, 1050 Banham, R., 1037, 1229 Barber, B. R., 1105, 1115-1116 Barnes, B., 76, 310, 518 Barney, D., 1385 Bartlett, N., 1059 basic function, 258-261, 266-268 basic formal ontology (BFO), 275-76, 280 basic principle, 125, 161, 193, 279, 316, 1341-1342, 1350 Batteux, A. C., 1236 Baudrillard, J., 1377-1378, 1383 Bauhaus, 416-418, 426, 429, 431, 1045, 1224, 1235-1236, 1240 Baumgarten, A. G., 1225, 1230-1231 Beardsley, C.W, 966 Beardsley, M. C., 1204 beauty, 952-953, 984-985, 1045-1046, 1049, 1051, 1054, 1057-1061, 1063, 1065, 1187, 1206-1207, 1229-1233, 1236-1237 Bechtel, W., 239, 1368 Becker, H. A., 802, 808 Beckmann, J., 36-37, 51, 161, 180181 Bedau, M., 1370 behaviour/behavior, 29-30, 175, 317, 358, 361, 377, 390, 413, 421,

1411

424, 426, 443-546, 559, 575576, 522-523, 680, 693, 702, 941-942, 945, 1220, 1222-1223 moral, 1153 Beitz, W., 235, 257-261, 266, 513, 516, 522, 529, 546, 554, 568, 985 belief, 303, 331, 380-381, 575, 929, 933 Bellucci, S., 1116 Ben-Zeev, A., 1385 Benedikt, M., 1239, 1382 Benjamin, W., 49, 1033 Bentham, J., 1153, 1167 Bernard, C., 1282 Bernoulli distribution, 857 Bernoulli equation, 126 beta distribution, 856-857 bias, 39, 78, 276, 315, 531, 559, 562, 895, 908, 901, 914, 965, 967, 1140, 1360, 1394, 1397 Biel, G., 1013 Bigelow, J., 36-37, 40, 146, 222, 227 Bijker, W., 6, 93, 96, 494, 537, 611, 893, 1117, 1177 Billington, D., 129, 1036, 1038, 1041, 1054-1056, 1058 Bimber, B., 1104, 1014 bioengineers, 1293 bioengineering, 730, 1204, 1334 biological function, 213-214, 221, 223, 226-227, 239, 248, 268, 869, 945, 1291, 1388 biological item, 925, 931, 943, 945 biological kind, 201, 204 biological mechanism, 1266 biological process, 749, 1301 biological system, 727, 749, 752, 869, 1307, 1308, 1316 biology, computational, 1307-1308 philosophy of, 13, 213, 238-39, 268-69, 294, 1195, 1310, 134344

1412

Marcel Scheele and Andreas Spahn

synthetic, 1163, 1306, 1308-1309, 1315, 1324, 1331 biomedical engineering, 634, 727-728, 749, 1276 biomedical ethics, 1164, 1169 biotechnology, 82, 149-150, 361, 885, 1142, 1196-1197, 1261, 12651266, 1280, 1301-1337 Bishop, K., 1042, 1367 Bjork-Shiley heart valve, 1283 Black, M., 650 bill of materials, 255 block diagram, 546, 586, 712 Bloom, P., 206 Boden, M., 509, 1365-1366, 1370 Boltzmann, L., 649 Borgmann, A., 493, 918, 1161-1162, 1182, 1259, 1380, 1382-1383, 1385, 1390 Borgo, S., 278 Borning, A., 1392, 1397 Bostrom, N., 44, 1182, 1389 bounded rationality, 465-66, 578, 583, 598 Boussora, K., 1231 Boyle’s law, 126, 784, 789, 791-792 Bratman, M., 398-400, 514, 531, 576 Brennan, G., 1390 Brey, P., 1364, 1382-1383, 1385, 1389, 1396-1397 Bridgman, P., 645, 784, 807-808 Briggle, A., 1166, 1182, 1384-1385 British Royal Society, 1013 Broens, R. C. J. A. M., 321 Brooks, R. A., 746, 1370, 1373-1374 Broome, J., 932-933, 949-951, 977 broad rationality, 573 Brummett, B., 1033 Brunelleschi, F., 123, 353, 639 Bucciarelli, L. L., 98, 106, 352, 364, 368, 466, 474, 570, 590, 594, 866, 870, 969 Buckingham, E., 635, 802 Buckingham’s theorem, 818-821

building code, 866, 871, 873, 953, 959, 960, 963, 964, 969, 1179 built in justification, 702 Bunge, M., 69-73, 83, 88, 166, 169, 309, 340-341, 347, 633 Burgess, J, 1043 Burke, E., 1230, 1232-1233 Burkhardt, J., 1202, 1260, 1265 business administration, 98, 106 Butler, T., 1360 Bynum, T. W., 1342-1343 Byrne, R. M. J., 652-653 Callon, M., 96-97, 148-149, 610 calculative rationality, 596-97, 1161 caloric theory of heat, 707-710, 715721, 723 Campanella, T., 653 Campbell, K., 146, 279 Campbell, N. R., 808 capacity, actual, 248, 302 attributed, 293, 294-297, 299-305 cognitive, 1055 functional, 204, 248-251 physical, 248-249, 364 space of, 291, 294-295, 299-300 capitalism, 54, 1157, 1245, 1260-1261, 1377, 1387 Carlat, D., 869 Carnap, R., 421, 568, 651, 656, 787788 Carnot engine, 329-330, 346, 634, 695, 703-723, 1022 Carnot, S., 132-133, 329-330, 1022 Carr, N., 1183 Carroll, L., 655 Carter, R., 1237 Cartwright, N., 72-73, 679, 754 Castells, M., 1310, 1377 causal dilution problem (in ethics), 1091-1097 cause-and-effect, 166, 169, 405-406, 1075, 1115

Index

causal explanation, 10, 304, 405 causality, 435, 444-451, 651 aristotelian theory, 422-423, 870, 1014 Cavendish, H., 1050 Certeau, M. de, 1175-1177, 1179 ceteris paribus, 609, 792-793, 878 CFC’s, 988-989, 993-994, 996 Challenger, 875-876 Chalmers, D., 1363 Channell, D. F., 24, 69, 71, 73, 79, 95, 143, 147, 150, 160 Charter of Athens, 1217, 1223 chemical engineering, 98, 693-694, 702, 802, 1045, 1051-1053, 1097 chemistry, 75, 86, 653, 783, 1044-1054, 1063-1067, 1315 agricultural, 74-75, 77 Chinese Room argument, 1352, 1367 Christian, G. J., 1015, 1316 Church-Turing thesis, 1347-1349 Churchland, P., 388-389, 1369 citizen, 111, 291, 617, 1003, 1042, 1065, 1110, 1114, 1116, 1119, 1121, 1125-1126, 1137, 1139, 11641165, 1176, 1258-1159, 1325, 1334, 1386-1387 Clapeyron, B. P. E., 132, 713, 716 Clark, A., 701, 1363, 1368-1369, 1375, 1389 classical/Newtonian mechanics, 126, 131, 133, 188, 327, 329, 347, 790-791, 793-796, 806, 809, 1059 classification system, 36, 47, 131 classification of scales and measurement, 833 classical artificial intelligence, 1366 classical/Newtonian mechanics, 126, 131, 133, 188, 327, 329, 347, 790-791, 793-796, 806, 809, 1059 Clausius, R. J. E., 132-133, 329, 714716, 1010

1413

clinical engineer, 1275, 1293 clinical trial, 1085, 1089-1090, 12821283 colocation, 191-192, 198, 210, 281 cochlear implant, 1286, 1289 Cockburn, C., 107-108 Cocking, D., 1384 Codd, E. F., 657-658 code of ethics, 953, 956-958, 961, 967, 970, 1172 codification, 336, 366 codified knowledge, 331 cognitive artifacts, 756, 1364 cognitive goal, 75, 157-158 cognitive science, 239, 356, 368, 640, 650-651, 653, 1142, 1351, 1362, 1365-1366, 1370 philosophy of, 1389 cognitivism, 928, 1362-1363 coherence, 159-160, 164, 170, 363, 391395, 531, 812 Colburn, T., 1345, 1357 Coleman, A. M., 1223 collective intentionality, 219-220, 231 collective rationality, 508 Collingridge, D., 1117, 1167 Collins, H., 56, 73, 112, 174, 334, 355 Collins, P., 1232 Colquhoun, A., 1208-1209 Colt, S., 642 combustion engine, 365, 389, 871, 1009, 1011 common morality, 954-956, 958, 960961, 970 commonsense knowledge, 277, 13681369, 1371 communication, 57, 105, 330, 332, 410, 429, 612, 615, 913, 1121-1122, 1197, 1294, 1310, computer mediated, 1383-1386, 1389, many-to-many, 1376, mass, 415, 429, 1377 neuron, 739-740, 742, 745, 748,

1414

Marcel Scheele and Andreas Spahn

753 risk, 1083, 1132 communication technology, 1141-1142, 1310-1311 communications infrastructure, 622 communicative rationality, 583, 597 compensatory technology, 1289, 12911292 competence, 97-99, 104-105, 159, 323, 325, 328, 333, 335, 477, 515, 597, 652, 911, 961, 1127, 1372 complex adaptive systems, 538, 613614 composition, 192, 506, 889-890, 1207, 1236, 1321-1322 computability, 702, 1342-1343, 13471349 computational biology, 1307-1308 computational complexity, 1342, 13491350, 1361 computational philosophy, 1343-1344 computational systems, 1341-1343, 1346, 1351, 1366 computational template, 702 computational turn, 1341, 1343-1344 computationalism, 1349, 1365, 1367 computer aided design (CAD), 238, 257, 535, 548, 559, 645-646, 1064 computer and information ethics, 1103, 1197, 1341-1342, 1345, 1360, 1384, 1390-1398 foundationalist debate, 1395-1396 computer programming, 105, 609, 106063, 1351-57 computer science, 68, 145-147, 434, 688, 1342, 1345, 1352-1364, 1366, 1372, 1391-1392 deflationary account, 1353-1354 computer science methodology, 105 computer simulation, 170, 637, 645, 1182, 1196, 1361-1362, 1370, 1382 Computer Supported Collaborative/

Cooperative Work (CSCW), 105, 1364 computer systems, 103- 106, 1362, 1364, 1366-1367, 1395, 1397-1398 computerized decision support, 1294 computing, 103-104, 146, 1062, 13451352, aesthetic, 1062 evolutionary, 1370 computing function, 547-561 conceptual design, 266, 393-394, 399, 593, 985, 1022, 1042 conceptual designing, 237-239, 257, 260, 267 conceptual innovation, 752 conceptual truth, 998 concinnitas, 1207 concrete entities, 279-280, 695, 702 conjoint measurement, 839-841, 849 conjoint structure, 794, 840-841 connectionism, 1368-1371 Connolly, C., 1234 consent, 618, 1095-1097, 1165, 1187, 1214, 1225, 1393 informed, 1167, 1173, 1282, consequentialism, 974, 1183 conservation laws, 259-260, 266, 268, 817, 1016 conservation of energy, 716, 768, 1010, 1014, 1016, 1020 consistency, 277, 392, 469, 476, 531, 562, 577-578, 1059, 1127, 1132, 1134, 1153, 1214, 1218, 1344, 1373, 1379 constitutive end, 383 constrained optimization, 608-610, 614, 616-617 constitution metaphysical/ontological, 274, 281, 289, 293-294, 298-299, 304305, 1053, 1106 legal/political, 38, 53, 289, 605, 619-620, 1025 constitutional economics, 620

Index

constraint, economic, 572, 986, 1224 engineering, 241-242, 253, 727, 752, 756 construction as activity, 30, 32, 37, 67, 70, 72, 76, 106, 128-129, 161-164, 316, 320, 443, 468, 470, 475, 721, 968, 1201 theory, 50, 220, 230, 695, 705, 724 construction as object, 30, 56, 66, 146, 250, 693, 740, 761, 771, 1133, 1235, 1313 social, 6, 1117 constructive technology assessment (CTA), 886, 1117-1118, 1167 consumer autonomy, 1168, 1330, 13331335 consumer economics, 45 contingent normativity, 898 contingent valuation, 1087 continual innovation, 136-137 contract theory, 1095 control-volume analysis, 317-318, 328 controllability, 509, 539, 603, 611, 618619, 622-623, 625 Cooper, D., 1032 Cooper, G., 900-902 Cooper, T., 785 coordinate system, 678-679 Cope, C., 1359 Copeland, B. J., 1348-1349, 1366-1367 Copernicus, N., 639, 647, 651 copied kinds, 201-204 Cornelius, P., 805, 809-810 corporate control, 1335-1337 correspondence, 84, 252, 318-319, 604, 625-626, 690, 698, 707, 731, 765, 772, 790, 1061-1062, 1314, 1373 cost-benefit analysis, 337, 621, 989, 991-993, 995, 1000-1001, 1071, 1090, 1133, 1139, 1140, 1154 Cotgrave, R., 641

1415

Cox, D. J., 646, 655 Craik, K., 651-652 Craver, C., 239 creativity, 44, 165, 352-354, 358, 362, 429, 445-446, 455, 514, 584586, 650, 653, 746, 913, 1179, 1187 critical epistemology, 1208 critical theory/critical social theory, 429, 1025, 1105, 1261, 1175 Cross, N., 170, 363, 431, 447-448, 450, 455, 461, 463-464, 471, 501, 516, 526, 529, 995 cultural value, 127, 984-85, 1024, 1227 culture, 85, 207, 360, 414-416, 427430, 984-985, 1161-1162, 1181, 1186-1187, 1209-1210, 12131214, 1286, 1316-1317, 1377, 1380 design, 409-411 material, 213-214, 216-217, 230231, 756, 1178 Cummins, R., 223, 226-228, 248, 268, 294, 322, 863, 867, 869-870 Cunningham, D. J., 1384 Curie, M., 1050 Curl, R., 149, 1047 Curtin, D., 1059 Curtis, W., 1210 customer requirements/needs, 407, 481, 489-511, 547-548, 555, 561, 566-567, 584, 587 cybernetics, 146, 746, 1290, 1383 cyberpolitics, 1341, 1386-1387 cyberspace, 1373, 1381-1387, 1390, 13951396 cyborg, 1162, 1290, 1317, 1341, 1363, 1388-1390 cyborg theory, 1162, 1388-1389 cyc/CYC, 277-278, 1371 Da Vinci, L, 123, 131, 532, 654 Daele, W van den, 128 Dahl, R., 1168

