Intentionality, Deliberation and Autonomy

INTENTIONALITY, DELIBERATION AND AUTONOMY This page intentionally left blank Intentionality, Deliberation and Auton...

Author: Christoph Lumer and Sandro Nannini

90 downloads 1603 Views 1009KB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

INTENTIONALITY, DELIBERATION AND AUTONOMY

This page intentionally left blank

Intentionality, Deliberation and Autonomy The Action-Theoretic Basis of Practical Philosophy

Edited by CHRISTOPH LUMER and SANDRO NANNINI University of Siena, Italy

© Christoph Lumer and Sandro Nannini 2007 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 permission of the publisher. Christoph Lumer and Sandro Nannini have asserted their right under the Copyright, Designs and Patents Act, 1988, to be identiﬁed as the editors of this work. Published by Ashgate Publishing Limited Gower House Croft Road Aldershot Hampshire GU11 3HR England

Ashgate Publishing Company Suite 420 101 Cherry Street Burlington, VT 05401-4405 USA

Ashgate website: http://www.ashgate.com British Library Cataloguing in Publication Data Intentionality, deliberation and autonomy : the action-theoretic basis of practical philosophy 1. Act (Philosophy) 2. Ethics, Modern I. Lumer, Christoph II. Nannini, Sandro 128.4 Library of Congress Cataloging-in-Publication Data Intentionality, deliberation and autonomy : the action-theoretic basis of practical philosophy / [edited by] Christoph Lumer and Sandro Nannini. p. cm. Includes bibliographical references and index. ISBN 978-0-7546-6058-3 (hardcover) 1. Practice (Philosophy) 2. Applied ethics. 3. Act (Philosophy) 4. Agent (Philosophy) 5. Intentionality (Philosophy) 6. Autonomy (Philosophy) I. Lumer, Christoph. II. Nannini, Sandro. B831.3.I58 2007 128'.4--dc22 2007005501 ISBN-13: 978 0 7546 6058 3

Printed and bound in Great Britain by MPG Books Ltd, Bodmin, Cornwall.

Contents Preface List of Contributors

Introduction The Action-Theoretic Basis of Practical Philosophy Christoph Lumer

vii ix

1

PART I

ACTION, INTENTION, INTENTIONALITY

1

Folk Concepts, Surveys and Intentional Action Frederick Adams and Annie Steadman

17

2

Action and Responsibility – A Second Look at Ascriptivism Ralf Stoecker

35

3

Action Theory and Cognitive Turn: How Can the Content of Intentions Contribute to Causing Actions? Sandro Nannini

47

4

What Do Deviant Causal Chains Deviate From? Geert Keil

69

5

The Double Failure of ‘Double Effect’ Neil Roughley

91

PART II

ACTION-THEORETICAL CONCEPTIONS OF PRACTICAL DELIBERATION

6

The Will and The Good Hugh J. McCann

119

7

The Grounds and Structure of Reasons for Action Robert Audi

135

8

An Empirical Theory of Practical Reasons and its Use for Practical Philosophy Christoph Lumer

157

Intentionality, Deliberation and Autonomy

vi

9

Anchors for Deliberation Michael E. Bratman

PART III

ACTION-THEORETICAL APPROACHES TO FREEDOM, AUTONOMY AND RESPONSIBILITY

10

Autonomy for Real People Michael Quante

209

11

Forming the Will Freely Gottfried Seebass

227

12

An Action Can be Both Uncaused and Up to the Agent Carl Ginet

243

13

Free Will: Action Theory Meets Neuroscience Alfred R. Mele

257

14

Belief and Moral Responsibility Carlos J. Moya

273

15

Autonomy and Weakness of Will Thomas Spitzley

289

Index of Names Subject Index

187

303 307

Preface In the philosophical tradition, from Aristotle to Hume, and, to a lesser extent, to Kant, as well as later, philosophy of action had a basing and ancillary function for normative practical philosophy. Philosophy of action had to clarify the basic concepts normative practical philosophy deals with, like ‘action’ and ‘intention’, to explain the practical relevance and the general value of the things so denoted, to clarify the modalities of our responsibility, like ‘intentional’ or ‘premeditated’, to inquire into the manner of our deliberating, in particular the origin of our decisions and the possible inﬂuence of all sorts of cognitions on them, to study the mechanisms of carrying out decisions, to determine what autonomous ways of deciding are, etc. The technical development of action philosophy since the 1960s has led to losing sight of this serving function of action theory for normative practical philosophy. On the other hand the present development in normative practical philosophy (as well as in metaethics), where action philosophy underpinnings are often presupposed but not provided (as in much of the unclear talk about “practical reasons”), makes the help from action theory even more necessary. The aim of this volume is to bring the serving function of action philosophy back into view, to strengthen the relation to practical philosophy in action theory, to promote those approaches, developments and theories that could provide important foundations for practical philosophy and to encourage normative practical philosophers to rely explicitly on action philosophy of this kind. The three parts of Intentionality, Deliberation and Autonomy stand for the three main parts of action philosophy: the conceptual analysis of notions like ‘action’ or ‘intentionality’, action and decision explanation, and advice regarding the basis and enforcement of decisions (by explicating what, for example, free decisions or actions of autonomous or strong-willed persons are). And they shall represent what these different parts of action philosophy have to offer normative practical philosophy. Methodologically this volume brings together a wide spectrum of approaches by distinguished philosophers and of analytical traditions from America and Europe. The authors were invited to contribute with the just sketched aim in mind, and their essays, all written for this special purpose and unpublished so far, were discussed during an intense conference (in March 2005 at the beautiful Certosa di Pontignano near Siena, Italy), which led to signiﬁcant alterations in many of the essays. I hope the reader will ﬁnd the results to be an important contribution to an action-theoretical foundation of practical philosophy and a stimulating new beginning of a discussion about the relation between action theory and practical philosophy. This volume has proﬁted from the help and contributions of many people and institutions. The Fritz Thyssen Stiftung (chieﬂy) and the University of Siena generously ﬁnanced most of the conference that gave rise to this volume. Sandro

viii

Intentionality, Deliberation and Autonomy

Nannini helped organize the conference at the University of Siena in his always kind, efﬁcient and obliging manner. The Conference Ofﬁce of the University of Siena offered invaluable help with the conference logistics. Paul Coulham from Ashgate welcomed and supported the book project with great enthusiasm. Duccio Pianigiani gave technical assistance. And, most important, all the authors did their best and were most cooperative, helpful and patient throughout. I thank all these persons and institutions for their support. Christoph Lumer Siena, November 2006

List of Contributors Frederick Adams is Professor and Chair of Philosophy at the University of Delaware. He is also Professor and Chair of Linguistics, Director of Cognitive Science and Director of the Delaware Interdisciplinary Ethics Program at the University of Delaware. His publications include the book The Bounds of Cognition, and the articles ‘Narrow Content: Fodor’s Folly’, ‘Fodorian Semantics’, ‘Thoughts and Their Contents: Naturalized Semantics’, ‘Names, Contents, and Causes’, ‘The Informational Turn in Philosophy’, ‘What’s in a(n Empty) Name?’, ‘Tracking Theories of Knowledge’, ‘Intentional Action in Ordinary Language: Core Concept or Pragmatic Understanding?’, among others. Robert Audi is an internationally known author in ethics, epistemology and philosophy of action. His books include Practical Reasoning and Ethical Decision (Routledge, 2006), The Good in the Right: A Theory of Intuition and Intrinsic Value (Princeton University Press, 2004), The Architecture of Reason (Oxford University Press, 2001) and Religious Commitment and Secular Reason (Cambridge University Press, 2000). He is a past president of the American Philosophical Association and presently Professor of Philosophy and David E. Gallo Chair in Ethics at the University of Notre Dame. Michael E. Bratman is U.G. and Abbie Birch Durfee Professor in the School of Humanities and Sciences, and Professor of Philosophy at Stanford University. He is the author of Intention, Plans, and Practical Reason (Cambridge: Harvard University Press, 1987; re-issued by CSLI Publications, 1999), Faces of Intention: Selected Essays on Intention and Agency (New York: Cambridge University Press, 1999), Structures of Agency: Essays (New York: Oxford University Press, 2007) and various other articles in the philosophy of action and related ﬁelds. Carl Ginet (b. 1932, Casper, Wyoming) received his BA from Occidental College in 1954 and his PhD from Cornell University in 1960. He taught at Ohio State University, the University of Michigan and the University of Rochester before joining the Sage School of Philosophy at Cornell in 1971, from which he retired in 1999. He has published two books, Knowledge, Perception, and Memory (1975) and On Action (1990) and many articles, mainly in epistemology and action theory but a few on (linguistic) meaning of one sort or another and on Wittgenstein. He resides in Ithaca, New York.

x

Intentionality, Deliberation and Autonomy

Geert Keil holds a chair in philosophy at the RWTH Aachen University, Germany. Born in 1963, he studied philosophy, linguistics and German literature at the Universities of Bochum and Hamburg. In 1991 he received his PhD from the University of Hamburg. From 1992 to 2004 he was Assistant Professor at the Humboldt-University of Berlin and a visiting scholar in Norway, USA (Stanford) and Switzerland. His areas of specialization are the philosophy of mind and action, philosophy of language, metaphysics and philosophy of science. He is the author of Kritik des Naturalismus (Berlin and New York, 1993), Handeln und Verursachen (Frankfurt am Main, 2000), Quine zur Einführung (Hamburg, 2002) and Willensfreiheit (Berlin and New York, 2007), and the co-editor of six anthologies, including Fifty Years of Quine’s “Two Dogmas” (Amsterdam, Atlanta and New York, 2003). A full list of publications is available at . Christoph Lumer (b. 1956, Ratingen, Germany) is Professor of Moral Philosophy at the University of Siena (Italy). He has been Assistant and Associate Professor of Philosophy at the University of Osnabrück (Germany), visiting professor at the University of Urbino (Italy) and visiting scholar at the University of Michigan, Ann Arbor (USA). His areas of specialization are ethics, action theory, theory of practical rationality and argumentation theory. He has published the books Praktische Argumentationstheorie (‘Practical Theory of Argumentation’, Braunschweig, 1990), Rationaler Altruismus (‘Rational Altruism’, Osnabrück, 2000), The Greenhouse: A Welfare Assessment and Some Morals (Lanham, New York and Oxford, 2002) and numerous articles in his areas of specialization. He has edited The Epistemological Approach to Argumentation (= Informal Logic 3.2005 and 1.2006). He is now preparing the monograph Kognitive Handlungstheorie (‘Cognitive Theory of Action’). Web page: . Hugh J. McCann is Professor of Philosophy at Texas A&M University. His publications in action theory include The Works of Agency, as well as numerous papers including ‘Settled Objectives and Rational Constraints’ (American Philosophical Quarterly, 1991), and ‘Intentional Action and Intending: Recent Empirical Studies’ (Philosophical Psychology, 2005). He also writes on philosophy of religion, and related topics in ethics and metaphysics. Alfred R. Mele is the William H. and Lucyle T. Werkmeister Professor of Philosophy at Florida State University. He is the author of Irrationality (Oxford, 1987), Springs of Action (Oxford, 1992), Autonomous Agents (Oxford, 1995), Self-Deception Unmasked (Princeton, 2001), Motivation and Agency (Oxford, 2003) and Free Will and Luck (Oxford, 2006). He also is the editor of The Philosophy of Action (Oxford, 1997), co-editor (with John Heil) of Mental Causation (Oxford, 1993), co-editor (with Piers Rawling) of The Oxford Handbook of Rationality (Oxford, 2004) and coeditor (with Mark Timmons and John Greco) of Rationality and the Good (Oxford, forthcoming).

List of Contributors

xi

Carlos J. Moya (b. 1952, Alcoi, Spain) is presently a member of the Department of Metaphysics and Theory of Knowledge at the University of Valencia, Spain, where he lectures in philosophy. He used to belong to the Steering Committee of the European Society for Analytic Philosophy and has been president of the Spanish Society of Analytic Philosophy. His main ﬁelds of research are the philosophy of mind and the philosophy of action, with a special interest in the problems of free will and of moral responsibility. His publications include The Philosophy of Action (Polity, 1990) and Moral Responsibility: The Ways of Scepticism (Routledge, 2006). He has also published several articles in philosophical journals and a number of contributions in edited books. Sandro Nannini (b. 1946, Siena, Italy) studied philosophy at the University of Florence and was Professor of Moral Philosophy at the University of Urbino. Since 1992 he has been Professor of Theoretic Philosophy at the University of Siena, Italy. After working for many years in the ﬁeld of the philosophy of social sciences, metaethics, action theory and the logical structure of historical explanations he has been working for ﬁfteen years in the ﬁeld of the philosophy of mind and cognitive sciences. Main publications: Naturalismo cognitivo. Per una teoria materialistica della mente (Macerata: Quodlibet, 2007), Seele, Geist und Körper. Historische Wurzeln und philosophische Grundlagen der Kognitionswissenschaften (Frankfurt am Main: Peter Lang, 2006), L’anima e il corpo. Una introduzione storica alla ﬁlosoﬁa della mente (Roma-Bari: Laterza, 2002), Naturalism in the Cognitive Sciences and the Philosophy of Mind (with H.J. Sandkühler [eds], Frankfurt am Main: Peter Lang, 2000), Il Fanatico e l’Arcangelo. Saggi di ﬁlosoﬁa analitica pratica (Siena: Protagon, 1998), Cause e ragioni. Modelli di spiegazione delle azioni umane nella ﬁlosoﬁa analitica (Roma: Editori Riuniti, 1992), Il pensiero simbolico. Saggio su Lévi-Strauss (Bologna: Il Mulino, 1981). Web page: . Michael Quante (b. 1962) currently holds a full professorship for philosophy at the University of Cologne and has formerly been Professor for Philosophy of Law and Social Philosophy at the University of Duisburg-Essen. He is Director of the Institut für Ethik in den Lebenswissenschaften and Co-Director of the Husserl Archiv in Cologne and an associated editor of the journal Ethical Theory and Moral Practice. Books (in English): Hegel’s Concept of Action (Cambridge University Press, 2004), Enabling Social Europe (Springer, 2005; co-authored with Bernd v. Maydell et al.), Moral Realism (Helsinki, 2004 [= Acta Filosoﬁca Fennica vol. 76]; co-edited with Jussi Kotkavirta), Pragmatic Idealism (Rodopi, 1998; co-edited with Axel Wüstehube) and Hegel’s Phenomenology of Spirit (Cambridge University Press 2007, co-edited with Dean Moyar). Books (in German): Personales Leben und menschlicher Tod (Suhrkamp, 2002), Einführung in die Allgemeine Ethik (Wissenschaftliche Buchgesellschaft, 2003), Ethik der Organtransplantation (Harald Fischer, 2000; co-authored with Johann S. Ach and Michael Anderheiden), Hegels Begriff der Handlung (frommann-holzboog, 1993) and Person (De Gruyter 2007).

xii

Intentionality, Deliberation and Autonomy

Neil Roughley is presently Visiting Professor for Practical Philosophy at the University of Muenster, Germany. His areas of specialization are action theory, philosophical anthropology, philosophy of mind, ethics and aesthetics. Main book publications: as editor, Being Humans: Anthropological Universality and Particularity in Transdisciplinary Perspectives (Berlin and New York: de Gruyter, 2000), Anthropologie und Moral. Philosophische und soziologische Perspektiven (Wuerzburg: Koenigshausen und Neumann, 2000; co-edited with M. Endress) and Wanting and Intending: Elements of a Philosophy of Practical Mind (Dordrecht: Springer, forthcoming). Gottfried Seebass is Professor of Philosophy at the University of Konstanz, Chair for Practical Philosophy. He studied at the universities of Tübingen, Mainz, Zürich, Heidelberg and Harvard. 1978: doctorate, University of Heidelberg. 1978−80: research scholar, Max-Planck-Institut, Starnberg. 1979: Lecturer, LMU Munich. 1980−91: Assistant Professor, FU Berlin. 1991: Habilitation, FU Berlin. 1991: Substitute Professor, Chair for Analytical Philosophy, FU Berlin. Since 1993: full professor at Konstanz. Head and/or member of various philosophical and interdisciplinary research projects (SFB 511 ‘Literatur und Anthropologie’; SFB 485 ‘Norm und Symbol’; Center ‘Intentionality’; DFG-research-group ‘Limits of Intentionality’; Exzellenzcluster EXC 16 ‘Kulturelle Grundlagen von Integration’). His publications include numerous papers on various subjects in theoretical as well as practical philosophy and ﬁve books: Das Problem von Sprache und Denken (Frankfurt am Main: Suhrkamp, 1981), Social Action (Dordrecht: Reidel, 1985; co-authored with Raimo Tuomela), Wollen (Frankfurt am Main: Klostermann, 1993), Handlung und Freiheit (Tübingen: Mohr-Siebeck, 2006), Willensfreiheit und Determinismus, Band I: Die Bedeutung des Willensfreiheitsproblems (Berlin: Akademie, 2007). Thomas Spitzley is Professor for Philosophy at the University of Duisburg-Essen. He studied philosophy, sociology and psychology at Hamburg University, obtained his PhD in philosophy from Hamburg University, and his Habilitation from the University of Erlangen-Nürnberg. His research interests concern philosophy of mind, philosophy of action, epistemology and philosophy of language. Selected publications: Handeln wider besseres Wissen. Eine Diskussion klassischer Positionen (Berlin and New York: de Gruyter, 1992), Facetten des „ich“ (Paderborn: Mentis, 2000) and, as editor, Willensschwäche (Paderborn: Mentis, 2005). Annie Steadman is a graduate student in Philosophy at the University of Arizona, with interests in metaphysics, epistemology and the philosophy of mind. She has published papers on intentional action in ordinary language, including ‘Intentional Action in Ordinary Language: Core Concepts or Pragmatic Understanding?’ and ‘Intentional Action and Moral Considerations: Still Pragmatic’, and continues to work in these areas.

List of Contributors

xiii

Ralf Stoecker has taught in Bielefeld, Hamburg, Erlangen and Magdeburg, and is now Professor for Applied Ethics at Potsdam University, Germany. Stoecker’s publications cover a wide range of topics from ontology and epistemology to medical ethics. Action theory is one of his areas of specialization. During the last ﬁfteen years Stoecker has worked on the development of a non-standard account of agency, which found expression in about twenty articles in this ﬁeld and should condense within the near future into a monograph on agency. Book publications: Was sind Ereignisse? (de Gruyter, 1992), Der Hirntod (Alber, 1999), Handlungen und Handlungsgründe (Mentis 2002), Menschenwürde – Annäherung an einen Begriff (öbv&hpt, 2003).

This page intentionally left blank

Introduction

The Action-Theoretic Basis of Practical Philosophy Christoph Lumer

In the philosophical tradition ever since Aristotle theory of action has been regarded as an essential and indispensable basis of the core disciplines of practical philosophy, which should explain, among others, how prudential rationality, moral behaviour, responsibility, freedom or autonomy are possible. We ﬁnd such explicit action theory underpinnings of their respective ethics or theories of rationality in Aristotle, in Epicurus, in Stoic philosophy or later, in modern times in Locke, in Hume or, to a minor degree, in Kant, Hegel and Schopenhauer – to name only a few. One main reason for such endeavours to give an action-theoretic foundation to practical philosophy seems to be this: before we can make ethical prescriptions and recommendations or give rational advice on how to act, we must know what actions are, what intentional actions are, how they function and what causes them, in particular what their deliberative origins are and how such prescriptions and recommendations can inﬂuence our decisions. Subsequently theory of action lost its importance in philosophy with the development of scientiﬁc empirical psychology and with the upsurge of rationalist and a priori conceptions of philosophical research during the nineteenth century. Only in the 1950s was it embraced again as an independent branch of philosophical research. And much research has been done since then concerning the concepts of ‘action’, ‘intention’, ‘intentionality’, ‘explanatory reason’, about mechanisms of control and self-control, as well as on the nature of mental states, on mental causation, the grounds of our decisions and autonomy and many other related questions. An important feature of philosophy of action since the 1950s has been its reestablishment as a discipline of its own and its – compared to previous theorizing – relative independence from the rest of practical philosophy. The latter feature, however, has also had negative consequences: often abstract questions seem to have made themselves independent as interesting puzzles, and nowadays, at the beginning of the twenty-ﬁrst century, the fundamental function of action theory for practical philosophy is rarely recognized. On the other hand these days in ethics, as well as in theory of practical rationality, there is a strong and growing tendency to support the latter theories with (mostly implicit) action-theoretic hypotheses, and there is much talk about “(motivating) practical reasons”. As shall be shown below, giving an action-theoretic foundation to normative theories of practical rationality is quite right and necessary for their sound justiﬁcation. However the action-theoretic hypotheses presently used not only mostly remain inadequately explicated or

2

Intentionality, Deliberation and Autonomy

completely implicit but are also usually not much reﬂected, ad hoc and cannot stand up to critical scrutiny. So both practical philosophy as well as philosophy of action, nowadays suffer from deﬁcits due to insufﬁcient consideration of the other side. The aim of this volume is to recall the relation of philosophy of action to practical philosophy and to develop it by working out some pieces of philosophy of action that are particularly interesting in this respect. The aim of the rest of this introduction is to explain the signiﬁcance of the various parts of philosophy of action for practical philosophy in more detail and to place the contributions to this volume into this framework. 1. The three parts of the philosophy of action Philosophy of action does not have the one or only a handful of questions. It comprises several types of theories differing in the respective kinds of propositions they are aiming at and even in the methodologies as well as the sorts of arguments used to substantiate them. In this respect at least three areas or parts of philosophy of action can be distinguished. 1. Conceptual analysis: The ﬁrst part of action theory is conceptual analysis, as the result of which concepts like ‘action’, ‘intention’, ‘intentional’, ‘voluntary’, ‘premeditated’, ‘deliberate’, ‘unconscious action’, ‘automatic action’ and so on are deﬁned. These deﬁnitions serve also for setting the stage, that is for determining what action theory is about. Sometimes ordinary language analysis or reconstruction of our intuitive understandings of these concepts are proposed as the right methods for this part of action theory. Although ordinary language and intuition harbour a wealth of experience and reﬂection, these two methods seem to be too simplistic because there is already too much theorizing in this part of action philosophy, far beyond what is implicit in that treasure. The action-theoretic concepts listed above share a particular feature; they all designate something that is very valuable in a certain respect in contrast to their opposites: action vs mere behaviour, intentional action vs unintentional action and so on. And the theorizing often deals with exactly what makes these things valuable and why. So a more adequate methodological understanding of this part of action philosophy proposes that the concepts in question have to be deﬁned in such a way as to capture what is valuable about their objects. The method for doing this could be to scrutinize the intuitions about the respective objects and new hypotheses about their nature, that is whether they capture what is valuable or makes a value difference and then to stick with those that are most revealing in this respect. This method may be called “idealizing hermeneutics” because it tries to understand our intuitions, but ﬁlters out those which best capture the particular value. The particular value of action, for example, could consist in controlling the environment from the (our) inside; this control in turn explicates the idea of accountability. Or, to give another example, instead of regarding wayward realizations of one’s intentions as an intellectually pretentious puzzle, which roughly has to be resolved by reconstructing common sense intuitions, one could deﬁne them in a way that explains why the particular value of intentional action is no longer realized.

The Action-Theoretic Basis of Practical Philosophy

3

2. Action and decision explanation: Another part of action theory tries to clarify how actions are to be explained, in particular: what intentions are empirically, how they “produce” actions, how we decide, what explanatory practical reasons are, how they “work”, how they come into being, what their (potential) content is, and so on. If one accepts a Humean kind of causality and a Hempel–Oppenheim type of covering-law conception of action explanation, this explanatory part of action theory has to rely heavily on scientiﬁc psychology of action and empirical decision research. The differences vis-à-vis the respective scientiﬁc psychologies are the following. (i) Explanatory action philosophy does not undertake empirical observations but uses the material provided by scientiﬁc psychology. (ii) It systemizes this material and develops explaining theories on its basis (thus explanatory action philosophy is nearer to theoretical psychology than to empirical psychology). (iii) Explanatory action philosophy has its own aims of theorizing, namely perhaps to provide the practically signiﬁcant information, on a not too microscopic level, needed in normative practical philosophy, in particular all possible inﬂuences of cognitions on our actions (it does not aim at providing the most detailed and technically universally exploitable laws, as scientiﬁc psychology does). If, on the other hand, one does not accept the orthodox conception of action explanation, the ﬁrst and currently principal task of explanatory action philosophy is to develop a philosophically and methodologically satisfactory heterodox concept of ‘explanation’. (This, of course, was the main topic of the explanation-understanding debate.) 3. Advice regarding the basis and enforcement of decisions: A further part of philosophy of action should explicate and determine what free, self-controlled, autonomous or authentic actions and decisions are and explain how they are possible: What are good mechanisms of self-control? If acting autonomously consists in not being forced but realizing some sort of real self, core or essence of the self, what does this core etc. consist of and how does it translate into actions? If free decisions are more than undetermined decisions, what are the standards they should meet? And so on. An action philosophy that gives answers to these questions is already normative (in a broad sense), or more precisely: it, at least implicitly, recommends certain ways of deciding and acting. Such recommendations should be based on identifying and developing different ways of deciding and acting, on methodically assessing them and on choosing as well as proposing a good or the best among them. By offering (implicit) recommendations this part of action philosophy is immediately practical; and the way of developing these proposals resembles inventing instruments and techniques. So this part of philosophy of action may be called “practical-technological”. Most of normative (in a broad sense that includes the advisory) practical philosophy is also practical-technological in the sense just explained. Because of this methodological and functional community, the practical-technological, advisory part of action philosophy can already well be considered a part of normative practical philosophy. Conceptual analysis and action explanation on the other hand have more of an ancillary function for normative practical philosophy. The claim that action theory is an indispensable basis of practical philosophy holds for them to a stronger degree.

4

Intentionality, Deliberation and Autonomy

The three parts of this book correspond to this tripartition of action philosophy. And the three keywords of the main title, Intentionality, Deliberation and Autonomy, which take up central concepts of the respective parts, shall stand for these three parts of action philosophy. 2. Part I: action-theoretical conceptual analysis and practical philosophy What can action-theoretical conceptual analysis contribute to practical philosophy? According to the above sketched method of an idealizing hermeneutics, conceptual analysis in action philosophy ﬁrst of all has to understand the particular meaning and value of action, intention and so on. Though explicitization of these values sometimes seems trivial because they regard deep-rooted self-understandings, it nonetheless may be helpful in orienting, for example, the basic conceptualization of ethics or the theory of practical rationality. A case in question is this. One fairly general understanding of the sense of action says: by acting intentionally we, or more precisely the central parts of our mental personalities, exert control over our bodily behaviour (and some mental events) from within and thereby, via its effects and implications, shape – of course to a limited degree – what is happening in the world according to our aims, desires, intentions and so on. This implies an instrumentalist view of actions: actions may have intrinsic value too but they are mainly and ﬁrst of all instruments for shaping the world. Now this understanding of intentional action leads ethics roughly into a consequentialist1 direction. If the central task of ethics is to prescribe or recommend actions (or to propose prescriptions for actions) and if the sense of actions is to shape the world according to our intentions, then ethics makes the most adequate use of these instruments if it makes its prescriptions and recommendations with the intention to shape the world according to moral ideals, with the help of the prescribed actions. Of course the consequentialist predisposition of actions as such does not exclude a kind of deontologism that sees the aim and content of morals in certain kinds of actions themselves (and in preventing other kinds of actions for their own sake) or an ethics of good will, where the good will is the only thing that is unreservedly good (cf. Kant, GMS BA 1). However, such ethics makes inadequate use of actions and squanders their potential. Another, much more direct, contribution of action-theoretical conceptual analysis to practical philosophy is the analysis of those terms that denote the various types of ascribability of actions as well as their consequences to persons and thus differentiate various types of responsibility as well as the (possible) reach of morals: ‘action’, ‘omission’, ‘intentional’, ‘voluntary’, ‘knowing’, ‘premeditated’, ‘reckless’, ‘negligent’, ‘unconscious’ and so on. The practical importance of these analyses is underlined by the fact that (criminal) law has to deal with the same questions and does so in a highly differentiated manner, often giving answers similar to those of philosophers. 1 “Consequentialist” here is meant in the general sense of: regarding a theory that stresses the importance of actions’ consequences (and measures their value in terms of their consequences’ value), – and not in the narrow sense of the misnomer introduced by Schefﬂer: regarding utilitarianism or ethics that prescribe welfare maximization (cf. Schefﬂer 1982).