1416

Marcel Scheele and Andreas Spahn

Dahlbom, B., 104 Dancy, J., 927, 929, 932, 938, 947, 949-951, 974, 997, 1004 data model, 711, 1357-1358, 1360-1361 data modelling, 470, 1341-1342, 1353, 1357-1361 database, 556-568, 1352, 1357-1360, 1393 database theory, 568 Davidson, D., 573, 650 Davis, M., 43, 103, 956-957, 11721173 Davis, A.B., 1275-1276, 1278-1279 de Meijere, A., 1047 de minimis risk, 1085 death, 825, 876, 966, 1070-1071, 1083, 1092-1093, 1170, 1182, 1212, 1225, 1283-1285 decision theory, 391-392, 929, 10701072, 1097, 1111, 1154 decision making problem, 589, 1137 decision making process, 176, 421, 516517, 566, 589, 608, 965, 1103, 1112-1113, 1116-1117, 1121, 1137, 1142, 1152 decision making under certainty, 580, 589 decision making under uncertainty, 580 decision making procedures, 174, 176, 1105, 1109, 1139 Decker, M., 1113, 1120-1121, 11231124 decomposition, 192, 315, 442-443, 445, 548-549, 1269 decoration, 1042, 1206-1208, 1212, 1218, 1229, 1236 Dedekind, R., 640 deductive-nomological, 863-864, 869, 877 deductive reasoning, 434, 1356 defeasance problem, 1097 DeGrazia, D., 1389 deliberative democracy, 1105, 11151117, 1126, 1135, 1139

deliberative rationality, 605, 618, 619 demarcation, 327-330 Dement, C., 236, 239, 254-256 democracy, 894-895, 1105, 1115-1116, 1136, 1168, 1172 deliberative 1105, 1115-1117, 1126, 1135, 1139 Dennett, D. C., 525, 1370 deontic, 439, 883, 927-928, 940, 943, 949-950 deontology, 1183 dependence/dependency, 194-198, 207208, 227, 281, 289, 305, 343345, 475-476, 479-480, 500, 602, 616, 700, 855-856 Descartes, R., 124, 766, 1151, 1363 design function-behaviour-structure (FBS) model, 464, 544, 547, 554 satisficing, 598 social context of, 436, 439 design constraints, 467, design culture, 409-411 design decisions, 439, 459, 468, 471, 475-476, 480, 498, 505, 553, 556, 568, 575, 589-592, 598, 602, 604-606, 609, 616-617, 621-622, 624-625, 926, 1032, 1236-1237 design evaluation, 397 design goals, 165, 399, 462, 617-625 design knowledge, 67, 336, 338, 468, 528-529 design methods movement, 426, 430448, 450 design methodology, 240, 257, 408411, 414, 421, 426, 428, 431433, 435-438, 445-448, 450451, 459, 461, 501-502, 516, 565, 588 design paradigm, 536-539, 556 design practice, 407, 412, 414, 420, 430, 443, 455-486, 509, 514517, 528, 584-592, 959

Index

engineering, 407, 514-515, 517, 519, 525, 527, 539, 565, 569, 579, 584-592, 599, 913, 970 design principle, 40, 129, 131, 422, 516, 586, 601, 615, 1240 engineering, 40 design problem, 362, 364, 368, 425, 428, 432, 437, 440-441, 444, 448-449, 456-457, 461-467, 474, 480, 515-518, 569, 572, 579, 588-589, 591-593, 610, 617618, 995-996, 1022, 1024 design process, 170-173, 257, 295, 361366, 368, 375-378, 385, 389390, 393-395, 398, 400, 424425, 437-439, 456-460, 463471, 473-477, 479-480, 501502, 510, 518, 523, 526, 529530, 565-572, 586-589, 601602, 912, 977, 981, 985-989, 999, 1002, 1022-1024, 1063 design requirements, 432, 489, 553, 985-988 design specification, 107, 566-567, 570, 591, 1023 design team, 515, 589-590 design as a decision making process, 517 designer, industrial, 429, 432 intelligent, 534 web, 1392 designer fallacy, 1220 determinism, 52, 370, 1090, 1097, 1222, 1397 historical, 411 technological, 499, 607, 611, 1161, 1296, 1397 Dewey, J., 44, 50, 55, 418-423, 425427, 430, 434, 451, 1158-1159, 1166-1167 diagnostic technology, 1276-1282 diagram block, 546, 586, 712

1417

phase, 529, 586-588 pattern, 440, 443 Dierkes, M., 1118 Dieter, G. E., 385, 389, 393 difference measurement, 836-837, 849 Digges, T., 639, 647, 651 digital divide, 1387, 1394 Dijk, J. van, 1377, 1387 Dijkstra, E., 1357 dimensional analysis, 645, 675, 759760, 785-787, 791, 794-796, 799-822, 1051 dimensional equation, 802-803, 813, 815-816, 819-820 dimensionless parameter, 799, 801-804, 814-822 dimensionless number, 635, 645, 759, 769, 776, 791, 793, 795, 1051 dimensionless quantities, 814-815, 821 dimensionless ratio, 772, 803, 815-818 Dipert, R. R., 191, 194-195, 213, 220221, 229-231, 291-292, 301, 924 Dirac, P., 681, 1051 direct reference, 52, 199-200, 210, 291 direction of fit, 378, 380-381 disability, 1290-1291 discourse ethics, 1115, 1126 disease, 749, 1085, 1183-1184, 1196, 1270, 1276-1291 disposition, 294, 299, 322, 355, 376, 869, 1056, 1280, 1323 distributed cognition, 1363-1364 Djabri, F., 1062 Dobson, P. J., 1360, Dodig-Crnkovic, G., 1349, 1395 DOLCE, 274, 276, 279-281, 284-291, 293-294, 296-300, 303-306 Dreyfus, H. L & S.E., 355, 465, 467467, 477, 583, 596-597, 1363, 1367-1369, 1371-1373, 13791381, 1383-1385, 1390 Du Plooy, N. F., 1359 dual nature, 9, 13, 216, 295, 323, 535,

1418

Marcel Scheele and Andreas Spahn

545, 618, 1326 Ducheyne, S., 908 Duhem P., 649-650 Duncan, W. J., 808, 819, 1079 Durbin, P., 1168 Dym, C. L., 352, 362, 375-376, 384385, 389-390, 393-395, 397400, 513, 516, 519, 523, 526528, 532, 535 dynamical systems theory (DST), 1370 Eades, P., 1060 Eardley, A., 1217 eclectic, 428, 1041, 1200, 1202, 1208, 1237-1238 Eco, U., 1049 economic value, 620, 903, 981-982, 991992, 995, 1003-1005 economic constraints, 572, 986, 1224 economics/economy constitutional, 620 consumer, 45 domestic, 47 information, 1057 institutional, 54 knowledge, 351 political, 51-52, 1156-1157, 12591261, 1309 socio- , 336,543-547, 555, 562, 614, 617, 1178, 1330 Eden, A. H., 1345, 1354-1357 Eder, W. E., 164, 170-171, 257, 338, 516, 522, 528, 535 Edison, T., 137 education of engineers, 93, 97, 99100, 102 Edvardsson, K., 391, 393, 395, 578 Edwards, A., 1234 effectiveness, 51, 71, 157, 376, 447, 456, 577-578, 989-991, 1011, 1014-1016, 1027, 1153, 1347 efficiency, 30, 41, 125, 131, 531, 624, 704, 707, 722, 989-991, 10071027, 1054, 1172

definitions, 618-621, 884, 10081010 economic, 415, 618, 1120, 1133 engineering, 1009, 1010, 1027 mechanical, 1008, 1023 efficiency of design, 459 efficient cause, 422, 635, 1008, 1014 Ehn, P., 105, 108 Ehrenfest-Afanassjewa, T., 802, 808 Einstein, A., 1163 Eisenman, P., 1203, 1231 Elder, C., 191, 201-202, 204, 207, 290291, 293 electric power/energy, 38, 257, 267, 537-538, 550, 556-557, 673, 889, 891, 910, 1140 electrical engineering, 27, 40, 48, 98, 136, 179, 244, 565, 693-694, 740, 1049, 1313 elegance, 618, 1032, 1041, 1050, 10541058 Eliade, M., 1215 eliminativism, 210 ELIZA, 1375 Ellis, B. D., 786, 790, 807-808, 810811 Ellul, J., 45, 493, 1025, 1161 embeddedness, 604, 606, 1201, 1261, 1313 embodiment, 280-281, 407, 433-434, 535, 700, 712, 739, 743-749, 985, 1026, 1054, 1059, 10611062, 1178, 1212, 1237, 1373, 1390 emergence, 353, 358, 370, 536-539, 603, 605-607, 613, 615-616, 625, 739, 1367, 1370, 1397 empirical adequacy, 67, 337, 339, 518, 1059 empirical content, 635, 759, 786-791, 793-796, 813 empirical model, 69, 697, 828 enabling technology, 480, 1276, 12901292

Index

ends constitutive, 383 final, 376, 378-379, 381, 384-386, 392, 864 instrumental, 392, 578 maieutic, 384-385 typology of, 382 endurant, 243-247, 258-260, 266, 275, 280-281, 285-290, 293-295, 299300, 303, 305 energy electric, 910 nuclear, 145, 173, 494, 902, 905907, 916, 1001, 1069-1070, 1097, 1109, 1155, 1330 solar, 262, 900, 910 wind, 109-110 energy infrastructure, 607-608 von Engelmeyer, Peter, 39 Engerman, S., 1019 engine Carnot, 329-330, 346, 634, 695, 703-723, 1022 combustion, 365, 389, 871, 1009, 1011 heat, 132, 316, 329-330, 634, 695, 703-723 Newcomen, 138, 538-39, 711-12 petrol, 167 steam, 9, 29, 36, 71, 126, 132-33, 139, 161-62, 316, 329, 532, 680, 705-707, 714, 717-21, 1018-19, 1022, 1155, 1185 engineering, bio, 730, 1204, 1334 biomedical, 634, 727-728, 749, 1276 chemical, 98, 693-694, 702, 802, 1045, 1051-1053, 1097 electrical, 27, 40, 48, 98, 136, 179, 244, 565, 693-694, 740, 1049, 1313 genetic, 148, 150, 1131, 1166, 1257, 1265, 1267, 1301-1306, 1308, 1310, 1315, 1320, 1330-1331

1419

interdisciplinary, 727 knowledge, 1358-1359, 1364, 13711372 mechanical, 1008-1009, 1027 safety, 1076-77, 1082-83, 1090, 1195 science based, 4-5, 14 software, 103-04, 321, 385, 646, 665, 885, 1033, 1054-65, 1341, 1352-60, 1397 engineering constraints, 241-242, 253, 727, 752, 756 engineering curriculum, 42, 93-94, 97100, 106, 317, 405, 425, 430, 1173 engineering design practice, 407, 514515, 517, 519, 525, 527, 539, 565, 569, 579, 584-592, 599, 913, 970 engineering design principle, 40 engineering education, 94, 98, 101102, 106, 1060 engineering efficiency, 1009, 1010, 1027 engineering ethics, 379, 886, 969, 1069, 1119-1120, 1169, 1172-1175, 1392 engineering forensics, 874-875 engineering knowledge, 93, 96, 106107, 112, 235, 238, 260, 267, 317, 356, 369, 634, 760-761, 763, 1053, 1174 engineering practice, 99, 159, 188, 242, 255, 313, 352, 377, 379-380, 393, 397, 400, 406, 516, 522, 525-526, 529, 532, 534-535, 543, 566, 574, 584, 634-635, 762, 775, 779, 886, 954, 956, 959-961, 963-964, 968-970, 1070, 1120, 1171 engineering standards, 169, 884, 953, 958, 960-964, 1120, 1179 engineering theory, 309, 316-317, 320, 322, 325-326, 330, 346, 1022 enhancing technology, 1276 Enlightenment, 1012, 1015, 1152, 1157,

1420

Marcel Scheele and Andreas Spahn

1181, 1211, 1216, 1236 entropy, 133, 319, 716, 813 Entscheidungsproblem, 1347-1348 environmental ethics, 1166, 1228 environmental compatibility, 1130, 11331134, 1139 environmental effect/consequence/ impact, 101, 558, 593, 989, 996, 1043, 1072, 1113, 1130 1134, 1173, 1182, 1221, 1229 environmental issue/concern/hazard, 958, 1075, 1138, 1332 environmental mechanism, 1276 episteme, 34, 1275 epistemic emancipation, 309-313, 319322, 326-328, 331-332, 334335, 340, 342, 344-345 epistemic function, 701, 722-723, 1032 epistemic justification, 941 epistemic normativity, 940 epistemic rationality, 577 epistemic standards, 568, 1054 epistemic value, 34, 330, 518, 695, 699, 703, 711, 755, 885, 1031, 1045-1046, 1378 epistemic value of models, 695, 699, 703, 711, 755 epistemic virtue, 319, 343 epistemology artefact/artifact, 67, 189 critical, 1208 internet, 1378-1381 social, 509, 1381 equation, Bernoulli, 126 Euler, 769 Navier-Stokes, 791 Escher, M. C., 655 Ess, C., 1384, 1396 ethical codes, 1392 ethical gap, 1086 ethical evaluation, 1219 ethics actualism, 1091-1092

architectural, 1219, 1228 applied, 1103, 1127, 1154, 1163, 1169, 1197, 1391, 1395-1396 biomedical, 1164, 1169 causal dilution problem, 1091-1097 code of, 953, 956-958, 961, 967, 970, 1172 common morality, 954-956, 958, 960-961, 970 computer and information, 1103, 1197, 1341-1342, 1345, 1360, 1384, 1390-1398 consequentialism, 974, 1183 deontological, 1093-1095, 11531154, 1183, 1335 discourse, 1115, 1126 engineering, 379, 886, 969, 1069, 1119-1120, 1169, 1172-1175, 1392 environmental, 1166, 1228 good life, 976, 983, 1004, 1161, 1170, 1181, 1184, 1201, 1258 medical, 1103, 1169, 1227, 1275, 1283, 1286, 1396 pre-modern, 1148-1151 professional, 957, 1147, 1169, 117172, 1175, 1228-29, 1360, 139293 utilitarianism, 978, 992, 1001, 1089, 1091-92, 1139, 1155, 1167, 1181,1245, 1396 virtue ethics, 1050 ethics of risk, 1090-1100 ethical and normative standards, 884, 953-958, 1004, 1139, 1223 ethical truth, 928 ethnography, 106, 352, 728-729 ethos, 1149-1150, 1160, 1226 ethos of science, 1311-1312 etiological theories of function, 248 Euclid of Alexandria, 119-120, 123, 405, 654 Euler equation, 769

Index

evaluation, design, 397 ethical, 1219 normative, 1123, 1127 numerical 390-92, 397 rational, 572, 577, 586 social, 1393 technology assessment, 12-13, 31, 174-77, 885-86, 1103-1143, 1166, 1229, 1264, 1302 technology/technological, 1103, 1120 evaluation criteria, 1123, 1125-1126, 1134-1135 evaluation and selection methods, 165 evaluative judgement, 926, 934, 936, 940, 946, 1127 evaluative statement, 894, 925, 928930, 934-935, 937, 943, 946, 974 Evans, R., 112 evidence-based medicine, 1283 evidence, scientific, 218, 1076 evolutionary computing, 1370 exemplification, 667, 745, 754-756 existence questions, 191-193, 196-198 existentialism, 1155, 1159 expectation value, 1070-1071, 1074, 1083 expected utility, 391, 570, 576, 579580, 585, 1071, 1092-1093, 1098-1099 experiment, 79-84, 122, 160, 170, 334, 346, 518, 736-737, 771-772, 776, 1050, 1062, 1362 thought, 329, 534, 1343, 1349, 1367, 1370 experimental action, 70, 79, 85 experimental process, 80, 86-87, 366 experimentation, 70, 78-81, 84-86, 160, 343, 346, 730, 756, 770, 10491051, 1053, 1314 expert system, 273, 335, 1364, 13711372

1421

expertocratic, 1105, 1115-1116 explanation, causal, 10, 304, 405 deductive nomological, 863-864, 869, 877 mechanistic, 239, 269 psychological, 333, 864, 683, 738, 861, 864, 1064, scientific, 2, 10, 635, 638, 691, 861-865 technical 2, 10, 13, 342, 635, 861878 teleological, 864 explanatory modelling, 672, 682 explicit knowledge, 331, 333, 355, 357, 891 extended lab, 148-149 extensive measurement, 825, 833-835, 841-850, 853, 857 external audience, 861, 866-868, 871, 873 externalism, 578, 1363 fact-value distinction, 345 Fallis, D., 1380-1381 Faraday, 648, 652, 654 Farmelo, G., 1059 Faulkner, W., 321-323, 325-326 fault tree, 1083-1084 feedback, 132, 146, 171, 365, 367, 393, 557, 561, 567, 613-614, 623, 743, 745, 868, 877, 1058, 1063, 1362, 1388 feedback mechanism, 1062, 1077 Feenberg, A., 494, 499, 888, 909, 914, 1011, 1057, 1179, 1385 Feinstein, A., 1278 Fellbaum, C., 276-278 Ferguson, E. S., 123, 131, 168, 352353, 356, 367, 642, 1056, 1060 Ferrario, R., 292 Fetzer, J. H., 1356, 1366 Feynman, R., 148-149, 669-670, 672, 674-675, 677-679, 681, 684,

1422

Marcel Scheele and Andreas Spahn

687, 1315 fictitious model, 314 final ends, 376, 378-379, 381, 384386, 392, 864 final rationality, 583 finalization theory, 74-78, 82, 88 Fisher, E., 1167 Fisher, F., 1107, 1137 Fisher, S., 1227 Fishwick, P., 1061-1062 fixation of ends, 572, 577-578, 582 fixation of means, 577-578 Floridi, L., 1342, 1350-1351, 1358, 1374, 1378, 1380, 1395-1396 Florman, S., 100, 309, 979 Fodor, J., 640, 1351, 1363, 1367, 1369 Fogel, R. W., 1012, 1019 forensics, 874-876 engineering, 874-875 social, 874-76 formal ontology, 188, 273-306 Forman, P., 24, 87, 147 Forster, E. M., 1234 Foucault, M., 905, 1161, 1183, 1216 foundational ontology, 273-275, 277279, 282, 284-285, 304-305 Fox, R., 127, 134 Fox, W., 1221, 1227-1228 Fraassen, B. van, 67, 668, 696 frame problem, 1386 Frampton, K., 429, 1213 Frankfurt school, 429, 1024-1025, 1105, 1157 Franssen, M., 205, 229, 376-377, 517, 590, 592, 930, 945, 949, 994 Frege, F. L. G., 656 Freud, S., 1210, 1234 Friedman, B., 621, 987, 1002-1003, 1167, 1177, 1392, 1397 Frigg, R., 634, 637, 651, 696-697, 1362 Frontinus, S. J., 638 Froude, W., 634, 644-645, 759, 765, 771, 779-780, 782, 784-788, 791-793, 795

Froudes extrapolation method, 760, 779-780, 782, 785, 787, 793, 796 Froudes law of similarity, 765, 791794 Froude number, 645, 769, 778, 786, 792, 815 Fuchs, V., 1295-1296 Fukuyama, F., 1389 function, artefact/artifact, 188-189, 214-232, 314 322, 341, 376, 497, 548, 870, 890, 892, 913, 935, 981 basic, 258-261, 266-268 biological, 213-214, 221, 223, 226227, 239, 248, 268, 869, 945, 1291, 1388 causal role/system, 248-249, 251, 268 etiological theory, 248 intended, 205-207, 294, 440, 574, 751, 897-898, 915, 990-991, 1175, 1120 intentionalist theory of, 226 user intentionalist approach of, 249 proper, 201-205, 221-232, 248, 446, 496, 498, 693-94, 707, 722, 889, 938 function-behaviour-structure (FBS) model of design, 464, 544, 547, 554 function-means tree, 376, 384, 389, 393, 526 functional analysis, 42, 424, 680, 869 functional composition, 240-248, 250, 253, 267 functional decomposition, 235-269, 534, 538, 548, 646 functional description, 235-240, 243, 247-252, 254, 256-260, 267269, 523-526, 528, 532, 534, 552, 567 functional kind, 188, 930-931, 935936, 938-939, 948

Index

functional modelling, 257, 259, 266, 268, 535, 665-691 functional nature, 323, 325, 496-499, 501, 504-505, 507, 509, 1224 functional organisation, 236, 240-248, 251, 267 functional part-whole relationship, 235236, 240, 251, 253-254, 256, 267-268 functional reasoning, 235, 238-240, 269, 527, 535, 1214 functional requirement, 405, 517, 566567, 570-572, 584, 587, 591, 612, 936, 944, 947, 985-986, 990 functional system, 247, 887-888, 890, 892, 896-897, 903, 911, 918 functionalism, 418, 612, 1033, 1036, 1042, 1065, 1214, 1216, 1240 functionality, 156, 158, 165, 173, 176, 188, 215, 217, 220, 448, 472, 543, 545, 548-549, 553-554, 558-560, 574, 612, 889-890, 914, 1033, 1054, 1057, 1236, 1283 fundamental design concepts, 324, 363364 fusion, nuclear, 76-77, 82 G¨ ardenfors, 294 Galileo, 125, 130, 414, 644, 647, 759, 762, 801, 1013 Galison, P., 67, 89, 143-144 Gartman, D., 1033 Gauldie, W. S., 1236 Geach, P., 292 Gelder, T. van, 1370-1371 Gelernter, M., 1038 Genard, J. L., 1225 gender, 108, 910, 1235, 1258 general formal ontology (GFO), 276 general technology, 156, 163-164, 179181

1423

generic abstraction, 752-754 genetic code, 150, 1306, 1309, 13261327 genetic engineering, 148, 150, 1131, 1166, 1257, 1265, 1267, 13011306, 1308, 1310, 1315, 1317, 1320, 1330-1331 genetic enhancement, 1386, 1288 genetic essentialism, 1281 genetic exceptionalism, 1280 genetic modification, 148, 150, 1131, 1166, 1257, 1265, 1267, 13011306, 1308, 1310, 1315, 1317, 1320, 1330-1331 genetic programming, 1310 genetic resources, 1307 genetic technology, 1277, 1286 genetic tests, 1280-1281 genetic use restriction technology (GURT), 1197, 1266 genetically modified/engineered organism (GMO), 1108, 1125, 1197, 1257, 1264-1266, 1317, 1319, 1325, 1329 Gentner, D., 651-652 geometric generalizations, 857 geometric similarity, 772, 789-790, 800801, 803, 821 geometrical ratio, 1013 Georgeff, M., 292 Gero, J., 464, 521, 544-545, 547, 556, 561 Gert, B., 929, 954-956 Gibbons, M., 77, 885, 1103, 1114 Gibbs, R. W., 1062 Giddens, A., 95, 1360 Giedion, S., 1035 Giere, R., 691, 696-698, 701, 711, 754, 803 Gillett, G., 1389 Gillies, D., 1372 Gilpin, W., 1233 Giorgi, 809-810 global village, 1385

1424

Marcel Scheele and Andreas Spahn

globalization, 1025-1026, 1109, 1114, 1140-1141, 1153 God, 121-122, 144, 352, 457, 459, 534, 1007, 1011-1014, 1124, 1150, 1203, 1209, 1303, 1317, 13311333 Goethe, 1234, 1268 golden ratio, 1231 Goldman, A., 1157, 1378, 1381 good life, 976, 983, 1004, 1161, 1170, 1181, 1184, 1201, 1258 Goodman, N., 755-756, 803 Gorniak-Kocykowska, K., 1396 Gorp, A. van, 176, 960, 965, 988, 1108, 1119, 1122, 1127 Grahn, W., 1047 graphical analysis, 129 gravitational force, 543, 777-778, 815 Gray, C., 1388-1389 green revolution, 1257, 1261-1264, 1266, 1270 Greyhound experiment, 765, 771, 785, 791, 793 Griffiths, P. E., 213, 222-224, 227, 229-230, 1310-1311 Grim, P., 1343-1344 Grin, J., 1118, 1138 Gropius, W., 416-418, 426-427, 1240, 1249 Gruber, T. R., 1358 Grunwald, A., 155, 167, 169, 175, 965, 1103-1104, 1107-1109, 1112, 1116, 1122-1128, 1132, 11381139, 1141 Guarino, N., 278, 1359 Guha, R.V., 277, 1371 Gundisalvo, D., 121 Guston, D., 1117, 1167 Gutenberg, J., 641 Habermas, J., 45, 79-80, 83, 87, 583, 596-598, 894, 1025, 1105-1106, 1115, 1126, 1139, 1142, 1157, 1199, 1360 , 1389

Hacking, I., 67, 160, 319, 406, 518, 695, 1314 halting problem, 1348-1349 handicap, 940, 1279, 1290-1292 Hansson, S. O., 10, 15, 42, 391, 393, 395, 578, 926, 1069, 1071, 1073, 1076, 1085-1086, 1089, 1091, 1099, 1195, 1268, 1275, 1279-1280, 1282, 1284, 1290 Haraway, D., 83, 1162, 1290, 1313, 1317, 1388-1389 Harel, D., 1348-1350 Hargittai, I., 1047 Harnad, S., 1351 Harr´e, R., 24, 80 Harries, K., 1226-1228 Harvey, D., 1377 Harvey, W., 648, 1282 Haselager, W. F. G., 1373 Hatamura, Y., 396 Haug, W. F., 1033 Haugeland, J., 1363, 1366 Hausdorff, 851, 854, 856 Hausdorffs Theorem, 851-852, 854 Hayek, F., 1158 Hayes, P., 1371 Hayles, K., 1389 hazard, 1069, 1073, 1077-1078, 1082, 1110, 1115, 1131, 1305, 1332 healthcare, 243, 904, 1161, 1163, 1167, 1186, 1275-1276, 1283, 12921295 health code, 959 health effect, 1085, 1090 health, public, 953, 956 Health Technology Assessment (HTA), 1133 Hearn, F., 1042 heat caloric theory of, 707-710, 715721, 723 conception of, 715-716 heat engine, 132, 316, 329-330, 634,

Index

695, 703-723 Heelan, P., 67 Heidegger, M., 25, 28, 58, 83-84, 87, 187, 493, 1160, 1203-1204, 1211-1215, 1226, 1237, 1307, 1360, 1363 Heidelberger, M., 24, 80 Heisenberg, W., 75, 858, 1054, 10571059 Helmont, J. B. van, 1050 Hempel, C., 86, 209, 789, 863, 866, 869, 877 Hendricks, V., 314-316, 338 Herdeg, K., 1240 hermeneutics, 57, 95, 1118, 1360 Herodotus, 1229, 1235 Hertog, P., 1111, 1119 Hesse, M. B., 650, 781, 783 heuretica, 643 Hickman, L., 55, 1159 hierarchical typology, 457, 460 Hilbert, D., 655, 1347 Hill, P. H., 385 Hillier, B., 1222-1223 Himma, K. E., 1379, 1391, 1395-1396 Hintikka, J., 303 historical determinism, 411 historically proper placement, 202-203 historicism, 1204, 1207-1211, 1217 history of design, 409-410, 501-502 history of technology, 5, 8, 52, 55, 163, 633 Hoare, C. A. R., 671-672, 680, 684, 1354 Hobbes, T., 124, 1181-1182 Hodges, W., 273, 665, 668, 685, 688 Hoff, J. H. vant, 1052 Hoffmann, R., 1046, 1047 homo oeconomicus, 1124 homomorphism, 697, 827-828 Hooke, R., 648 Horkheimer, M., 1024-1025, 1157 Horn sentence, 688 Hottel, H. C., 802

1425

Houkes, W., 15, 207, 214, 218, 224226, 228-229, 239, 247, 249, 291, 304, 309, 341-342, 455, 530-531, 567, 887, 912-915, 934, 939, 981 Hoven, J. van den, 1388, 1396 how-questions, 582, 866, 868, 870 Howard, A., 1269-1270 Hippocrates, 1169 Hubka, V., 164, 170-171, 257, 338, 516, 521-522, 528, 535 Hugh of St. Victor, 44, 121, 1150 Hughes, J., 15, 375-377, 388, 527 Hughes, T.P., 6, 87, 97, 137, 362, 605, 608, 611, 1197 Hughes, S. A., 821 human behaviour/behavior, 29, 47, 95, 465, 1219, 1222-1224, 13741375, 1392 human computer interaction (HCI), 105, 1341, 1352-1353, 13601364 human enhancement, 1286, 1389 human intentionality, 194, 196-197, 207-208 human nature, 1142, 1182-1183, 1259, 1288-1289, 1388-1389 human reasoning, 389, 576, 1369, 1371, 1375 human responsibility, 176, 414, Humberstone, I. L., 381 Hume, D., 377-379, 577, 1153, 12301231 Humphreys, P., 702, 1361-1362 Huntingtons disase, 1281 Husserl, E, 274, 1159-1160 Hutchins, E., 367, 1364 hypercomputation, 1349 hypertext, 1380, 1385 hypothetical device, 708-711, 713, 721 hypothetical retrospection, 1098-1100 hypothetical system, 700, 711 iconography, 1061, 1215

1426

Marcel Scheele and Andreas Spahn

ideal form, 1047, 1204-1207, 1249, 1383 ideal heat engine, 710-712, 714, 716717, 721, 722-723 ideal model, 700 idealization, 120, 314-315, 329-330, 346, 697, 701, 704, 724, 727, 752, 754-755, 817 idealized device, 722 identity criteria, 290, 295-297, 300 Ihde, D., 50, 76, 83, 343, 1058, 1220, 1313 Iivari, J., 104 Ikonomou, , 1234-1235 ill-structured problem, 437, 462, 464, 466, 570, 572 ill-defined problems, 364, 448-449, 490, 570 Illich, I., 1151, 1162, 1168, 1183-1184, 1296 Illies, C., 1199, 1218, 1227 imaginary entity, 704, 711, 724 imaginary phenomena, 710 imaging, 356, 738 imitation, 423, 1206, 1250 implicit knowledge, 166-167, 285, 324, 333, 335 implicit learning, 355, 357 incommensurability, 510, 598, 977-979, 994-995, 998, 1000, 1087-1088 incomplete information, 608, 610, 618 indetectable effects, 1084-1086 individualism, 1167, 1261, 1344, 1387 industrial culture, 416, 427-429, 1012 industrial society, 1007, 1012, 1024, 1026, 1259, 1377 industrial design, 409, 429, 431-432, 434, 451, 458, 475, 946, 1032 industrial designer, 429, 432 industrial process, 51, 82, 510, 10511052, 1303-1304 industrialization, 39, 45, 51, 118, 128, 130, 139, 149, 413-416, 426, 1007-1008, 1014, 1017, 1019, 1153, 1160, 1166, 1184, 1260,

1311 industry-based science, 118, 134-136 indwelling, 354, 359, 363, 368-369 inference, 156, 169, 277-278, 354, 441, 652, 695, 698-699, 701-702, 711, 727, 729, 749-751, 756, 781, 800, 1361 logical, 438 model based, 750-754 practical, 527-528, 909 tacit, 354, 358-59, 362-63, 367369 similarity based, 816-818 inforgs, 1378 information, 1364, 13681371, 1396 concept, 1310 general definition of (GDI), 1351 internet, 1341-1342, 1378-1380, 1394 philosophy of, 9, 13, 1342, 1345, 1351, 1360 information economy, 1057 information ethics, 1103, 1197, 13411342, 1345, 1360, 1384, 13901398 information infrastructure, 1394 information management, 1379 information overload, 1376, 1379 information processing, 48, 147, 424, 578, 744, 751, 754, 1093, 1362, 1368-1369, 1373 information society, 615, 1327, 1341, 1376-1378 information system, 11, 103-105, 107, 273-274, 424, 439, 1341-1342, 1353, 1359-1360 information technology, 9, 68, 11661167, 1197, 1310, 1341, 1359, 1388 philosophy of, 9, 13, 1342, 1345, 1351, 1360 informatisation of life, 1306-1311 informed consent, 1167, 1173, 1282, infrastructure