Introduction

5

However, one point should be underlined again. If the conceptual part of action theory also has an ancillary function for (normative) practical philosophy, then it should meet this requirement by considering, analysing and evaluating the practical relevance of the respective objects. And the ﬁnal deﬁnitions should reﬂect this value (cf. the example of wayward realizations of intentions mentioned earlier). The chapters in the ﬁrst part of this book contribute in different ways to an actiontheoretical conceptual analysis relevant to practical philosophy. Fred Adams and Annie Steadman discuss the concept of ‘intentional action’ and conﬁrm the simple view, that is, the thesis that intentionality presupposes the respective intention, by rejecting an objection from folk judgments. Thereby they also criticize (folk) intuitionism and relativize the importance of such folk judgements for the philosophical endeavour. The simple view itself is one part of the view that intentional action shapes the world according to our intentions. It also captures the legal distinction between ‘intentionally’ and ‘knowingly’. Ralf Stoecker develops a (neo-)ascriptivist conception of actions, that is the view that action sentences ascribe responsibility to a person for certain events, where this responsibility consists in the fact that the event in the end can be explained by the person’s deliberation. Stoecker defends this view against several objections and emphasizes its merits. Of course, this theory has an immediate practical philosophical relevance in that it conceives actions in terms of the practical philosophical notion of ‘responsibility’: acting is being responsible. Sandro Nannini defends a causalist understanding of actions from a naturalistic viewpoint. In particular he discusses the problem: how can the mental contents (of intentions) as such cause respective actions as opposed, for example, to a wayward causation of the same behaviour? The practical interest behind this question, of course, is to explain freedom of action, that is, how our thoughts can control the world. Nannini proposes an adverbial theory of mental content and an explication of the content’s inﬂuence through neuronally different ways of causing. Geert Keil discusses the problem of deviant realization of intentions, which is at the heart of deﬁning ‘intentional action’: when can we say that a certain course of events does not only correspond to our intentions but was also controlled by us? Keil gives a new, double answer to this problem. There is no general way to characterize the “right” causal genesis of intentional action because the world is ontologically denser than our descriptions of the intended world in our intentions. And deviant causal chains depart from what the agent would ex post qualify as an intended course of events. Neil Roughley discusses a particular question of the relevance of action-theoretical distinctions to ethics: has the difference between doing something knowingly (collateral consequences) and doing it intentionally any signiﬁcance for the moral appraisal of this action? In particular, might doing something morally bad only knowingly sometimes lead to its being permitted – as the doctrine of double effect maintains? Or are there other moral uses of this distinction? Roughley develops a new criterion for distinguishing between knowingly and intentionally and then answers all these questions in the negative.

6

Intentionality, Deliberation and Autonomy

3. Part II: the use of action and decision explanation for practical philosophy That normative practical philosophy has to be based on the explanatory part of action theory, or more precisely on the decision psychology developed in action theory, for many is still more obvious and today better visible than the foundational function of action-theoretical conceptual analysis. The most general argument for this dependency is that normative theories of deliberation and practical reason for being reasonable need information about the empirically necessary structure and the open spectrum of different ways of deliberation and practical reasoning. On the other hand, this kind of dependency is strongly contested by foundational externalists in ethics and in the theory of practical rationality. In opposition to the latter and sustaining the former view, the following dependency thesis for normative ethics shall be defended: A sound justiﬁcation of morals depends heavily on substantial empirical actiontheoretical information about human decision-making (motives, ways of deciding and so on). Before substantiating this claim with some more speciﬁc arguments let me introduce some distinctions of several forms of internalism and externalism in ethics (and analogously in the theory of practical rationality). “Effective internalism” here shall mean the claim that an adequate justiﬁcation of morals (perhaps only under certain, though usually fulﬁlled conditions) among others has to produce the motivation to adopt these morals and some motivation to follow them; effective externalism says that such motivation is not necessary for an adequate justiﬁcation of morals. So effective internalism is an adequacy condition for justiﬁcations of morals. “Foundational internalism” here shall denote the claim that the motivational conditions of effective internalism can only be fulﬁlled by the justiﬁcation’s recourse to motives that precede the justiﬁcation; foundational externalism says that for meeting the requirements of effective internalism no such recourse to pre-existing motives is necessary. In particular, foundational externalists assume that an externalist justiﬁcation of morals (that is a justiﬁcation that does not make recourse to pre-existing motives) can provide the necessary motivation via the power of reason. David Brink, for example, is an effective externalist; Kant is an effective internalist but a foundational externalist; and Bernard Williams is an effective as well as foundational internalist. (In addition to the two forms of internalism mentioned so far there is semantic internalism, which, however, does not concern us here.) The argument for the dependency thesis for normative ethics proceeds in three steps. 1. Effective externalism is inadequate. There are four possibilities for effective externalism. (i) The justiﬁcation in accordance with effective externalism does not motivate, and consequently the justiﬁed moral is not (practically) embraced and consequently not put into practice. In this case the justiﬁcation is pragmatically useless; we can do well without it; and from a practical viewpoint there is no need to discuss it. (If somebody is theoretically interested in it and has nothing better to do

The Action-Theoretic Basis of Practical Philosophy

7

he may discuss it from a theoretical standpoint. But this is not practical philosophy.) (ii) The justiﬁcation of moral is not motivating in the required sense but the proposed moral is motivating nonetheless because of other reasons. In this case too the justiﬁcation is pragmatically useless and we can ignore it from a practical viewpoint because it is superﬂuous for the moral subjects’ practical adoption of this moral; subjects ignore it in practice. (iii) Even though the justiﬁcation in accordance with an effectively externalist framework has not been designed to meet the standards of effective internalism, it is motivating because its author by chance has invented a justiﬁcation that actually is effectively internalist. This would be the case where the justiﬁcation is only subjectively (that is for the arguer) but not objectively effectively externalist. (iv) The justiﬁcation is motivating by chance, independently of its content, because some other particular (circumstantial) condition is fulﬁlled – the arguer might, for example, be recognized as an authority. But then again, as in case (ii), the justiﬁcation in its content is practically useless and can be ignored. To sum up, justiﬁcations of morals on the basis of effective externalism, that is without aiming at motivation to moral action, are not sound because they are practically irrelevant and thus miss the essential function of practical justiﬁcation. 2. A justiﬁcation of morals on the lines of foundational internalism, obviously and by deﬁnition, recurs to pre-existing motives for moral action and therefore needs empirical information about them. 3. Foundational externalism, ﬁnally, tries to achieve the required motivation to act morally by externalist justiﬁcations, which do not recur to pre-existing motives. Foundational externalism (at least implicitly) assumes that the externalist justiﬁcation or the belief justiﬁed by it, that is a belief that the moral in question has a certain justifying property F, by themselves provide the necessary motivation, where the F in question has nothing to do with one’s pre-existing motives and might for example be ‘is just’, ‘is rational’ or ‘is universalizable’. However, that a certain kind of cognition, that moral m is F, leads to a motivating type of adoption of moral m is an empirical, causal relation and not analytical. So it is a question of our empirical motivational makeup whether the foundationally externalist assumption is true. And in order to follow the foundationally externalist justiﬁcation strategy one needs empirical action-theoretic information as to whether such an F exists and what its exact content is. Please note that the third step of the argument just given does not say that foundational externalism is false but only that it relies on empirical action-theoretical information about our decisions, too. Whether the foundationally externalist assumption is true or whether foundational internalism is right in denying this is yet another, empirical action-theoretical question. Stressing this point still more: the very question of whether foundational internalism or externalism is right is an empirical action-theoretical question about human motivational and decisional makeup. A further, independent argument for the dependency thesis is this. Christine Korsgaard has distinguished between motivational skepticism about practical reason and content skepticism about practical reason, where the latter refers to doubts that pure practical reason can establish (the content of) moral duties or desirabilities (Korsgaard 1986, 311). A fairly general argument for content skepticism is that

8

Intentionality, Deliberation and Autonomy

reason as such can establish truths about possible actions or even about complete moral systems, but it cannot establish relevances or importances; it cannot say which of those indeﬁnitely many truths is the practically relevant one in the sense that if this truth about an action or moral system holds, the action shall be executed and the moral system be adopted respectively. From the standpoint of reason all truths have the same value, which consists in their being true. Practical relevance in an action-directing way can only derive from practical attitudes like desires, intentions or feelings. If this is so then even for establishing the material content of morals we need empirical action-theoretic information about which practical attitudes humans have and how they inﬂuence decisions. A dependency thesis similar to the one for normative ethics also applies to the theory of practical rationality; and it can be defended rather analogously in three steps. 1. Effective externalism about theories of practical rationality leads to practically irrelevant systems and criteria of practical rationality and therefore is inadequate. 2. Foundational internalism about theories of practical rationality may not, at a ﬁrst glance, seem to lead to the claimed dependence because very weak forms of such theories do not seem to need such empirical information. One might argue: subjective expected utility theory for example, of course refers to motivating preferences but it does not need to know them; it can simply recommend a way of deﬁning utilities over such preferences – whatever their content is – and then recommend maximizing the subjective expected utility. However, this argument forgets the immensely rich empirical action-theoretical information about the structure of our decisions behind the advices of subjective expected utility theory, namely that we are beings with basic and non-basic preferences, which can be quantiﬁed, that we can integrate utilities and probabilities in the recommended way, that we are able to choose and then to realize the action with the highest expected utility and so on. Sometimes these presuppositions are even contested, and exactly for such empirical reasons a different criterion of practical rationality is proposed. Thus even weak theories of practical rationality try to adjust their advices to rather general structural features of human decision-making and therefore need the respective empirical information. Stronger theories of practical rationality need even more information about our decision-making. They try to ﬁt the recommendations for rational decision much more on the structural reality of our empirical decision-making. 3. The most important form of foundational externalism in the theory of practical rationality is value objectivism, that is the theory that there are values in the world whose worth is independent of anybody’s preferences or (implicit) criteria for such preferences. Many philosophers contest value objectivism. But this is not the place to enter into this discussion. So let us assume value objectivism to be true and people to be able to truthfully assign objective value predicates V1, V2, …, Vn to certain objects. Now in order to fulﬁl the motivational requirements of effective internalism the respective value judgements, that is judging ‘object a is Vi’, must be motivating to some degree in correspondence with the objective value. However, motivating is an empirical, causal relation. And in order to be motivated in the right way by such value judgements our motivational system must be made up as to be

The Action-Theoretic Basis of Practical Philosophy

9

“sensitive” exactly to judgements of the kind ‘a is Vi’; for each value predicate Vi there must be at least one motivational counterpart, that is a motivational structure that uses Vi as the criterion for assigning motivation to the realization of the objects that are Vi. If our motivational system is sensitive in this way to some value predicate Vi is an empirical action-theoretical question and is not settled by reason or by some analytical implication of the truth of ‘a is Vi’. (From a motivational viewpoint, value objectivism and value subjectivism here are on a par. The difference between them regards the issue as to whether or not values can be attributed independently of such motivational counterparts.) Therefore, for ascertaining the foundations of our practical valuing, for establishing the practical, motivational reach of our value judgements (there may be value judgements without corresponding motivation and vice versa) and for being motivationally effective, objective value theories have to use empirical action-theoretical information about human motivation. There is again a second general argument for the dependency thesis about the theory of practical rationality, analogous to the argument for content scepticism about practical reason. 1. Value subjectivism, roughly, is the theory that values are always relative to a value-subject, whose structures or criteria deﬁne which things have which value, such that the values’ content is determined by these structures. The structures that deﬁne these contents may be motivational, emotional/feeling-related or simply the organic functioning of the subject, that is in the end the subject’s survival. 1.1. If the subjective values are motivation-based, that is if they are deﬁned by our motives, empirical action-theoretic information about our motives is required to determine their content. 1.2. In case of values deﬁned by emotional structures or by survival the values’ content is not determined directly by our motivational structure. However, if these values are to have some practical bearings, as is required for their being relevant in a theory of practical rationality, the non-motivational values must be “taken up” or “represented” by our motives or desires, which simply “repeat” those values or interpret them. So for establishing and understanding the content of such values in the version relevant for practical rationality we again need empirical action-theoretical information about how our motivational structure deals with them. 2. Value objectivism holds that values are non-relational but “deﬁned” or given by the value object itself. Therefore these values are completely independent of subjective structures. However if these values are also to be values for us and if they are to have some practical bearings, then what has been said about non-motivational subjective values (case 1.2) also applies to them: they must be represented in some way by our motives; and for establishing their practically interpreted content we need information about how our motivational system deals with them. Currently there is a fashionable overﬂow of talk about unclear “practical reasons”. Taking into account the two dependency theses just defended should give more substance to the theorizing about practical reasons and hopefully bring back to earth accounts that presuppose obtaining the practical and motivational relevance of their unclear “practical reasons” for free.

10

Intentionality, Deliberation and Autonomy

The chapters in the second part of this book examine the nature of explanatory practical reasons (that is the reasons that actually lead to our decisions), and in particular how they can contribute to the rationality and autonomy of our actions. Hugh McCann presupposes an objectivist conception of the good and proposes a strong solution to the motivational problems such a conception might have. According to his proposal, desires motivationally represent what is objectively good. Objective values exercise some tug on us, and we apprehend them. Conative desires, but not cognitive judgements, are the receptors of these values and they represent them in motivational terms. Only on this basis can desires be rational, and objective values practical. One consequence of this conception is that an examination of desires leads us to the content of objective values. Robert Audi compares theoretical and practical reasons and examines the (possible) rationality and objectivity of the latter. Or more precisely, in both cases he distinguishes between (objective) reasons proper and reason states, that is beliefs and desires, where the reasons are the content of the reason states. Some results of Audi’s comparison are the following. Although beliefs and desires have different directions of ﬁt, also desires and desirability mostly are grounded in (for example hedonic) experiences. However the rationality of (instrumental) desires depends on the rationality of beliefs but not vice versa. A further asymmetry is that beliefs have propositional contents whereas desires have inﬁnitive contents and therefore are neither true nor false. Christoph Lumer ﬁrst sketches a comprehensive empirical theory of practical reasons: What are intentions? How are they connected to (probabilistic) beliefs about action implications and intrinsic desires? What kind of intrinsic desires do we have and how do they develop? What kind of motives for acting morally do we have? In the second part of his chapter he outlines very strong implications of this empirical theory for the theory of practical rationality and for ethics. Michael Bratman develops a normative model of particular practical reasons, which he calls “anchors of deliberation”. He characterizes these anchors functionally from a normative viewpoint, namely as such valuings in a deliberation that can provide its agential authority and autonomy. For fulﬁlling this function such anchors must be rather stable, though potentially revocable; they may differ interpersonally but stand in a complex relation to judgements about the best. Bratman empirically identiﬁes personal policies as one type of such anchors, where personal policies are commitments to certain standards (for example to act as a professional philosopher) or to certain types of reasons (for example not to act from envy) in one’s deliberation. 4.

Part III: theory of autonomy and practical-technological philosophy

The third part of action philosophy gives advice regarding the basis and enforcement of decisions by saying what free, self-controlled, autonomous or authentic actions and decisions are. Because of its practical-technological aims and methodology and because it deals with actions in general it is already part of normative (in the broad sense) practical philosophy. On the other hand, like the latter it must be based on the ancillary parts of action theory, in particular on action-theoretical empirical

Introduction

11

information about our ways of deciding and about how actions come into being. The general reason for this claim is again that whoever wants to give reasonable advice – in this case advice regarding the basis of our decisions – must be informed about the possible courses to follow and their relevant implications – in our case the possible bases of decisions and the consequences of their use. Let me show this in a more concrete way for the theory of practical freedom. The ﬁrst part of the theory of freedom is the theory of freedom of action. Freedom of action, roughly, consists in the capacity to do what one intends (Leibniz <1704>/1996, II, xxi, 8). This capacity means that the intentions produce the respective basic action in a controlled way. For clarifying this capacity further empirical action theory is required to explain what intentions are and by which control mechanisms they produce the action. The other part of the theory of practical freedom is the theory of freedom of decision. Freedom of decision, formulated deliberately vaguely, consists in the fact that the decision is reached in the right way. There are at least three main approaches to what counts as “right” in this context: (1) incompatibilism, which demands that the decision not be determined, (2) rationalism, which requires the decision to be rational, and (3) autonomy theory, saying that a free decision may not be determined by constraints or coercion extraneous to the subject but must have its origin in the subject’s essence, core and so on. Ad (1): As a minimum answering the question of whether incompatibilist freedom is possible or actually realized in our world requires information about the origins and possible causes of our decisions. Furthermore, incompatibilism can be differentiated into two main directions. Monistic incompatibilism holds that indeterminacy of decision is necessary and sufﬁcient for the decision to be free. Monistic incompatibilism is a rather implausible position because it cannot explain why undetermined, and therefore completely contingent, decisions without any differentiation should be good (ibid., II, xxi, 15; 50). Therefore most incompatibilist maintain pluralist incompatibilism, which says that indeterminacy of decision is necessary but not sufﬁcient for the decision to be free, further conditions must be fulﬁlled; and such conditions may be those of rationalism and autonomy theory. Ad (2): If free decisions are (more or less) identical to rational decisions then for developing such a theory of free decisions everything that has been said in the last section about the action-theoretic underpinnings of the theory of practical rationality holds true, in particular the dependency thesis. Ad (3): Finally, an autonomy theory of free decisions needs strong information about the possible ways and sources of our decisions, in order to be able to decide which of them are deep-rooted and are part of or express the subject’s core or essence. Beyond the theory of practical freedom there are further branches of research in the practical-technological part of action philosophy: theory of self-control, of responsibility or of authenticity. But what has been said about the necessary empirical action-theoretical underpinnings of the theory of practical freedom holds for these theories analogously. They all (at least implicitly) give advice: What are good forms of self-control? When are we responsible? What does authenticity require? And because these theories give advice they need information about the possible paths to follow in the respective ﬁelds and about the consequences of the possible courses. And, of course, giving such justiﬁed recommendations about the basis and enforcement of decisions is already a good piece of normative practical philosophy.

12

Intentionality, Deliberation and Autonomy

All the chapters in the third part of this book discuss the analytic or empiric conditions of freedom of decision, mostly in terms of autonomy. So they all try to contribute to determining or to establishing the conditions of one of the most fundamental aspects of good decisions, which in turn are quasi necessary for good actions. Michael Quante ﬁrst argues that incompatibilist theories of free decisions can only be plausible if they are what above has been called “pluralist” so that incompatibilist and compatibilist theories of freedom both have to develop a satisfying concept of autonomy. He goes on to contribute to such a conception by offering a solution to Kane’s manipulation argument, according to which a person who is under covert non-constraining control allegedly cannot be autonomous. Quante’s solution is that education is a form of covert non-constraining control and that the right form of education with a gradual release of control, exactly in contrast to what Kane claims, makes autonomy possible. Gottfried Seebass defends what is usually called “incompatibilism”2 and criticizes compatibilism with a wide range of arguments. His main argument is that determinacy, in the sense of all events being ﬁxed, does not permit alternative courses of action; as a consequence there would not be different options between which we could decide and hence no freedom of choice. An argument against easy attacks on incompatibilism is that the latter may appear in the form of advocating a partial indeterminacy. Carl Ginet defends incompatibilism against a particular and strong objection, namely that indeterminism implies that actions cannot be up to the agent. He attacks in detail four more or less strong arguments in favour of this objection. As a result he claims the contrary, that is U: an action can be uncaused and up to the agent. Finally, he makes the even stronger claim, that only uncaused actions can be up to the agent. Alfred Mele defends the possibility of free will against Libet’s respective attacks that experimental ﬁndings show that unconscious preparation of actions precede the forming of a conscious intention, such that the latter has no decisive role or only the function to eventually veto what has already been unconsciously prepared. Mele criticizes Libet’s interpretation of his data; in particular he presents several strong arguments that what Libet takes to be intentions actually are only urges to act. Carlos Moya develops an objectivist theory of free decisions. Accepting ultimate control as a necessary condition for free decisions he tries to show that determining freedom in conative terms must lead to an inﬁnite regress. Therefore he proposes a cognitivist, rationalist approach to deﬁning ‘freedom of decision’, where freedom is reached by recognizing objective evaluative truths. Such cognitions guarantee authorship, control of the value judgements and responsibility for them, where the authorship is independent of the originality of the value judgement. Thomas Spitzley examines the implication relations between autonomy and weakness of will and defends the thesis that a weak-willed person is not autonomous 2 Seebass himself ﬁnds the terms ‘compatibilism’ and ‘incompatibilism’ misleading because they distract our attention from the main problem, namely to develop positive and sufﬁcient conditions of freedom of choice.

The Action-Theoretic Basis of Practical Philosophy

13

with respect to the weak-willed action she performs. For proving this claim Spitzley develops some necessary conditions of autonomy. Then he shows that the conditions of several approaches to weakness of will (those of Hare, Watson, Davidson, Ainslie and Ursula Wolf) imply the violation of one or the other, surprisingly different, conditions of autonomy. References Korsgaard, Christine M. (<1986>/1996), ‘Skepticism about Practical Reason’, in Christine M. Korsgaard, Creating the Kingdom of Ends, Cambridge: Cambridge University Press, pp. 311–34. Leibniz, Gottfried Wilhelm (<1704>/1996), New Essays on Human Understanding (Nouveaux essais sur l’entendement humain), trans. and ed. Peter Remnant and Jonathan Bennett, Cambridge and New York: Cambridge University Press. Schefﬂer, Samuel (1982), The Rejection of Consequentialism, Oxford: Clarendon.

This page intentionally left blank

PART I Action, Intention, Intentionality

This page intentionally left blank

Chapter 1

Folk Concepts, Surveys and Intentional Action Frederick Adams and Annie Steadman

This chapter considers important empirical ﬁndings about folk judgments of intentional actions and intentions. We discuss empirical ﬁndings of Joshua Knobe, Hugh McCann and our own. In these ﬁndings, there is considerable data to show that folk judge some actions to be intentional, but not intended. We consider whether these ﬁndings support the view that some actions can be intentional, but not intended. We argue that they do not support this, and that there is something pragmatic in intentional language that better explains folk judgments. We go on to discuss the role of surveys in doing philosophy. We show in our own ﬁndings that there is an important distinction in the mind of the folk (and in the law) between acting knowingly and acting intentionally. We discover that the folk sometimes override this distinction when they wish to blame an actor for a wrong or harm. … if there is a widely shared concept of intentional action … a philosophical analysis of intentional action that is wholly unconstrained by that concept runs the risk of having nothing more than a philosophical ﬁction as its subject matter. (Mele 2001, 27)

1. Introduction In a recent paper Al Mele (2003) suggests that the Simple View of intentional action is “ﬁction” because it is “wholly unconstrained” by a widely shared (folk) concept of intentional action. ‘The Simple View’ (Adams 1986; McCann 1986) states that an action is intentional only if intended. As evidence that the Simple View is not in accord with the folk notion of intentional action, Mele appeals to recent surveys of folk judgments by Joshua Knobe (2003a, 2003b, 2004). Knobe’s surveys appear to show that the folk judge unintended but known side effects of actions to be performed intentionally. In this chapter we will reject Mele’s suggestion that the Simple View is “ﬁction”. We will also discuss the relationship between surveys and philosophical theories, and the abilities of surveys to access folk core concepts. We will argue that considerations of both fail to support Mele’s suggestion.

Intentionality, Deliberation and Autonomy

18

2. Knobe’s surveys In a series of experiments Joshua Knobe (2003b) asked subjects questions about intentional actions. In Knobe’s ﬁrst experiment, he handed out surveys to 78 people spending time in a Manhattan public park. Subjects were randomly assigned to one of two conditions: a “harm” condition or a “help” condition. Subjects read vignettes about actions that differed only by whether an actor helped or harmed the environment. The exact harm vignette was as follows: The vice president of a company went to the chairman of the board and said, “We are thinking of starting a new program. It will help us increase proﬁts, but it will also harm the environment.” The chairman of the board answered: “I don’t care at all about harming the environment. I just want to make as much proﬁt as I can. Let’s start the new program.” They started the new program. Sure enough, the environment was harmed.

In the “help” vignette, Knobe gave the same scenario, replacing the word “harm” with the word “help” (“helping”). In both the “help” and “harm” conditions, subjects were asked to rate the amount of blame (or praise, respectively) the chairman deserved for harming or helping the environment on a scale from 0 to 6 and to say whether the chairman intentionally harmed or helped the environment. The two conditions “elicited radically different patterns of responses” (ibid., 4). In the harm condition 82 per cent of the subjects said the chairman intentionally harmed the environment. In the help condition 77 per cent said the chairman did not intentionally help the environment. The difference was highly statistically signiﬁcant and stunning. Why the asymmetry? In the harm condition the folk judgments seem to accord with the view that one can intentionally do something without intending it, while in the help condition, the folk judgments seem to accord with the Simple View. This difference in judgments needs to be explained. Knobe conducted a second experiment in order to validate his results. Knobe explained why he ran the second experiment, saying: “Perhaps the results obtained in experiment 1 can be explained in terms of some highly speciﬁc fact about the way people think about corporations and environmental damage” (ibid., 4). The second experiment was structurally identical to the ﬁrst, but instead the vignettes are about sending soldiers to their possible doom. In Knobe’s second experiment he surveyed 42 people spending time in a Manhattan public park. They were again assigned randomly into “harm” and “help” conditions and given the following vignette: A lieutenant was talking with a sergeant. The lieutenant gave the order: “Send your squad to the top of Thompson Hill.” The sergeant said: “But if I send my squad to the top of Thompson Hill, we’ll be moving the men directly into the enemy’s line of ﬁre. Some of them will surely be killed.”

Folk Concepts, Surveys and Intentional Action

19

The lieutenant answered: “Look, I know that they’ll be in the line of ﬁre, and I know that some of them will be killed. But I don’t care at all about what happens to our soldiers. All I care about is taking control of Thompson Hill.” The squad was sent to the top of Thompson Hill. As expected, the soldiers were moved into the enemy’s line of ﬁre, and some of them were killed.

In the help condition the difference in vignette is signiﬁcant: A lieutenant was talking with a sergeant. The lieutenant gave the order: “Send your squad to the top of Thompson Hill.” The sergeant said: “If I send my squad to the top of Thompson Hill, we’ll be taking the men out of the enemy’s line of ﬁre. They’ll be rescued!” The lieutenant answered: “Look, I know that we’ll be taking them out of the line of ﬁre, and I know that some of them would have been killed otherwise. But I don’t care at all about what happens to our soldiers. All I care about is taking control of Thompson Hill.” The squad was sent to the top of Thompson Hill. As expected, the soldiers were taken out of the enemy’s line of ﬁre, and they thereby escaped getting killed.

Again subjects were asked to determine blame (in the harm condition) or praise (in the help condition), on a scale from 0 to 6 and to say whether the lieutenant intentionally placed the soldiers in the line of ﬁre (harm condition) or moved them out of the line of ﬁre (help condition). And again the results were similar. In the harm condition 77 per cent said that the actor intentionally placed the soldiers in the line of ﬁre. In the help condition, 70 per cent said the actor did not intentionally move them out of the line of ﬁre. The results where highly statistically signiﬁcant. 3. Knobe’s explanation of his data Initially Knobe (2003b, 7) reported that overall subjects said “the agent deserved a lot of blame (with a mean of 4.8 on the 0–6 scale) in the harm condition, but very little praise (mean of 1.4) in the help condition, and the total amount of praise or blame … was correlated with their judgments about whether or not the side-effect was brought about intentionally.” In other words, Knobe surmised that the asymmetry in praise and blame correlated well with the asymmetry in judgments of the intentionality of actions. “… they seem considerably more willing to say that a side effect was brought about intentionally when they regard the side effect as bad than when they regard it as good.” That is where Knobe left the matter in this particular paper. In another paper Knobe (2003a) says: “evaluative considerations do play some role in people’s concept of intentional action.” He believes his data show folk core concepts of intentional action entail that actions can be intentional, though not intended. They would be deemed intentional because of moral considerations. If true, there is a semantic marker for whether an action is morally right or wrong in one’s concept of intentional action. And this marker can override other markers in one’s folk concept for whether the action is intended. If true, this would indeed be surprising.