Index

communications, 622 energy, 607-608 information, 1394 social, 537 socio-technical, 612, 617-618, 626, 861, 1261 technical/technological, 49, 959, 1261 infrastructure systems, 537, 602-603, 612, 616, 618, 623, 625, 626 inherent normativity, 12, 884, 886, 896-898, 903, 908, 910, 915 inherent safety, 1077-1078 inheritance, 277-278, 282, 460 innovation, conceptual, 752 continual, 136-137 technical/technological, 174, 250, 323, 331, 335, 964, 982, 996, 1001-1002, 1106-1107, 1112, 1117, 1119, 1141, 1180, 1263 innovation process, 1119 intelligent designer, 534 institutional arrangement, 603, 606, 609-610 institutional economics, 54 institutional entities, 197, 1383 institutional environment, 604-605 instrumental ends, 392, 578 instrumental means, 383 instrumental rationality, 341-342, 572, 577, 579, 581-583, 585, 588589, 593-594, 596, 598, 1024, 1157 instrumental value, 315, 376, 378, 883, 932-933, 946-949, 975-976, 980982, 985-986, 1010-1111, 10501051, 1060, 1375 instrumentalism, 50, 318-320, 376-380, 382-386, 395, 398, 400 intellectual property, 1166, 1180, 1197, 1319, 1325, 1328-1330, 1336, 1394-1395 intelligibility, 1058-1061, 1064-1065

1427

intended function, 205-207, 294, 440, 574, 751, 897-898, 915, 990991, 1175, 1120 intentional fact, 928 intentional fallacy, 1204 intentional selection, 221-223, 275, 292293, 297, 299-300, 302 intentionalist theory of function, 226 intentionality, human, 194-197, 207-208, 214, 218-221, 230, 1218 collective, 219-220, 231 original, 1351 intentions of designers, 188, 222 intentions of makers, 208 intentions of speakers, 200-201 intentions of users, 249, 251, 1398 interactional expertise, 112 interaction of ethics and technology, 1147 interdisciplinarity, 93, 98-99, 101-102, 104, 110, 509, 1112 interdisciplinary engineering, 727 International Organization for Standardization (ISO), 525, 658, 959-961, 1071, 1134 internet, 629 philosophy of the, 1345, 13761390 internet aided modelling, 174 internet epistemology, 1378-1381 internet information, 1341-1342, 13781380, 1394 interpersonal compensability, 1089 intrinsic value, 975-976, 992, 1166, 1286, 1396 Introna, L., 1360 invariance theorem, 635 Inwagen, P. van, 1, 193, 195 irrational use, 913-915 isomorphism, 315, 688-691, 696-698, 794, 826-829 isomorphism of models, 826-828

1428

Marcel Scheele and Andreas Spahn

Jacobs, J., 1039, 1045 Jacobsen, K., 100 Jacobson, N., 1287 Jahn, H. A., 1046 Jakobsen, A., 314-315 Jameson, F., 1377 Jasanoff, S., 1179, 1305, 1316-1318, 1330 Jencks, C., 1039-1040, 1210 Joerges, B., 89, 611, 900-902 Johnson, B. 337 Johnson, D., 1374, 1387, 1391, 1395 Johnson, E.C., 206 Johnson, C.W. 538 Johnson-Laird, P. N., 651-653 Jonas, H., 886, 1105, 1107, 1132, 1150, 1182-1183 Jones, J. C., 415, 432-433, 435-437, 439, 444, 446, 448-449 Jones, M.E. 102 Jones, W., 28 Joss, S., 176, 1104, 1110, 1116, 1126 judgement/judgment, 974 aesthetic, 1043, 1056, 1230-1231, 1236 evaluative, 926, 934, 936, 940, 946, 1127 normative, 341, 927, 939, 946, 951, 1122, 1128, 1219 value, 1119 Justi, J. H. G. von, 648 justification built in, 702 epistemic, 941 moral, 955 K¨ onig, W., 42, 118, 134, 136, 158-159 Kahn, L. I., 1204, 1239-1242, 1249 Kahn, P. H. J., 1002, 1392, 1397 Kant, I., 586, 975, 1001, 1031, 1046, 1153, 1180, 1181, 1231, 1236 Kapp, E., 39 Kardon, J. B., 962 Karlsson, P., 1062

Kass, L., 44, 1162, 1170, 1389 Keeble, L., 1034 Keeley, B. L., 1370 Kelvin (William Thomson), 132, 649, 715-716, 813, 1010 Kierkegaard, S., 1159 kind artefact/artifact, 198-210, 930-931, 936, 938-939, 944 artefactual/artifactual, 198-210 biological, 201, 204 copied, 201-204 functional, 188, 930-931, 935-936, 938-939, 948 natural, 199-200, 207-209 kinematic similarity, 766, 772 King, B., 507 King, J., 1059 Kl¨ uver, L., 1136 Klein, U., 653 Kline, R., 30, 117 Kline, S. J., 95 Knapp, F. L., 161 know-how, 73, 155, 161, 163-166, 168171, 323-324, 328, 332-333, 335-336, 510, 528-529, 728, 730, 891-892 knowledge base, 68, 95, 98, 103, 105, 181, 235, 238-240, 260, 266267, 277, 1358, 1371 knowledge economy, 351 knowledge engineering, 1358-1359, 1364, 1371-1372 knowledge management, 330, 335-336, 357, 1141 knowledge representation, 273, 278 knowledge background, 597, 717, 1369 design, 67, 336, 338, 468, 528529 engineering, 93, 96, 106-107, 112, 235, 238, 260, 267, 317, 356, 369, 634, 760-761, 763, 1053, 1174

Index

implicit, 166-167, 285, 324, 333, 335 personal, 167, 353 prescriptive,337-42, 347-48 scientific, 3, 7, 25, 66, 76-79, 85, 95, 162, 174, 321, 326-31, 335-348, 493, 516, 1155, 1163, 1170 tacit, 11, 73, 166-69, 330-37, 351370, 867, 1371 technical/technological 6-9, 11, 13, 25, 33, 40, 77-78, 88, 98, 106, 118, 155, 158-61, 164-71, 174, 179, 181, 309-348, 368, 416, 513, 529, 892, 839, 913, 1031, 1151, 1156, 1160, 1179, 1216, 1250 knowledge society, 1114, 1141 Knuth, D., 146, 1090, 1354, 1357, 1375 Koen, B., 7, 56, 181 Koepsell, D., 1383 Koetter, F., 1238 Kohler, J., 1322-1324 Kolnai, A., 379, 396-398 Koolhaas, R., 763, 1202, 1204, 12151216, 1233, 1242-1247 Kornblith, H., 199-201 Korsgaard, C., 975, 978, 1001, 1181 Koyr, A., 310 Kranakis, E., 97, 128-129, 1328-1329 Krantz, D. H., 803, 808, 810, 840, 842, 847, 856 Kroes, P. A., 9, 13, 15, 67, 205, 207, 219, 290-291, 295, 316-317, 320, 330, 405, 496, 513, 518, 521, 526, 534, 537, 539, 544, 565, 567, 602, 605, 614, 703, 801, 865, 985, 1022, 1225 Krohn, W., 74, 77, 82, 128 Kroto, H., 149, 1047 Kruft, H.-W., 1207, 1210, 1236 Kuhn, T. S., 74-76, 118, 1262, 1357 kuhnean paradigm, 75-78, 82 Kummer, E. E., 654

1429

Kurzweil, R., 1171 Kutz, C., 231 Laat, P. de, 1384 Ladikas, M., 1113, 1120-1121, 1123 Lakoff, G., 196, 207, 355, 1062 Langhaar, H. L., 645, 784, 802, 808, 819 lanthanic disease, 1278 Larkham, P. J., 1042 Latour, B., 6-7, 68, 83-84, 87, 93, 96, 143, 343, 909, 1160, 13131314, 1317, 1360 Law, J., 1177 law, Boyle’s, 126, 784, 789, 791-792 Froude’s, 765, 791-794 model, 635, 759, 765, 767, 771772, 774, 776-779, 781, 784785, 790-793 patent, 53, 605, 1307, 1311, 1314, 1319-24, 1326-28 laws of nature, 40, 157, 167, 202, 354, 358, 435, 445, 863, 1326-1327 Laudan, R., 9, 71, 117 Laymon, R., 314-315, 320 Layton, E., 7, 40, 67, 71, 94, 117118, 131, 310, 361, 633, 892, 1171 Le Corbusier, 641, 1037-1038, 1040, 1042, 1064, 1200-1201, 1206, 1210, 1215, 1217, 1236-1238, 1245 Le´sniewski, S., 270 Leach, N., 1214, 1226 Leibniz, G. W., 124, 129, 133, 643644, 768 Leiss, W., 1157 Leitbild assessment, 1118-1119 Lelas, S., 15, 83-87 LeMessurier, W., 963-964, 969 Lenat, D., 277, 1371 Leslie, S. W., 138, 143, 147 Lethaby, W., 1200

1430

Marcel Scheele and Andreas Spahn

levelling-out effect, 1093 levels of expertise, 477-479 Levinson, J., 207 Levold, N., 96 Levy, D., 1286, 1279 Lewens, T., 214, 224 liberty, 1000, 1158, 1164-1167 Lida, S., 1222 life cycle analysis/assessment (lca), 1123, 1128, 1130, 1133-1134, 1139 lifeworld, 493, 597, 1159-1160, 1162, 1180, 1184 limiting case abstraction, 745, 752 limits of rationality, 407, 596-99 Lipman, A., 1223 Little, A., 260 Little, P., 375-376, 384-385, 389-390, 393, 395, 397, 399-400, 516, 523, 526, 528, 532 Llull, R., 179 Locke, J., 30-31, 1259-1261 Lodge, A., 635, 805-806, 810 logic first-order, 276-278, 686, 688, 1347 formal, 274, 437, 650, 1365 fuzzy, 1135 mathematical, 681 predicate, 273 logical inference, 438 logical malleability, 1391, 1395 logical quality, 286-88, 294-300 logical reasoning, 652, 720 Lorenz, E. N., 357, 645 Lovejoy, A. O., 1206 Lowe, E., 280 Luce, R. D., 392, 785, 794-795, 847, 1072 Lucivero, F., 1389 Luegenbiehl, H., 957 Lyotard, J.F., 31, 1377-1378, 1385 Macarthur, J., 1233 Mach number, 801, 814 Machamer, P., 239

Machiavelli, N., 1151 machine abstract, 1357 molecular, 1049 Turing, 1346-50, 1362 virtual, 1357 MacLennan, B. J., 1031, 1055, 1058 Magnus, H. G., 654, 1278 Mahoney, M. S., 1354 maieutic ends, 384-385 Malcolm, D. G., 645 malfunction, 217-218, 223-224, 229, 231, 295-297, 303-305, 359, 376, 406, 520, 925-927, 931, 934-940, 942, 945, 947-948, 1356 malfunctioning artefact/artifact, 297, 305, 494, 936 Mallgrave, H.F., 1234-1235 management, information, 1379 knowledge, 330, 335-336, 357, 1141 Mandelbaum, M., 1208 Maner, W., 1391, 1395 Manning, R., 1372 Marcuse, H., 1105, 1157 Maritain, J., 33, 1151 market pull, 481-482, 521 market push, 481-483 Martini, F di Giorgio, 123, 643 Marx, K., 44, 51-52, 161, 230, 415, 418, 429, 1033, 1105, 11551156, 1260-1261 Marx, L., 49 Masolo, C., 276, 285-287, 289, 296, 301, 303-304 Massey, D., 1214 material culture, 213-214, 216-217, 230231, 756, 1178 material flow, 176, 261, 264, 554-555, 1121, 1123 mathematical modelling, 165, 430, 1045, 1051-1052 mathematical models, 12, 436, 528,

Index

634, 638, 650, 653-659, 665691, 701, 714, 750, 771, 803, 1051, 1053, 1056 mathematical ratio, 1008 mathematical truth, 789 Mathiassen, L., 104 Matthews, S., 1390 Maun, C., 1389 Maxwell, J. C., 40, 130, 132, 648, 650, 715, 804-805, 809 May, W. F., 954 Mazlish, B., 1375, 1389 Mazouz, S., 1231 McAdams, D., 260, 547, 552 McCarter, R., 1239 McDermott, D., 1371 McKirahan, R., 800 McLaughlin, B.P., 294, 534, 1357 McLaughlin, P., 220-222, 228, 230 McLuhan, M., 1385 Meadows, D., 1105 Meagher, S. M., 1199 means constitutive, 383, 386 instrumental, 383 necessary, 388-389, 527-528, 1388 optimal, 52, 388-390, 397 sufficient, 376, 388-391, 1366, 1379 means-end rationality, 7, 367, 527, 579, 581-82, 1010-11 means-end reasoning, 11, 377-380, 384, 386, 391, 395, 400, 519, 526528, 579, 581-582 means of production, 982, 1156-1157 measurable parameters, 704, 710 measuring instrument, 289, 644, 835 measuring technique, 165, 178 mechanical discipline, 1015 mechanical efficiency, 1008, 1023 mechanical engineering, 1008-1009, 1027 mechanical philosophy, 124 mechanical theory of heat, 132, 714715 mechanical work, 329, 712, 715, 717,

1431

722, 1010 mechanics, 1034 applied, 1016 classical/Newtonian, 126, 131, 133, 188, 327, 329, 347, 790-791, 793-796, 806, 809, 1059 quantum, 72-73, 144, 147-148, 358, 814, 858, 1195 structural, 675-677 mechanism, 721 biological, 1266 environmental, 1276 feedback, 1062, 1077 social, 470, 623 mechanistic explanation, 239, 269 mechanization, 126, 1292 medical ethics, 1103, 1169, 1227, 1275, 1283, 1286, 1396 medical diagnosis, 1170, 1277 medical technology, 12, 1275-1296 medicalization, 1183, 1197, 1296 Meijers, A., 1, 9, 13, 15, 290-291, 295, 304, 359, 496, 1106 Meinecke, F., 1208 Meinel,C., 653 mental attitude, 274, 291-292, 303 mental model, 651-653, 727, 737, 746747, 749-751, 756 mental process, 576, 1362 mereology, 13, 235-269 Merleau-Ponty, M., 1160, 1213, 1363 Merricks, T., 192, 196 Merton, R.K., 54, 336, 1013 method, analytic and synthetic, 405-407 evaluation and selection, 165 method of weighted objectives, 993994 methodological individualism, 1344 methodological requirement, 174-175, 1134 methodology computer science, 105 design, 240, 257, 408-411, 414,

1432

Marcel Scheele and Andreas Spahn

421, 426, 428, 431-433, 435438, 445-448, 450-451, 459, 461, 501-502, 516, 565, 588 scientific, 56, 71, 375, 646, 648, 784, 792, 1126, 1128, 1131 methods of design, 165, 415, 435, 461 methods of research, 165 methods of implementation, 165 Meyer, H., 418, 1036, 1224, 1235-1236 Mikellides, B., 1223 Mill, J. S., 31, 51, 1153, 1167 Miller, C.W., 595 Miller, C.O., 1074 Miller, D., 230 Miller, F.G., 1287 Miller, S., 1050 Millgram, E., 377-381, 391, 393, 395 Millikan, R., 204, 223-224, 228, 230 mimesis, 1206 mind-dependence, 193-195, 198 Minsky, M., 1346 Mitcham, C., 3-5, 27, 33, 39, 55, 87, 155-156, 163, 187, 313, 352, 410, 516, 885, 888, 892-893, 896, 1011, 1013, 1108, 1119, 1126, 1147, 1151, 1166, 1168, 1174, 1183 modality, 191, 201, 421 modes of technology, 4-5 Mode 1/Mode 2 approach, 77, 885 model, analogue, 650, 667, 675, 781 Bohr, 651 data, 711, 1357-1358, 1360-1361 empirical, 69, 697, 828 epistemic value of, 695, 699, 703, 711, 755 ideal, 700 isomorphism of, 826-828 fictitious, 314 Froudes, 765, 778-779, 791-793 mathematical, 12, 436, 528, 634, 638, 650, 653-659, 665-691, 701, 714, 750, 771, 803, 1051,

1053, 1056 mental, 651-653, 727, 737, 746747, 749-751, 756 Newtons, 767, 774, 776, 790-791, 793 notion of, 656, 685, 723-724 phase, 393, 529 physical, 649, 655, 727, 729, 731, 736-737, 739-741, 747, 749751, 756, 763 pictorial, 674-675 Reynolds, 776-777 semantic, 634, 657-658, 667-669, 677, 684-689 scale, 11-12, 130 624, 633 644645, 650, 667, 675, 699, 701, 759-796 textual, 675-677 theoretical, 75, 86, 333, 650, 697, 723, 783-785 theory of, 656 model as epistemic tool, 699-703 model-based inference, 750-754 model-based reasoning, 722-723, 727757 model building, 72, 643 model class, 687-691 model construction, 703, 722-723, 752 model description, 704, 723-724 model for education, 653 model law, 635, 759, 765, 767, 771772, 774, 776-779, 781, 784785, 790-793 model system, 558, 634, 682, 700, 702, 723, 727-731, 733-734, 736751, 753-756 modelling, data, 470, 1341-1342, 1353, 13571361 internet aided, 174 explanatory, 672, 682 functional, 257, 259, 266, 268, 535, 665-691 mathematical, 165, 430, 1045, 1051-

Index

1052 modellus, 638, 641 modernism, 39, 1035-1041, 1064-1066, 1218, 1229, 1233 modernity, 85, 87, 1007, 1106, 1181, 1209, 1212-1213, 1215, 1247, 1389 modulus, 638-639, 641, 647 Mokyr, J., 55, 347-348 molecular, 1049 Moneo, R., 1242 monster, 655, 872, 1316-1319 Moor, J., 1342-1343, 1365, 1374, 1389, 1391-1392 Mooradian, N., 1383 Moore, A.D., 1280 Moore, G.E., 31, 950, 976, 1126 Moore, H., 1213-1214 moral agency, 957, 1374 moral behavior, 1153 moral considerations, 1006, 1007 moral ideals, 955, 1223 moral justification, 955 moral obligations, 942, 949, 978, 997, 1150 moral practice, 1148 moral responsibility, 943, 964, 1000, 1097, 1108,1372, 1374, 1390 moral rules, 955-956, 958, 960, 1226 moral supererogation, 997-999 moral value, 44, 885, 946-948, 960, 978, 981-984, 987-988, 992, 997-999, 1004, 1056, 1091, 1397 Morgan, M. S., 72, 610, 634, 657, 696, 699-700, 723, 783, 1361 Moritz, K. P., 1236 Morrison, M., 72, 634, 657, 682, 696, 699-700, 723, 783, 1361 Moses, R., 899-902, 916, 983 Mostowski, A., 657 motion control, 1015-1016 motive power, 705-707, 717-721 mould, 639, 641-642, 646

1433

Mullins, P., 1380 multi-level systems, 603 multi-stakeholder, 602, 615, 617, 623 multiple criteria analysis, 993, 1000 multiple realizability, 215, 218, 227228, 231, 528 multiple utilizability, 215-216, 218, 227228, 231 multiplicative approach, 281, 285, 293, 304-305 Mumford, L., 49-50, 611, 1024, 1033, 1161, 1068, 1227 Mumford, S., 294 Murphy, A.H., 1086 Murphy, C., 1360 Muschamp, H., 1216 Nagel, T., 1217 nanotechnology, 10, 14, 142, 147-149, 160-161, 1048-1049, 1053, 1108, 1138, 1142, 1197, 1311, 1313, 1318 Nash, J., 656 natural kind, 199-200, 207-209 natural kind terms, 199-200, 206-209 natural object, 13, 193-95, 341, 407, 537, 923-25, 930, 938, 94243, 946, 949, 1151 natural science, 1, 11-12, 102, 110, 180, 192, 381, 413, 437, 509, 801, 1149, 1313 in relation to design, 421, 423, 430, 523 in relation to engineering sciences/ technical sciences, 42, 15660, 162-63, 166, 345, 633, 863 in relation to other sciences, 23, 37, 43, 47, 50, 96, 98, 191, 194, 208-10, 510, 802, 1130, 1353 in relation to technology, 24, 42, 65-90, 94-95, 162, 187, 316, 318, 1275, 1282

1434

Marcel Scheele and Andreas Spahn

nature, 3, 34, 41, 84, 118, 121-132, 362, 534, 943, 946, 1016, 1046, 1051-53, 1149-1153, 1169, 1203, 1257, 1306, 1314-1333 engineering and ∼, 41-42, 121 human, 56, 417, 954, 1064, 1107, 1142, 1182-1183, 1259, 1288, 1388-1389 laws of ∼, 40, 157, 167, 202, 354, 358, 435, 445, 863, 1326-1327 of engineering (sciences), 9, 10, 155, 156-160, 162-163, 352, 356 of technology, 39, 181, 521 of technical artefacts/artifacts 9, 13, 188, 204-210, 282-306, 414, 423, 496-498,509, 887, 902 of technical/technological knowledge 6-8, 188, 309-350 in relation to technology, 52, 58, 85, 155, 1108, 1149-1153, 1163, 1166, 1169, 1219, 1221, 1224, 1257-1260, 1267, 1304, 13141333 Navier-Stokes equation, 791 Neander, K., 221-222, 224, 228-229, 248, 376 necessary means, 388-389, 527-528, 1388 Nelson, D., 1026 Nelson, J., 199, 201 Nelson, R.R., 356, 358, 365 Nersessian, N. J., 8, 634, 652-653, 676, 702-703, 728, 751, 754, 757 Neumann, J. von, 146, 645, 655, 1347 neural network, 741, 1354, 1368-1370 neuron communication, 739-740, 742, 745, 748, 753 neutrality thesis, 53, 865, 883, 900, 947, 979-980, 1177 Newcomen steam engine, 132, 532533, 705-706 Newcomen, T., 706

Newell, A., 146, 652, 1290, 1354, 13651367 Newman, J.H., 45 Newman, O., 1223 Newman, W., 124 Newtonian mechanics, 126, 131, 133, 188, 327, 329, 347, 790-791, 793-796, 806, 809, 1059 Nietzsche, F., 28, 1159, 1209-1211, 1238, 1246-1250 Nissenbaum, H., 1381, 1384, 1388, 1394, 1397 non-logical constant, 656-658, 678 Norberg-Schulz, C., 1213, 1215 norm, 887, 893-896, 912, 916-917 norm of efficiency, 531, 588, 916 normal configuration, 364 normal distribution, 857 Norman, D. A., 108, 652, 1058, 1364 normative evaluation, 1123, 1127 normative fact, 928, 931-939 normative judgement, 341, 927, 939, 946, 951, 1122, 1128, 1219 normative requirement, 897, 932-933, 949 normativity, 341, 893-898, 923, 927930, 1122 contingent, 898 epistemic, 940 inherent, 12, 884, 886, 896-898, 903, 908, 910, 915 practice / practical, 894, 916 social, 898, 902 normativity of artifacts/artefacts, 8, 520, 883, 896, 923 normativity of technology, 883, 888, 896-898, 902, 909, 1120-1127 notational framework, 457, 459, 467470, 476-477, 484-485 Nowotny, H., 93, 147 Nozick, R., 574-576, 1093-1094, 1158, 1164 NP-problem, 1350 nuclear energy, 145, 173, 494, 902,

Index

905-907, 916, 1001, 1069-1070, 1109, 1155 risk of nuclear energy, 1069-1070, 1097 nuclear fusion, 76-77, 82 nuclear war/weapons, 142-144, 1093, 1163, 1168,1196 nuclear waste, 905, 916, 1080, 1082 numerical evaluation (matrices), 390392, 397 numerical representation, 825-827, 836, 839, 841, 852 N´ un ˜ez, R. E., 1062 object, natural, 13, 193-195, 341, 407, 537, 651, 923-926, 930, 938, 942, 943, 946, 948, 1151 physical object, 13, 29, 235, 281, 286-299, 301, 317, 525, 534535, 543, 647-648, 681, 751, 771, 787, 828, 833, 844, 847, 852, 979, 981, 1148, 1175, 1351, 1382 object world, 570, 594-599 objective function, 558, 608-609 objective probability, 1070, 1086 objective risk, 1071, 1083 objectivity, 397, 698, 1129, 1226, 1245 observable, 470, 572, 609, 679-680, 686, 696, 704, 710, 721, 724, 799, 1064 Ogburn, W. F., 54, 1181 Olschki, L., 654 Oltramari, A., 292 Ong, W., 1183 Onsager, L., 1052 ontological reproduction, 1382 ontology, 122, 188, 192, 202, 204, 273, 275-281, 363, 666, 1205, 1357 analytic, 1 applied, 7-8, 11, 273 artefact/artefact, 10, 13, 188-189, 202, 204, 208, 213, 282-306,

1435

1188 basic formal (BFO), 275-276, 280 descriptive ∼ for Linguistic and Cognitive Engineering (DOLCE), 276, 285-291 formal, 188, 273-281 foundational, 274, 278, 282 general formal (GFO), 276 ∼ in relation to epistemology, 84, 187, 302 of science, 84 of social reality, 274-275, 281, 307 of virtual reality, 1196, 1345, 1353, 1381-1383 suggested upper merged (SUMO), 278, 283 technology and ∼, 84 Oost, E. van, 909-911 operational principle, 14, 323-324, 358359, 363-368, 586, 616, 931, 938 optimal design, 989-991, 994-995, 1023 optimal means, 52, 388-390, 397 optimality, 389-390, 551, 598 optimizing, 994-1004 ∼ functions of artefacts/ artifacts, 694-695 difficulties of, 995-996 orectic rationality, 577-578 Ormrod, S., 107-108 Ortega y Gasset, J., 5, 8, 1160 Osmond, H., 1223 Ostwald, W., 179 output/input relationship, 260, 522525, 553-556, 561, 567, 990, 1008, 1012-1014, 1016, 1161, 1368 Owen, R., 694 Pacioli, L. B. de, 654 Pahl, G., 235, 257-261, 266, 513, 516, 522, 529, 546, 554, 568, 985 Palacios, J., 784, 802, 807-808, 819 Palchinsky, P., 970

1436

Marcel Scheele and Andreas Spahn

Paley, W., 534 Pallasmaa, J., 1212-1213 Pankhurst, R. C., 802, 808, 819 Panofsky, E., 1200, 1206 Pappus of Alexandria, 179 Paquette, L., 1050 paradigm design, 536-539, 556, 565 kuhnean, 75-78, 82 neoclassical, 54 physics as ∼, 192, 329 paradigmatic examples, 329, 338, 364 parallelism, 754 parameter variation, 79, 130, 132, 140, 761 pareto-optimal, 508, 551, 598, 621 Pargetter, R., 222, 227 Parker, W., 1361 Parnas, D., 361, 1356, 1374-1375 participation, 105, 108, 1384, 1387 in design, 444, 447, 1157 in social decision processes, 435, 1104, 1106, 1111, 1113, 11161117, 1126, 1136, 1157, 11641166, 1168, 1187 logic ∼, 274, 289 Pascals wager, 1333 Pasteur, L., 1050 patent law, 53, 605, 1307, 1311, 1314, 1319-1324, 1326-1328 patentability, 1319-1324, 1327-1328 pattern diagram, 440, 443 Pavitt, K., 356-357, 361-362, 365 Pearce, D., 1389 Pedersen, S. A., 314-315 Pellegrino, E. D., 1169 Pepper, D., 1221 perdurant, 243-247, 258-259, 260, 266, 275, 280-281, 285-289 Perlman, M., 15, 213, 222, 534 personal knowledge, 167, 353 Persson, D., 1280, 1395 petrol engine, 167 Petroski, H., 7, 43, 203, 521, 990

Pettit, P., 998, 1384, 1390 phantom function, 217-218, 223, 227, 229-231 phase diagram, 529, 586-588 phase model, 393, 529 phenomenology, 1058, 1155, 1159, 1213, 1363 phenomenon, 556-567, 650, 701-704, 707-712, 722-724, 783-786, 816822 Phillips, A., 1042 Philips Natuurkundig Laboratorium, 490, 492, 494 philosophy of artificial intelligence, 1342, 1364, 1373 philosophy of biology, 13, 213, 238239, 268-269, 294, 1195, 1310, 1343-1344 philosophy of chemistry, 9 philosophy of cognitive science, 1389 philosophy of computation, 1343-1344 philosophy of computer science/computing, 1341-1392 philosophy of history, 1258 philosophy of information (technology), 9, 13, 1342, 1345, 1351, 1360 philosophy of (the) internet, 1345, 13761390 philosophy of new media, 1342, 13761390 philosophy of science, 1, 4, 5, 9-10, 15, 56, 79, 239, 319, 321, 327, 339, 347, 355, 368, 375, 427, 568, 575, 635, 649, 694696, 723, 799, 803-804, 1053, 1103, 1195, 1343-1344 Anglo-Saxon, 76 in relation to philosophy of technology, 28, 361, 1195 photoelectric lighting kit, 910-912, 916 photography , 354, 415, 648, 1213, 1245, 1376 clinical photography, 1277

Index

physical capacity, 248-249, 364 physical chemistry, 353, 1045, 1051 physical model, 555, 649, 655, 727756, 763 physical object, 13, 29, 235, 281, 286299, 301, 317, 525, 534-535, 543, 647-648, 681, 751, 771, 787, 828, 833, 844, 847, 852, 979, 981, 1148, 1175, 1351, 1382 physical quantity/quantities, 317, 679, 760, 772, 784-788, 792, 801820, 858 physical similarity, 769, 801-804, 814, 821-822 physical simulation, 729, 746, 751, 756 physical symbol system, 423-424, 13661367 physical system, 254, 522, 254, 605, 607, 681, 688, 733, 751, 753, 801-803, 814-821 physically similar systems, 802-803, 808, 817-820 Piaget, J., 651 Pianesi, F., 305 Pickering, A., 343, 1313 picturesque (theory of the ∼), 12301233, 1249 Pitt, J. C., 635, 865-866, 883, 900, 979 ucker, J., 654 Pl¨ plan, production, 530-532, 1109 typical, 1245-1246 use, 207, 225, 341-342, 364, 455, 514, 530-531, 567, 887, 912916, 934, 936, 981-986, 1176 planning, 42, 43, 154, 171-174, 398400, 427, 435-439, 616, 623, 625, 1025, 1048, 1105, 11141121, 1133, 1264 agent, 220, 514 environmental, 674 financial, 1371