20

Intentionality, Deliberation and Autonomy

Knobe (2003a) calls his hypothesis a “dissociation” hypothesis. Normally, intentionally doing an action is associated with intending it and being skilful at doing it. One who is good at playing darts may be deemed to have intentionally thrown a bull’s eye when wanting to, intending to, and being skilful at doing so. Knobe looks at a series of examples which vary the degree of skill of an actor. As the skill diminishes, fewer people tend to say the action was performed intentionally. The one exception is in cases where there are moral considerations. If one lacks skill in shooting a gun and shoots at a rival from a great distance, people still judge the success as intentional shooting. Moral considerations override the usual association of skill and intentional outcome. Similarly Knobe would say that there is a dissociation of the intention from intentional action in cases where there are moral considerations about the outcome. The case of the CEO harming the environment is a paradigm case of dissociation. The CEO is deemed to have harmed the environment intentionally, though he did not intend the harm. We are inclined to agree that moral considerations cause subjects to dissociate intentional actions from skill and intention. For us the question is why? Is it because there is a fully articulated folk core concept of intentional action that includes a semantic marker for moral consideration? Or is there another explanation? We believe there is another, better explanation of the dissociation phenomenon. We believe that there are pragmatic factors inﬂuencing folk judgments. We do not think these pragmatic factors are part of a fully articulated folk core concept of intentional action, as we shall explain. 4. Surveys and folk core concepts We ﬁnd Knobe’s surveys to be important and interesting, as well as surprising. In particular, we ﬁnd surprising the asymmetry of judgment in the “help” and “harm” conditions. We agree that he is accessing something in people’s judgments, but we doubt that he is accessing fully articulated core concepts. We suspect that folk notions of intentional action are not clearly articulated. There are many factors required for an action to be performed intentionally. One factor involves the causal relation between an intention and the intended action. Not many folk would have very clear notions of counterfactual causal dependency of action upon intention necessary for intentional action. For instance, if an intention is connected by causal deviance to its conditions of satisfaction, the action is not done intentionally. Few folk would have clear notions of the exact relations of dependency between action and intention to block such causal deviance. Indeed, the exact relation of dependency is still in dispute among philosophers and cognitive scientists. Another factor involves skill (Mele 2001; Mele and Moser 1994). Is it possible for a non-skilled player to intentionally make a basket from the half-court line? Even among philosophers, there is a difference of opinion about not only how skilled one must be to make such a shot intentionally, but how conﬁdent one must be. Paul Grice (1971) claims one must believe that one will make the basket in order to intend and, ultimately, do it intentionally. Robert Audi (1973) believes that one’s chances must be at least as good at making it as not making it in order to intend to make the

Folk Concepts, Surveys and Intentional Action

21

shot. Anything less and one is merely trying with the hope of success. Myles Brand (1984) believes that even with “serious doubts” about success, one may intend to do something (such as make the basket), and Hugh McCann (1998, 212) maintains that any goal that guides one’s planning and acting behaviour is one that is intended (regardless of how low one’s subjective estimates of success may be). These philosophers are paid to have articulated concepts of intentional action and they do not agree. Surely it is unlikely that the folk share a single articulated concept of intentional action, in light of this. Even the folk are not consistent as to what constitutes the elements of intentional action. Malle and Knobe (1997) did an elaborate survey of the folk concept of intentional action. They found there to be at least ﬁve aspects to intentional action in the minds of the folk: belief, desire, skill, intention and awareness. They also found that no subjects indicated all ﬁve aspects (and the missing item kept changing). This supports our view that the folk do not normally possess a clearly articulated theory of the mental mechanisms of intentional action.1 A possible reason for this is that ordinary life seldom requires one to have a theory of intentional action ﬁnely articulated to the point of being able to differentiate acting knowingly from acting intentionally. In some circumstances, whether one’s act is intentional or not is irrelevant. Using mouthwash kills bacteria. No one worries about whether or not they are killed intentionally. Driving one’s car wears down the rubber on the tyres. No one thinks about whether this is done intentionally (in most circumstances). Of course, in the right context, both of these could have signiﬁcance, but that is not the norm. In other circumstances it is obvious that people act on purpose, and the folk do not need to articulate the cognitive or motivational conditions on the nature of intentions. If one robs a bank, we know this was done on purpose or intentionally, even if we do not know what it takes to intend to rob the bank. Did one merely hope to rob it, though she was not conﬁdent that she would succeed? These sorts of considerations seem not to apply. We realize that in the law ﬁner distinctions are made (such as ﬁrst degree murder, second degree murder and manslaughter). The average person only makes these distinctions in the courtroom setting.2 So the “folk” concept is not necessarily in synch with legal concepts of “intent”. We have now given some reasons to think that the folk are not likely to have fully articulated concepts of intention and intentional action. We will now give alternative interpretations of Knobe’s survey data. Later we will report data from our own surveys which tend to support our interpretation of Knobe’s data.

1 If the folk always indicated the same 4 out of 5 features, then we might be willing to accept that there was something approaching a universal folk concept. The fact that there was signiﬁcant variation in the missing features suggests there may be no single universal folk concept of intentional action. 2 We discuss these legal distinctions further in Section 6 below.

22

Intentionality, Deliberation and Autonomy

4.1 Our interpretation: it’s pragmatics Even if the folk do not have a fully articulated core concept of intentional action, almost everyone knows clearly that bad acts done intentionally are morally worse than bad acts done unintentionally. And almost everyone knows that saying “you did that on purpose” or “you did that intentionally, didn’t you?” are social ways to assign blame and of discouraging actions that one disapproves. Hence, it is very likely that folk concepts of the pragmatic dimension of intentional talk are more richly understood than the core notions of the cognitive machinery that underlies intentional action. Good actions deemed intentional are more highly praised (and encouraged). Bad actions deemed intentional are more severely blamed (and discouraged). The praise and blame associated with intentional action is part of the pragmatics of the concept, not part of the core. This is because the truth conditions for “S did act A intentionally” do not include praise or blame. It is not necessary for act A to be good or bad for the action to be intentional.3 However, the folk may associate intentionality with judgments of praise and blame due to social or evolutionary pressure (Cosmides and Tooby 1994; Mithen 1996).4 Folk may be more inclined to judge “intentional” an act they want to strongly blame and discourage. We believe that something like this is a very plausible explanation of Knobe’s ﬁndings. If the chairman in Knobe’s vignette literally “does not care at all about the environment”, then he does not intend to help or harm it. Yet in the harm condition folk largely judge that the chairman’s harming the environment was intentional. We suspect that what is going on in the minds of the folk is that they disapprove of the chairman’s indifference to the harm of the environment. They want to blame that indifference and they know that their blame is stronger and more effective at discouraging such acts, if the chairman is said to have done the action intentionally. They associate blame with intentional action (and “blame” with “intentional”). They likely do not consider whether the chairman actually intends to harm the environment or not. If it were pointed out to them that they were judging that one could do an action intentionally without intending it, it may well confront them with a cognitive disconnect and inconsistency.5 When subjects deem the chairman’s act is intentional in the harm condition, they are more likely accessing the pragmatic features of the intentional talk, than accessing an articulated core concept. One of the stunning features of Knobe’s study is the asymmetry of judgments. While the folk may judge that the chairman (and lieutenant) intentionally acted in the harm condition, they judged that they did not intentionally act in the help conditions. Yet these conditions are structurally isomorphic. What could explain this asymmetry? 3 A morally neutral act such as setting one’s watch can be perfectly intentional and yet be worthy of neither praise nor blame. 4 Consider Cosmides and Tooby’s “cheater-detection” modules which work on purposive behaviour, or Mithen’s claim that the modern human mind evolved for the purpose of tacking social facts. 5 We conducted surveys in which we tested to see whether subjects display cognitive inconsistency when confronted with it. The results are discussed at length below.

Folk Concepts, Surveys and Intentional Action

23

Since subjects judge the actions to be done intentionally in the harm conditions, why not in the help conditions? For pragmatic reasons, in the help conditions, the folk may ﬁnd the attitudes of the chairman and lieutenant so despicable that to say their actions were “intentional” would be to praise them. The language of the harm conditions seems natural (if uncaring), but the language in the help conditions seems highly strained. We cannot picture a lieutenant saying such things as “look I know we will be taking them out of the line of ﬁre and I know that some of them would have been killed otherwise. But I don’t care at all about what happens to our soldiers.” We understand that Knobe wanted to keep the vignettes in the help and harm conditions parallel. However, in the help conditions when one says something to the effect that either “I don’t care if I help the environment” or “I don’t care if I save the soldier’s lives” there is something pragmatically odd about these utterances (not to mention downright cold). Subjects might wonder why the actors do not care if the good consequence came along with what the actors did intend. Subjects surveyed might even take this indifference in the help condition to express a negative attitude about the good side effects. Not wanting to praise those who are indifferent to good outcomes, the folk are understandably reluctant to deem the agent’s acts to be intentional. Pragmatics may thus be able to help explain why the actions were not judged to be intentional (without importing praise or blame into the folk core concept of intentional action). Pragmatic forces are at work in the help condition, and they yield results consistent with the view that intentional actions are intended. So in the help condition people may be making the right judgment, but for the wrong (pragmatic) reason. Or, due to pragmatics in play in the help condition, the folk may well see that if one is indifferent to the outcome of an action, one is not intending that outcome, and not doing the action intentionally. If this is the case, the real mystery in Knobe’s results is why subjects judged the actions to be intentional in the harm conditions. To explain this we have argued that it is due to the pragmatics of intentional language and blame. Judging the actions to be “intentional” in the harm condition pragmatically implies strengthened blame. The subjects surveyed want to levy blame and they are likely not doing a mental check for consistency upon an articulated core concept of intention or intentional action. That is, they are likely not accessing an articulated core concept of intention or intentional action at all. So we suspect that the pragmatic forces at work in the harm condition do run against the view that intentional actions are intended, but mainly because no clearly articulated core concept of intention or intentional action is being consulted. In his ﬁrst study Knobe did not ask the folk whether the actors intended to harm or help the environment (rescue or put in harm’s way the soldiers). It is at least possible that in the minds of the folk, the actors did intend the respective outcomes. In that case, in agreement with Malle and Knobe (1997) we might explain the asymmetry as due to the fact that: … people may distinguish between intentions and doing something intentionally more for positive behaviours, than for negative behaviours, because it is easy (and common) to have positive intentions but harder to fulﬁll them intentionally, whereas a person’s negative intention is already deviant (and threatening to others) even before fulﬁlling it intentionally. (Malle and Knobe 1997, 116)

24

Intentionality, Deliberation and Autonomy

The subjects may blame in the harm condition on the basis of their disapproval of the attitudes of the actors alone. The subjects may take the indifference to be an intention to harm. Whereas they do not take the indifference in the help condition to be an intention to help. Since this explanation is also consistent with the view that intentional actions are intended, it offers another way of explaining the asymmetry of Knobe’s data without abandoning the view that intentional actions are intended. 4.2 McCann’s interpretation McCann (forthcoming) noted that Knobe did not ask subjects in the chairman vignette whether or not the chairman intended to harm the environment.6 Later Knobe (2004) did ask some subjects in a follow-up experiment whether the chairman intended to harm the environment, and asked others whether the chairman harmed the environment intentionally. However, McCann remained unsatisﬁed that Knobe did not ask the same subjects both questions. The thinking was that the same subjects who said the chairman harmed the environment intentionally might be the very ones who would attribute to the chairman the intention to harm the environment (if asked). McCann (forthcoming, 7) surmised that, to be consistent, subjects would say yes or no to both questions, speculating “… if respondents had to face the alleged contradiction directly [of judging one to have acted intentionally without intending it], they would be more inclined to treat intentionality and intention as going together …”. So with Knobe’s help McCann devised a survey of his own, concentrating exclusively on the “harm condition”. They divided the pool of subjects into two groups. One group would receive both questions, and the other group would receive only one question (either the question “Did the chairman intend to harm the environment?” or “Did the chairman harm the environment intentionally?”) McCann surveyed 106 subjects. Of those asked one question, 63 per cent said the chairman’s act of harming the environment was intentional. Meanwhile, 27 per cent said the chairman intended to harm the environment. Of those asked both questions, 80 per cent said the chairman’s act of harming the environment was intentional. Only 12 per cent said that the chairman intended to harm the environment. According to McCann, the results for those asked both questions were “completely unanticipated” (ibid., 9). McCann and Knobe repeated a slightly modiﬁed experiment, but found strongly similar results. In the modiﬁed experiment they asked not whether the chairman had the intention of harming the environment (which suggests that he had only one main intention), but whether he intended to harm the environment (suggesting that he may have had that intention, among others). In this case there were 99 subjects. Of those who were asked one question only, 64 per cent said the chairman’s action was intentional, and 42 per cent said his action was intended. Yet for those asked both questions 75 per cent said the chairman’s action was intentional and only 31 per cent said it was intended.

6 We too (Adams and Steadman 2004a) worried about the fact that Knobe did not survey to see if the “folk” were attributing intentions to the CEO.

Folk Concepts, Surveys and Intentional Action

25

McCann, a supporter of the view that intentional actions must be intended, took solace in the result that subjects are “more willing to say the chairman ‘intended’ to harm the environment than that it was his intention to do so” (ibid., 10). He took solace in this because he thinks that change in phrasing of the questions has a signiﬁcant impact on the subjects’ responses. When asked about “the intention” of the chairman, fewer attribute to him the intention to harm the environment, than when asked about “an” intention (even though this is not the exact wording McCann and Knobe chose). Although this represents a signiﬁcant change, the asymmetry in the condition where subjects were asked both questions remains stunning. As McCann says, subjects do not appear to “shy away” from any contradiction implicit in their verdicts (as would be true if their folk concept of intentional action required that it be intended). In the end McCann does not offer a convincing explanation of this asymmetry in folk judgment. He does, however, say the results one obtains in such surveys depend upon the way the questions are worded and that different wordings may not be correctly accessing what the folk really think (for a variety of reasons). For our part, we are happy that McCann acknowledges that in these survey situations there is something that “has partly to do with the pragmatics of the situation – with the fact that to deny the chairman’s action was intentional is to suggest it was not blameworthy … then factors that appear to tip the results … may for the most part be pragmatic, rather than anything to do with the semantics of the terms ‘intentional’ and ‘intend’” (ibid., 13). Yet McCann thinks this still does not explain well the asymmetry in the harm and help conditions. It does not explain why the CEO was perceived as blameworthy. McCann does not think that it can be as simple as that the CEO is perceived to knowingly (but not intentionally) have done something wrong. The CEO is not praised for knowingly helping the environment in the help condition. So McCann thinks there is more at stake.7 McCann wants to say that, for the harm condition, the folk are judging the act itself (ibid., 15). They believe there is a “perfect duty” not to harm the environment that cannot be overridden or set aside for proﬁts. McCann adds that, since the CEO goes forward anyway, “there is … an additional dimension of intending in the harm vignette. In setting aside the requirements of morality the chairman resolves – that is, forms the intention – to put his own and his company’s projects above moral principle, and in moving ahead he intends to do exactly that” (ibid., 16). For McCann this means there is a “hidden variable” in the vignette … an extra intention. The CEO does intend to harm the environment, after all. But why, then, do the subjects not see ﬁt to say “yes” to the “intend/intention” questions, if they see ﬁt to say “yes” to the “intentional” questions? If the Simple View that intentional acts are intended is what is prompting these results in the harm condition, as McCann suggests, then why does the same mechanism not bring the exact same numbers of responses for both questions, when both are given to the same subjects? 7 We do not think there is more at stake because we attribute lack of praise to the CEO’s admission that “he doesn’t care at all about helping the environment”. Subjects do not wish to praise such an attitude, even if it leads to helping the environment knowingly.

26

Intentionality, Deliberation and Autonomy

First, McCann explains this away by saying “the intention to harm the environment is not part of the content of the chair’s decision in the harm vignette” (ibid., 16). True, but remember this second intention is “hidden” (not explicitly mentioned in the wording of the vignette). One need not ﬁnd it in the vignette itself, on McCann’s view. Hidden variables still have causal inﬂuence on his view, so that cannot explain why subjects do not answer both questions the same. Second, McCann renews his claim above that “we are all taught as children that when you have knowingly done wrong, you don’t get to not mean it. Just the opposite: you did wrong on purpose” (ibid., 17). So, again, the CEO does intend the harm after all. But if subjects see the hidden variable (see the hidden intention), why do they not show it by marking the questions about intent and intentional action of the CEO the same? McCann has no good explanation of this. Now, again, we think the lesson is not that if one does wrong knowingly one intended to do wrong, but rather that if one does wrong knowingly, one is blameworthy. Being blameworthy, the best way to make it stick is to say “you did that on purpose” or “you did that intentionally”. And since, as we will see below, people can make the “knowingly” vs “intentionally” distinction in general, we maintain that there is in addition some sort of socialized pragmatic override of the semantics going on in these morally charged cases. Since McCann acknowledges the possibility of something pragmatic going on, our main disagreement is over just how much is attributed to the pragmatics. McCann thinks the folk judgments are consistent with the view that intentional actions are intended because the folk are judging the character of the actors in light of their actions themselves. We see this point as consistent with our claim that it is the pragmatics of praise and blame that is at work in Knobe’s vignettes and the surveys. The difference between our account and McCann’s is that we give a mechanism – the pragmatics of intentional language and judgments – to explain the phenomena. McCann explains the asymmetry of folk judgment by giving up on the distinction between acting knowingly and intentionally only when the known consequences are bad (harm condition). We think this is unprincipled, to put it mildly. On his view, you have the result that the folk blur the judgment that something is done wrong knowingly with the judgment that something is done wrong intentionally. This appears to be McCann’s view of the folk and part of his explanation of why they deem the chairman’s acts to be harming the environment intentionally. In what follows we will present data that show the folk do make the discrimination between acting knowingly and acting intentionally. So the asymmetry of judgment in the harm vs help conditions cannot be explained solely in these terms. McCann too needs some explanation to play the role of our pragmatic override in the harm condition. Merely saying one does “not get not to own it”, does not fully explain. It does not distinguish between the CEO being blameworthy because he caused harm knowingly and the CEO being blameworthy because he caused harm intentionally. In both cases, you don’t get not to own it. Thus, McCann’s explanation is still incomplete.

Folk Concepts, Surveys and Intentional Action

27

5. Our new data In previous papers (Adams and Steadman 2004a, 2004b) we maintained that Knobe’s surveys did not give respondents an opportunity to distinguish an act’s being done knowingly from its being done intentionally. We suggested that if given the option of saying the chairman harmed the environment knowingly versus harmed the environment intentionally, a signiﬁcant number would opt for the former. We thought this was important because in Knobe’s surveys he gave the folk only the option of saying the chairman’s act was intentional or not intentional. Since subjects were inclined to accept that it was not the chairman’s intention to harm the environment, their judging that the chairman harmed the environment intentionally (though the only option given by Knobe) may actually create cognitive tension. We surmised that subjects would opt for judging that the chairman harmed the environment knowingly, but not intentionally, if given the option. This choice would relieve that cognitive tension, and, at the same time, allow subjects to blame the chairman for knowingly harming the environment. We believe that a large number of the respondents in Knobe’s original experiments only said the chairman harmed the environment “intentionally” because they wanted to blame his actions. We maintain that acting knowingly versus intentionally is a real distinction that most folk know how to make, and do make. A simple example is one such as the following. When we drive our cars we wear down the tyres and know that we do. However, it is neither a desire of ours nor a plan that we act upon to wear down the tyres. We do so knowingly, but not intentionally. Were we to do so intentionally we might squeal the tyres, engage in fast starts and stops, corner hard, and so on. A more exotic example is the following. Trapped on a narrow ledge with no way to escape but to jump over a chasm, Al might decide to try to jump to escape. Al knows that his chances of success are slim and knows that his jumping may ultimately cause his death. If Al jumps and does not make it, he causes his death knowingly, but he does not commit suicide (that is intentionally cause his own death). We were convinced that the folk can and do make this distinction. So we designed a survey of our own to test this. With the above considerations in mind, we designed and conducted a new experiment. The goals of the study were a) to see what would happen if subjects were offered the opportunity to distinguish an action performed “knowingly” from an action performed “intentionally” in their responses, and b) to test subjects’ usage of intentional language in a morally neutral scenario (thus removing the praise/blame pragmatic implication issues from the equation). We wanted to remove the praise/ blame aspect of the judgment because it is our hypothesis that the desire to blame the chairman for his morally bad actions can interfere with one’s ability to discriminate actions done knowingly from those done intentionally. After all, they will judge an action to be intentional even if not intended, when there is blame to be levied. We surveyed 49 summer undergraduate researchers at the University of Delaware. The subjects were randomly divided into two groups. The ﬁrst group received Knobe’s “harm” vignette, identically reproduced. They were then asked:

Intentionality, Deliberation and Autonomy

28

Please select the best answer: “The chairman harmed the environment knowingly, but not intentionally.” “The chairman harmed the environment knowingly and intentionally.”

Twenty per cent of respondents answered “knowingly, but not intentionally”, indicating that the folk are capable of making such a distinction when offered the chance. Still, 80 per cent of subjects said that the action was done “knowingly and intentionally”, telling us that their desire to blame is hard to override. These results demonstrate that the folk can and do make the distinction between acting “knowingly” and “intentionally”, as well as showing just how difﬁcult it is to override the folk’s pragmatic programming to use intentional language to levy blame. The second group received the same vignette with the following modiﬁcations designed to make the scenario morally neutral while maintaining Knobe’s structure: The vice-president of a company went to the chairman of the board and said: “We are thinking about starting a new program. It will help us increase proﬁts, but starting the new program will tip off our competitors.” The chairman of the board answered: “I don’t care at all about tipping off our competitors. I just want to make as much proﬁt as I can. Let’s start the new program.” They started the new program. Sure enough, this tipped off their competitors.

The subjects were then asked: Please select the best answer: “The chairman tipped off the competitors knowingly, but not intentionally.” “The chairman tipped off the competitors knowingly and intentionally.”

For this question, 71 per cent of respondents chose “knowingly, but not intentionally”, whereas only 29 per cent went for “knowingly and intentionally”. Note that the results in this morally neutral scenario are exactly the opposite of what one receives in the “harm” condition. Table 1.1 provides a summary of our data. Table 1.1

Summary of our data Environment “harm” Vignette

Tipping off the competition Vignette

Knowingly and intentionally

80%

29%

Knowingly, but not intentionally

20%

71%

Folk Concepts, Surveys and Intentional Action

29

In all of these vignettes the chairman clearly states that he “does not care at all” about the suggested side effect. If one truthfully does not care at all about a certain possible outcome, it makes no sense to say that he or she intended it. Even in Knobe’s experiments, few of the folk attribute the intention to harm the environment to the chairman. Our results demonstrate that the folk are capable of making the distinction between a result caused intentionally and a result caused knowingly. It is only in a morally neutral case, however, that the majority of respondents are freed of their pragmatic programming so that they can make judgments purely on what they think about intentional action; and it seems that here their thinking is in keeping with the view that intentional actions are intended. We attribute the difference between the high number of subjects who pick “knowingly and intentionally” in the “harm” case versus the high number who pick “knowingly but not intentionally” in the “morally neutral” case to pragmatic override. That is, we believe that even if one can make the distinction between acting knowingly and acting intentionally (as our data clearly indicates), the life-long habit and social training of pragmatic use of intentional language to assign blame is hard, if not impossible, to override. Both Mele and McCann had much earlier suggested to Knobe that subjects might be instructed that a person can be blameworthy for an act that one did not perform intentionally. And Mele and McCann both believed that this “instruction” might cause subjects to respond differently in the “harm” kinds of cases. It was even suggested that instead of “instructing” subjects, one might draw them to this realization by a type of “Socratic questioning”. However, Knobe (2003a) conducted experiments in which the “instruction” had no signiﬁcant effect on the responses in the “harm” conditions of his surveys. This shows how deeply ingrained is the pragmatic use of intentional language, even when it has been recently brought to a subject’s attention that blame does not require attribution of intention. Attribution of intention is the usual way, the habitual way of assigning blame – a habit that is very hard to break or override. We suspect that part of the explanation for this is that subjects compartmentalize. In one cognitive compartment they keep the information that it is indeed possible to blame without attributing intention or intentionality. Yet when the situation for blaming arises, from another cognitive module, their tendency to blame kicks in and overrides information stored in the other module. Compartmentalization is not an uncommon phenomenon. For example, in an introduction to philosophy course, students learn to critique the arguments for the existence of God. In an exam where they give these critiques of the arguments, if asked “Do you believe that God created the world?” very many will say: “Yes, the world had to have come from somewhere.” This answer discounts their own critique of the Cosmological Argument earlier in the same exam. Even if they appear to ﬁnd the critique of the Cosmological Argument convincing, many students tend to compartmentalize and not let their critique interact with their religious convictions formed prior to taking the class.

30

Intentionality, Deliberation and Autonomy

6. Surveys and philosophy The challenge from the opening quote by Mele seems to suggest that one’s philosophical theory of action should conform to folk judgments (concepts) of action. If they do not, the philosophy is ﬁction. We see several problems. First, we deny that the folk have clearly articulated concepts or theories of intentional action. We have given reasons above to believe that even philosophers disagree about these matters, when they try to articulate such concepts (theories). Second, why think that a philosophical theory should conform to folk judgments? We would not think that relativity physics or mathematics of orders of inﬁnity should conform to folk judgments, so why should philosophical theories of action? The only answer we can think of is that Mele must not think philosophical theories go that deep. The nature of intentional action is not hidden (in the way the fundamental nature of reality might be) from the reasoning of the common man. He seems to think that whatever intentional action is, the folk ought to be able to ﬁgure it out. What is surprising to us is that Mele has spent his philosophical career working on articulating concepts and theories about the nature of intentional action. It cannot be so easy that the folk all do it. Third, we suspect that the correct role for surveys of folk judgments is to measure what people think about reality (intentional action), not what reality (intentional action) is. We agree that there can be philosophical interest in ﬁnding out what the folk think about intentional action, and why they think it. That is why we have used surveys ourselves. Nonetheless, we have a deeper concern whether or not surveys can access the kinds of subtle distinctions we are appealing to in our explanations of folk judgments. That is, we are not sure that either Knobe’s or our own surveys are sensitive enough to determine whether subjects are making judgments based upon the semantics of intentional language or the pragmatic implications of the use of that language. We suspect that none of the surveys we have seen up to now are able to make these discriminations, and we have not yet decided how to go about testing for them in this context. We have tried to begin to get at such differences by seeing if subjects can distinguish doing something knowingly from doing it intentionally. Still, how can a subject’s checking a box on a survey tell us whether it is the semantics or the pragmatics of the terms they are employing that is responsible for the answers? Until we have an answer to this question, we do not think the dispute between us and Knobe, Mele, McCann and others will be able to be decided by surveys. We do think that it is the pragmatics of intentional language that is responsible for the results that Knobe, McCann and others are getting on the surveys. We have given philosophical reasons why we think this. But we do not think surveys alone can decide (at least, not those employed so far). While we have argued that the folk do not have a fully articulated concept of intentional action, we do believe that the seeds of the distinction we test for between acting intentionally and knowingly do exist in the minds of the folk. This is of practical importance for the law. Thus our philosophical view of intentional action is not going to be “ﬁction” by Mele’s lights. We say this because we ﬁnd reﬂected in the law this very distinction between acting knowingly and intentionally. In the

Folk Concepts, Surveys and Intentional Action

31

Delaware State Code, Chapter Two, Section 231, the following distinctions are made concerning the state of mind of agents committing offences: (a) “Intentionally”, – A person acts intentionally with respect to an element of an offense when: (1) If the element involves the nature of the person’s conduct or a result thereof, it is the person’s conscious object to engage in conduct of that nature or to cause that result; and (2) If the element involves the attendant circumstances, the person is aware of the existence of such circumstances or believes or hopes that they exist. (b) “Knowingly”, – A person acts knowingly with respect to an element of an offense when: (1) If the element involves the nature of the person’s conduct or the attendant circumstances, the person is aware that the conduct is of that nature or that such circumstances exist; and (2) If the element involves a result of the person’s conduct, the person is aware that it is practically certain that the conduct will cause that result.

The Code goes on to specify even more differences in state of mind for acts done “recklessly” or out of “criminal negligence”, or simply “negligence”, as well. Concentrating mainly on the distinction between acting intentionally and knowingly in the law, we see that the Code identiﬁes an intentional act with one where the agent has the act consciously as a goal object, and believes or hopes that the circumstances necessary for carrying out the action exist. Whereas, the agent who acts knowingly is aware that his conduct is of such a nature that it may bring about the result and is aware that it is practically certain that his conduct will cause the result. What is missing in the latter case is the conscious goal. We would further point out that the distinction is clearly written to categorize acts that may be illegal. This shows that at least the law can override the pragmatics of intentional talk and realizes that some things may be blameworthy and punishable, even when done “knowingly” but not “intentionally”.