1437

social, 410, 423, 439, 674, 1137 technical, 170-174, 398-400, 985, 1105 urban, 41, 409, 430-431, 442, 885, 1033-1044, 1064-1065, 1197, 1217, 1219, 1223, 1230, 1232 planning theory, 14, 377, 399-400 Plato, 3, 28, 31, 33-35, 44, 119, 123124, 187, 410, 441, 654, 799, 1046, 1149, 1199, 1205-1207, 1229 Platonism, 1046, 1051-1053, 1061, 1204, 1207, 1234, 1239, 1383 Platts, M. de Bretton, 380 Poel, I. van de, 16, 176, 407, 508, 919, 946-947, 949, 959, 965, 986, 988, 1024, 1108, 1119, 1122, 1128, 1130, 1133-1135, 1167 Poincar´e, H., 665 Polakov, W., 1020 Polanyi, M., 167, 330, 333-334, 353370 Polhem, C., 131, 179 political conflicts, 95 political constitution, 38, 53, 289, 605, 619-620, 1025 political economy, 51-52, 1156-1157, 1259-1261, 1309 political norms, 894 political science, 50, 103, 413, 606, 620-621 political theory, 58, 1161 politics, 439, 617, 619, 648, 871-872, 875-876, 1021-1022, 1025, 10331034, 1042-1044, 1065, 11031104, 1107-1116, 1121-1133, 1140-1141, 1147-1148, 11501152, 1154-1159, 1164, 1167, 1178-1181, 1202, 1227, 1232, 1250-1251, 1317, 1328-1331, 1386-1388 political aspects of technology, 5, 77, 138, 413, 898-903, 1008 political concerns about technol-

1438

Marcel Scheele and Andreas Spahn

ogy, 45 political economy, 51-52, 12591267, 1309 totalitarian, 45, 418 politics of artefacts/artifacts, 504, 887, 898-903, 907, 911 politics of cyberspace, 1341-1345, 13861388 polytechnical, 98-101, 104, 106, 110 Popper, K., 56, 584, 649 population, 744 of cells, 732 of humans, 120, 139, 1085-1086, 1117, 1129, 1136, 1245, 1260, 1262, 1270, 1286, 1289, 1305, 1336, 1378 population growth, 1257, 1262-1263, 1289 Port, R. F., 1370 Posener, J., 1222 positivism, 1, 79, 353, 421, 429, 1114, 1238 Posner, M. I., 356 Posner, R. A., 1026 post-metaphysical, 1105, 1204, 1215 Poster, M., 1377-1378 postmodernism, 78, 87, 1204, 1216, 1309-1312, 1307, 1345, 13771380, 1385, 1388, 1390 poststructuralism, 1345, 1380, 1390 power, electric, 38, 257, 267, 537-538, 550, 556-557, 673, 889, 891, 910, 1140 motive, 705-707, 717-721 nuclear, 145, 173, 494, 902, 905907, 916, 1001, 1069-1070, 1109, 1155 political, 138, 619, 1152 solar, 262, 900, 910 wind, 109-110 Pr´evot, L., 276-278 practical goal, 157-160, 377, 1315 practical inference, 527-528, 909

practical rationality, 377, 519, 531, 565, 573-578, 583, 930, 951, 976, 1076 practical reasoning, 1188-1189, 340, 367, 375-401, 530, 569, 575, 583, 795, 976, 978 practical syllogism, 378, 386-389 practice, best, 1173 design, 11, 13, 42, 325, 368, 407, 412, 414, 416, 420, 426, 430, 443-444, 455-485, 509, 514517, 525-528, 565-579, 584589, 593, 599, 913, 959 engineering, 2, 8, 12, 99, 159, 188, 242, 255, 313, 352, 373, 377, 379-380, 393, 397, 406, 516, 522, 525-529, 532-535, 566, 574, 584, 634-635, 757, 762, 775, 779, 886, 956, 960970, 1077, 1120, 1171 moral, 1148, 1153, 1168 professional, 42, 104, 334, 451, 594, 958, 1164, 1168, 1173 scientific, 72-75, 81, 86, 88, 375, 568, 694-700, 703, 724, 795, 804 social, 6, 158-159, 174, 324, 757, 893, 927, 1089, 1097 technical/technological, 24, 40, 66, 155, 158, 166-169, 173-174, 311, 313-315, 330, 887-891, 898, 911, 916, 926, 1015, 1069, 1169-1170, 1196 , 1267 pragmatism, 32, 418, 429-430, 11541155, 1172 pre-modern ethics, 1148-1151 precaution, 1073, 1107-1108 precautionary principle, 1075-1076, 1103, 1113, 1132-1133, 1180, 13321333 predicate logic, 273 Presbyter, T., 1150 prescriptive knowledge, 337-342, 347-

Index

348 Preston, B., 13, 188, 204-205, 215, 224-230, 235, 247-248, 322, 339, 914, 1367 preventive technology, 1276 Price, U., 1233 principle of dimensional homogeneity, 816-820 primacy of practice, 83, 158-159, 1195 primary prevention, 1077 privacy, 885, 960, 976, 983, 987, 1157, 1164-1167, 1182, 1187, 1197, 1280, 1293-1294, 1324, 1342, 1387, 1390-1397 probability, 667, 681, 828, 834, 844, 847, 1077-1098, 1129, 1131, 1333, 1372 objective, 1070, 1086 of error, 825, 849 of truth, 681 of unwanted events/risk, 10701073, 1077, 1081-1086, 1092, 1098 qualitative ∼ distribution, 849859 subjective, 849 probability estimation, 1086-1090 problem-solving, 95-96, 107, 111, 164, 165, 355, 364, 517, 584, 645, 727, 751-756, 1113 rational, 12, 165, 461-464, 469, 565, 569, 596 technical, 164, 175 problems of legitimisation, 1109-1110, 1128 procedural rationality, 7, 578-579, 583, 588 process, biological, 749, 1301 cognitive, 333, 412, 729, 10321033, 1056, 1362-1263 control, 161, 274, 605 decision, 439, 589, 1065 design, 170-173, 257, 295, 361-

1439

366, 368, 375-378, 385, 389390, 393-395, 398, 400, 424425, 437-439, 456-460, 463471, 473-477, 479-480, 501502, 510, 518, 523, 526, 529530, 565-572, 586-589, 601602, 912, 977, 981, 985-989, 999, 1002, 1022-1024, 1063 engineering, 163, 392, 401, 547, 556 experimental, 80, 86-87, 366 historical, 568, 754, 1264, 1312 industrial, 51, 82, 510, 1051-1052, 1303-1304 innovation, 1119 learning, 489-490, 1118 mental, 576, 1362 physical, 258, 635, 693, 722, 782, 1354 political, 439, 1034, 1251, 13861387 production, 156, 337, 352, 361, 494, 914, 985, 1260, 1304 rational, 7, 565, 585, 1010, 1230, 1363 social, 7, 231, 435, 447, 466, 469, 593-599, 622, 625, 899, 904, 1054, 1118, 1156, 1195, 1377 technological, 81, 88, 156, 161, 1175, 1186 process modelling, 470 process regulation, 165 product design, 586, 884, 1002, 1031 product patents, 1322-1328 profession-based appropriation, 93, 102112, profession, 96, 380, 478, 894, 957, 968, 1041, 1168-1172, 1220, 1227, 1392 engineering, 38, 41-43, 93, 97-98, 102-103, 495, 953, 956, 1119 profession code/codes, 1147, 1172 professional discretion, 961 professional ethics, 957, 1147, 1169,

1440

Marcel Scheele and Andreas Spahn

1171-1172, 1175, 1228-1229, 1360, 1392-1393 professional responsibility, 953, 1168, 1345 professional standards, 2, 13, 953-971, 1168 programming, computer, 105, 609, 1060-1063, 1351-1357 genetic (or DNA), 1310 programming language, 146, 680, 683, 1054, 1061, 1341, 1352-1358 proper function, 201-205, 221-232, 248, 446, 496, 498, 693-694, 707, 722, 889, 938 propositional knowledge, 323, 333, 342, 347 proteomics, 1281 prototype, 165, 300, 361, 366, 414415, 424, 434, 505, 514, 566, 579-580, 587, 589, 597, 637, 644, 674, 765, 770-795, 802, 818, 821-822, 1133 design, 464, 478 novel, 224, 227, 231 prototype theory of concepts, 29, 33 psychological condition (of humans)/ factors, 230, 381, 443, 571, 828, 891, 896-897, 1099, 1219, 1280 psychological experience(s)/experiment, 417, 836 psychological explanations/theories, 333, 864, 683, 738, 861, 864, 1064 psychological phaenomenon/phaenomena, 276, 469, 828, 869, 976, 1222-1237, 1288 psychological process, 424, 474, 568 psychology, 30, 45, 100, 282, 331, 334, 355, 380, 423, 426, 451, 509, 648, 872, 1062-1063, 1068, 1170, 1223, 1234, 1362, 1365, 1384 public health, 953, 956

public values, 602, 616-624 Pufendorf, S. von, 1181 Pugin, A. W. N., 1202, 1208, 1229, 1237 Putnam, R., 1385 Pylyshyn, Z., 355, 1365, 1368-1369 Pythagoras, 802, 1231 qua problem, 200 quality, 175-176, 280, 474, 494, 499, 961-962, 964, 967-970, 1120 aesthetic, 1035, 1043-1044, 1063 logical, 286-288, 294-300 of design, 108-109, 266, 416, 455, 470 of knowledge, 167, 1177, 1371, 1378-1381 of life, 602, 953, 1043, 1057-1058, 1120, 1130, 1175, 1267, 1284, 1292, 1396 of technology, 125, 174, 361, 415, 961, 968 quality-adjusted life years (QALY), 1087 quality assurance, 165, 173, 178, 361 quality control, 164, 1133 quality function deployment (QFD), 499-511, 546-547 quantification, 1013-1017, 1050, 1073, 1115, 1124, 1128-1131 quantity equation, 635, 804-812 quantum mechanics, 72-73, 144, 147148, 358, 814, 858, 1195 Quine, W. V. O., 192-193, 198, 407, 786 Radder, H., 2, 5, 7, 10, 13-16, 24, 57, 67, 73, 78, 80, 82, 85, 86, 88, 95, 104, 117, 143, 160, 169, 173, 346, 883, 888, 893, 901, 903, 907-909, 912, 915, 926, 979, 983, 1177, 1314, 13221324, 1327 Radio Frequency Identifiers (RFID), 481, 1378

Index

Raiffa, H., 392, 1072 Railton, P., 379, 388 Rankine, W. J. M., 127-128, 130, 132134, 1016 Rao, A., 292 Rapoport, A., 1200, 1220 Rasmussen report, 1071 ratio, geometrical, 1013 golden, 1231 mathematical, 1008 ratio scale, 786-796, 810, 843, 977, 994 measurable, 592, 760, 787-796, 830-832 rational choice theory, 576-585, 589, 1344 rational evaluation, 572, 577, 586 rational problem solving, 12, 165, 461464, 469, 565, 569, 596 rational process, 7, 565, 585, 1010, 1230, 1363 rational reconstruction, 468, 470, 475477, 568-569, 588, 780, 955 rational supererogation, 998 rational use, 913-915 rationalism /rationalist, 449, 894, 976, 1153-1155, 1205, 1211, 1213, 1225, 1230-1231, 1356, 1360, 1363, 1367, 1369 rationality, bounded, 465-466, 578, 583, 598 broad, 573 calculative, 596-597, 1161 collective, 508 communicative, 583, 597 deliberative, 605, 618, 619 epistemic, 577 final, 583 instrumental, 341-342, 572, 577, 579, 581-585, 588-589, 593598, 1024, 1157 limits of, 407, 596-599 means-ends, 7, 367, 527, 579, 581-

1441

582, 1010-1011 narrow, 573 normative or descriptive, 569 orectic, 577-578 practical, 377, 519, 531, 565, 573575, 578, 583, 930, 951, 976, 1076 procedural, 7, 578-579, 583, 588 social, 597 substantive, 7, 578-579, 583 technical/technological, 565, 451, 1179 telical, 583 theoretical, 565, 578, 951 rationality in design, 529, 565-599 rationalization, 45, 914, 1017, 1022, 1024-1025, 1157 raven and swan experiment, 756 Ravetz, A., 73, 1039, 1103 Rawls, J., 1002, 1095, 1139, 1164, 1388 Ray, N., 1196, 1202, 1227, 1250 Rayleigh, Lord (JohnWilliam Strutt), 645 Raymond, E., 1377 Raz, J., 929, 941, 949-950, 977-978, 1003-1004 Rea, M., 201, 275, 293, 296 realism/realist, 33, 191, 197, 207-208, 318-320, 360, 787, 808, 811, 1242, 1247, 1251, 1322, 1358, 1360, 1382 reason, 410, 576-583, 929-951, 1025, 1157 artificial intelligence and ∼, 1365 critique of ∼, 407, 596-599, 1209, 1211, 1216, 1377 enlightenment (theory of), 1157 faith vs. ∼, 122 human (capacity of), 450, 576, 1150 instrumental, 341-342, 572, 577, 579, 581-585, 588-589, 593598, 1024, 1157

1442

Marcel Scheele and Andreas Spahn

means-end, 7, 367, 527, 579, 581582, 1010-1011 practical, 367, 583, 929 prima-facie, 1335 reasonable care, 965, 967, 969 reasoning, 375-401, 437-438 analytic and synthetic, 405 ∼ about ends, 582 deductive, 434, 1356 dialectical, 443 economic, 618 explanatory, 239 functional, 235, 238-240, 269, 527, 535, 1214 human, 389, 576, 1369, 1371, 1375 logical, 652, 720 means-ends, 11, 377-380, 384, 386, 391, 395, 400, 519, 526-528, 579, 581, 582 model-based, 8, 12, 633-634, 722723, 727-757, 781 monotonic, 469 of designers, 421, 437-438, 447, 461, 475, 519 of engineers, 375-401, 519 plausible, 434 practical, 340, 375-401, 530-31, 569, 575, 1000 practical ∼ in engineering (science), 375-401 practical ∼ /theoretical ∼ (relation), 188, 375 rational, 568-569, 575-576 scientific, 519, 569, 695, 705, 719, 724 similarity-based, 799-800, 803 surrogate, 698 syllogistic, 389-390 theoretical, 375, 391 reciprocity principle, 767 recombinant DNA technology, 150, 12641265, 1302-1303, 1306, 1312, 1330-1331 reconciled functional basis project

(RFB), 257, 259-267 recycling, 215-218, 227-231, 499, 558, 1221 Redtenbacher, F. J., 127-128, 162 redundancy, 198, 210, 1080-1081 reflective equilibrium, 1002 reflective modernity, 1106 regulation, 604,606, 622, 926, 964, 1092, 1117, 1140, 1196, 1267, 1318, 1330, 1345, 1386-1387, 13911392, 1395 legal, 616, 1138, 1305 political/social, 76, 606, 625, 905, 911, 1178-1180, 1330, 1342 price, 619 process, 165 self-regulation, 622 regulatory standard, 954, 961, 964965, 968 Reiss, J., 1265, 1362 relational database, 657-658 relational structure(s), 668, 794, 838 relational system, 668 reliable experience, 161 Renaissance, 35, 102, 123, 409, 414, 432, 646, 654, 802, 1015, 1064, 1147, 1199-1200, 1202-1204, 1206, 1208, 1380 Renn, O., 1116, 1126, 1139 Renssen, A. van, 7 representation, knowledge, 273, 278 indirect, 700 inferential conception of, 698 invariance theorem of, 825, 829 semantic account of, 697 representational means, 701-705, 711714, 721-724 representational force, 698 representational relationship, 696, 711 requirement(s), customer, 3, 11, 13, 407, 481, 489-511, 547-548, 555, 561, 566-567, 584, 587