7. Conclusion In this chapter we have offered a defence of the Simple View of intentional action in the face of Mele’s claim that it is “ﬁction”, if not in accord with empirical folk surveys such as those ﬁrst used by Knobe. We have offered an alternative explanation of Knobe’s results (that it is the pragmatics of intentional language) that is consistent with the truth of the Simple View that intentional actions are intended. We have replied to Knobe’s objections to our alternative explanation of his results. We considered McCann’s further experimental results and his own explanation of them. We have given reasons to think McCann’s explanation is insufﬁcient to explain the asymmetry of folk judgments in the “help” and “harm” conditions. We have offered

32

Intentionality, Deliberation and Autonomy

results of our own surveys that support our explanation that it is the pragmatics of intentional language that is accounting for the asymmetry. We discovered that the “folk” can and do distinguish acting knowingly from intentionally. We suggested that they do this because they feel the cognitive pull of the Simple View (that intentional actions are intended). Still we recognize that we need to appeal to mechanisms such as pragmatic override and cognitive compartmentalization to explain why the folk abandon the “knowingly/intentionally” distinction, when it comes to the harm condition. If there is further empirical work to be done in this area, we think it will be in trying to ﬁgure out the cognitive mechanisms of compartmentalization and pragmatic override.8 References Adams, F. (1986), ‘Intention and Intentional Action: The Simple View’, Mind & Language, 1, 281–301. —— and A. Steadman (2004a), ‘Intentional Action in Ordinary Language: Core Concept or Pragmatic Understanding?’, Analysis, 64, 173–81. —— (2004b), ‘Intentional Action and Moral Considerations: Still Pragmatic’, Analysis, 64, 268–76. Audi, R. (1973), ‘Intending’, Journal of Philosophy, 70, 387–403. Brand, M. (1984), Intending and Acting: Toward a Naturalized Action Theory, Cambridge, MA: MIT Press. Cosmides, L. and J. Tooby (1994), The Adapted Mind, Oxford: Oxford University Press. Grice, H. (1971), ‘Intention and Uncertainty’, Proceedings of the British Academy, 57, London: Oxford University Press. Knobe, J. (2003a), ‘Intentional Action in Folk Psychology: An Experimental Investigation’, Philosophical Psychology, 16, 309–24. —— (2003b), ‘Intentional Action and Side Effects in Ordinary Language’, Analysis, 63, 190–93. —— (2004), ‘Intention, Intentional Action and Moral Considerations’, Analysis, 64, 181–7. Malle, B. and J. Knobe (1997), ‘The Folk Concept of Intentionality’, Journal of Experimental Social Psychology, 33, 101–21. McCann, H. (1986), ‘Rationality and the Range of Intention’, Midwest Studies in Philosophy, 10, 191–211. —— (1998), The Works of Agency: On Human Action, Will, and Freedom, Ithaca, NY: Cornell University Press. —— (forthcoming), ‘Intentional Action and Intending: Recent Empirical Studies’, Philosophical Psychology.

8 We would like to thank the University of Delaware’s Ofﬁce of Undergraduate Research for support for this project. We also thank Al Mele, Joshua Knobe, Hugh McCann and Shaun Nichols for helpful discussions.

Folk Concepts, Surveys and Intentional Action

33

Mele, A. (2001), ‘Acting Intentionally: Probing Folk Notions’, in B. Malle, L. Moses and D. Baldwin (eds), Intentions and Intentionality: Foundations of Social Cognition, Cambridge, MA: MIT/Bradford. —— (2003), ‘Intentional Action: Controversies, Data, and Core Hypotheses’, Philosophical Psychology, 16, 325–40. —— and P. Moser (1994), ‘Intentional Action’, Noûs, 28, 39–68. Mithen, S. (1996), The Prehistory of the Mind, London: Thames and Hudson.

This page intentionally left blank

Chapter 2

Action and Responsibility – A Second Look at Ascriptivism Ralf Stoecker

Ascriptivism is the thesis, put forward most prominently by H.L.A. Hart, that action sentences are not descriptive but ascriptive sentences which ascribe responsibility to the agent. The ﬁrst part of this chapter lists eight valid objections against ascriptivism that explain why it never became popular in action theory. The second part defends a modiﬁed version of ascriptivism, which admits that action sentences are descriptive (contrary to ascriptivism) but which adopts the idea that responsibility is essential for agency (contrary to the received views in action theory). According to this conception, to say that an agent acts is to say that something (the result of what she does) is explainable by reference to the agent’s being responsible for it. As it turns out this conception avoids the difﬁculties of traditional ascriptivism and also promises solutions for some of the most persistent problems in modern action theory (for example the inclusion of negative acts, the individuation of actions and the problem of wayward causal chains). 1. Ascriptivism Ascriptivism is an offspring of the so-called linguistic turn in philosophy, applied to the central question of action theory, namely: what are actions? It was put forward by H.L.A. Hart in his classic article ‘The Ascription of Responsibility and Rights’ (Hart 1948/49). According to Hart it is not surprising that the philosophers did not succeed in ﬁnding a satisfactory answer to the question what actions are, since the whole question is due to a misunderstanding of the logic of agency ascriptions. As with other problems in philosophy, a too simple model of the functioning of language creates Scheinprobleme (as Wittgenstein had argued with respect to many philosophical questions of the form “What is X?”). With regard to action theory, according to ascriptivism the mistake lies in the implicit assumption that action sentences like “He did it” are descriptive sentences, while in fact they are what Hart calls ascriptive sentences, sentences ascribing responsibility to a person.

Intentionality, Deliberation and Autonomy

36

From the premise 1. It is essential for action sentences that they ascribe responsibility to agents, ascriptivists infer the following statements: 2. One should better not regard action sentences as descriptive; 3. Action sentences do not describe actions; 4. It is pointless to ask what actions are. Ascriptivism never became very popular though. The Philosopher’s Index gives fewer than a dozen entries, mostly from the 1960s and 1970s, and even Hart himself later on retracted ascriptivism and wrote his monumental Causation in Law (together with Tony Honoré) without mentioning it at all. Moreover, there are indeed a number of excellent reasons to be sceptical about ascriptivism. Still, it seems to me that there is much truth in ascriptivism which has not been sufﬁciently noticed by action theorists. In this chapter I shall ﬁrst list a number of serious objections against ascriptivism and then argue that in spite of these difﬁculties the most attractive conception of human action could be regarded as a modiﬁed version of ascriptivism which I shall call neo-ascriptivism.

2. Objections against ascriptivism The ﬁrst group of objections is directed against the semantic thesis that action sentences are non-descriptive (2). (I)

As Peter Geach has emphasized, reading action sentences as non-descriptive creates problems for an adequate understanding of their role in logically complex sentences, for example in conditionals (Geach 1960). Prima facie the sentence: “If the gardener killed the king, then he also killed the queen.”

(II)

has a truth value which depends on the truth values of the antecedent and the consequent. Yet if these two, being action sentences, are non-descriptive, they do not have truth values. Hence the whole sentence could not be truthfunctional in the familiar way, nor could it be an ascriptive sentence itself (since as it stands it does not ascribe any responsibility to the gardener). Its semantic status remains mysterious. Normally, when we talk about people’s actions what we have in mind is not responsibility ascription at all but simply the description of what happens. “The waiter seated me near the entrance, and then I waited for you.”

Seating someone is an action, but it seems absurd to deny that the sentence provides a description of a particular scene in a restaurant.

Action and Responsibility – A Second Look at Ascriptivism

(III)

37

The claim that action sentences are not descriptive is particularly odd in explanatory contexts: “Because he felt ill, Peter took a pill which immediately put him to sleep.”

We certainly work on the assumption that our actions are part of the world’s history, triggered by things that happen to us and in turn exerting inﬂuence on future events. The objections of the next group concern the ascriptivist’s claim 1, that is, that there is a tight connection between action and responsibility. (IV)

Ascriptivism seems to take for granted that responsibility is a unique relationship, while in fact there are different kinds of responsibility, for instance moral, legal and causal responsibility, which are not even coextensional. As we all know, a politician could be morally responsible for the death of many people without being legally responsible. On the other hand there may be legal systems (perhaps ours is one of them) which hold people legally responsible for things they are not morally responsible for. Causal responsibility, ﬁnally, has a much larger extension and also allows far more things to be responsible than moral or legal responsibility do. (The tsunami of December 2004 was causally responsible for excessive destruction in southeast Asia without even being capable of being morally or legally responsible.) Hence, if action sentences ascribe responsibility, as ascriptivists believe, one would like to know which kind of responsibility they ascribe. (V) In any case, not everything we are responsible for is an action. Lee Oswald was responsible for John F. Kennedy’s death; Christo and Jeanne-Claude were responsible for the appearance of New York’s Central Park in 2005; in the former German government Joschka Fischer was responsible for foreign affairs; a school teacher is responsible for his pupils. Neither deaths or appearances nor foreign affairs or children are actions though. One may therefore wonder what the difference is between those responsibility ascriptions that constitute agency and others that do not. (VI) While there are many things we are responsible for that are not actions, there also seem to be actions for which we are not responsible. At least prima facie, no one is always and some people are never responsible for what they do. A driver who suddenly and unexpectedly suffers a heart attack and drives her car into the window of a shop is not responsible for breaking the window; still it was she who smashed it. Small children are notoriously irresponsible for what they do, but they certainly are not inactive. Therefore, it seems, responsibility could not be a necessary condition for agency, and so not all action sentences could be construed as responsibility ascriptions. (VII) It is tempting to argue, in response to objection (VI), that doings like driving into a window or the activities of children are not really actions, that they are only behaviour of some sort, while real actions are always something the agent is responsible for. Since it is often assumed as almost a triviality that actions are what we are responsible for, it is worth noting that this is not

38

Intentionality, Deliberation and Autonomy

the way we usually talk about actions.1 Oswald killed Kennedy but nobody would say that Oswald was responsible for his killing of Kennedy; what he was responsible for was Kennedy’s death (which is not an action). Given our ordinary way of speaking, our own actions are almost the only things in the world we are never responsible for. So why should we assume that action sentences are in fact ascriptions of responsibility? The last objection is directed against Hart’s claim (4) that ascriptivism could account for the failure of philosophical attempts to answer the question what actions are: (VIII)

On the one hand it is not obvious that providing necessary and sufﬁcient conditions is in general the only or even the best way for answering the question what something is. After all, there have not been many successful attempts in philosophy to give such an answer to a what-question. (Compare the analogous difﬁculties of giving necessary and sufﬁcient conditions for knowledge in epistemology.) Hence the whole enterprise of elucidating central philosophical concepts by way of specifying necessary and sufﬁcient conditions could be ill-advised from the beginning, irrespective of whether these concepts are descriptive or not. On the other hand it is not clear to what extent Hart’s ascriptivism really could spare us the question what actions are. In the passage where he introduces the term “ascriptive” he writes that the principal function of action sentences is “to ascribe responsibility for actions” (Hart 1948/49, 171, my emphasis). But then one may wonder what these actions are that the agent is responsible for, that is what actions are at all. The central problem of action theory would then return.

In view of these difﬁculties it is not surprising that ascriptivism did not ﬂourish. So why bother about an obsolete theory? 3. Hopes conjoined with ascriptivism The main reason for reconsidering ascriptivism is that it might solve a number of obstinate problems that accompany the standard conception of actions. According to the standard view, actions are events (usually bodily movements) which are set off by mental phenomena (propositional attitudes, acts of will, decisions) that also reveal those events as being reasonable (to a certain degree). There are several notorious problems for the standard view, including:

1 This observation was already made by Georg Pitcher (1960). The usual assumption is exempliﬁed in Paul Edwards’ Encyclopedia of Philosophy, vol. 7, 183: “Persons are normally judged morally responsible for their actions. But they may be judged responsible for almost anything – events, processes, their own psychological characteristics.” There follows a list of examples for the latter kind of responsibility ascription, but there are none for the former.

Action and Responsibility – A Second Look at Ascriptivism

39

A. difﬁculties of action individuation (choosing between ﬁne-grained and coarse-grained conceptions, the time-of-a-killing-problem, a satisfactory account of “by”-sentences); B. difﬁculties of giving an adequate account of omissions and commissions; C. the problem of wayward causal chains; D. problems of mental causation. These and other problems have been tackled by defenders of the standard conception in numerous ways, leading to various brands of the standard view. What I want to suggest, though, is that they are evidence for some fundamental ﬂaws in the standard conception and that actions should be understood rather along the lines proposed by ascriptivism. The remainder of this chapter develops this suggestion. 4. What is right and what is wrong with ascriptivism? What is wrong with ascriptivism is the claim that action sentences are not descriptive (2). To be sure, we employ action sentences for different purposes, and one of these purposes is to praise or blame an agent for what she has done. Yet we do this by pointing out something about the agent that may be true or false. A boy who tells his mother that his sister has eaten all the cookies may thereby express admiration or outrage, but in any case he describes something concerning his sister. Objections (I) and (II) are sound. What is special about action sentences is not their mode but their content. In fact not only action sentences are descriptive; the same holds for the ascription of responsibility too. The statement that Oswald was responsible for Kennedy’s death is true or false, no matter if anybody blamed him for his deed. His responsibility is a fact belonging to US history. But given that responsibility ascriptions are also descriptive, one can reject the idea that action sentences are not descriptive and still stick to claim 1 that action sentences attribute responsibility to the agent. Such a new, semantically conservative ascriptivism would retain the close relationship between action and responsibility and build it into an account of agency. The easiest way to do this would be to say: it is essential to action sentences that they ascribe responsibility to agents because an action is just the behaviour an agent is responsible for. But such a claim would still be confronted with the other objections listed earlier. It must therefore be amended. The key for a more promising ascriptivism is objection (VII), the surprising observation that we rarely if ever speak of an agent’s being responsible for his action. Oswald was not responsible for his killing of Kennedy; he was only responsible for Kennedy’s death. Why then is it so intuitively plausible that actions are at the core of what we are responsible for, although we do not attribute responsibility for them? A ﬁrst, provisional answer is that by describing an agent as being responsible for some non-action we attribute agency to him. Actions are not what the agent is responsible for; they are his being responsible for things other than actions, like states or events.

Intentionality, Deliberation and Autonomy

40

This could not be quite correct though, since strictly speaking there are no ‘being-responsibles’. An agent’s being responsible is nothing in the world. This leads to a second respect in which ascriptivism was on the right track (besides the insistence on responsibility): the persistent question what actions are is based on a misunderstanding of the logical form of action sentences. Ordinary action sentences (like “Oswald pulled the trigger and killed Kennedy”) do not talk about actions – they do not refer to actions – they only talk about people (Oswald, Kennedy), things (the trigger), maybe events (a death) (and other action sentences may also talk about states, conditions, facts); and they attribute agency to some of the people (Oswald).2 What makes them action sentences is not what they talk about but what they say about it: they describe the agent as being responsible. Objection (VII), that we never hold agents responsible for actions, and objection (V), that we hold them responsible for various other things, turn out to be two sides of the same coin. Holding people responsible for things is attributing agency to them. (In fact this claim is much too strong and will be modiﬁed soon.) In addition to objections (VII) and (V) the proposal also meets objection (VIII), since it rejects the ontological presumption that actions are entities of a special kind, without leaving any room for worries about what kind that is. (Strictly speaking, that is ontologically strictly speaking, the answer to the question what actions are is: there are none. Actions do not belong to the entities that make up the world.) Perhaps the most serious challenge to ascriptivism is objection (III). As it was stated earlier, objection (III) was only directed against the thesis that action sentences are non-descriptive: if action sentences could function as explananda or explanantia, how could one ever doubt that they are descriptive? As soon as it is agreed that action sentences are descriptive the problem is solved. But the explanatory role of action sentences is not just one of many roles they play. It is commonly agreed that in some sense or other it is essential for actions to be explainable in certain ways. Actions are due to agents. No construal of agency is adequate if it does not account for actions as a kind of product of agents. Moreover, action explanations frequently refer to the agent’s reasons, intentions or mental states, the function of which seems to be essential for actions as well: actions usually are performed because of how things appeared to the agent. An adequate explanation of agency has to take this in account too. The standard view can accommodate these aspects since it regards actions as being caused by the agents’ reasons. (Although some defenders of agent causality doubt that such a relationship between actions and reasons already captures the essence of actions as being brought about by agents.) But for ascriptivism the situation looks worse, since responsibility seems to be independent of any kind of production or causation. (This is obvious in cases where people are responsible for other people, as, for example, the teacher is for his pupils.) So how could one maintain that actions essentially are due to agents and at the same time hold that what is special to actions is that the agent is responsible for them? In response to these worries one could be inclined to fall back on the distinction between different sorts of responsibility. A teacher may be morally and legally 2

I have said more about the logical form of action sentences in Stoecker 2004.

Action and Responsibility – A Second Look at Ascriptivism

41

responsible for his pupils but certainly he is not causally responsible for them. One could therefore assume that ascriptivism has to be read as the thesis that action consists in an agent’s causal responsibility for something. Oswald killed Kennedy since he was causally responsible for Kennedy’s death. Such a reading would preserve the idea that action is to be construed relationally, as a relation between an agent and what von Wright has called the result of the action, but it would reduce the responsibility component of ascriptivism to a terminologically extravagant twin of the causal condition of the standard view. Instead of demanding that agents or their intentions should cause the actions, one would require them to be causally responsible for the actions, which would not be of any advantage. Moreover, if one were to conﬁne ascriptivism to causal responsibility one would forgo any hopes to specify what is peculiar to actions in contrast with other doings. But there is a better ascriptivistic answer to the objection that it is essential for actions to be in some sense elicited or generated by the agent, an answer which in a sense turns matters upside down. The ascriptivist should maintain that in cases of agency the result of the action is explainable with recourse to the agent in so far as the agent was responsible for it. To ascribe an action to a person is not simply to say that the person is responsible for something (the result of the action), it is to state that the result can be explained from the fact that the person was responsible for it. Oswald killed Kennedy since he was responsible for Kennedy’s death and his being responsible explained Kennedy’s death. But how can we explain something with recourse to an agent’s responsibility for it? To see what I have in mind it is necessary to step back and start at the very core of responsibility, investigate its explanatory role for actions, and then try to develop the account from the inside out. 5. The role of responsibility ascriptions in action explanations The core of responsibility, etymologically and historically, is the request for a ‘response’, more particularly the request for a justiﬁcation of what one has done. A person is asked to account for her doings by giving reasons that speak in favour of them. Responsibility in this sense is a particular role in a social or communicative practice. The responsible person has to defend what she has done to fend off the thread of being found guilty and blameworthy or to earn praise and reward. It is quite easy to see how the fact that a person is responsible in this sense (that is in the sense of being called to account) for having done something could also explain what the person has done: in order to avoid being blamed and scolded by the others or in order to be praised and rewarded by them the agent takes care only to do such things she could successfully defend afterwards. As everybody knows, being examined may be an unpleasant event if one could not give satisfactory answers. Hence the thread to be confronted with such a situation, together with the positive expectation of reward, may easily explain the appropriate behaviour. So far responsibility only accounts for a small minority of actions. Only very few of our actions are performed in the expectation of a consecutive critical examination. But this used to be different. When we were small children a lot of what we did

42

Intentionality, Deliberation and Autonomy

was accompanied by a running commentary of others, usually our parents. With their praise and encouragement as well as with their complaints and perhaps even punishment parents establish responsibility relations. They ascribe some of the things that happen (usually bodily events) to the child as his or her act. (These are not always the typical examples for bodily movements we know from modern action theory, but also acts like belching, ﬁlling the diapers and so on.) To that extent the parents’ linguistic reactions are on a par with other positive and negative reinforcements and lead to similar conditioning effects. But soon (the sooner the better) these bold evaluations of the very early responsibility ascriptions broaden into an evaluative discourse: parents add reasons for their judgements, they weigh practical arguments and draw conclusions from them. The children learn that part of what happens is subject to a form of discourse that is centred on agents: “You should have eaten breakfast, then you wouldn’t be hungry now”, “Why did you beat your sister?” Moreover, the adults themselves accompany what they do by explicit reasoning: “I’d better give you the warm coat, it’s chilly outside”, “I was sure you’d like spaghetti more than potatoes.” And the children experience such dialogues also between adult people (“You should’ve given her a warm cap, too, she just had an inﬂammation of the middle ear”). Moreover, as soon as the children learn to talk they learn to participate in such communications (“She ate all the cookies!”), and then they tirelessly practise them in role-playing with other children and with dolls and puppets. But the point of these exercises is not merely to enable the children to take part in communication with their parents or with other people; the point is to enable them to act in accordance with the demands of such a communication (if it were to take place). And the way they achieve this aim is by accompanying whatever they do by such a running commentary. A two-year old toddler for example who does something shady would typically accompany her doing by mumbling the words she would expect her mother to utter in the situation. The children feign being in constant dialogue while they act, and thereby they adjust their behaviour to the standards of such a dialogue. They slowly start to act responsibly. The habit of mumbling practical dialogues to oneself in order to act better is not conﬁned to children though. It is what we always do when we deliberate about our actions, although we are much better now in doing it silently (and it is usually more difﬁcult to identify our mother’s voice in it). Thinking, in particular practical thinking, is ﬁrst and foremost talking,3 sometimes in combination with other devices that help us in real discussions too, for example drawing schemas, pacing up and down, staring into the air. Our established habit of performing a kind of Loya Jirga in our heads allows for the explanation of far more action results than the prospect of being taken to account afterwards: since we are disposed to act on the judgements of the advisory hearing we play through in our thoughts, the results of our actions can often be explained with recourse to these dispositions. The results obtain because we are responsible for them and because our habit of reasoning disposes us to act responsibly. 3 No doubt we rarely use complete well-formed sentences; we talk elliptically, in hints, as we do when we communicate with a very close mate.

Action and Responsibility – A Second Look at Ascriptivism

43

Still the vast majority of our actions are not deliberate in this sense, nor could or should they be. Talking with others and feigning such discussions in our hearts are slow, unstable procedures that could never direct us through our multifarious active lives. This is as true for small children as it is for elderly philosophers. While the children learn how to act in accordance with the practical deliberations of the authorities, they simultaneously learn many other things too: they learn how to handle building bricks and spoons, how to keep one’s distance from ﬁres and knives, and perhaps even how to treat their baby sister. They know a lot about correct behaviour which they normally need not reconsider. When they get older they increase both capabilities, acting according to practical considerations and doing the right things without consideration, until ﬁnally they achieve the level of competence we have, which is characterized by two skills: ﬁrst to act frequently as if we had deliberated about what to do, without actually haven done so, and secondly, to engage in real, practical deliberations, when the ﬁrst skill threatens to fail. These two capacities constitute the very core of human agency. With respect to our domain of responsibility we take care that things accord with our practical deliberations, sometimes by actually performing such a deliberation (our silent Loya Jirga), but mostly because we are disposed to behave as if we had performed it. These capacities allow for a unique explanatory strategy with respect to agents: one may explain things from the fact that the agent is responsible for them. Her being responsible for them makes them an apt object of her practical deliberation, which in turn allows us to explain them by referring to aspects that would speak in favour of them in her deliberation. To say that an agent acts is to point out that a certain event or fact is in this way explainable – explainable because she is responsible for it. 6. Discussion of the neo-ascriptivistic account It is one of the advantages of this modiﬁed ascriptivism that it may easily embrace so called negative actions, omissions and commissions. (This was difﬁculty B for the standard account, mentioned earlier.) A large part of what we do consists in non interfering, in letting things happen, which to my mind entails that we are responsible for these occurrences and moreover have them under control. They continue, because they meet our plans and demands. Unfortunately, though, we are not always perfectly reasonable agents. Sometimes we loose control and things happen in our domain of responsibility which do not agree with our practical deliberations (whether explicit or potential). They may still constitute actions, but in this case the actions are not intentional. If at breakfast table I reach for my newspaper and knock over my wife’s coffee mug, reaching for the paper is an intentional action of mine, since the arm movement is an expression of my skill to behave according to my practical reasoning (potential practical reasoning, because I certainly would not waste a thought on groping for the paper). Knocking over the mug, on the other hand, is not intentional, since the mug does not topple because practical reason recommends spilling coffee. Nonetheless it is an action of mine, an unintentional action, and it can be explained by referring to my

44

Intentionality, Deliberation and Autonomy

disposition to act according to practical reason: I have done it by doing something else intentionally (namely reaching for the paper). As is well known, agency constitutes so called “action trees”, built up by the words “by” or “thereby”. Most things we do we do by doing something else. Now the coffee example makes vivid that “by” relates explanations, or better: statements of explainability – which is just what action ascriptions are. I was reaching for the newspaper and thereby knocked over the coffee mug. This is true because ﬁrst my arm movement was explainable by referring to me, and secondly because the fact that my arm movement was explainable by referring to me justiﬁes the claim that the toppling of the mug was explainable by referring to me too. In this way the neoascriptivist proposal of construing action ascriptions as statements of explainability based on responsibility can account for the peculiarities of action trees that are so tricky for the standard theory (difﬁculty A).4 As stated, the proposal sketched here can also cope with the other problems for the standard account listed above. In particular, it can be shown that the explainability of a result with recourse to an agent’s being responsible for it is still a particular kind of causal explainability. By achieving the two capacities that lie at the ground of agency we achieve a causal power, the power that within our domain of responsibility the world usually accords to our wishes. The notion of a causal power was already employed by Davidson in the context of action explanation (Davidson 2001, 64). It has the advantage to allow for an attractive combination of neo-ascriptivism and causalism that solves problem D of the standard account. Moreover it can be shown that all cases of causal powers exhibit phenomena similar to the standard examples of wayward causal chains and that they are due to a peculiarity of these kinds of explanation, which allows for a solution of the problem of waywardness (problem C) as well.5 But there are still some difﬁculties for ascriptivism that have not yet been tackled here. Objection (VI) points out that we sometimes say that an agent has done something but is not responsible for it. Children in particular are very active, although we seem to exempt them from responsibility. Even if this is not quite true, since, as stated earlier, it is a core part of children’s learning history that they are involved in practical discourse that ascribes responsibility to them, it points to the fact that our usage of “doing” and “acting” is much looser than described here. In December 2004 a tsunami killed thousands of people in Asia. Tsunamis can not think. What the tsunami did was not an action. But there is a striking parallel between saying that a tsunami killed people and that Oswald killed Kennedy. As with Oswald, saying that the tsunami killed people entails that one can explain the death of these people by the tsunami. To say that someone or something does something is to state that one may ﬁnd an explanation in the doer. What is special to actions is only the kind of explanation in question. It seems to me that this parallel may be a heritage from animistic science, where waves and other natural phenomena literally were blamed for what they did. Evidence for this presumption is our use of the expression “causal 4 I have argued this point in more detail in Stoecker 2005. 5 With respect to problem C I have tried to accomplish this in Stoecker 2003a, and with respect to D in Stoecker 2003b.

Action and Responsibility – A Second Look at Ascriptivism

45

responsibility”. Causal responsibility is what things or persons doing something have. This leads to objection (IV), the question what sort of responsibility is at issue in ascriptivism. According to neo-ascriptivism, agency is based on the communicative practice of taking agents to account. Yet there is no communicative practice of taking someone to account causally. If it is essential for agency that the so-called result of the action is explainable by reference to the agent’s responsibility, “responsibility” has to be read in a stronger sense than causal responsibility. This observation still leaves the choice between moral and legal responsibility. In contrast with causal responsibility one can easily be held morally as well as legally responsible for something. In fact it would be better to say: one can be held responsible for moral grounds as well as for legal grounds. Morality and law do not establish different kinds of responsibility, they only provide two sources for establishing what people are responsible for. And both sources may equally justify the assumption that an agent will behave responsibly. Hence ascriptivism need not distinguish whether it is moral or legal responsibility that allows for the explanation of the result of the agent’s action. 7. Résumé It was my intention to dispel the worries that let traditional ascriptivism look so unattractive, so that my basic claim – that to say that an agent acts is to say that something (the result of what she does) is explainable by reference to the agent’s responsibility for it – could gain at least some initial plausibility.6 References Davidson, Donald (2001), ‘Freedom to Act’, in Donald Davidson, Essays on Actions and Events, 2nd edn, Oxford: Clarendon, pp. 63–81. Edwards, Paul (ed.) (1967), The Encyclopedia of Philosophy, New York and London: Macmillan. Geach, Peter (1960), ‘Ascriptivism’, The Philosophical Review, 69, 221–5. Hart, H.L.A. (1948/49), ‘The Ascription of Responsibility and Rights’, Proceedings of the Aristotelian Society, 44, 171–94. —— and Tony Honoré (1985), Causation in the Law, 2nd edn, Oxford: Clarendon. Pitcher, Georg (1960), ‘Hart on Action and Responsibility’, Philosophical Review, 69, 226–35. Stoecker, Ralf (2003a), ‘Climbers, Pigs and Wiggled Ears: The Problem of Waywardness in Action Theory’, in Sven Walter and Dieter Heckmann (eds), Physicalism and Mental Causation, Exeter and Charlottesville, VA: Imprint Academic, pp. 295–322.

6 I want to thank Armin Berger, Rüdiger Bittner, Jens Kulenkampff, Christoph Lumer and Thomas Spitzley for their critical support.

46

Intentionality, Deliberation and Autonomy

—— (2003b), ‘First Person Authority and the Merits of Minimal Monism’, in Andreas Baechli and Klaus Petrus (eds), Monism, Frankfurt am Main and New York: Ontos, pp. 235–53. —— (2004), ‘Intra-sententielle Deixis und die logische Form von Handlungssätzen’, in Mark Siebel and Mark Textor (eds), Semantik und Ontologie, Frankfurt am Main and Lancaster: Ontos, pp. 153–68. —— (2005), ‘By “by”’, in João Sàágua (ed.), A Explicação da Interpretação Humana, Lisbon: Edições Colibri, pp. 335–42.

Chapter 3

Action Theory and Cognitive Turn: How Can the Content of Intentions Contribute to Causing Actions? Sandro Nannini

The causal theory of action can be better defended if mental states, in the framework of a ‘cognitive turn’, are considered as natural phenomena that are implemented by brain processes. Intentionalists usually object to causalists that the reduction of intention-action relations to mere cause-effect relations does not take into account the fact that intentions explain actions only thanks to their content, that is, thanks to the Intentionality relation (in Brentano’s sense) that they have with the corresponding actions. And they maintain that Intentionality is not reducible to a cause-effect relation. However according to the ‘adverbial theory of perception’ the content of a mental state can be reduced to a monadic property of that state and, in the light of neurosciences, this property can be considered identical with the synchronization of certain brain processes. Intentionality is reducible to a brain property that is causally efﬁcacious. Therefore application of the cognitive turn renders the causal theory of action clearer and more plausible.

1. Action theory after the cognitive turn This chapter discusses action theory from the standpoint of a philosopher of mind and claims that the causal theory of action can be better defended if it is viewed as part of a more general naturalistic conception of the mind. Let us start from a historical remark. In the middle of the twentieth century many philosophers conceived philosophy of mind as an application of a “linguistic turn” (Rorty 1993) to psychological terms. The same applies for action theories.1 Neowittgensteinians such as G.E.M. Anscombe (1957) or G.H. von Wright (1971, ch. 3), by developing some passages of Wittgenstein’s Philosophische Untersuchungen (1953, sect. 621), claimed that the reason for which (or the intention with which) 1 Cf. for example about action theories: Searle 2001; Holmström-Hintikka and Tuomela 1997; Mele 1992; Ginet 1990; Moya 1990; Bishop 1989; Brand 1984; Davidson 1980; Tuomela 1977; Goldman 1970.

48

Intentionality, Deliberation and Autonomy

an action is executed cannot be the cause of that action by virtue of the logical grammar of common expressions such as ‘explaining an action by the reasons the action was executed for’ or ‘explaining an action by the intention the action was executed with’.2 In the ﬁeld of philosophy of mind the linguistic turn was followed by a ‘cognitive turn’ in the 1960s as mental states were no longer considered mere behavioural dispositions but also as authentic subjective states capable of being the inner causes of behaviour. Such a return from philosophical behaviourism to mentalism did not usually imply a rehabilitation of the Cartesian ‘body/mind dualism’ since the mental states seen as inner subjective states were either directly identiﬁed with brain processes (identity theory) or viewed as functional states of the brain (functionalism). In both cases philosophers no longer inquired into the nature of mental states by a mere conceptual analysis of psychological terms but revised the psychological concepts of common sense and philosophical tradition in the light of the cognitive sciences that were gaining ground in the 1960s. The success of the cognitive sciences deeply changed the approach to the study of mind on the part of many contemporary philosophers usually called ‘naturalists’.3 According to them the mind is a part of the natural world. Its study is a normal empirical enterprise to be conducted by the usual methods of natural sciences and not a ﬁeld reserved to the speculation of philosophers. Therefore naturalists usually agree with W.v.O. Quine on the existence of a strong ‘continuity’ between philosophy and sciences: this is the core of their ‘naturalism’.4 Naturalism was also present in the ﬁeld of action theories (see for example Brand 1984). Moreover the upholders of the causal theory of action – let us call them ‘causalists’ – have always accepted a stronger or weaker form of naturalism unlike their critics (let us call them ‘intentionalists’).5 According to causalists, even a typically philosophical task like the conceptual analysis of those explanations that explain an action by mentioning the intention with which it was executed cannot be achieved without referring to psychological (and possibly neurophysiological) studies on human behaviour and its motivation. The cognitive turn had less inﬂuence on the debate on action theories than on the philosophy of mind. Therefore, by means of naturalistic arguments in favour of the causal theory of action, I shall try to show that in this ﬁeld as well, joining traditional conceptual analysis with empirical ﬁndings and theories that derive from the cognitive sciences is the most fruitful approach. In order to argue in favour of applying such a naturalistic approach to the causal theory of action I shall freely refer to well known arguments and thought experiments belonging to the dispute

2 I have presented a historical reconstruction of the debate about ‘causes and reasons’ in Nannini 1992, 121–281. 3 I refer to philosophers such as D.C. Dennett, J.R. Searle and the Churchlands. Cf. about naturalists and a historical reconstruction of the ‘cognitive turn’: Nannini 2002, 94–203; 2006, 149–275. 4 Cf. about this kind of naturalism for example: Walsh 2001; Nannini and Sandkühler 2000; Keil and Schnädelbach 2000; Marconi 1999; Agazzi and Vassallo 1998. 5 This terminology is used in von Wright 1971.

Action Theory and Cognitive Turn

49

about action theories – and the related topic of the so-called ‘wayward causal chains’ – that has been ongoing between causalists and intentionalists for ﬁfty years. In this context, in order to privilege a naturalistic approach, I shall also refer to recent studies of cognitive neuroscience and neuroinformatics. However, my aim is neither to offer a historical reconstruction of the debate between causalists and intentionalists nor to present an up-to-date survey of the action theories that are more widespread nowadays, nor shall I present a critical discussion of the various interpretations of wayward causal chains.6 Furthermore, although I will be referring to neurosciences and neuroinformatics I shall not provide a complete report of the new studies on the neurophysiology of voluntary movements.7 I shall limit myself to using some of these scientiﬁc studies only in order to offer further support to the causal theory of action. More speciﬁcally my most important aim in this chapter will be to show – by drawing freely on the above mentioned philosophical theories and scientiﬁc studies – that causalists’ replies to the objections of intentionalists are unable to explain the difference between the cases in which intentions and beliefs cause actions independently of their having certain contents and cases in which their contents are causally efﬁcacious. Secondly I shall try to combine the ‘adverbial theory of perception’ with some scientiﬁc studies on the dynamics of neural processes in order to give a cue for a possible solution to the above problem.

2. The ‘killer nephew’ The causal theory of action I am trying to defend can be shortly formulated as follows. If I explain A’s action of doing X by saying that A did X because she wanted to get Y and believed that in her present situation doing X was the most expedient way to get Y, my explanation is a causal explanation, that is, the ‘because’ included in my explanation refers to a cause-effect relation of the same kind as the causal relation that may obtain between two physical events. A’s intention of getting Y and A’s belief that she will get Y if she does X are two mental states which, if combined, cause A’s action of doing X just as ﬁre causes smoke (let us call this kind of causality ‘natural causality’ and indicate it by the symbol ‘’).8 This theory has been criticized by several arguments. In order to best understand their common essential core let us start from a well known counter-example against the interpretation of intention-action relations as cause-effect relations: the ‘killer nephew’ (Chisholm 1966, 30ff.). A man wants to come into possession of his uncle’s inheritance at any cost and he thinks he can do that only by killing his uncle. Therefore he decides to kill him, takes a pistol and drives to his uncle’s house with the intention 6 The bibliography about the wayward causal chains is very large. Cf. for example: Enç 2004; Mele 1987; Peacocke 1979. Cf. also Keil’s essay in this volume and Lumer (forthcoming). Conditions designed to capture what counts as an intentional action can be found in the essays cited in the footnote n. 1 and also in Mele and Moser 1994; Mele 1987; Bach 1978; Audi 1973. 7 Cf. for example: Libet, Freeman and Sutherland 1999; Roth 2001; Wegner 2002. 8 Cf. for example Stoutland (1976) about the causal theory of action.

Intentionality, Deliberation and Autonomy

50

of killing him. This decision – a consequence of his intentions and beliefs – agitates him so much that while driving to his uncle’s house he loses control of his car, runs over a pedestrian and kills him. By chance the pedestrian is his uncle! The nephew killed his uncle because of his intention to kill him. Nevertheless our common sense intuitions tell us that this action was not intentional. Therefore, there are cases in which we cannot describe the action of doing X as intentional although the cause of that action is the intention of doing X. In order to evaluate the implications of this example correctly let us distinguish ‘intentional explanations’ from ‘intentional descriptions’. The sentence “The nephew killed his uncle because he intended to come into possession of his uncle’s inheritance” is an intentional explanation. The general form of such explanations is: “She did X because she intended to get Y.” In this case the intention of getting Y is the intention with which the action X was done; let us call it an ‘intention with’. The sentence “The nephew killed his uncle because he intended to kill him” is instead an intentional description that describes the nephew’s action as an intentional action (in this case as a murder). The general form of such descriptions is: “She did X because she intended to do X.” In an intentional description the intention of doing X is the intention in (or under) which a certain behaviour is described as the intentional action of doing X. Let us call it an ‘intention in’ (or an ‘intention under’).9 The example of the ‘killer nephew’ aims at proving the inadequacy of a causal interpretation both of intentional descriptions and intentional explanations. Moreover it is noteworthy that every intentional explanation presupposes an intentional description. The intentional explanation “A did X because she intended to get Y” is valid only if it can more precisely be stated as: “A did X because A intended to get Y and believed that doing X would bring about her getting Y.” This explanation is equivalent in turn to the following sentence: “A’s intention of getting Y [IA(Y)] and A’s belief that her doing X would cause her getting Y [BA(XY)] led A to do X”: (1)

IA(Y) ∧ BA(XY) ⇒ X 10

(1) is implied in its turn by these two sentences: (2)

IA(Y) ∧ BA(XY) ⇒ IA (X)

(3)

IA(X) ⇒ X

Sentence (2) is to be read as: “A’s intention of getting Y and A’s belief that her doing X would cause [] her to get Y led [⇒] A to intend to do X”; (3) however

9 As for the distinction between ‘intention with’ (or ‘intention of’) and ‘intention in’ cf. for example Tuomela (1982, 18) and, although by means of an opposite terminology, Searle (2001, ch. 2, sect. 5). 10 I refer to natural causality, that is, to ordinary causal relations like the relation between ﬁre and smoke by the symbol ‘’. By the symbol ‘⇒’ I refer instead either to the relation of an intention (plus a belief and possibly other mental states) to its corresponding action or to the relation between two mental states if the former ‘induces’ the latter, a relation whose logical nature is at stake here.

Action Theory and Cognitive Turn

51

is to be read: “A’s intention of doing X led her to do X.”11 Consequently if (1) is deducible from (2) and (3) the validity of the intentional explanation (1) presupposes the correctness of the intentional description (3) that is equivalent under another formulation to the statement “A’s action X was intentional because A intended to do X”. Therefore the argument of the ‘killer nephew’ can be restated as an argument that aims at showing that the relation ‘⇒’ cannot be causal (that is, identical to ‘’) because there are cases in which the statement IA(X)X is true but the statement IA(X)⇒X is false. 3. Three alternatives to the causal theory of action Three interpretations have been given by intentionalists to the relation ‘⇒’ as an alternative to the causal one proposed by causalists. The ﬁrst interpretation is based on Wittgenstein’s concept of language game, explaining human actions and explaining natural events are two distinct language games. For example if I am explaining why you have opened a window by saying that you intended to air a room, I am not formulating the hypothesis that a certain mental state of yours (that is your intention of airing the room) has caused your action of opening the window. It is certainly true that some processes in your brain caused a certain movement of your body by means of which the window was opened. However, speaking of brain processes and bodily movements belongs to a language game different from the language game in which one speaks of intentions, beliefs and actions. By describing the movement of your body as the action of opening the window I am not stating that your body was moved by certain brain processes. I am instead subsuming your bodily movement under your intention of opening the window (‘intention in’). Accordingly, by explaining your action of opening the window by your intention of airing the room (‘intention with’) I am reinterpreting your action in the light of your intention. Thus the same action that I had described as the action of opening the window is re-described as the action of airing the room. In both cases descriptions or redescriptions are based on a conceptual relation between the subsuming intention and the subsumed action, a relation that is incompatible with the logical independence usually required for the concepts of two events for the existence of a cause-effect relation between them. This argument against the causal theory of action was developed by the Neowittgensteinians and is known as the “Logical Connection Argument” (Stoutland 1970). It can assume two major forms that aim at proving that the intention of doing X (or the intention of obtaining Y by doing X) cannot, for logical reasons, be the cause of the intentional action of doing X. The ﬁrst version of this argument is based on the alleged fact that if an intentional explanation is seen as a causal explanation, the description of the alleged cause (that is the intention of doing X) mentions the alleged effect (that is the action of doing X) and therefore – it is maintained – an unallowable logical tie between cause and effect arises (Melden 1961). 11 It is noteworthy that, with regard to the action of doing X, IA(Y) is an ‘intention with’, IA (X) is instead an ‘intention in’.

52

Intentionality, Deliberation and Autonomy

In this ﬁrst version the Logical Connection Argument runs into the trivial objection of falsely assuming that, since one’s intention of doing X mentions the action of doing X, then such intention is not logically independent of one’s doing X and therefore it cannot be its cause. However, the mere fact that a certain description of an event C mentions an event E does not prevent C from being the cause of E. The unknown cause of a certain illness X is the effective cause of X even if we can describe it only as ‘the cause of X’. The second version of the Logical Connection Argument avoids this objection (von Wright 1971, ch. 3, sect. 3). I see a man who is running to the rail station. Why is he doing so? Let us assume that the circumstances suggest that his running to the station can be reasonably explained only by assuming that he intends to catch a train. His intention of catching a train does not mention the action of running. But I can infer that he intends to catch a train only from his very running. The veriﬁcation of the explanation coincides with the veriﬁcation of the presence of the alleged cause (this is the reason why this version of the Logical Connection Argument is called ‘the Argument from Veriﬁcation’). If my explanation of the action were causal – as the upholders of the Argument from Veriﬁcation argue – I would infer from the alleged effect the existence of the cause by means of which I explain the very effect. Therefore, if my explanation “He is running because he intends to catch a train” were a causal explanation it could sound very odd. However, it sounds right! Why? It is so because my explanation of the reason why the man is running is no causal explanation at all but an interpretation and redescription of his action in the light of a certain context (he is running to the rail station and so on). However, neither is the Argument from Veriﬁcation as sound as its upholders thought it was. Consider this example of causal explanation.12 An astronomer explains the dark lines of an absorption spectrum of a particular star by saying that they are caused by the fact that the star contains certain chemical elements. How can the astronomer verify that there are those elements in the atmosphere of the star? The very dark lines are the only evidence! The explanation is ‘self-evidencing’ because the explanandum is the only evidence of the explanans. Does it mean that this explanation is either circular or non causal? No, since although it is true that in this particular case the presence of the cause can be tested only by certain features of its effect, the causal relation between certain chemical elements and a certain absorption spectrum on which the explanation is based can be conﬁrmed in many laboratory situations where causes of that kind and effects of that kind are independently testable. Can we also apply this conclusion to the case of the man who is running to catch a train? The answer is deﬁnitely yes! Although we have no other evidence of his intention to catch a train except his running, we are able to see the connection between the intention and the action thanks to our broad experience of human beings which teaches us that in many other situations in which the presence of such an intention and of the corresponding action were independently testable, that intention was indeed the cause of that action.

12 The example is drawn on Hempel 1965, 372. See a more detailed discussion of this criticism on von Wright in Nannini 1999.

Action Theory and Cognitive Turn

53

Therefore the Logical Connection Argument is false in both its versions. The second possibility of interpreting the relation ‘⇒’ between an intention and the corresponding action (or between an ‘intention with’ and the corresponding ‘intention in’) as different from the causal relations ‘’ that may obtain between natural events is to introduce a new kind of causality by means of a concept such as ‘agent causation’.13 In the ‘killer nephew’ the death of the uncle was caused by the nephew’s intention of killing him in the same sense in which a natural event can cause another natural event. The fact that such an intention was not simply a mere event that by chance obtained in the mind of a certain agent, but was the intention of an agent, is irrelevant in that example. Although the nephew intended to kill his uncle and such intention brought about the uncle’s death, it is false that he killed his uncle. His intention caused the death of his uncle but properly speaking he did not kill him. Actions are intentional not if they are directly caused by the intention of doing them, but if they are ‘caused’ by agents who intend to do them. The relation ‘⇒’ (this time interpreted as a relation between an agent that intends to do X and her doing X) is certainly causal, but it is so according to a concept of causality that is different from the concept of natural causality. This interpretation of the difference between ‘⇒’ and ‘’ presupposes the possibility of clearly deﬁning what agent causation is. However although we have knowledge by acquaintance of our being the authors of our voluntary actions, we have no proof whatsoever that this agency feeling is not illusory (Wegner 2002). It is likely indeed that my very intention of doing X is identical to a brain process that automatically triggers those movements of my body that are called my action of doing X. But if it is so then the concept of agent causation is useless since the statement “The agent A did X” is always translatable, at least in principle, into the statement “A’s brain processes caused those bodily movements of which X is a redescription”. An upholder of the agent causation concept could avoid this conclusion only if one could plausibly argue that the relation of agent causation between agents and actions is not reducible – even in principle – to a relation of natural causality between brain processes and bodily movements. Do such arguments in favour of the impossibility of reducing agent causation to natural causality exist? Only arguments such as the ‘killer nephew’ seem to play in favour of the existence of agent causation. But unfortunately the interpretation of such arguments is exactly what is at stake here. A way to mitigate the strength of the objections that can be raised against the concept of agent causation could be to replace it with the concept of ‘purposive causation’ (Tuomela 1977) or something similar, that is, with the concept of a particular kind of causality – obtaining only between mental states and actions – that implies natural causality but is not reducible to it. One could not have done X if one’s brain processes did not cause those bodily movements that implement X – the upholders of this third alternative to causalists’ theories concede – but it is also true that “A did X” means something more than “A’s brain processes caused those bodily 13 About agent-causation besides Chisholm’s (1966) and R. Taylor’s (1966) classical studies see also: O’Connor 1995; Clarke 1996.

54

Intentionality, Deliberation and Autonomy

movements of A of which X is a redescription”. This ‘something more’ is usually interpreted as a teleological dimension: A did X in order to get Y. The nephew did not intentionally kill his uncle – although he brought about his death because of his intention of killing him – since it is not the case that he ran over him in order to kill him. Purposive causation is natural causality plus teleology. The concept of purposive causation can be interpreted in two distinct ways according to the meaning given to its teleological component. Is this component reducible to a particular causal mechanism of the brain? Then in the ﬁnal analysis purposive causation is reducible to natural causality.14 Is this teleological component instead not reducible to any brain processes? Then it is not essentially different from agent causation and it faces the same objections. To sum up, none of the three interpretations of the relation ‘⇒’ between intentions and actions (conceptual tie, agent causation relation or purposive causation relation) is a satisfying alternative to the causal one. However, this conclusion would, in turn, be insufﬁciently justiﬁed if causalists were unable to explain how they can admit that the nephew’s action was unintentional and at the same time claim that it was caused by the ‘right’ intention. The best argument causalists have found to reject all counterexamples against their theory such as the ‘killer nephew’ is based on the assumption that they are trivial examples of ‘wayward causal chains’ and not a serious refutation of the causal theory of action. Although the intention of killing the uncle caused his death, this event happened in a different manner than the nephew had planned. Therefore the ‘killer nephew’ shows only that if the intention of doing such and such causes the action of doing such and such differently than the agent had planned, then the action is not intentional. They do not prove that in normal cases the description of an intentional action by the intention in which the action was executed is not equivalent to the statement of a causal relation between the intention and the action. Speciﬁcally, if the nephew had killed his uncle as he had planned to do, the causal chain from his intention to his action would have been of this kind: (planned course of events) (4) ‘intention with’ of coming into possession of his uncle’s inheritance and belief that killing him with a pistol is the best way to come into possession of his inheritance ‘intention in’ of previously driving to his uncle’s house and later killing him with a pistol (…) intentional action of killing him with a pistol.

14 Cf. ﬁrst of all Nagel (1961, 401–27) on the reduction of teleological explanations to causal explanations and Wright (1976) on teleological explanations at large. The naturalistic approach to teleology defended nowadays by the ‘teleonomists’ is based on the reduction of the biological concept of function to the result of natural selection processes. About the debate on teleology and biology cf. for example: Ariew, Cummins and Perlman 2002; Allen, Bekoff and Lauder 1998; Buller 1999; Sober 1993; Bedau 1991; Rescher 1986; Millikan 1984.

Action Theory and Cognitive Turn

55

As a matter of fact things went like this: (real course of events) (5)

‘intention with’ of coming into possession of his uncle’s inheritance and belief that killing him with a pistol is the best way to come into possession of his inheritance ‘intention in’ of previously driving to his uncle’s house and later killing him with a pistol (…) state of excitement unintentional action of killing him by running over him with a car.

The action of killing the uncle is intentional in the planned course of events (4) and unintentional in their real course (5), not because the intention-action relation is not a natural causality relation in the former case and is so in the latter, but because the course of events is different in the two cases.

4. The ‘killer mountaineer’ However, it seems that the strength of such a reply can be questioned by a reﬁnement of the ‘killer nephew’, a reﬁnement in which the executed action is still unintentional although the planned course of events and their real course are by hypothesis identical. Such a reﬁnement can be found in the counter-example of the ‘the killer mountaineer’ (Davidson 1973).15 Two mountaineers are climbing a mountain. One of them hates the other for private reasons and for many years has been waiting for the opportunity to kill him. While they are facing up to a difﬁcult manoeuvre the former understands that the time to act has come: it is sufﬁcient that he feigns losing hold of the rope that prevents the latter from hurtling down and his criminal project will be realized! He decides to let the rope go in that moment. But this intention agitates him so much that he unintentionally lets the rope go in the same moment and in the same way in which he had decided to do so intentionally. Therefore his action is unintentional although it was caused by the intention of executing it exactly in the way and in the moment in which the agent had planned to execute it. In this case it seems that the ‘wayward causal chain reply’ is no any longer at the causalists’ disposal. How are we to judge the ‘killer mountaineer’ argument and similar replies to the objections against the causal theory of action? Surely this argument is contrived in such an artiﬁcial way that our intuitions vacillate! Its conclusion seems to be drawn in such a strained way that it is highly implausible to transfer it to normal intentional actions. However, considerations of this kind are merely a way to reject the example and not to reply to it. Therefore the safest strategy for a causalist is to concede the logical possibility of the ‘killer mountaineer’ and try to prove that even after this concession the example is still invalid as an objection to the causal theory of action. The best starting point to prove that, in spite of all appearances, even the ‘killer mountaineer’ is only a wayward causal chain case is to remark that the most 15 Cf. a broader discussion of such counterexamples especially in: Enç 2004; Brand 1984, 17ff.; Moya 1990, 115–28.

Intentionality, Deliberation and Autonomy

56

plausible interpretation of the whole story is that the intention of letting the rope go did not directly cause the action of letting it go; it caused it by bringing about an intermediate state of excitement. Therefore in this case, as in the ‘killer nephew’, the real course of events differs from the course that would have led to the execution of an intentional action: (planned course of events) (6)

‘intention with’ of killing his friend and belief that letting the rope go now is the best way to kill him ‘intention in’ of letting the rope go now intentional action of killing him by letting the rope go now.

(real course of events) (7)

‘intention with’ of killing his friend and belief that letting the rope go now is the best way to kill him ‘intention in’ of letting the rope go now state of excitement unintentional action of killing him by letting the rope go now.

An intentionalist might object that by such an interpretation of the ‘killer mountaineer’ the causalist takes back what he previously said to concede, that is, the fact that in this example the real course of events and their planned course are by hypothesis identical. The intentionalist might also add that the causalist’s interpretation of the ‘killer mountaineer’ requires that the intention of letting the rope go is seen as a mental state different from the state of excitement that brings about the unintentional movement of letting it go. However this distinction does not emerge at the level of a common sense analysis of the killer mountaineer’s action: since he intended to kill his friend he was also very excited of course! He would have been very excited even if he had intentionally killed him! He was in a complex mental state that included both intentions and emotions. This complex unanalysed mental state precedes the same action, that is, the release of the rope both in the planned and the real course of events. Therefore, by hypothesis, the difference between the two cases cannot lie in the complex mental state that precedes the action but in the nature of its relation to the action: a relation that is a natural causality relation () in the real course of events and a relation of another kind (⇒) in their planned course. What can a causalist reply? First of all he can refuse to continue to concede that in the ‘killer mountaineer’ the complex mental state that precedes the release of the rope is really the same in both courses of events. In the planned course of events the action of letting the rope go would have been directly caused by the corresponding intention. Even if one assumes that this intention would have agitated the killer and put him in a state of excitement, this state would have been a byproduct of the events’ course and not the proxy cause of the action. In the real course of events the state of excitement was instead an essential intermediate step in the causal chain that connected the intention to the action. However, such a reply obliges causalists to engage a dispute with intentionalists and the outcome is always highly controversial insofar as the discussion develops at the level of a conceptual analysis based on common sense. A causalist can of course decide to abandon this level of analysis and shift to considerations based on cognitive sciences. There are many cues offered

Action Theory and Cognitive Turn

57

by neurosciences that voluntary actions are planned and controlled by cortex areas different from the areas that implement emotions (Roth 2001, 284–377 and 474ff.) This shift is indeed advisable as I shall later try to show. However if a causalist were to make such a move too early, that is before having developed the discussion with intentionalists at the level of common sense up to the ﬁnal consequences, he would give the impression that he wants to upset the chessboard instead of trying to win the game. Therefore let us continue to assume that in the ‘killer mountaineer’ the mental states that, both in the planned course of events and in their real course, precede the action are prima facie identical. How can a causalist then continue to state that the difference between the two cases does not lie in the nature of the relation between intention and action but in the respective causal chains? First of all the causalist can show that even at the level of common sense analysis, without applying to neurophysiological hypotheses, admitting the existence of a difference between the planned course of events and their real course in the part of the causal chain inside the mind of the agent is very plausible. Let us see why it is so. Causal relations imply counterfactual conditionals. If C caused E then, if C did not happen, E would not have happened either.16 This is compatible with the hypothesis that the ‘killer mountaineer’ is a wayward causal chain case in which the causal chain is deviant in the part inside the killer’s mind. Let us assume the hypothesis most favourable to intentionalists: a complex but unique mental state made up of the intention of letting the rope go and a state of excitement caused the death of the friend. Also in this case there is an obvious difference between the psychological mechanism that would have produced the action in the planned course of events and the psychological mechanism that produced their real course. Even if such mechanisms are reconstructed only at the common sense level, the difference between them is clearly shown by the distinct counterfactuals implied by the two causal chains: (8)

Planned course of events: The action of killing the friend is intentional because the killer would cause the death of the friend even if he was not excited, provided he intends to kill him.

(9)

Real course of events: The action of killing the friend is unintentional because the killer would cause the death of the friend even if he did not intend to kill him, provided he is sufﬁciently excited, quite apart from what the cause of his state of excitement is.

An intentionalist could object that the claim that such a distinction between the planned and real courses of events lies in the part of the causal chain inside the agent’s mind is a mere hypothesis. In a behaviouristic (and Wittgensteinian) vein the intentionalist could reply that the agent’s mind must be considered as a ‘black box’ and that the only possible distinction between the planned course of events and their real course depends on the fact that in the former case one describes the release of 16 The literature on counterfactuals and causality is immense. Cf. particularly: Lewis 1973, 1979, 2000.

58

Intentionality, Deliberation and Autonomy

the rope under the ‘excited intention’ of letting it go, whereas in the latter case one considers the movement of the killer’s hands as the effect of that very mental state of ‘excited intention’. The previous mental state is just the same in both cases. The difference lies only in the kind of relation it has with the action. What can a causalist reply this time? Either she thinks that the time to upset the chessboard has come or she modiﬁes her previous argument a little. Let us examine the second possibility. Even if one concedes that the same single mental state of ‘excited intention’ precedes the action of letting the rope go both in the planned course of events and in their real course, one can remark that there is a difference in the description under which this state is causally efﬁcacious.17 The ‘excited intention’ induces the killer to let the rope go in the former case qua intention, in the latter case qua state of excitement. The same mental state of ‘excited intention’ (let us call it M) can be considered as a state that has both the property Int (intending to kill the friend) and the property Exc (being excited). However, only the former property is a causally necessary condition of bringing about the friend’s death in the planned course of events, whereas this role of unique causally necessary condition is assumed by the latter property in the real course of events. In other words, a causalist can say that in the planned course of events the following statements would have been true: (10) The killer caused the death of the friend. (11) M was Int and Exc. (12) M would have caused the death of the friend even if it had not been Exc provided it was Int. In the real course of events (10) and (11) are still true but (12) is false and is substituted by (13): (13) M would have caused the death of the friend even if it had not been Int provided it was Exc. To sum up, in the ‘killer mountaineer’ the real course of events differed from the planned one either because the action of letting the rope go was caused by a state of excitement and not directly by the intention of letting it go or because the property Exc and not the property Int of the agent’s mental state was causally efﬁcacious.