Index

design, 257, 432, 474, 489, 553, 936, 985-988, 995-996, 9991000 functional, 405, 517, 566-567, 570572, 584, 587, 591, 612, 936, 944, 947, 985-989, 990 methodological, 174-175, 1134 normative, 897, 932-933, 949 technological, 899 user, 107, 112, 393 responsibility, 176, 968, 1000, 1097, 1108, 1132, 1139, 1167-1179, 1283, 1286, 1293, 1356 blame and ∼, 862, 1176 human, 176, 414 moral, 943, 964, 1000, 1097, 1108, 1372, 1374, 1390 of architects, 426, 965 of designers, 428, 1117 of engineers, 513, 965, 967, 1069, 1163-1165, 1167-1168, 11731174 of (engineering) science, 174-175, 1312 professional, 953-954, 958, 961, 1168, 1345, 1393, 1395 role, 1150 social, 175, 415, 428, 1128, 1312 Reuleaux, F., 131, 179, 181, 1016 reverse engineering, 235, 237-239, 267, 524, 543, 547-552, 554-556, 559, 874 Reynolds model, 776-777 Reynolds number, 645, 769, 776-780, 793, 795, 801, 814 Reynolds, J., 1231 Reynolds, O., 645, 766, 779-780 Reynolds, T.S., 130, 1022 Riabouchinsky, 802 Rices theorem, 1348 Rice, H. G., 1348 Richardson, H., 381-382, 386-389, 396397, 583, 1249 Ridder, J. de, 10, 237, 239, 342, 863

1443

Rifkin, J., 1303, 1306, 1389 rights-based moral theories, 1094 Rio Declaration, 1075, 1332 Rip, A., 175, 886, 1117 risk, 173, 176, 1069-1100, 1165, 1184 acceptable, 1085,1108-1109, 1265 communication of, 1083, 1104 decision under uncertainty or ∼, 580-581, 1070-1072, 1111 definition of, 1069-1071 distribution of ∼ and benefits, 960, 969, 1089, 1138 ethics of, 1090-1100 objective, 1071, 1083 subjective, 1071, 1083 risk of nuclear energy, 1069-1070, 1097 risk analysis, 176, 579, 907, 1080, 10821090, 1121, 1130 risk assessment, 176, 1085, 1121, 1124, 1131-1133, 1139 risk communication, 1083, 1104 risk perception, 1083 Risk Society, 1114 risk-benefit analysis, 1082, 1087-1090, 1154, 1161 Rittel, H., 170, 430, 432, 435-439, 444, 446-450, 517 Roberts, J., 1231-1232 Robinson, A., 656-657 Robinson, M., 102 Robinson, R., 31 robot, 290, 728, 743, 745-749, 1104, 1291, 1295, 1370, 1372-1375 robotics, 164, 1311, 1354, 1364, 1396 Rogers, G., 10, 352, 356 Rogers, H., 48 Rogers, K., 83 Rooksby, E., 1388 Root-Bernstein, R., 1052 Ropohl, G., 42, 171, 181, 321-324, 326, 601, 612 Rosenbaum, T., 1026 Rosenberg, N., 95, 129, 330, 353, 1073 Rousseau, J.-J., 1153

1444

Marcel Scheele and Andreas Spahn

Rowe, C., 1238 Ruskin, J., 44, 415, 1035-1037, 1202, 1229, 1231, 1327 Russell, B., 30, 1163 Russell, C., 646 Russell, J., 102, 134, 785 Russell, P., 534 Rutherford, E., 1050 Rykwert, J., 1240 Ryle, G., 31, 332-334, 347-348, 528 Rzchtarikova, J., 1223 Saalman, H., 642 Sabatier, P. A., 606 safety, 12, 13, 46, 81, 156, 158, 176, 245, 257, 324, 396, 469, 474, 481, 497, 518-519, 597, 618619, 795, 884-885, 905, 916917, 953, 956, 958-961, 964, 967-971, 981, 983, 987, 989, 992-993, 996, 1000-1001, 1031, 1069, 1074-1099, 1120, 1131, 1159, 1164-1168, 1172-1173, 1178-1179, 1187, 1195, 1197, 1200, 1219, 1221, 1276, 1334, 1356, 1393 inherent, 1077-1078 safety barrier, 1077, 1080-1082 safety engineering, 1076-1077, 10821083, 1090, 1195 safety factor, 324, 1077, 1079-1080 safety margin, 1079 safety standards, 961, 967, 1131 Sage, A. P., 516, 536, 612 Sagoff, M., 1265, 1316 Saha, S., 1276, 1283 Saint-Simon, H. de, 1156 Salmon, J., 53 Salmon, W., 863 Salter, A., 331 Sanders, J. W., 1374, 1395-1396 Sandin, P., 1075-1076 Sarewitz, D., 1117, 1167 Sarkar, S., 1310

Sarlemijn, A., 502-505 Sartoris, A., 1037 satisficing (design), 389, 425, 578, 583, 598, 996-1000, 1003-1004 Sawhney, H., 611-612 Sawyer, R. K., 614 scaffolding, 701, 735, 752, 762 scale factor, 767-794, 818, 821 acceleration, 773 frictional force, 776 inertial force, 773-774 linear, 772-773, 793 pressure force, 774, 776 time, 773, 778 velocity, 773 scale model, 11-12, 130, 633, 644-645, 650, 675, 699, 701, 759-796 scales of measurement, 830-831 scaling laws dynamic, 791 Froudes, 793, 796 geometrical, 789-790 Scanlon, T. M., 929, 946-947, 949950 Scerri, E. R., 358 Schacter, D.L., 355 Sch¨ afer, W., 74 Scharff, R.C., 5 Schatzberg, E., 7, 30, 888, 1147 Schaub, J. H., 174, 176 Scheele, M., 225-56, 228-231, 293, 301 Scheler, M., 1160 Scheutz, M., 1347 Schierbeek, A., 648 Schiffer, M. B., 215, 229-230 Schildt, G., 1250 Schiller, F., 1237 Schl¨ ozer, A.L., 648 Schmidtz, D., 379, 383-385, 998 Schmitter, P. C., 622 Schneider, J. 162 Schneider, V. 612-613 Scholderer, V., 641 Scholfield, P. H., 1231

Index

Schomberg, R. von, 1103, 1108, 1132 Sch¨ on, D. A., 1201 Schopenhauer, A., 1237 Schot, J. W., 6, 508, 886, 1117, 1167 Schr¨ odinger, E., 150, 1306 Schummer, J., 884, 946, 1033, 10441045, 1048-1050, 1197 Schwartz, S., 199-200 Schweninger, E., 1295 science see also: natural science see also: philosophy of science applied, 1, 29, 40, 48, 54, 66, 6974, 77, 127, 150, 162, 188, 309-312, 340, 344-345, 347348, 413, 418, 516, 518-519, 802, 863, 1082, 1344, 1366 cognitive, 239, 356, 368, 640, 650651, 653, 1142, 1351, 1362, 1365-1366, 1370 computer, 68, 145-147, 434, 688, 1342, 1345, 1352-1364, 1366, 1372, 1391-1392 political, 50, 103, 413, 606, 62021 social, 1, 4, 6, 11, 13, 23-25, 27, 37, 43-44, 47, 50-58, 65, 93112, 160, 205, 209, 339, 351, 355, 357-358, 369, 421, 509510, 601, 610, 650, 831, 1041, 1082, 1129, 1154, 1166 techno∼, 25, 83, 119, 142-149, 343-345, 1044, 1147, 1160, 1311-1318 science and modernity, 85 Science and Technology Studies (STS), 2, 6, 8, 343, 346, 611, 695, 900, 1113 science-based engineering, 4-5, 14 science of the artificial, 146, 367, 465, 520 science-as-technology thesis, 84-87 science-based industry, 118, 134-35, 137

1445

science-technology relationship, 25, 53, 65-91 see also: natural science, in relation to technology sciences of making, 157 scientific evidence, 218, 1076 scientific explanation, 2, 10, 635, 638, 691, 861-865 scientific knowledge, 3, 7, 25, 66, 7679, 85, 95, 162, 174, 321, 326-331, 335-348, 493, 516, 1155, 1163, 1170 application of ∼, 160, 311, 347, 1282 in relation to technical knowledge, 188, 225, 311-313, 322, 326348, 364, 368, 634 in relation to design, 361, 427430, 528, 985 new ∼, increase of ∼, 107, 703 social ∼, social implication of ∼, 138, 1103, 1133, 1141, 1195 scientific management, 1017, 1021 scientific methods/methodology, 56, 71, 375, 646, 648, 784, 792, 1126, 1128, 1131 scientific practice, 72-75, 81, 86, 88, 375, 568, 694-695, 697, 699700, 703, 724, 795, 804, 811, 894 scientific realism, 318 Sclove, R., 1157, 1168 Scott, D., 843 Scott, G., 1202 Scott-Brown, D., 1218 Screpanti, E., 1013 script theory of design, 887 script of technological artefacts/artifacts, 909-916, 1177 Scruton, R., 985, 1200, 1230 Searle, G. R., 1017 Searle, J. R., 195, 219-220, 228-231, 274, 290, 355, 358-360, 578, 584, 1017, 1352, 1365, 1367,

1446

Marcel Scheele and Andreas Spahn

1383 Sebestik, J., 36 secondary prevention, 1077-78 security, 171, 173, 178, 481, 618-621, 624, 1069, 1074-1078, 1106, 1179, 1186, 1388, 1390-1394 Sedov, L. I., 808 Segerberg, K. , 527, 575 Seligman, E. R. A., 53 semantic model, 643, 657-658, 667669, 684-689 semantic theory, 657, 676, 696-697 Senker, J., 331, 334 sensitivity analyses, 1135 Shagrir, O., 1347, 1349, 1398 Shakespeare, W., 124, 639 Shapiro, S., 961, 963, 968, 1365 Shenk, D., 1379 Shew, A., 862 Shinn, T., 89, 97, 127 Shipley, J., 638 Shrader-Frechette, K., 992, 1131, 1165 Shute, G., 1357 SI system, 786, 808-14, 820 side effects, 175, 476, 544, 982-84, 11031108, 1111-118, 1125, 1135, 1282 similar systems, 801-803, 808, 817820 similarity Froudes law of ∼, 765, 791-794 geometric, 772, 789-790, 800-801, 803, 821 kinematic, 766, 772 physical, 769, 801-804, 814, 821822 Simon, G., 8 Simon, H. A., 146, 170, 181, 337, 352, 367, 381, 389, 392, 418, 420427, 430, 433, 437, 451, 462466, 517, 519-523, 531, 536, 565, 572, 577-578, 582-83, 652, 701, 995-998, 1354-1355, 1365-1367

Simons, P., 239, 252-256 simulation, 164-165, 174, 347, 466, 549-550, 562, 635-636, 673, 676, 729, 731-734, 737, 745, 751-752, 754, 756, 1121, 1365, 1382-1383 computer, 72, 170, 347, 422-425, 637-638, 644-645, 652, 705, 743, 746-747, 755, 1182, 1196, 1341, 1353, 1361-1362, 1370, 1382-1383 Singer, C. J., 5, 55 Sinsheimer, R. L., 1301 Skempton, A. W., 1015 Slokarnik, M., 802 Slote, M., 998-999 Slovic, P., 1126, 1132 Smart, N., 57 Smeaton, J., 41, 79, 130, 132-134, 784, 1015, 1022, 1171 Smith, A., 1156, 1259 Smith, B., 274 Smith, C., 97, 132, 1016, 1019 Smith, M., 381 Smith, R. J., 41-42, 51 Smits, M., 1317 Smits, R., 1110-1111, 1119 Snow, C. P., 510 Sober, E., 251 social artefact/artifact(s), 275, 291292, 301-303, 1164 social construction, 6, 1117 social constructivism, 865, 1113 social epistemology, 1381, 509 social evaluation, 1393 social explanation, 7, 861, 864-865 social forensics, 874-876 social infrastructure, 537 social mechanism, 470, 623 social networking, 1376, 1384, 1396 social philosophy, 429, 1344, 1370, 1384 social planners, 603, 616-617, 623 social process, 7, 231, 435, 447, 466, 469, 593-599, 622, 625, 899,

Index

904, 1054, 1118, 1156, 1195, 1377 social reality, 219, 274-275, 281, 302 social responsibility, 175, 415, 428, 1128, 1312 social sciences, 1, 4, 6, 11, 13, 2325, 27, 37, 43-44, 47, 50-58, 65, 93-112, 160, 205, 209, 339, 351, 355, 357-358, 369, 421, 509-510, 601, 610, 650, 831, 1041, 1082, 1129, 1154, 1166 social theory, 1157, 1345 Society for Philosophy and Technology (SPT), 8, 10 socio-technical infrastructure, 612, 617618, 626, 861, 1261 socio-technical system, 12, 407, 537538, 593, 601-626, 893, 944, 987 socio-economy, 336, 543-547, 555, 562, 614, 617, 1178, 1330 socio-technical arrangement, 607, 620, 625-626 sociology, 6, 50, 52-54, 76, 97, 106, 163, 222, 337, 343, 362, 418, 435, 457, 509, 610, 945, 1245 software engineering, 103-104, 321, 385, 646, 665, 885, 1033, 10541065, 1341, 1352-1360, 1397 Sokal, A., 1314 solar energy/power, 262, 900, 910 Solli, J., 109 Sombart, W., 52-54 Sorabji, R., 925 Srensen, K. H., 14, 24, 65, 71, 93-94, 96-100, 105-106, 108, 111, 673, 905 Soulsby, B.H., 457 Sowa, J., 278 specification, design, 107, 566-567, 570, 591, 1023 technical, 11,13, 407, 489-511, 985-

1447

986, 1267 Spender, J.C., 356 Spur, G., 16, 157, 161, 181 Srzednicki, J., 252 Stahovich, T.F., 549 stakeholders, 438-439, 458-459, 468, 470, 474-476, 602, 605-606, 610, 615, 617-618, 622-623, 884, 959, 986-987, 1034, 11211122, 1125, 1136-1137, 1139, 1220, 1230, 1398 Stanberry, R., 1293 standards, 953-971 aesthetic, 1057, 1231 engineering, 169, 884, 953, 958, 960-964, 1120, 1179 engineering education, 101 epistemic/rational, 568, 1054 ethical/normative, 884, 953-958, 1004, 1139, 1223 ISO, 14, 658, 959 national/international, 616, 620, 625, 802, 830, 959-960, 12671269 professional, 2-3, 13-14, 884, 953971, 1168 safety, 967, 969, 1079 scientific, 177, 326, 811 technical/technological, 111, 604, 958, 986, 996, 1173 Starr, P., 613 status, 39, 94, 97-98, 302, 918, 1031 social, 119, 226, 288, 604-605, 61011, 613, 861, 873-875, 1098 Staudenmaier, J., 7, 9, 309-310, 312, 316, 521, 634 steam engine/power, 9, 29, 36, 71, 126, 132-133, 139, 161-162, 316, 329, 532, 680, 705-707, 714, 717-721, 1018-1019, 1022, 1155, 1185 Stehr, N., 1107, 1141 Steiner, R., 1268, 1270 Stenning, K., 653, 676