5. Naturalism and the ‘adverbial theory of perception’ This possible reply of causalists to intentionalists is on the whole correct. However it poses a problem that causalists usually neglect: it works only if having an intention is considered a natural property of the agent’s mental states on a par with being 17 For example, if a white billiard ball hits a red ball the mass of the white ball is causally efﬁcacious because it contributes to determining the velocity of the red ball but its colour is completely irrelevant.

Action Theory and Cognitive Turn

59

excited. In other words, the analysis sketched above shows that the causal theory of action requires that intentions be considered as mental states implemented by brain processes on a par with emotions or, more broadly speaking, it requires that all mental states be seen as natural events capable of naturally causing other natural events. And this is the reason why intentions can be seen as the alleged natural causes of intentional actions in the same sense in which natural events can be the cause of other natural events. However, this is the key point that intentionalists do not concede: there is no way – they object – to consider Int as a natural property. And this precludes the possibility that intentions are natural causes of actions or of other intentions. According to intentionalists, examples such as the ‘killer nephew’ or the ‘killer mountaineer’ clearly show that considering the intention-action relation a normal causal relation between natural events is counterintuitive. Intentions do not explain actions (or describe them as intentional) because they are their alleged causes but because of the similarity between their contents and the descriptions of the respective actions. Showing that the action of doing X was executed because of the intention of doing X is not sufﬁcient to justify the assertion that such an action was intentionally executed. In order to justify this assertion one needs to show that the action was executed only because the intention with which it was done had a certain content, that is, only because it was the intention of doing X that led the agent to do X. Why – intentionalists ask – do we intuitively perceive a difference between killing the friend and accidentally bringing about his death although this event is a consequence of the same intention in both cases? It is very simple: in the latter case any other mental state could have indirectly caused the friend’s death provided that mental state directly caused a sufﬁciently strong state of excitement. The intention of killing the friend was only indirectly efﬁcacious as it brought about an unspeciﬁc state of excitement, and not because of its speciﬁc content. The fact that the intention of killing the friend was an ‘Intentional state’ (in Brentano’s sense of Intentionality)18 played no role in the real course of events, whereas it would have been essential if the planned course had been realized. Therefore causalists seem to fall into the following dilemma. Either the causal explanations of intentional actions do not take into account the fact that the alleged causes of such actions are Intentional states causally efﬁcacious only thanks to their content or they are forced to distinguish two distinct kinds of causality: on the one hand natural causality (), that is, a certain kind of relation between natural events, and on the other hand ‘Intentional causality’ (⇒), that is, a certain kind of relation between intentions and actions. In the ﬁrst case their interpretation of intentional actions’ explanations as causal explanations misses the point, since causalists are not able to distinguish between the explanations in which the content of the intention plays an essential role and the explanations in which such a content is as indifferent as the colour of billiard balls in the explanation of their movements by the laws of mechanics. In the second case they do distinguish these two kinds of explanations but at the price of renouncing the essential point of their theory, that is, 18 I am writing ‘Intentionality’ (‘Intentional’ and so on) with a capital ‘I’ to distinguish Brentano’s Intentionality from ‘intentionality’ in common language.

60

Intentionality, Deliberation and Autonomy

they are obliged to admit that, although human actions can be ‘causally’ explained, the ‘Intentional causality’ (it does not matter if it is identical to agent causation, purposive causation or something else) by means of which actions are intentionally explained is different from natural causality. Being directed to an object does not seem to be a property reducible to a natural property that can be implemented by brain processes. For example, one’s intention of killing one’s friend seems to set up an internal teleological relation between a mental state and an action that cannot be reduced to any cause-effect relation. What can a causalist reply? Within the limits of conceptual analysis a good strategy might be to extend the ‘adverbial theory of perception’ to all mental states, insofar as they are Intentional states (see for example Tye 1984). According to this theory “I see a red spot” means for example, as it were, “I see ‘red-spot-wise’ (or ‘red-spot-ly’)”, “I see a dog” means “I see ‘dog-wise’”, and so on. In other words, the inner content by means of which a perception is directed to a certain external object is considered a property of that perception expressible by means of a monadic predicate and not by means of a relation to an object. If such a theory is extended from perceptions to all mental states, including Intentional states such as intentions or beliefs, then the content of the mountaineer’s intention (that is its being the intention of killing his friend) can be reduced to a particular way of intending. The difference between the intention of doing X and the intention of doing Y is a difference between two distinct ways of intending, just as being intensely excited is different from being mildly excited. The content of Intentional states is reduced to an aspect of their form. The adverbial theory of perception was formulated by philosophers such as Ducasse (1942) and Chisholm (1957) to argue against the existence of sense data as ‘objects’ of conscious perceptions (Moore 1910) by means of which both bodies and minds are allegedly constructed (Russell 1921). Brieﬂy, the upholders of this theory argued in favour of an antirealistic metaphysics according to which a perception is an Intentional state not because it refers to a certain object but because it has a certain ‘quale’: perceiving X means feeling X-ly. Going beyond Ducasse and Chisholm, I am applying the adverbial theory to mental representations at large (and not only to conscious representations) in order to naturalize Intentionality. If the content by means of which a representation refers to an object is a monadic property of that representation, then one can formulate two naturalistic hypotheses of this kind: 1. The perception of object O is identical to an intermediate step in the brain process P that connects perception M of O to some adequate actions on O (adequate according to the agent’s needs and purposes). 2. The monadic property of M by means of which it, qua representation of O, refers to O is implemented by a certain dynamics of P. M represents O for a certain agent A only if M has the ‘proper function’ of leading A in normal conditions to act on O in such a way that A maximizes the achievement of her purposes with regard to O. For example, if I want to grasp an object O, a certain mental state of mine is a perception of O only if the neural implementation of that perception causes the movements of my hands necessary to grasp O in

Action Theory and Cognitive Turn

61

normal conditions thanks to the way electro-chemical signals are propagated in my central nervous system. In other words I am suggesting unifying the adverbial theory of perception and the ‘teleological theory of content’ (called also ‘consumer’s semantics’) in a single naturalistic theory of Intentionality according to which what a mental representation represents depends on the function that the very representation has in the interaction between an agent and her environment.19 If we apply this naturalistic approach to the ‘killer mountaineer’ the difference between the respective causal chains in the planned course of events and in their real course becomes obvious. The killer’s mental activity before bringing about the friend’s death had the property Int and the property Exc exactly as a billiard ball can be white and heavy. That mental activity was causally efﬁcacious in the real course of events only so far it was Exc. Its being Int was as irrelevant as the billiard ball’s being white is irrelevant to its movements. Exactly the opposite would have been the case if the action of killing the friend had been intentional. The parallelism of the two cases shows that Int is a monadic property of mental states as well as Exc. Therefore, since the contents of Intentional states are reducible to monadic properties of such states, there is no longer any difﬁculty in understanding how such contents can be causally efﬁcacious without assuming the existence of a special kind of causality different from natural causality. However, an intentionalist might remark that the adverbial theory of perception cannot answer the following objection (Jackson 1977). I see a brown square and a green triangle. According to the adverbial theory I see brownly and squarely and greenly and triangularly. However, even if I saw a green square and a brown triangle I would continue to see brownly and squarely and greenly and triangularly. Is the adverbial theory capable of discriminating the two cases? It seems that it is not. Therefore it is false. And if it is false for perceptions it is false for intentions as well. The form of an Intentional state is always too generic to individuate and express its content exactly. No Int property will ever be able to make ‘intending-in-an-Int-way’ exactly equivalent to ‘having the intention-of-killing-the-uncle’. A causalist can object that simultaneously seeing brownly-squarely and greenlytriangularly is different from simultaneously seeing brownly-triangularly and greenly-squarely. Therefore also the property Int can, at least in principle, be so cleverly formulated that a mental state of the killer will be the cause of the intentional action of killing the friend if and only if such a state will be Int. However, so far causalists accept to remain prisoners of the battleﬁeld chosen by intentionalists, that is, the common sense level of analysis ﬁnding a watertight argument against them will be impossible. For example an intentionalist might continue to object that at the level of a simple conceptual analysis the upholder of the adverbial theory can offer no argument at all to justify the pairing of ‘brownly’ with ‘squarely’ if I see a brown square, and the pairing of ‘brownly’ with ‘triangularly’ if I see a brown triangle. 19 Cf. Millikan 1984, 1993, 2005; and Dretske 1981, 1995. On Intentionality naturalized see also Nannini 2004.

62

Intentionality, Deliberation and Autonomy

To be sure, ﬁnding a deﬁnitive proof in favour of the adverbial theory of perception will be impossible even if one sees the whole debate on the nature of Intentionality in the light of the ‘cognitive turn’. Nevertheless, at least with regard to the example mentioned above, cognitive neurosciences offer a theory that permits a better understanding of how the perception of forms and the perception of colours can be combined to make up the perception of a single coloured object – this is the so called ‘binding problem’ – although the brain processes that implement the perception of forms are different from the processes that implement the perception of colours. When we see a brown square, the neurons that detect forms ﬁre at the same frequency as the neurons that recognize colours so that there is synchronization between them.20 Therefore one can give a plausible neurophysiological interpretation of the adverbial theory of perception that avoids Jackson’s objection mentioned above. Let us assume that I see a brown square and a green triangle and that my seeing these two objects is equivalent to my seeing brownly-squarely and simultaneously greenly-triangularly. Moreover let us formulate the hypothesis that a certain brain process C1 having the physical property P1 (so that the statement “C1 is P1” is true and means “C1 is a brain process that propagates electro-chemical signals according to the dynamics P1”) is the neurophysiological implementation of my seeing squarely, that is, of my seeing a square. The brain process C2 having the physical property P2 implements my seeing brown. In a similar way the brain processes C3 and C4 (endowed with the respective properties P3 and P4) implement my seeing a triangle and my seeing green. Now it is plausible to introduce the further hypothesis that if I see a brown square and a green triangle, C1 is connected to C2 and C3 to C4 thanks to certain physical relations of P1 to P2 and of P3 to P4 (for example equal frequency), whereas if I see a green square and a brown triangle, C1 is connected to C4 and C3 to C2, thanks to a relation of P1 to P4 and of P3 to P2. Can one argue that even the reduction of the causal efﬁcacy of the content of an intention to a monadic natural property would be more plausible if this property could be implemented by certain complex features of the brain dynamics? So far such an assertion is bold and rash. Not even the neurophysiological interpretation of the adverbial theory of perception mentioned above can be considered directly proved. Nevertheless the physicists’ and mathematicians’ studies on the theory of complex systems (Kaplan and Glass 1995) provide the possibility of interpreting mental representations (and, broadly speaking, mental states) as intermediate stages (that is, intermediate patterns of activation of neural groups) in the dynamics that transfer information from the sensory input to the motor response. Such a hypothesis becomes still more plausible if it is combined on the one hand with the simulations executed by means of artiﬁcial neural networks (Rumelhart and McClelland 1986) and with their implication for the philosophy of mind (see for example P.M. Churchland 1995; P.S. Churchland 2002; and Churchland and Sejnowski 1992), and on the other hand with a philosophical exploitation of a NeoDarwinian perspective on the evolution of human cognitive capacities (see for example Dennett 1995). 20 For this solution of the binding problem see: Crick and Koch 1990; Singer and Gray 1995; Engel 2003, 2005.

Action Theory and Cognitive Turn

63

Within this naturalistic theory of the mind it is at least plausible to individuate the neural correlate of what is described, at the level of analysis of folk psychology, as Intentional content in a certain ‘modulation’ of the dynamics of the brain processes which control our body’s interaction with its environment. In order to bring this interaction to a successful conclusion our brain needs an inner model of the external world. And since our brain cannot do anything but produce electrochemical currents, it is plausible that the only means at its disposal for representing external objects (or events and so on) is to modulate these currents. Intentionalists might object that not even this neurobiological approach makes it possible to avoid the usual objection against all attempts to naturalize Intentionality: how can a brain process be related to a certain object only because it possesses a certain monadic property? However, if the monadic property in question is a certain dynamics of the brain process, a causalist can this time reply to this objection. Let us see how. First, the dynamics of brain processes at time t is the way in which electrochemical signals propagate in the brain to trigger the motor responses executed at t. If such propagation is modulated in such and such a manner, its behavioural effects will change accordingly. Second, it is plausible to assume that the perception of an object is correct if it coincides with that functional state of the brain that allows acting on the very object in such a way that the achievement of the agent’s goals is maximized. For example, my perception of redness is implemented by the processes in my brain that trigger the movement of my right foot on my car’s brake pedal, if I see that trafﬁc lights are red. Therefore it is reasonable to formulate the hypothesis that the brain processes that bring about the actions on a certain ‘object’ (or situation) are so modulated that they trigger those actions that allow an agent to realize her projects with regard to that object (or situation). If trafﬁc lights in front of me are red, my seeing that they are red is identical to the modulation of the signals coming from my eyes necessary to bring about the movement of my foot on the brake pedal. If instead I see that the trafﬁc lights are green, the brain processes that implement my seeing that they are green would be so modulated that such brain processes would bring about the movement of my foot on the accelerator pedal. Therefore, although the adverbial theory of perception was ﬁrst formulated by antirealists who wanted to reduce the relation of a perception with its external object to a inner property of the very perception, the naturalistic interpretation that I have proposed of this theory does not renounce the existence of a true relation between the mind and the world by means of perceptions. However, I assume that this relation is ﬁrst of all pragmatic. For example, Mother Nature selected our ability to see colours because this was useful for survival. Therefore the perception of colours is identical to those brain processes that control those actions in which the ability to discriminate objects by means of their colours is on the average useful. One can therefore advance the hypothesis that all actions based on the identiﬁcation of the colour X are triggered and controlled by brain processes that share the feature Y (for example a certain dynamics). If this is the case then one can say that a certain perception is the perception of X because the brain process that implements it has feature Y.

Intentionality, Deliberation and Autonomy

64

To sum up, the following hypotheses are the best ones we can currently formulate on intentional actions in support of the causal theory of action: 1. Mental states are implemented by brain processes. 2. Intentional actions and unintentional actions are caused by different brain processes. 3. The content of mental states is implemented by the dynamics of the neural processes that implement such states. This dynamics controls the motor response necessary on the average to improve the probability of survival. 4. Therefore the Intentionality of mental states can be naturalized and plays an essential role in the causal chain that connects our brain processes to our bodily movements (that is, in a naturalistic perspective, the causal chain that connects our mental states to our actions, including our intentional actions). Therefore application of the cognitive turn renders the causal theory of action clearer and more plausible. References Agazzi, E. and N. Vassallo (eds) (1998), Introduzione al naturalismo ﬁlosoﬁco contemporaneo, Milano: Angeli. Allen, C.; M. Bekoff and G. Lauder (eds) (1998), Nature’s Purposes, Cambridge, MA: MIT Press. Anscombe, G.E. (1957), Intention, Oxford: Blackwell. Ariew, A.; R. Cummins and M. Perlman (eds) (2002), Functions: New Essays in the Philosophy of Psychology and Biology, Oxford: Oxford University Press. Audi, R. (1973), ‘Intending’, Journal of Philosophy, 70, 387–403. Bach, K. (1978), ‘A Representational Theory of Action’, Philosophical Studies, 34, 361–79. Bedau, M. (1991), ‘Can Biological Teleology be Naturalized?’, Journal of Philosophy, 88, 647–57. Bishop, J. (1989), Natural Agency, Cambridge: Cambridge University Press. Brand, M. (1984), Intending and Acting: Toward a Naturalized Action Theory, Cambridge, MA: MIT Press. Buller, D. (ed.) (1999), Function, Selection, and Design, Albany, NY: State University of New York Press. Chisholm, R. (1957), Perceiving: A Philosophical Study, Ithaca, NY: Cornell University Press. —— (1966), ‘Freedom and Action’, in K. Lehrer (ed.), Freedom and Determinism, New York: Random House, pp. 11–44. Churchland, P.M. (1995), The Engine of Reason, the Seat of the Soul: A Philosophical Journey into the Brain, Cambridge, MA: MIT Press. Churchland, P.S. (2002), Brain-Wise: Studies in Neurophilosophy, Cambridge, MA: MIT Press.

Action Theory and Cognitive Turn

65

—— and T.J. Sejnowski (1992), The Computational Brain, Cambridge, MA: MIT Press. Clarke, R. (1996), ‘Agent Causation and Event Causation in the Production of Free Action’, Philosophical Topics, 24, 19–48. Crick, F. and C. Koch (1990), ‘Towards a Neurobiological Theory of Consciousness’, Seminars in the Neurosciences, 2, 263–75. Davidson, D. (1973), ‘Freedom to Act’, in T. Honderich (ed.), Essays on Freedom and Action, London: Routledge, pp. 137–56. —— (1980), Essays on Actions and Events, Oxford: Clarendon. Dennett, D.C. (1995), Darwin’s Dangerous Idea: Evolution and the Meanings of Life, New York: Simon & Schuster. Dretske, F.I. (1981), Knowledge and the Flow of Information, Cambridge, MA: MIT Press. —— (1995), Naturalizing the Mind, Cambridge, MA: MIT Press. Ducasse, C.J. (1942), ‘Moore’s Refutation of Idealism’, in P. Schilpp (ed.), The Philosophy of G.E. Moore, Chicago, IL: Northwestern University Press. Enç B. (2004), ‘Causal Theories of Intentional Behavior and Wayward Causal Chains’, in . Engel, A.K. (2003), ‘Temporal Binding and the Neural Correlates of Consciousness’, in A. Cleeremans (ed.), The Units of Consciousness: Binding, Integration and Dissociation, Oxford: Oxford University Press. —— (2005), ‘Neuronale Synchronisation und Wahrnehmungsbewusstsein’, in C.S. Hermann, M. Pauen, J.W. Rieger and S. Schicktanz (eds), Bewusstsein. Philosophie, Neurowissenschaften, Ethik, München: Fink, pp. 216–41. Gibbons, J. (2001), ‘Knowledge in Action’, Philosophy and Phenomenological Research, 62, 579–600. Ginet, C. (1990), On Action, Cambridge: Cambridge University Press. Goldman, A.I. (1970), A Theory of Human Action, Englewood Cliffs, NJ: PrenticeHall. Hempel, C.G. (1965), Aspects of Scientiﬁc Explanation and Other Essays in the Philosophy of Science, New York and London: The Free Press and Macmillan. Holmström-Hintikka, G. and R. Tuomela (eds) (1997), Contemporary Action Theory, Dordrecht: Kluwer. Jackson, F. (1977), Perception: A Representative Theory, Cambridge: Cambridge University Press. Kaplan, D. and L. Glass (1995), Understanding Nonlinear Dynamics, New York: Springer. Keil, G. and H. Schnädelbach (eds) (2000), Naturalismus. Philosophische Beiträge, Frankfurt am Main: Suhrkamp. Lewis, D. (1973), ‘Causality’, The Journal of Philosophy, 70, 556–67. —— (1979), ‘Counterfactual Dependence and Time’s Arrow’, Noûs 13, 445–76. —— (2000), ‘Causation as Inﬂuence’, The Journal of Philosophy, 97, 182–97. Libet, B.; A. Freeman and K. Sutherland (eds) (1999), The Volitional Brain: Towards a Neuroscience of Free Will, Thorverton: Imprint Academic. Lumer, C. (forthcoming), Kognitive Handlungstheorie.Empirische Handlungsgesetze,

66

Intentionality, Deliberation and Autonomy

Freiheit und die Grundlagen praktischer Rationalität. Marconi, D. (ed.) (1999), Naturalismo e naturalizzazione, Vercelli: Mercurio. Melden, A.I. (1961), Free Action, London: Routledge. Mele, A.R. (1987), ‘Intentional Action and Wayward Causal Chains: The Problem of Tertiary Waywardness’, Philosophical Studies, 51, 55–60. —— (1992), Springs of Action, Oxford: Oxford University Press. —— and R.K. Moser (1994), ‘Intentional Action’, Noûs, 28, 39–68. Millikan, R.G. (1984), Language Thought and Other Biological Categories, Cambridge, MA: MIT Press. —— (1993), White Queen Psychology and Other Essays, Cambridge, MA: MIT Press. —— (2005), Language: A Biological Model, Oxford: Oxford University Press. Moore, G. (1910), ‘Sense Data’, reprinted in G. Moore, Selected Writings, London: Routledge. Moya, C.J. (1990), The Philosophy of Action: An Introduction, Cambridge: Polity Press. Nagel, E. (1961), The Structure of Science, New York: Harcourt, Brace & World. Nannini, S. (1992), Cause e ragioni. Modelli di spiegazione delle azioni umane nella ﬁlosoﬁa analitica, Roma: Editori Riuniti. —— (1999), ‘The Logical Connection Argument Again’, in R. Egidi (ed.), In Search of a New Humanism: The Philosophy of Georg Henrik von Wright, Dordrecht: Kluwer, pp. 103–12. —— (2002), L’anima e il corpo. Una introduzione storica alla ﬁlosoﬁa della mente, Roma-Bari: Laterza. —— (2004), ‘Mental Causation and Intentionality in a Mind Naturalising Theory’, in A. Peruzzi (ed.), Mind and Causality, Amsterdam and Philadelphia: John Benjamins, pp. 69–95. —— (2006), Seele, Geist und Körper. Historische Wurzeln und philosophische Grundlagen der Kognitionswissenschaften, Frankfurt am Main: Peter Lang. —— and H.J. Sandkühler (eds), 2000, Naturalism in the Cognitive Sciences and the Philosophy of Mind, Frankfurt am Main: Peter Lang. O’Connor, T. (1995), ‘Agent Causation’, in T. O’Connor (ed.), Agents, Causes, Events: Essays on Free Will and Indeterminism, Oxford: Oxford University Press, pp. 173–200. Peacocke, C. (1979), Holistic Explanation: Action, Space, Interpretation, Oxford: Clarendon. Rescher, N. (ed.), (1986), Current Issues in Teleology, Lanham, MD: University Press of America. Rorty, R. (1993), The Linguistic Turn: Essays in Philosophical Method; With Two Retrospective Essays, Chicago, IL: University of Chicago Press. Roth, G. (2001), Fühlen, Denken, Handeln, Frankfurt am Main: Suhrkamp. Rumelhart, D.E and J.L. McClelland (eds), (1986), Parallel Distributed Processing: Exploration in the Microstructure of Cognition, Cambridge, MA: MIT Press. Russell, B. (<1921>/1995), The Analysis of Mind, new edn, London: Routledge. Ryle, G. (1949), The Concept of Mind, London: Hutchinson. Searle, J. (2001), Rationality in Action, Cambridge, MA: MIT Press.

Action Theory and Cognitive Turn

67

Singer, W. and C.M. Gray (1995), ‘Visual Feature Integration and the Temporal Correlation Hypothesis’, Annual Review of Neuroscience, 18, 555–86. Sober, E. (1993), Philosophy of Biology, Boulder, CO: Westview Press. Stoutland, F. (1970), ‘The Logical Connection Argument’, American Philosophical Quarterly, 7. —— (1976), ‘The Causal Theory of Action’, in J. Manninen and R. Tuomela (eds), Essays of Explanation and Understanding, Dordrecht: Reidel. Taylor, C. (1964), The Explanation of Behavior, London: Routledge. Taylor, R. (1966), Action and Purpose, Englewood Cliffs, NJ: Prentice-Hall. Tuomela, R. (1977), Human Action and its Explanation: A Study on the Philosophical Foundation of Psychology, Dordrecht: Reidel. —— (1982), ‘Explanation of Action’, in G. Fløistad (ed.), Contemporary Philosophy: A New Survey, The Hague: Nijhoff, vol. 3, pp. 15–43. Tye, M. (1984), ‘The Adverbial Theory of Visual Experience’, Philosophical Review, 93, 195–225. Von Wright, G.H. (1971), Explanation and Understanding, Ithaca, NY: Cornell University Press. Walsh, D.M. (2001), Naturalism, Evolution, and Mind, Cambridge: Cambridge University Press. Wegner, D.M (2002), The Illusion of Conscious Will, Cambridge, MA: MIT Press. Wendel, H.J. (1997), Die Grenzen des Naturalismus, Tübingen: Mohr Siebeck. Wittgenstein, L. (<1953>/1984), Werkausgabe: Philosophische Untersuchungen, vol. I, Frankfurt am Main: Suhrkamp. Wright, L. (1976), Teleological Explanations, Berkeley, CA: University of California Press.

This page intentionally left blank

Chapter 4

What Do Deviant Causal Chains Deviate From? Geert Keil

The problem of deviant causal chains is endemic to any theory of action that makes deﬁnitional or explanatory use of a causal connection between an agent’s beliefs and pro-attitudes and his bodily movements. Other causal theories of intentional phenomena are similarly plagued.1 The aim of this chapter is twofold. First, to defend Davidson’s defeatism. In his treatment of deviant causal chains (DCCs), Davidson makes use of the clause “in the right way” to rule out causal waywardness, but he regards any attempt at specifying ‘right’ sorts of causal histories as hopeless and even harmful. To my mind, Davidson’s defeatism contains a valuable insight, so I shall try to explain the reasons for it. Second, I shall try to answer a question that has often been ignored or passed over in the literature; namely the question of what it is that DCCs deviate from.

1. The problem of deviant causal chains What is the problem of deviant causal chains? According to Davidson, the problem consists of the fact that “not just any causal connection between rationalizing attitudes and a wanted effect sufﬁces to guarantee that producing the wanted effect was intentional. The causal chain must follow the right sort of route” (Davidson 1973, 78). Deviant cases have the status of counterexamples which threaten the adequacy of a causal account of the underlying intentional relation. DCCs are best considered as affecting the causal theorist’s deﬁnition of intentional action. Davidson himself makes two attempts at such a deﬁnition. In his paper ‘Agency’, he says that a person performs an action “if and only if there is a description of what he did that makes true a sentence that says he did it intentionally” (Davidson 1971, 46). This explication is a pre-analytic one, since it appeals to the notion of doing something 1 Deviant causal chains (DCCs) occur in causal theories of reference, knowledge and perception. A synoptic account of DCCs in the various theories of intentional phenomena is still missing. Most philosophers have concentrated on the case of action. The closest analogy is probably found in the causal theory of perception. Here I am thinking of cases of “veridical hallucination”, as described by David Lewis (1986). In such cases the scene before my eyes happens to cause a hallucination of that very scene, such that the visual experience matches the scene before my eyes, and I still do not see that scene.

70

Intentionality, Deliberation and Autonomy

intentionally. This is, in Davidson’s words, “to analyse the obscure by appeal to the more obscure – not as pointless a process as it is often thought to be, but still disappointing” (ibid., 47). His second attempt, therefore, aims at ﬁnding “a mark of agency that does not use the concept of intention” (ibid.). We should keep in mind, however, that Davidson does not dispute the adequacy of equating “performing an action” with “doing something intentionally”. What he seeks is a further analysis in terms of beliefs and desires. In “Intending”, he arrives at the following analytic deﬁnition: An action is performed with a certain intention if it is caused …[2] by attitudes and beliefs that rationalize it. (Davidson 1978, 87)

Note that Davidson’s analysandum is not “performing an action” tout court, but performing an action “with a certain intention”. Now strictly, there can be no counterexamples to a deﬁnition. Deﬁnitions may be more or less useful, but they are neither true nor false. Things are different if the causal theorist’s deﬁnition is meant to be an analysis of “doing something intentionally”. Only by accepting this analysandum are DCCs counterexamples to the causal theorist’s deﬁnition. If the causal theory of action does not ﬁnd a way of ruling out deviant causal chains, it classiﬁes a sequence consisting of a belief-desire pair, a bodily movement and a wanted effect as a case of intentional action which none of us would accept as being intentional. If the culprit sincerely declares “I didn’t mean to do it like that”, a theory of action should be capable of providing an explanation that accounts for this evidence. We are well advised to reject any deﬁnition or explanation that ties the agent down to an intention that he did not have. It seems desirable to rule out deviant cases by reﬁning the conditions that have to be fulﬁlled for a wanted effect to be intentional. Unfortunately only a few philosophers have turned this coin over to take a look at the other side. On closer examination, there are good reasons for not trying to exclude DCCs by reﬁning the conditions. Davidson thinks that the price for excluding them would be too high, amounting to establishing strict (that is exceptionless) intentional laws. Davidson’s well-known arguments for the anomalism of the mental will not be restated in this chapter. The arguments will however be touched upon, DCCs themselves being falsifying instances for such laws. I call the view that right sorts of causal histories cannot be speciﬁed Davidson’s defeatism. Calling DCCs “unavoidable” or “ineliminable”, as many authors do, is ambiguous; hence one word of clariﬁcation. The issue is not whether nature is capable of thwarting our plans or not. Of course even our best intentions and plans can fail. The question under discussion is not whether DCCs can be avoided, but whether they can be ruled out by the causal theorist’s analysis of “doing p intentionally”, that is, whether the conditions the causal theory sets can be made strict enough to exclude all deviant cases, without excluding too much.

2 I have skipped the phrase “in the right way”, which anticipates the problem of causal deviancy.

What Do Deviant Causal Chains Deviate From?

71

2. Some examples A person wants to commit suicide. She decides to play it safe. She swallows an overdose of sleeping pills, walks onto a river bridge, puts a rope around her neck, ties it to the handrail, loads her pistol and holds it to her forehead. Then she jumps down, inadvertently jerks the pistol, pulls the trigger, shoots through the rope, plunges into the icy water, is just saved from drowning, brings up the pills – and dies of pneumonia three days later. Has she achieved her aim? Yes and no. The suicide had not intended to die in this way. Nevertheless, in this case one could perhaps reckon on some posthumous remark such as “So be it”. Proceeding in such a circumspect way, the suicide has shown that her weariness with life clearly outweighed her preference for a particular way of dying. So, if her intention was to die no matter how, she has reached her goal. If her intention was to die in one of the four ways planned, she has not reached it. Perhaps this case calls for special treatment. We might introduce a class of ‘stoic’ intentions characterized as follows: “to bring about the state p while accepting any conceivable incident, provided it helps to reach p”. For such a class of intentions, the problem of DCCs would not arise. It is doubtful, however, whether such intentions deserve their name. The classical DCC examples are cases where the interfering event was unforeseen, perhaps even unforeseeable. Hence the event could not have been represented in my intention or plan. The retrospective “So be it”-style rationalization is a mere rationalization. It merely covers up the fact that what happened was not intended at all. For a causal theory of action, moreover, only those attitudes and intentions that an agent has formed beforehand are of use. The corresponding mental events are assumed to be the cause of the agent’s bodily movement, and no later change in the agent’s attitudes can contribute to that cause. What is done is done, and what was intended was intended. The class of intentions that I was tempted to introduce in order to save the suicide (her rationalization, not her life) is presumably empty. We cannot allow such stoic intentions that embrace DCCs. Not bothering about unforeseeability would blur the distinction between things we do and things that happen to us. Let us consider three familiar examples of DCCs. (i) The sniper: Daniel Bennett devised the case of an unpractised gunman who intends to shoot someone. His shot misses the victim, but startles a herd of wild pigs that trample the victim to death (cf. Davidson 1973, 78). (ii) The driver: A reckless car driver runs over a pedestrian who turns out to be the driver’s uncle whom he wanted to kill (cf. Chisholm 1966, 37). (iii) The mountaineer: A mountaineer wants to get rid of the second man on the rope. This thought unnerves him and causes him to loosen his grip (cf. Davidson 1973, 79). In all these cases a person’s beliefs and desires cause and rationalize a bodily movement. It seems that all the conditions set by the causal theory of action are met, and yet we are not inclined to admit that the agent produced the effect intentionally.

72

Intentionality, Deliberation and Autonomy

The would-be sniper neither intended to startle the wild pigs, nor did he alert them intentionally; how could he if he did not even notice them? He wanted the victim’s death, but he did not want this death. The pre-analytic judgement and the causal theory’s analysis diverge.

3. Basic and nonbasic deviance The cases hereto presented fall roughly into two groups. DCCs have been classiﬁed according to the stage in which the chain degenerates. In the wild pig case the causal chain degenerates outside of the agent’s body, or, to put it in terms of temporal succession, after the agent has made his contribution. In the mountaineer case, the interfering event occurs inside of the agent’s body. The example shows that short as well as long causal chains may degenerate. DCCs do not only occur in dominostyle actions, or in extended, planned activities where several steps have to be taken. DCCs are undemanding; basic actions are affected as well. Davidson speaks of “internal” and “external” DCCs. Bishop distinguishes “basic” from “non-basic” deviance, Mele “primary” from “secondary” deviance, Brand “antecedential” from “consequential waywardness”. The distinctions are not completely equivalent. In particular, the internal/external distinction seems of limited interest, since meanderings that happen inside the body may be just as unintended and beyond one’s control as external ones.3 I shall adopt Bishop’s terminology: in cases of “basic deviance”, the deviance “affects the causal link between mental states and basic action” (Bishop 1989, 133), while nonbasic deviance affects the causal link between the basic action and its further effects, as in the wild pig case. This distinction cuts across the agent’s bodily limits, since a causal chain that degenerates inside an agent’s body might do so after he has begun his basic action. Set out in this way, the distinction marks the difference between cases where, despite the DCC, something was done intentionally, whatever it was, and those cases where no action was performed at all. Both the sniper and the driver cases are clear cases of intentional action. Having startled the wild pigs unintentionally, the unpractised gunman can resort to a further, more basic description of his deed under which it was intentional, for example “shooting at the victim”, or “pulling the trigger”. Things are different with the uncomradely climber. Loosening one’s grip as a result of nervousness is not intentional under any description, and thus cannot count as an action (according to Davidson’s criterion, which is widely accepted in the debate). Both cases, basic and non-basic deviance, are counterexamples to standard causal analyses of “doing p intentionally”, that is performing an action of a speciﬁed type. If however the analysandum is “acting intentionally”, instead of “doing p 3 To Stoecker, “it is hard to see why such a difference [that is, whether the causal chain goes wayward inside or outside the agent] should matter much. Imagine Bennett’s would-be killer as a would-be doctor instead, who wants to self-cure his nausea with an injection of a substance that is in fact totally ineffective. Yet, clumsy as he is, he hurts himself so badly with the needle that the pain makes him vomit, which immediately cures his stomach. Although the causal chain leading to his recovery from nausea goes wayward inside of him, the example is obviously of the same type as the pigs case” (Stoecker 2003, 302).

What Do Deviant Causal Chains Deviate From?

73

intentionally”, only the cases of basic deviance count as counterexamples (cf. Brand 1984, 18). The most comprehensive classiﬁcation of deviant chains was suggested by Christoph Lumer (see Lumer forthcoming, ch. 4.2). Lumer, incidentally, does not speak of “deviant causal chains”. He makes it clear that strictly the term is a misnomer, since it does not capture cases where the deviation from an agent’s intention does not rest in some causal link in a chain, but is due to a certain faulty assumption of the agent’s. If, for example, the agent reaches his aim by mistaking someone for somebody else, what goes astray is not a causal process from his intentional state to his bodily movement or its further effects. To cover such cases of “tertiary waywardness” (Mele’s term), Lumer suggests the collective term “deviant realizations of intentions”. While the clariﬁcation is helpful, I shall stick to the inherited label “DCC” for the sake of convenience.

4. Some suggested solutions 4.1 Further speciﬁcations of the mental cause When Davidson brought up the DCC problem, he was in search of an analysis of “doing something intentionally” that does not use the concept of intention. Davidson never did dispute that performing an action amounts to doing something intentionally, but until the mid-1970s he sought a further analysis that manages with beliefs and desires. The climber case may suggest that substituting intentions for belief/desire pairs will rule out basic DCCs. But this is not the case. Even if in Davidson’s original example the climber’s beliefs and desires immediately cause his nervousness, the story can easily be modiﬁed so that he loses control over his hands after he has formed a proximal intention. The question as to what it is that DCCs deviate from remains a hard one even if we extend our conceptual resources and allow ourselves to resort to intentions. Several authors have proposed to specify in greater depth the mental cause of an action in order to rule out DCCs. They have taken the step from intentions to plans. As a matter of fact, many of our intentions are not merely directed towards the desired outcome of an action, but they include a detailed anticipation of how this state of affairs should be reached. These plans comprise, we might say, meansdirected intentions, in addition to the end-directed intentions. So the suggestion is that something being brought about in the right way often means in the way planned and anticipated, so that DCCs would deviate from an agent’s plans. In discussing Chisholm’s driver example, Jerome Shaffer suggests the following treatment: It is a case in which the end (the death of the uncle) requires a means. … To deal with this, we must add the further condition that where it is necessary to employ some means in achieving an end, to bring about the end intentionally one must also bring about the means intentionally. (Shaffer 1968, 105)

This condition is not met in Chisholm’s case, therefore, according to Shaffer, “the purported counterexample fails” (ibid., 106).

74

Intentionality, Deliberation and Autonomy

True, achieving an end requires, in all cases of non-basic action, employing some means, and some ends involve taking more than one step. Taking the appropriate steps, however, may not always sufﬁce to achieve one’s aim. Over and above taking the appropriate steps, the agent must rely on the causal chains he has triggered running their usual course, that is, running as expected. And if they do not? If something unforeseen happens to interfere, there are three possible cases to be distinguished: (a) The action fails, that is, the state of affairs that the agent wanted to bring about is not reached. (b) The agent manages to react to the incident. He succeeds in including the interfering event in his plan, or in making the requisite adjustments to the new situation in such a way that the desired result is attained nonetheless. Neither (a) nor (b) are cases of DCCs. A deviant chain only develops if the agent cannot ﬁt the incident into his action-plan, and if: (c) the desired state of affairs does materialize, but not in the way planned or foreseen. In this last case the agent is merely, so to speak, nature’s handyman. He works as a medium that nature uses to reach a state of affairs which, by pure chance, coincides with the state of affairs the agent tried to bring about in his way. Now, had we questioned Chisholm’s driver about his intentions before the incident, his answer probably would have been “I want to kill my uncle.” Well, there you are then! What else can we do but check whether the outcome of the action matches what the agent had sincerely declared to be his intention? Moreover, beyond what he sincerely declared to be his intention is there room for the question as to what he really intended? Shaffer’s suggestion is that DCCs deviate from an agent’s plans, though not from his end-directed intentions. Insofar as action plans encompass a series of steps that have to be taken, and a reliance on foreseeable regularities, the way the agent caused the uncle’s death obviously deviated from his plan. This deceptively simple answer is, however, not satisfactory for the following reason. On closer examination, every course of events deviates slightly from our plans and expectations, while not every such mismatch constitutes a DCC. To make things worse, our action plans do not anticipate the course of events down to the last detail, so that it is often not possible to cite a speciﬁc element in a plan that an unexpected incident deviates from. We shall return to this point in sections 6 and 7 below. 4.2 Gricean deference Goldman, Armstrong and Mele employ a strategy which has been called “Gricean Deference”, in allusion to a proposal that Paul Grice has made for the causal theory of perception.

What Do Deviant Causal Chains Deviate From?

75

Gricean Deference is an analytic technique. It is supposed to allow a philosopher to assert, say, … that perception involves a particular causal relationship, while freeing him from any responsibility for an account of the cause’s or the causal relationship’s precise nature. For this account the philosopher defers at a certain point in his analysis to the specialist in the relevant special science, who is competent to speak on the causal relationship in question. (Cooper 1976, 91)

Applied to action theory, the proposal is to supplement the condition “if caused by beliefs and desires” with the clause “in the right way”, or “in the normal way”, while deferring, or delegating, any further speciﬁcation of that normality clause to some special science. This is exactly what Goldman does: [P]recisely what is this ‘characteristic’ mode of causation by which wants and beliefs cause intentional action? … A complete explanation of how wants and beliefs lead to intentional action would require extensive neurophysiological information, and I do not think it is fair to demand of a philosophical analysis that it provide this information. (Goldman 1970, 62)4

The ﬁrst step in this move is to add a normality clause. It is indeed tempting to say that DCCs deviate from the “usual” or “normal” way things go. But what does this mean? The mere words “normal”, “usual”, “right”, “wrong”, “deviant” or “wayward” are of no help here. Mother Nature draws no distinction between normal and deviant causal chains. Nature knows no right or wrong ways to bring things about. Whatever may go wrong with our actions, the course of events will not violate laws of nature. Various participants in the debate have observed that “the notion of deviance has not been given a sense in relation to causality” (Mitchell 1982, 353).5 A comparison with the difference between the desired and the undesired effects of a medicine might be illuminating. Nature does not recognize this difference either. The standard phrase to appear in the instructions reads: “Besides their main effect, medicines may have unwanted effects, so-called side effects.” This phrase is very instructive. The difference between effects and side effects is exclusively that between wanted and unwanted effects. The property of being unwanted does not correspond to a distinct physical property. Side effects do not constitute a natural kind in any physical science. But perhaps a DCC’s being deviant does? Armstrong tries to reinforce Goldman’s line of argument by using a computing analogy. His solution is very simple: “P does not bring about Q as a result of standard computing practices”, due to a “disorganization of the computer’s internal processes …. In the same way, we have within ourselves certain ‘mechanisms’ with certain powers”, and these mechanisms may malfunction (Armstrong 1981, 84–5). While Armstrong employs the computer model of mind, Ginet (1990, 41) and Lumer simply invoke “reliable” or “sufﬁciently match-ensuring” mechanisms. 4 Surprisingly, Myles Brand imputes the view “that specifying causal normalcy is a scientiﬁc matter” (Brand 1984, 19) to Davidson. One wonders which passage in Davidson may have abetted this misreading. As far as I can see, Davidson has never encouraged scientiﬁc investigations into right sorts of causal histories. 5 Cf. Kenny 1975, 121; Searle 1983, 139; Brand 1984, 19; and Føllesdal 1985, 322.

76

Intentionality, Deliberation and Autonomy

But it is hard to see in which way the “mechanisms” of practical reasoning should have malfunctioned in the example Armstrong discusses (the climber case). Up to the output of the deliberation system everything works ﬁne. Why should neurophysiologists sort out physiological processes that show no eccentricities whatsoever, except that they sometimes form links in causal chains that are a headache for some philosophers of action? And how could neurophysiologists sort such processes out? Our brains could not care less about DCCs.6 But the Gricean deferentist is not yet left empty-handed. Perhaps it is not the mechanisms of practical reasoning that malfunction, but some later physiological process. Mele suggests that in the climber case, there is an anomaly in the motor control system that can be identiﬁed physiologically. He devises a class of agents that are wired in such a way that they “can perform overt actions only in cases in which the acquisition of a proximal intention initiates the sending of motor signals to appropriate muscles”, so that “any alien intervention into normal processes of guidance simply shuts down the motor control system” (Mele 2003, 61 and 62). For such agents, the climber case would not threaten the causal theorist’s deﬁnition of intentional action, since the unnerved climber would either perform no action at all or at any rate not the action of dropping his partner.7 Mele admits that his solution is stipulative. He leaves it open whether the biological set-up of human agents is as he devises, that is, whether humans engage in overt action only if no alien intervention into normal guidance processes occurs. It is noted above that the strategy of Gricean deference is prone to underestimate the problem of identifying a chain as deviant in the ﬁrst place. It is one thing to recognize that Davidson’s climber has lost “control” or “guidance” of his movements, it is quite another thing to be able to specify this effect in a naturalistic way, that is, without invoking unanalysed agentive notions such as “guidance” and “control”. It should be clear that the causal theorist’s analysis “can be satisfactory only if it avoids any reference to actions, agent-causation, exercises of control, and the like” (Bishop 1989, 98).8 Now it is in the spirit of the Gricean deference strategy that the question of naturalistic speciﬁcability does not get deﬁnitively settled. According to Goldman (1970, 62), it is “not fair to demand of a philosophical analysis” that it provide the required neurophysiological information. But it seems that the task of specifying 6 For a more detailed criticism of Armstrong’s account, see Bishop 1989, 134–5. 7 Mele distinguishes various cases (2003, 59–60): (1) Either the agent’s intention does not result in motor signals being sent at all. (2) Or the motor signals do not reach his muscles. Instead they unnerve him and the relevant intention is quickly extinguished. (3) Or the intention is not extinguished, and in that case the agent’s nervousness temporarily deprives him of control over the motion of his hands. (4) The signals reach the muscles, his ﬁngers start moving. This unnerves him, with the result that his grip loosens. In this case the movement of his ﬁngers is not a “direct ballistic continuation” of his previous motions, hence no guidance, no control, no intentional action. In neither case, Mele claims, does the agent perform the action of dropping his partner. 8 This constraint is neglected in many answers to the DCC problem, for instance in Adams’s (1989) demand that the agent must not have the feedback control loop broken prior to his part in the action coming to a close.

What Do Deviant Causal Chains Deviate From?

77

a causal chain as deviant has two parts: ﬁrst, classifying a sequence consisting of a belief-desire pair, a bodily movement and a wanted effect as unintentional, and second, trying to correlate this feature with some anomaly in the physiological or physical chain. It seems that the ﬁrst part of the task can only be achieved by a philosophical analysis, and that the second question cannot even be addressed before the class of “deviant” cases has been speciﬁed in intentional vocabulary. And it might well turn out that any physiological predicate that covers the relevant malfunctionings would have to be wildly disjunctive. If Mele is right, however, the deviant sequences do constitute a well-behaved physiological type. Since this matter cannot be settled here, I conﬁne myself to the observation that if the naturalistic strategy of Gricean deference works, it does so only for cases of basic deviance, where the output of practical deliberation is prevented from causing the right kind of physiological process. It cannot work for tertiary waywardness, nor for nonbasic deviance, where all is ﬁne both with the practical deliberation and with its causing a physiological and behavioural process that constitutes a full-blown action, as in the wild pigs case. Incidentally, many participants in the debate, including Davidson, think that basic deviance is harder to handle than non-basic deviance. I submit that exactly the reverse is the case. It is nonbasic deviance that deﬁes a naturalistic solution. 4.3 The causal immediacy strategy The last proposal I will scrutinize is the “causal immediacy strategy”, as John Bishop calls it (Bishop 1989, 138). The idea is that DCCs spread in a region which, in normal cases, must not exist, that is, in a causal gap between the mental antecedent and the bodily process caused. Myles Brand, one of the champions of the causal immediacy strategy, insists that there be no such gap: An adequate Causal Theory must preclude the possibility of these types of interventions: there can be no causal space between the mental antecedent and the beginning of the physiological chain. (Brand 1984, 20)

As a remedy, Brand introduces the notion of “proximate causation” (ibid.). Roughly speaking, he simply disallows any intervening events. The only bodily movements that count as actions are those caused by immediately preceding mental events. It is obvious that the causal immediacy strategy is designed specially for basic deviance. (Resorting to action plans, on the other hand, was designed for non-basic deviance.) The main problem with the causal immediacy strategy is that it is too restrictive. It excludes many clear cases of doing something intentionally. Normally, a spatiotemporal gap between mental antecedent and bodily movement does not preclude the latter being performed intentionally. 9 Brand should be prepared to spell out exactly what must not happen in the meantime.10

9 See Keil 2001. 10 In addition, Brand’s account is dependent on a questionable notion of causal relata, as John Bishop draws attention to: “Have we … any reason to suppose that [the very idea of

78

Intentionality, Deliberation and Autonomy

A variation on the causal immediacy strategy demands that the agent’s intention accompany every stage of the action. John Searle demands that there be “continuous efﬁcacy of intentional content”: We can now state the conditions necessary to amend the account as to eliminate all the deviant causal chains we have considered. A ﬁrst condition is that there should be continuous efﬁcacy of Intentional content under its Intentional aspects. (Searle 1983, 138) It is not enough that one intended at some prior time, to perform the act; this intention must have persisted into the time when one acted, for the act to be intentional. (Mitchell 1982, 353)11

Again, the proposal has to be evaluated twice, for basic and for non-basic deviance. If applied to the causal link between mental antecedents and basic action, the proposal coincides with the immediacy strategy just discussed. If applied to the subsequent stages, it becomes questionable what “continuous efﬁcacy of intentional content” amounts to. How does the agent ensure such efﬁcacy? Perhaps our riﬂeman, watching the wild pigs do their bloody deed, says to himself: “Oh, obviously the intentional content of my mental state isn’t effective any more. Next time, I must equip it with more causal power!” The truth is that once the causal chain has left the agent’s body, there is nothing whatsoever he can do to make his intention “continuously effective”. The bullet ﬂies simply too fast. On a less absurd reading, Searle’s demand is that the agent keeps control of the consequences of his basic action. Such a condition, however, is forbidden fruit for the causal theory of action, which tries to get by with event causality. Unanalysed talk of control violates this constraint (see Bishop’s criticism quoted above). Furthermore, Searle’s condition is too strong. As shall be explained below, the density of actual courses of events precludes a perfect match with our intentions, anticipations or plans. And in many cases there is not even a fact of the matter about these incongruencies, since nature knows more than one way to fulﬁl our intentions. Hardly anything happens exactly according to our intentions and plans. This fact is not accounted for in any of the suggested remedies discussed hitherto. And it is this fact that makes our title question so hard to answer. While the ﬁrst two moves – further speciﬁcation of the mental cause and Gricean deference – still operate within the conceptual framework of the causal theory of action, this is not so obvious for the causal immediacy strategy and for Searle’s demand of continuous efﬁcacy of intentional content. Since I am interested in the challenge that DCCs pose to the causal theory’s original event-causal analysis of “doing something intentionally”, I do not consider here any proposals that go beyond the Davidsonian metaphysics of singular causation between events. This is why I neglect teleological accounts, such as Sehon’s (1997), as well as the “differential explanation” strategy, as developed by Peacocke (1979) and Bishop (1989). Referring to “the existence of a suitable pattern of counterfactual dependence of

proximate causation] makes any more sense than a corresponding notion of ‘proximate’ points on a line?” (Bishop 1989, 139). 11 Thalberg, Audi and Frankfurt have made similar suggestions.

What Do Deviant Causal Chains Deviate From?

79

resulting behavior upon the content of the agent’s basic intention” (Bishop 1989, 150) goes beyond the conceptual resources of an event-causal theory that views causality as an extensional relation. Besides, using the assessment as to whether, “had the agent’s intention differed in content, the resulting behavior would have differed correspondingly” (ibid.), as a means of identifying DCCs, would require a superior kind of knowledge which is not available to the agent in his prior practical reasoning. The causal theory of action can invoke only those attitudes that cause the agent’s movement. No later change in the agent’s attitudes and no third party assessment can contribute to that cause. Perhaps the differential explanation strategy can even be harmonized with my own account (see below), but it cannot be moulded into a condition a Davidson-style causal theory could use to improve its original deﬁnition.

5. Defending Davidson’s defeatism At this point it might be helpful to recall the problem that Davidson responds to when bringing in, in addition to the agent’s “primary reason”, the causal relation between the reason and the action performed. The problem is that, in Davidson’s words: a person can have a reason for an action, and perform the action, and yet this reason not be the reason why he did it. Central to the relation between a reason and an action it explains is the idea that the agent performed the action because he had the reason. (Davidson 1963, 9)

The agent’s rationalization must not, in other words, be a mere rationalization. The causal relation is brought into play in order to single out, among the reasons the agent had, the one that was effective in bringing about his behaviour. The appeal to causal efﬁcacy is indispensable, since the rationalization by itself “provides no reason for saying that one suitable belief-desire pair rather than another (which may well also have been present in the agent) did the causing” (Davidson 1987, 42). The next thing we can expect the causal theorist to do is to specify this effective reason, or to tell us how it can be identiﬁed. But Davidson disappoints us here. He does not take any further steps to specify the effective reason. Among the reasons the agent had, he picks out the one which yields the most plausible explanation, but instead of specifying its position in the causal network of the physical world, he just calls it the cause, which is not particularly informative. The only principle Davidson employs here is the rule of thumb best reason = strongest cause. For an analytic deﬁnition of “doing p intentionally”, he wishes to get by with the following elements: (1) the agent’s bodily movement, (2) his primary reason (citing a pair of belief and proattitude), (3) the fact that the primary reason rationalizes the action, and (4) the fact that the primary reason (or rather, its occurrence) causes the action. We may complete the last condition by demanding that the cause operate “in the right way”, but with this we run out of resources. The clause “all causal chains running normally” is nothing but a general ceteris paribus clause, serving to express our unwillingness to accept DCCs but giving no hint as to how to rule them out. More than these elements

80

Intentionality, Deliberation and Autonomy

we do not have, and, according to Davidson, we should not want to have. His reason is that improving on these conditions by further specifying them, in a way that should eliminate wrong causal chains, would also eliminate the need to depend on the open appeal to causal relations. We would simply say, given these (speciﬁed) conditions, there always is an intentional action of a speciﬁed type. This would be understood as a causal law, of course, but it would not need to mention causality. Unavoidable mention of causality is a cloak for ignorance; we must appeal to the notion of cause when we lack detailed and accurate laws. (Davidson 1973, 80)

With this clariﬁcation, we have a most interesting interim result. Davidson has turned out to be anything but a prototypical causal theorist. Among the agent’s rationalizing beliefs and pro-attitudes, he picks out a pair and calls it the cause of the action – and leaves it at that. A causal theory of action that makes explanatory use of the causal connection between the belief-desire pair and the action performed cannot leave it at that. Such a theory will have to specify the conditions required to eliminate the deviant cases. It will set out in search of a right sort of causal history argument. Davidson, however, is convinced that there cannot be such an argument. “What I despair of spelling out”, he confesses, “is the way in which attitudes must cause actions if they are to rationalize the action”. All we can say is that the effect has to be brought about “in the right way” (Davidson 1973, 79). Davidson’s despair in the face of DCCs is closely connected with his anomalism thesis of the mental. If we wanted to specify the conditions in such a way as to exclude DCCs, we would have to be in possession of something which does not exist, namely strict intentional or psycho-physical laws. If the price for excluding DCCs by strengthening the conditions is too high, we are facing a dilemma, if not a paradox. A causal chain has badly degenerated, and we witness the agent’s sincere declaration, “I didn’t mean to do it like that.” On the one hand, we are somehow convinced that this testimony has a truth-maker. On the other hand, given the agent’s prior beliefs and desires, it is hard to explain why he is so dissatisﬁed. Did he not reach his declared goal? The Davidsonian background of the dilemma is this: on the one hand, various kinds of complications must be allowed so that human action can be the kind of thing it is – viz. intentional behaviour holistically embedded in a person’s comprehensive pattern of interlocking beliefs and pro-attitudes, its interpretation constrained by normative considerations. Sequences of beliefs, pro-attitudes and bodily movements that fall under strict psycho-physical laws would no longer amount to the same thing, according to Davidson. Too many things can interfere with the anticipated courses of events to allow for a comprehensive list to be established. Due to the holism and to the normative character of the mental, there exist no true lawful generalizations of the form “given these (speciﬁed) conditions, there always is an intentional action of a speciﬁed type” (Davidson 1973, 80). If we found such a law, Davidson claims, we could be sure that we would have changed the subject. On the other hand, a desired effect that was brought about in too bizarre a way ceases to be a case of “doing p intentionally”. Now, the question of what it is that DCCs deviate from has to be distinguished carefully from the question of how the causal theorist’s deﬁnition can be reﬁned so that DCCs are excluded. Regarding the

What Do Deviant Causal Chains Deviate From?

81

latter question, I share Davidson’s defeatism. It is a deplorable shortcoming of the DCC debate that the former question has often been skipped in favour of the latter.

6. The deep problem The underlying problem, which makes the question of what DCCs deviate from so hard to answer, is the fact that hardly anything in the world happens exactly according to our intentions and plans. The reason is not that the world is populated with evil demons who keep playing tricks on us. Quite the contrary: “Only a world that is regular in its sequence of happenings is a world in which intentions can be formed and executed. Only actions that have anticipated consequences are actions that can be performed for reasons” (Beck 1975, 116). It is not only that the success of our undertakings depends on Mother Nature’s benevolence. Beck makes the stronger claim that in a chaotic world no beings could evolve that can form intentions in the ﬁrst place.12 However no exceptionless regularities or strict empirical laws are needed for that purpose. Defeasible regularities sufﬁce. The reason why the course of events typically deviates slightly from our expectations is that the world is thicker, or denser, than our mental representations of it. Our expectation of what will happen is, even in non-deviant cases, not detailed enough to be accurate. We do not anticipate our bodily movements right down to the last detail, let alone the way our movements will interact with the rest of the universe. Davidson is well aware of this fact: [A]n intention cannot specify all the characteristics of the intended act that are relevant to its desirability. No matter how elaborately detailed an intention is, there are certain to be endless ways in which it could be realized that are unwanted and unintended by the agent. (Davidson 1985, 196)

Davidson speaks of intentions here, but it should be clear that with respect to foreseeability, plans are no better off than end-directed intentions. Now our plans and intentions cannot only be thwarted in endless ways, they can also be fulﬁlled by more than one course of events, for “[o]ne’s intentions almost never specify an action so exactly that no action beside the particular action will satisfy them” (Morton 1975, 14). The important distinction to be drawn here is this: our intentions and plans are directed towards actions of a certain type, falling under a certain description, while the particular action performed is underdetermined by the propositional content of our prior intentions.13 Action plans are not based on accurate anticipations of the future, 12 Perry claims that even belief-desire psychology could not have evolved in a chaotic world. Human agents acting on beliefs and desires can rely on a certain benevolence on the part of Mother Nature, described by Perry as follows: “The actions we perform because we have certain desires and beliefs, are often of a sort that will promote the satisfaction of the desires if the beliefs are true” (Perry 1986, 194). 13 Cf. Davidson 1978, 96–7.