1448

Marcel Scheele and Andreas Spahn

Sternberg, R. J., 334, 355 Sterrett, S. G., 635, 780-787, 795, 802, 818 Stikkelman, R. M., 607 Stocker, M., 974, 997, 1000 Stokes, G. G., 134 Stone, R., 257, 260-261, 264-265, 548, 550, 552, 554 Strawbridge, Z., 266 structural mechanics, 14, 1079, 1197 structural description, 243, 523-528, 532, 534, 567 Sturtevant, S., 643-644 Surez, M., 696-699 subjective probability, 849 subjective risk, 1071, 1083 substantive rationality, 7, 578-579, 583 Suchman, L.A., 95, 105, 108, 465, 1363 sufficient means, 376, 388-391, 1366, 1379 Suh, N.P., 586 Sullins, J. P., 1295, 1374 Sulston, J., 1321, 1326 Summerton, J., 900 Sunstein, C. R., 995, 1381, 1386-1387 Suppe, F. R., 658, 684, 688-689, 696697 Suppes, P., 635, 650, 657, 659, 686, 697, 835-837, 843, 847, 849850, 859 surveillance, 481, 1007, 1016, 1023, 1182, 1360, 1393 Sussman, H., 46 sustainability, 458, 519, 619, 621, 885, 916, 960, 970, 983, 987-989, 992-994, 1004, 1023, 1104, 1112, 1125, 1128, 1133-1334, 1139, 1141, 1164-1166, 1179, 1227-1228, 1263, 1270 sustainable development, 1107, 11131114, 1141, 1165 Swift, K. G., 554 Sycara, K., 556, 1372-1373

synthetic biology, 1163, 1306, 13081309, 1315, 1324, 1331 Sypher, W., 46 system adaptive, 538, 613-614 artefactual/artifactual, 889, 892 biological, 727, 749, 752, 869, 1307, 1308, 1316 classification, 36, 47, 131 computational, 1341-1343, 1346, 1351, 1366 computer, 103-106, 1362, 1364, 1366-1367, 1395, 1397-1398 coordinate, 678-679 expert, 273, 335, 1364, 1371-1372 functional system, 247, 887-888, 890, 892, 896-897, 903, 911, 918 hypothetical, 700, 711 information, 11, 103-105, 107, 273274, 424, 439, 1341-1342, 1353, 1359-1360 infrastructure, 537, 602-603, 612, 616, 618, 623, 625, 626 multi-level, 603 physical, 254, 522, 254, 605, 607, 681, 688, 733, 751, 753, 801803, 814-821 social, 119, 226, 288, 604-605, 610611, 613, 861, 873-875, 1098 socio-technical, 12, 407, 537-538, 593, 601-626, 893, 944, 987 symbol, 420-21, 423-424, 652, 13661367, 1369 technical system, 237, 243, 248250, 255-258, 266-267, 324, 522, 604 system theory, 612, 614 Szucs, E., 769 Szykman, S., 260 tacit inference, 354, 358-359, 362-363, 367-369 tacit knowledge, 11, 73, 166-169, 330-

Index

337, 351-370, 867, 1371 Tamburrini, G., 1389 Tarski, A., 252, 655-658, 668-669, 676, 681-682 taxonomy/taxonomies, 66, 276 biological, 29, 457 of DOLCE-based categories, 286, 289 of flows, 259, 261-262 of (basic) functions, 258-263 of technological/technical knowledge, 7, 311, 321-30 Taylor, B. N., 809-810, 812-814 Taylor, C., 369, 1176, 1223 Taylor, F. W., 990, 1017, 1020-1021 Taylor, J., 121 Taylor, J. R., 682 Taylor, N., 1031, 1033-1035, 1038, 1041, 1065 techn´e, 3, 33-35, 44-45, 413, 892, 1149, 1169, 1275 technical code, 14, 884, 958-961, 986, 996, 1179 technical component, 237, 537, 1260 technical concept, 316, 346, 1002, 1071, 1310 technical/technological evaluation, 1103, 1120 technical/technological knowledge, 69, 11, 13, 25, 33, 40, 77-78, 88, 98, 106, 118, 155, 158161, 164-171, 174, 179, 181, 309-348, 368, 416, 513, 529, 892, 839, 913, 1031, 1151, 1156, 1160, 1179, 1216, 1250 technical/technological infrastructure, 49, 959, 1261 technical/technological innovation, 174, 250, 323, 331, 335, 964, 982, 996, 1001-1002, 1106-1107, 1112, 1117, 1119, 1141, 1180, 1263 technical rule(s), 166-171, 363 technical specifications, 11,13, 407, 489-

1449

511, 985-86, 1267 technical standards, 111, 604, 958, 986, 996, 1173 technical system, 237, 243, 248-250, 255-258, 266-267, 324, 522, 604 technocracy, 1172 technological change, 51, 351, 617, 865, 1156, 1168, 1180, 1182 technological determinism, 499, 607, 611, 1161, 1296, 1397 technological explanation, 2, 10, 13, 342, 635, 861-878 technological imperative, 1167, 12951296 technological normativity, 883, 888, 893, 896-98, 902, 909, 916918, 1120-1127 technological rule, 13-14, 70, 340-342, 346-347 technology assistive, 1291-1292 biotechnology, 82, 149-150, 361, 885, 1142, 1196-1197, 1261, 1265-1266, 1280, 1301-1337 communication, 1141-1142, 13101311 compensatory, 1289, 1291-1292 diagnostic, 1276-1282 enabling, 480, 1276, 1290-1292 enhancing, 1276 general, 156, 163-164, 179-181 genetic, 1277, 1286 history of ∼, 5, 8, 52, 55, 163, 633 information, 9, 68, 1166-1167, 1197, 1310, 1341, 1359, 1388 maintenance of ∼, 4, 11, 14, 41, 65, 78, 164, 214, 217-218, 255, 331, 339-340, 366, 455, 536, 558, 586, 636, 670, 891, 897, 900, 905, 910, 1054, 1077, 1081-1082, 1133, 1355, 1392 medical, 12, 1091, 1275-1296

1450

Marcel Scheele and Andreas Spahn

modern, 44, 52, 95, 103, 344, 519, 918, 1155, 1157, 1159-1160, 1182-1183, 1185-1186, 1212, 1226, 1240, 1276, 1284, 1311, 1389 preventive, 1276 reproducibility of ∼, 904, 906, 908, 911 symbolic meaning of ∼, 918-919, 1220, 1224 therapeutic, 1276, 1282-1283, 12901293 universal, 1290-1292 technology as applied science, 40, 54, 69-74, 77, 88, 310, 633 technology assessment, 12-13, 31, 174177, 885-86, 1103-1143, 1166, 1229, 1264, 1302 constructive, 886, 1117-1118, 1167 parliamentary,1106, 1113 participatory, 1115-1117, 1126 technology conflict, 1109-1110, 1120 technology design, 107, 1111, 1117, 1119 technology pull, 481-482 technology push, 159, 175-176, 48183, 492, 521 technology transfer, 31, 331, 335, 910, 1180 technoscience, 25, 83, 119, 142-149, 343-345, 1044, 1147, 1160, 1311-1318 telemedicine, 1293 teleological explanation, 864 teleology, 534, 1153, 1182, 1209, 1268 Teller, E., 698, 1046 tempio, 1207 Tertullianus, Q. S. F., 639 textual model, 675-677 Thagard, P., 16, 391, 393, 395, 1343, 1378, 1381 theology, 120-21, 457, 1012, 1171 theoretical rationality, 565, 578, 951 theoretical concept, 65, 85-87, 316,

489, 653 theory construction, 50, 220, 230, 695, 705, 724 therapeutic technology, 1276, 128283, 1290-1293 therapy, 82, 1122, 1277, 1282-85, 1288, 1293, 1295 thermodynamics, 29, 42, 132-133, 161, 179, 188, 317, 329, 330, 347, 502, 716, 721-722, 814, 10081010, 1015-1016, 1019-1020, 1022, 1027, 1052 Thomas Aquinas, 33-34, 1014, 1150 Thomas, R. J., 109 Thomasson, A., 1-2, 15, 188, 195-197, 200, 206, 290-291, 513, 939 Thompson, A., 809-810, 812-814 Thomson,W. (Lord Kelvin), 132, 648, 715-16, 1010 Thomson, J., 1097 thought experiment, 329, 534, 1343, 1349, 1367, 1370 thresholds, 842-843, 996-999, 1004, 1094, 1132, 1332, 1368 Tilli, S., 94 Tolman, C. C., 651 topology, 1382 Total Quality Management (TQM), 494, 499 Toulmin, S., 143, 650, 1154, 1164 tractability, 361, 636, 697, 702 tradition of testability, 129-130 traditional design paradigm, 536-539 transdisciplinary, 93, 101, 103, 105, 109-112, 637, 1103, 1112 Tredgold, T., 3, 41, 1172 Trevelyan, J., 94 trial and error, 108, 126, 136-137, 169, 222-224, 310, 342, 353, 361, 363, 365, 414-415, 436, 616, 1370 trope, 279, 278 trust, 597, 599, 954, 1083, 1266, 13841385, 1396

Index

truth and epistemic acceptability, 755 and success/usefulness, 158, 312, 312, 316, 318-319, 339, 345, 368 as aim of science, 67, 71-73, 312, 318-319, 518, 649 in science vs. in technology/in engineering, 70-71, 159, 337, 339 of theories, 84 conceptual, 998 definition(s) of, 657, 668, 681682 ethical, 928 mathematical, 789 relative vs. univeral, 1209, 12161217 scientific, 76, 1044, 1163 semantic theory of, 657, 681 Turing machine, 1346-1350, 1362 Turing test, 1365 Turing, A., 1349, 1365, 1373 Turkle, S.,1183, 1390 Turner, R., 1345, 1357, 1398 tuxedo syndrome, 1073 typology, 457-461 of design practice, 455-486 of ends, 382 of means, 386 Ulam, S., 645 undecidability, 1348 Unger, S., 965, 967, 1167 unit operation, 181 unit process, 181 universal technology, 1290-1292 urban planning, 41, 409, 430-431, 442, 885, 1033-1044, 1064-1065, 1197, 1217, 1219, 1223, 1230, 1232 Urquhart, A., 1349, 1398 use plan, 207, 225, 341-342, 364, 455, 514, 530-531, 567, 887, 912-

1451

916, 934, 936, 981-986, 1176 user needs, 96, 107, 257 utilitarianism, 978, 992, 1001, 1089, 1091-1092, 1139, 1155, 1167, 1181,1245, 1396 validation, 75, 168-169, 314, 771, 1354, 1356 value, 974-979 aesthetic, 7, 12, 14, 884, 946948, 984, 1004, 1031-1067, 1229, 1383 conceptual analysis, 974-979 cultural, 127, 984-85, 1024, 1227 epistemic, 24, 330, 518, 695, 699, 703, 711, 755, 885, 1031, 1045, 1378 instrumental, 315, 376, 378, 385, 883, 932-933, 946-949, 975976, 980-982, 985-986, 10101011, 1050-1051, 1160, 1375 intrinsic, 975-976, 992, 1166, 1286, 1396 moral, 44, 885, 897, 946-948, 960, 978, 981-984, 987-988, 992, 997-999, 1004, 1056, 1091, 1237, 1392, 1397 social, 126, 213, 903, 1121, 1287, 1318 value analysis (VA), 449, 501 value commensurability, 977-978, 992, 994-995, 998, 1000 value conflict, 973, 977-979, 987-989, 994, 1000-1004 value holism, 1002-1004 value judgement, 1119 value monism, 979, 992, 1001 value neutrality, 865, 883, 886, 979980, 1132 value pluralism, 994-995 value sensitive design, 1001-1004, 1390, 1392, 1397 value statements, 974 Varzi, A., 252, 254, 305 Vaschy, A., 802

1452

Marcel Scheele and Andreas Spahn

Vaughan, D., 875 Veblen, T., 30, 53-54, 1158 Vedder, A., 1379-1380 Velleman, J. D., 389 Venter, C., 150, 1308-1309, 1311, 1326 Venturi, R., 1040, 1216, 1218, 1240 verification, 394, 671, 1356 Vermaas, P. E., 15, 188, 207, 210, 214, 218, 224-230, 247, 249, 341-342, 359, 364, 408, 464, 530, 562, 677, 887, 912-915, 981 Verstehen, 209 Vesely, D., 1213, 1231 Vicos principle, 1315-1316 Vieu, L., 188, 254 Vincenti, W. G., 6-7, 9, 106, 129-130, 140, 167-168, 309-310, 313, 317, 321-324, 326, 328, 330, 352, 356, 363-368, 370, 483, 489-490, 495, 509, 515, 528, 566, 593, 634, 761 Virilio, P., 1377 virtual environment, 1295, 1345, 1364, 1376, 1382-1383 virtual identities, 1388-1390 virtual machine, 1357 virtual reality, 646, 674, 1048, 1196, 1295, 1381-1384 virtual subjects, 1388 virtual world, 223, 754, 1342, 1376, 1382, 1388 virtue, 42, 650, 954, 1003, 1042, 10491050, 1225 aesthetic, 1049-1050 epistemic, 319, 343 moral, 33, 101, 954, 976, 1151 of professionals/engineers, 956, 973 viscosity, 315, 702, 765, 767, 769, 775779, 784, 801, 820 vision assessment, 1138-1139 visualization, 411, 415, 420, 424, 742, 744, 748, 1048, 1361-1362 Vitruvius (Marcus Vitruvius Pollio),

41, 411, 638, 1033, 1064, 11991200, 1206, 1229, 1235 Vries, M. J. de, 68, 76, 107, 137-138, 321-326, 328, 407, 490, 494, 566 Walker, D., 1381 Wallach, W., 1375 waste, 602, 607, 693, 740, 905-906, 916, 923, 1007, 1080-1082, 1106, 1109-1110, 1116, 11331134, 1138, 1269, 1303 Watkin, D., 1202 Watson, J., 106, 149, 1302-1303, 1306, 1312, 1314 web designer, 1392 Webber, M., 170, 437, 517, 608, 1215 Weber, J., 147 Weber, Max, 52, 54, 1024, 1114, 1157 Weber, Moritz, 645 Webler, T., 1115-1116, 1126, 1139 Wechsler, J., 1059 Weckert, J., 1384, 1391 Weil, V., 970 Weiss, C. H., 95, 126, 1017 Weizenbaum, J., 1372, 1375 welfare, 43, 69, 609, 618, 621, 884, 953, 956, 961, 970, 982-983, 987, 992, 1001, 1004, 1110, 1151-1152, 1156, 1158, 1164, 1166, 1168, 1172-1173, 1264, 1130 well-being, 602, 956, 987, 1130, 1184, 1222, 1237 Welsch, W., 1225 Werner, A., 1046, 1048 Wiener, N., 146, 612, 842, 1391 Wiggins, D., 378, 392, 396 Wigley, M., 1212 Wilde, O., 1236 William of Ockham, 1013 Williams, R., 55, 102-103, 106-108, 578, 603

Index

willingness to pay (WTP), 992, 10871088, 1129-1130 Willis, R., 131 Wilson, H.W., 47 Wilson, P. N., 1295 Wilson, St. J., 1036-1038 Wilson, T. D., 373 Wimsatt, W. C., 239, 1204 wind energy/power, 109-110 Wingspread Statement, 1075 Winner, L., 50, 504, 865, 887, 898903, 907, 911, 916, 983, 1108, 1157, 1168, 1385-1387 Winsberg, E., 347, 1361 Wise, G., 138, 310 Wittgenstein, L., 31, 57, 495, 637, 1163, 1224 Wittkower, R., 1207 Woessner, M., 1212, 1214 Wohler, A., 1079 W¨ olfflin, H., 1234 Wood, K., 235, 257, 260-201, 264-65, 523, 547-548, 552 Wood, W. H., 235, 257, 547-548, 550, 552 Woods, J., 16 Woodward, R. B., 1046, 1050 Woolgar, S., 68, 900-902 Wright, F. L, 1224, 1240 Wright, G. H. von, 387-388, 527-528, 575, 940 Wright, P.H., 43 Wright, T., 765 Wynne, B., 906-908, 1265 X-ray, 147, 1046, 1170, 1277-1278 Young, A., 1260 Young, I., 653 Young, J., 1214 Young, S., 1389 Youngs modulus, 641 Zhai, P., 1382 Zhang, J., 701

1453

Zhang, K., 1060 Zwart, S., 8, 15, 16, 511, 645, 659, 665, 675, 801 Zwicky, F., 197

This page intentionally left blank