82

Intentionality, Deliberation and Autonomy

but on reasonable abstractions, which are made according to presumed relevance. In most cases we get by with these abstractions. Since we have not anticipated events in detail, we cannot compare them afterwards with our expectations point by point. Therefore it is, strictly, not correct to say that such details deviate from any element of our plan. If we did so, there would be no non-deviating chains left. As long as the countless unplanned details of an action performed are insigniﬁcant enough, there is simply no fact of the matter about “deviations”. Though the unplanned details exceed our plans, there is no speciﬁc element in the plan that they deviate from. Now, in our sample cases of DCCs the unplanned details that exceed our plans can no longer be ignored. But again, it is, strictly, not correct to say that the disturbance concerns a particular element of the agent’s intention or plan. The agent has, after all, simply not considered this complication before. He did not even notice the wild pigs, therefore he could not form an intention to avoid the incident.14 Planning his action and anticipating the future happenings, he did rely on some unspeciﬁc expectation that everything would go smoothly, but he had not drawn up a list of possible complications. In that sense all action plans include an implicit ceteris paribus clause. Since the causal route by which the result is reached might always meander in an unforeseeable way, the agent is not able to name a desired end state such that, having reached it, he would, whatever the case, have to be content with what has happened, and to admit to have brought it about intentionally. 7. Spots of indeterminacy Sometimes Mother Nature thwarts our plans, and some of these complications qualify as DCCs. But which ones? My leading question was not whether science can rid us of DCCs, but rather, what it is that DCCs deviate from. The fact that human agents are not Laplacean demons makes DCCs ineliminable, but it does not make them deviant. Hitherto our enquiries have not been very encouraging. It seems as if we have been collecting answers to the question as to what DCCs do not deviate from. We have learned that they do not deviate from nature’s way of bringing things about, since nature knows no right or wrong ways to bring things about. I have maintained that, strictly, DCCs do not even deviate from an agent’s prior intentions and plans. We can perhaps say that they deviate from certain assumptions of normalcy implicit in the agent’s intentions or plans, but as long as the notion of a normal course of events deﬁes analysis, this answer provides little illumination. It is typical of DCCs that we cannot cite a particular element of an agent’s intention that an interfering event clashes with. The deep problem about DCCs is that nature knows of more than one way to meet our expectations, so that in many cases events exceed our intentions and plans without coming into collision with them. This fact makes our title question so hard to answer. 14 Stoecker makes the same point for cases of basic deviance: “The climber had no plans or expectations about the origin of his loosening the grip and consequently no false ones” (Stoecker 2003, 301).

What Do Deviant Causal Chains Deviate From?

83

Now literary theory has coined the illuminating notion of “spots of indeterminacy”. All ﬁctional texts contain such indeterminate spots, or blanks, that is zones where there is no telling, on the basis of what the text says, whether a certain object or situation has a certain property or not.15 Take Emma Bovary. Does she have a mole on her right shoulder or not? If Flaubert does not tell us, nobody will. If the novel does not contain the information, the question must remain open. There is simply no fact of the matter. The real world, on the other hand, contains no such blanks. Every one of us either has a mole on the right shoulder or not (or more than one). Unlike ﬁction, the real world is fully determinate. Now our plans and intentions are like ﬁctional texts in that they contain indeterminate zones. The content of a future-directed mental representation is not as dense as the real world. The representation contains blanks which the real world can ﬁll in different ways, some of which constitute DCCs, some of which do not. But in both cases the ﬁlled-in details do not clash with any particular element of the intentional content of the agent’s prior attitude. The agent’s intentional attitude simply leaves the details open, just like Flaubert’s novel leaves open the exact constitution of Emma Bovary’s shoulder. 8.What deviant causal chains deviate from – and why this is bad news for the causal theory of action Having touched upon many related issues, we have still found no convincing reply to our title question. On the one hand, the climber, the driver and the riﬂeman have reasons to be somewhat dissatisﬁed with the course of events. On the other hand, every attempt has failed to specify conditions for a ‘normal’ course of events from which the eccentric killings were deviations. And worse still: as Davidson sees it, specifying conditions for a normal course of events that deviant chains deviate from would amount to the establishment of a strict intentional law of the sort “Given these (speciﬁed) conditions, there always is an intentional action of a speciﬁed type” (Davidson 1973, 80), and this venture is hopeless, according to anomalous monism. The question remains of how tight this package is tied up. The impression is hardly avoidable that there is a point to the agent’s sincere assertion that he did not mean to do it like that, and it is only fair to demand of a philosophical analysis that it accounts for this assertion. Otherwise we lack any reason to call the cases deviant. It seems desirable, though, to have an analysis which does not require a strict conceptual or nomic link. I dare say that the solution has been staring us right in the face all the time. It is worth noticing that the agent’s autobiographical report we wish to account for is always given in the past or perfect tense: “I didn’t mean to do it like that!” It is true: the riﬂeman did not kill his victim intentionally in this roundabout way, though nothing in his prior intention explicitly ruled out this course of events. When 15 “We ﬁnd such a place of indeterminacy wherever it is impossible, on the basis of the sentences in the work, to say whether a certain object or objective situation has a certain attribute” (Ingarden 1973, 50).

84

Intentionality, Deliberation and Autonomy

starting his action, the agent was not in a position to formulate an intention detailed enough to exclude every possible complication. In retrospect, he can tell whether this particular course of events was intentional or not. So I wish to propose the following answer: DCCs deviate from that counterfactual course of events which would have corresponded to the retrospectively speciﬁed formulation of the agent’s intention. Beforehand, the agent could take into consideration potential incidents only in the form of a general ceteris paribus clause. When the action is completed, his epistemic position is better. Now he knows (though not necessarily!) which of the vast number of potential complications has occurred, and he can substitute an explicit speciﬁcation of his intention for the general ceteris paribus clause. For the prior reservation “The course of my action will be intentional if the causal chain unravels normally”, he can now substitute: “The course of my action was unintentional because the wild pigs intruded.” What DCCs deviate from can only be told ex post actu. This result accords with the insight, which is not alien to Davidson (cf. 1963, 16), that explanation of action contains an irreplaceable ex post-element. We should not hesitate to apply this insight to the problem of DCCs. We cannot tell what they deviate from without referring to the agent’s retrospective speciﬁcation of his intention, which he was not in a position to give beforehand. I insist on the legitimacy to speak of a speciﬁed intention instead of an altered or a revised one. Surprised by the wild pigs’ interference, the agent narrows down his prior intention to kill the victim, but he does not withdraw it. If he did so, he would be lying about his past intentional states. In speaking of a retrospective speciﬁcation, I do not want to inaugurate a distinction between prior intentions and ex post actu intentions. There cannot be such things as ex post intentions, as long as we regard intentions as mental causes of actions, as the causal theory of action does. What is done is done, and what was intended was intended. The past cannot be undone, nor can the causal efﬁcacy of past events. The distinction to be drawn is that between intending to do p and doing it on the one hand and doing p intentionally on the other. The so-called “simple theory” equates both phenomena, but the DCC cases reveal that this cannot be right. The adverb “intentionally” qualiﬁes the agent’s attitude towards a given particular action, whereas the mental act of intending, as Davidson says, “cannot single out a particular action in an intelligible sense, since it is directed to the future” (Davidson 1978, 99). The ex post judgment, I wish to add, is de re and indexical, while the content of the prior intention is de dicto and descriptive. And only because nature can ﬁnd a way to place its complications into the unspeciﬁed parts of our plans and intentions, it is non-paradoxically true that not every action that arose from a prior intention has been done intentionally. This is bad news, though, for the causal theory of action. The causal theorist cannot gain from the suggested account. In need of reputable causal relata, he has restricted himself to prior mental episodes. The agent’s attitudes, or rather changes

What Do Deviant Causal Chains Deviate From?

85

in his attitudes,16 function as the cause of his bodily movement, and no subsequent modiﬁcation can contribute to that cause. For a Davidson-style causal theory, the agent’s prior attitudes perform a double duty: they play their role as causes, and they are supposed to explain the action. However, the explanatory force of the assertion that the agent acted exactly for that reason which caused his action was modest enough, since Davidson’s causal theory lacks any independent characterization of this cause. Davidson counters this objection with the claim that rationalization itself “is a species of causal explanation” (1963, 3), that is, that citing the appropriate belief-desire pair is per se causally explanatory. In view of our treatment of DCCs, however, it becomes hard to see in which way the ex post rationalization should pass for a causal explanation. Causal and rational explanations fall apart again. In deviant cases the belief-desire pair was still the cause of the action, and in cases of nonbasic deviance the beliefs and desires even rationalize the action under a certain description (for example “shooting at the victim”), but they are too unspeciﬁc to rule out DCCs. Nevertheless, we are not left empty-handed. We are capable of sifting out a particular causal chain as deviant. Our account is based on an ex post-speciﬁcation of the unspeciﬁed “right way”-clause. It is only the causal theorist who is left empty-handed. Davidson can stick to his view that “the propositional attitudes are by nature explanatory” (1986, 206) only if he parts company with the orthodox causal theory of action, which must comply with the agent’s prior attitudes. What enables us to sift out certain causal chains as deviant are the agent’s attitudes, but not the ones he had when he began his action.17 9. Three objections I would like to conclude with an attempt to counter three objections that may be raised against my proposal.

16 “[I]t is changes in the attitudes, which are events, which are the often unmentioned causes. … [T]he cause of the action was the advent of one or both of the belief-desire pair” (Davidson 1993, 288). 17 I have covered up a point of disagreement. When he introduces the DCC problem, Davidson says that “not just any causal connection between rationalizing attitudes[!] and a wanted effect sufﬁces to guarantee that the wanted effect was intentional” (1973, 78), while in other places he says that the beliefs and desires would have rationalized the action if they had caused it in the right way (1973, 79). In the latter, ‘hypotactic’ account he makes the rationalization dependent upon the existence of the right causal chain (rationalized if caused in the right way), while in the former, ‘paratactic’ account the rationality condition and the causality condition are being evaluated separately (rationalized and caused in the right way). Though Davidson does not seem to have noticed the tension between both formulations, the hypotactic account plausibly reﬂects his real opinion (for the details, see Keil 2002). But the hypotactic account is untenable. The riﬂeman’s practical deliberation was unobjectionable, and what he did, that is, aiming and pulling the trigger, was perfectly rational in the light of his beliefs and desires. The fact that the causal chain meandered does not affect the rationalization. His attitudes do rationally explain his basic action, even if they do not make it the case that he aroused the wild pigs intentionally.

86

Intentionality, Deliberation and Autonomy

(i) First, how can I account for hidden DCCs? We can well imagine DCCs which are only detected after years, or never, or not by the agent himself. Let us assume that our gunslinger has been successful. No wild pigs. The victim is hit right in the head. Many years later it turns out that the shot was a ricochet, which had bounced off a nearby rock. We might ask what this causal chain deviates from until the detour is discovered. And what if it is never discovered? Must we conclude that there is no deviancy in such cases? Well, I did not claim that DCCs are necessarily detected afterwards. I claimed that they can only be detected afterwards. Some DCCs may be hard to trace, but this has no bearing upon the asymmetry between what the agent can tell beforehand and what he can tell afterwards. Now imagine that the waywardness is detected by some observer, whereas the agent remains ignorant. In such a case the observer knows better what actually happened. This knowledge alone, however, does not put him in a position to judge that a DCC has occurred. Identifying a DCC is not merely a matter of discovering unknown physical facts. Ascertaining which meanderings count as DCCs requires a re-evaluation of the agent’s attitudes in view of the incident. Cases of insincerity and self-deception set aside, this re-evaluation must not contradict the agent’s own judgement. Such an investigation, however, concerns the veriﬁcation of a DCC, not its deﬁnition. Rejecting veriﬁcationism, as we should do, we must not infer from contingent problems of veriﬁcation that there is no fact of the matter about the phenomenon to be veriﬁed. (ii) The second objection stems from a further counterexample, which is another variation on the wild pig case. It goes like this: what if the riﬂeman had formulated his intention in a different way? What if he had declared beforehand that his intention was to kill the victim with a well-aimed shot to the head? Furthermore, let us assume that he had noticed the wild pigs, and had even considered the threatening complication, but then discarded it as too unlikely. He shoots, his assumption proves wrong, the wild pigs do their bloody deed. Would not, contrary to my claim, the course of events in this case clearly deviate from a speciﬁc element in the agent’s prior intention? And would not the agent have been in a position to tell in advance that, if the wild pigs should intervene, his intention would not be fulﬁlled?18 Both questions must be answered in the afﬁrmative. But I am not convinced. The objection is beside the point, since the present case would no longer be a DCC. For something to be a DCC, the state of affairs brought about must fulﬁl an agent’s prior intention. This condition is met in all the examples I used, but not in the present case. Here we are confronted not with a DCC, but with a case where the state of affairs that the agent intended to bring about is not reached (see above, section 4, case (a)). If the agent had considered the threatening complication beforehand, he would have intended either to include it in his plan or to avoid it. In both cases, the complication would not have constituted a DCC. It goes without saying that in these cases, as always, another complication might take place that bypasses even the more speciﬁc intention. 18 This objection was raised by Andreas Kemmerling. Discussing it with him helped me to clarify my position.

What Do Deviant Causal Chains Deviate From?

87

(iii) Third, one may raise doubts as to whether the agent’s retrospective reevaluation of his intention is grounded in fact. If nothing in his prior mental state ruled out the wayward chain, then which good reasons can the agent cite for the claim that he “didn’t mean to do it like that”? Does he enjoy a kind of ﬁrst person authority when he speciﬁes his intention in retrospect? Are there any objective constraints, or can he say just anything? Perhaps he has just changed his mind? Let us assume that the objection does not question the credibility of autobiographical reports in general. True, an agent who prefers to dissociate himself from his deed when seeing his victim in a pool of blood may be insincere. But such cases are of no interest for the present debate. If the agent has incorporated the wild pig complication into his secret plan, the complication would no longer constitute a DCC. We are dealing with cases instead where we have every reason to buy the agent’s tale. One fact is beyond dispute, viz. the agent’s prior intentional state. The past cannot be undone, and if the content of the agent’s intention was, say, “I’m going to kill this guy”, then no retrospective judgment can be sincere that renounces this intention. This is why I speak of a retrospectively speciﬁed intention instead of a revised one. Given my story about blanks in our anticipations, that is, given that futuredirected intentions can be fulﬁlled by a certain range of courses of events, the question obviously matters whether the deviation was a considerable deviation. In the literature, the case is discussed of a soccer forward who kicks the ball hard towards the opposing goal with the intention to score. The goalkeeper, however, manages to touch the ball, thus deﬂecting it slightly from a straight path, but not nearly enough to cause it to miss the goal. It does not seem that this unexpected deﬂection makes it wrong to say that S kicked the ball into the goal intentionally. (Ginet 1990, 79)

I agree. If the player’s intention to score the goal is fulﬁlled, then the exact trajectory of the ball is a matter of indifference to him. But this holds only within certain limits. What are those limits? What kind of mismatch is so signiﬁcant that it makes the result unintentional? Ginet simply demands that “one must not be too lucky” (ibid., 78). Lumer argues as follows: it is generally permissible to make probabilistic assumptions about one’s prospects of success. Humans act in a world where success is never guaranteed, hence intentional action must not be incompatible with probabilistic calculations. But some complications are beyond that statistical range. Lumer offers the following criterion: would the agent, had he known about the threatening complication before, have proceeded in the same way, or would he have changed his plan and tried in another way? The soccer player would arguably not have proceeded differently, hence he scored the goal intentionally (see Lumer forthcoming, 222). I feel free to adopt this idea, since it is in the spirit of my own proposal anyway. None of us knows the future. The piece of counterfactual reasoning that Lumer makes use of – Would knowledge of the incident have made a difference? – is only available ex post actu. Beforehand, in his practical deliberation, the agent is simply not in a position to judge whether knowledge of this deviation would affect his action plan

88

Intentionality, Deliberation and Autonomy

or not. And if the agent had considered the threatening complication before, it would not have constituted a DCC (see my reply to the second objection). His foresight being limited, the agent can assess the particular course of events only ex post actu, and decide of it, hence de re, whether he brought it about intentionally or not. But he might have said in advance which course he approves of, if only a Laplacean demon had drawn his attention to the threatening complication. And this is why the piece of counterfactual reasoning Lumer offers resolves the doubt as to whether the agent’s retrospective speciﬁcation of his prior intention is grounded in fact. Though there is, strictly, no fact of the matter about whether his prior attitudes ruled out the deviancy, there is a fact of the mind, as it were. The gap between both facts can be closed by applying the principle of the supervenience of the mental: a world in which the assertion “I didn’t mean to do it like that” is insincere and hence false, must contain a physical difference as well, wherever it is located. In a somewhat ethereal sense of “fact of the matter”, we may even say that there is such a fact ante actum, namely in the agent’s dispositions. Nobody asked him then, but this does not mean that the answer would have been arbitrary. I submit that the mental state he was in when deliberating his action ﬁxed which answer he would have given, had a being endowed with foreknowledge questioned him about the complication. This prior disposition serves as a constraint for his retrospective re-assessment of his intention. References Adams, Frederick (1989), ‘Tertiary Waywardness Tamed’, Crítica, 21, 117–25. Armstrong, David M. (1981), ‘Acting and Trying’, in David M. Armstrong, The Nature of Mind, Brighton, Sussex: Oxford University Press, pp. 68–88. Beck, Lewis White (1975), The Actor and the Spectator, New Haven, CT and London: Yale University Press. Bishop, John (1989), Natural Agency: An Essay on the Causal Theory of Action, Cambridge: Cambridge University Press. Brand, Myles (1984), Intending and Acting: Toward a Naturalized Action Theory, Cambridge, MA: Cambridge University Press. Chisholm, Roderick M. (1966), ‘Freedom and Action’, in K. Lehrer (ed.), Freedom and Determinism, New York: Random House, pp. 11–44. Cooper, W.E. (1976), ‘Gricean Deference’, Metaphilosophy, 7, 91–101. Davidson, Donald (1963), ‘Actions, Reasons, and Causes’, in Davidson 1980, pp. 3–19. —— (1967), ‘Causal Relations’, in Davidson 1980, pp. 149–62. —— (1970), ‘Mental Events’, in Davidson 1980, pp. 207–25. —— (1971), ‘Agency’, in Davidson 1980, pp. 43–62. —— (1973), ‘Freedom to Act’, in Davidson 1980, pp. 63–81. —— (1974), ‘Psychology as Philosophy’, in Davidson 1980, pp. 229–44. —— (1978), ‘Intending’, in Davidson 1980, pp. 83–102. —— (1980), Essays on Actions and Events, Oxford: Oxford University Press. —— (1985), ‘Replies’, in Bruce Vermazen and Merrill Hintikka (eds), Essays on Davidson – Actions and Events, Oxford: Oxford University Press, pp. 195–229, 242–52.

What Do Deviant Causal Chains Deviate From?

89

—— (1986), ‘Judging Interpersonal Interests’, in J. Elster and A. Hylland (eds), Foundations of Social Choice Theory, Cambridge, MA: Cambridge University Press, pp. 195–211. —— (1987), ‘Problems in the Explanation of Action’, in P. Pettit, R. Sylvan and J. Norman (eds), Metaphysics and Morality, Oxford: Oxford University Press, pp. 35–49. —— (1990), ‘Representation and Interpretation’, in K.A.M. Said et al. (eds), Modelling the Mind, Oxford: Oxford University Press, pp. 13–26. —— (1993), ‘Replies’, in R. Stoecker (ed.), Reﬂecting Davidson: Donald Davidson Responding to an International Forum of Philosophers, Berlin and New York: de Gruyter and Hawthorne. —— (1995), ‘Laws and Cause’, Dialectica, 49, 263–79. Føllesdal, Dagﬁnn (1985), ‘Causation and Explanation: A Problem in Davidson’s View on Action and Mind’, in E. LePore and B. McLaughlin (eds), Actions and Events: Perspectives on the Philosophy of Donald Davidson, Oxford and New York: Blackwell, pp. 311–23. Ginet, Carl (1990), On Action, Cambridge: Cambridge University Press. Goldman, Alvin I. (1970), A Theory of Human Action, Princeton, NJ: Princeton University Press. Ingarden, Roman (1973), The Cognition of the Literary Work of Art, Evanston, IL: Northwestern University Press. Keil, Geert (2000), Handeln und Verursachen, Frankfurt am Main: Klostermann. —— (2001), ‘How Do We Ever Get Up? On the Proximate Causation of Actions and Events’, Grazer Philosophische Studien, 61, 43–62. —— (2002), ‘Über eine Inkonsistenz in Davidsons Handlungsdeﬁnition’, in A. Beckermann and C. Nimtz (eds), Argument und Analyse – Sektionsvorträge, Paderborn: Mentis, pp. 668–78, . Kenny, Anthony (1975), Will, Freedom and Power, Oxford: Oxford University Press. LePore, Ernest and Brian P. McLaughlin (eds) (1985), Actions and Events: Perspectives on the Philosophy of Donald Davidson, Oxford and New York: Blackwell. Lewis, David (1986), ‘Veridical Hallucination and Prosthetic Vision’, in David Lewis, Philosophical Papers, vol. II, New York and Oxford: Oxford University Press, pp. 273–86. Lumer, Christoph (forthcoming), Kognitive Handlungstheorie, ch. 4.2: ‘Abwegige Absichtsrealisierung’. Mele, Alfred R. (1987), ‘Intentional Actions and Wayward Causal Chains: The Problem of Tertiary Waywardness’, Philosophical Studies, 51, 55–60. —— (2003), Motivation and Agency, Oxford: Oxford University Press. Mitchell, Dorothy (1982), ‘Deviant Causal Chains’, American Philosophical Quarterly, 19, 351–3. Morton, Adam (1975), ‘Because He Thought He Had Insulted Him’, Journal of Philosophy, 72, 5–15.

90

Intentionality, Deliberation and Autonomy

Peacocke, Christopher (1979), Holistic Explanation, Oxford: Oxford University Press. Perry, John (1986), ‘Circumstantial Attitudes and Benevolent Cognition’, in John Perry, The Problem of the Essential Indexical and Other Essays, New York and Oxford: Oxford University Press, pp. 193–204. Searle, John R. (1983), Intentionality: An Essay in the Philosophy of Mind, Cambridge, MA: Cambridge University Press. Sehon, Scott R. (1997), ‘Deviant Causal Chains and the Irreducibility of Teleological Explanation’, Paciﬁc Philosophical Quarterly, 78, 195–213. Shaffer, Jerome A. (1968), Philosophy of Mind, Englewood Cliffs, NJ: Prentice Hall. Stoecker, Ralf (ed.) (1993), Reﬂecting Davidson: Donald Davidson Responding to an International Forum of Philosophers, Berlin and New York: de Gruyter and Hawthorne. —— (2003), ‘Climbers, Pigs and Wiggled Ears’, in S. Walter and D. Heckmann (eds), Physicalism and Mental Causation, Exeter and Charlottesville, VA: Imprint Academic, pp. 295–322. Vermazen, Bruce and Merrill B. Hintikka (1985), Essays on Davidson – Actions and Events, Oxford: Oxford University Press.

Chapter 5

The Double Failure of ‘Double Effect’ Neil Roughley

The ‘doctrine of double effect’ claims that it is in some sense morally less problematic to bring about a negatively evaluated state of affairs as a ‘side effect’ of one’s pursuit of another, morally unobjectionable aim than it is to bring it about in order to achieve that aim. In a ﬁrst step, this chapter discusses the descriptive difference on which the claim is built. That difference is shown to derive from the attitudinal distinction between intention and ‘acceptance’, a distinction that is in turn claimed to ground in a feature of the decisions that generate the attitudes in question. The resulting analysis is then plugged into two different normative principles that may each be thought to specify the intuitions behind the doctrine of double effect, but which have frequently been conﬂated. The ﬁrst concerns the permissibility of bringing about the merely accepted state of affairs, the second its reduced attributability. It is argued that examination of the intuitions behind the two principles supports neither version of the doctrine. Rather, the intuitions are best captured in an attribution principle based on subjective probabilities and a principle of attitude evaluation, neither of which make explicit reference to the attitude of intending. In the 1950s Elisabeth Anscombe claimed that there could be no substantial progress in moral philosophy until clarity had been established on the central concepts of philosophical psychology, in particular on the concept of intention (Anscombe 1958, 26). She believed that establishing such clarity would reveal much of modern moral philosophy, in particular utilitarianism, to be built on shaky attitudinal foundations and the Catholic ‘doctrine of double effect’, in contrast, to be well founded (Anscombe 1961, 58–9; 1982, 23–4). This chapter will illustrate one way in which Anscombe’s ﬁrst claim is on the right lines: moral theory cannot get by without clarity on a whole set of issues that belong to philosophical psychology or action theory. Moreover, one such issue does indeed have consequences for the ‘doctrine of double effect’ and the intuitions it attempts to channel. However, I will be arguing against Anscombe that a precise understanding of the nature of intention reveals the doctrine to be without attitudinal foundation. The discussion takes place in ﬁve steps. It begins by examining the form the doctrine is ofﬁcially given within contemporary Catholic moral theology. The action theoretic resolution of a number of confusions and ambiguities permits

92

Intentionality, Deliberation and Autonomy

the identiﬁcation of a core normative claim. According to this claim, a particular distinction in philosophical psychology is of such signiﬁcance that it grounds an important difference in the moral evaluation of actions. The second step advocates a speciﬁc understanding of the relevant attitudinal distinction, between intending and ‘accepting’, an understanding provided by what I call the ‘upstream theory of intention’. Having clariﬁed the principle’s psychological basis, the discussion then, thirdly, goes on to distinguish two ways in which the attitudinal distinction may be thought to be normatively signiﬁcant, two forms of normative signiﬁcance that have frequently been conﬂated by advocates of the doctrine. These are best articulated in two distinct principles. Sections four and ﬁve of the chapter discuss whether the psychological distinction can play either of the roles that it is assigned by the two principles. My answer will be negative. Nevertheless, I hope to show that a combination of attention to the psychological basis of the principle and careful consideration of its possible function in the structure of moral evaluation can disentangle the valid intuitions that are mistakenly thought to support it. 1. From the doctrine of double effect to the principle of collateral consequences The so-called doctrine of double effect has been most frequently discussed in the context of ‘applied ethics’. In these discussions pairs of similar examples are juxtaposed in order to mobilize intuitions that the minimal descriptive difference between the cases grounds an important moral difference. Three such repeatedly discussed cases are the following: (i) Analgesia/euthanasia A doctor administers a dose of morphine in order to relieve pain, knowing that it will kill the suffering patient in the process. This is contrasted with an otherwise identical deliberate mercy killing by the same means. (ii) Hysterectomy/abortion A doctor saves the life of a pregnant woman, in the ﬁrst case, by removing her cancerous womb, thus also killing the foetus, in the second case, by crushing the skull of the foetus trapped in the birth canal. (iii) Disarmament bombing/demoralization bombing Bombs are dropped on a munitions depot, in spite of the deaths of children in a nearby school this will inevitably cause; in the second, behaviourally identical case, the target is chosen in order to kill the children and thus demoralize the enemy.1 1 For an early critical discussion of (i), see Hart 1968, 122–5. More recent discussions include Powers 1995, 341f. and Frey 2003, 465f. Example (ii) is given prominence in Jonathon Bennett’s Tanner Lectures on Human Values (1981), 106–9; Davis 1984, 109–16; Quinn 1989, 336–47 and Connell 2003, 880f. Bennett’s third Tanner Lecture focuses primarily on (iii) (1981, 95–116). Anscombe and Finnis both use the doctrine of double effect as a standard for the criticism of the “obliteration bombing” of cities, particularly as practised by the allies

The Double Failure of ‘Double Effect’

93

In each of these pairings there is a descriptive difference between the two cases that grounds in the way in which the agents view the dual results of their behaviour. Put succinctly, in each of the former examples, the death brought about by the agent is an event he views as a ‘mere side effect’ of his action; in each of the latter examples it is an effect he aims to bring about in order to achieve some further goal: relieving the patient’s pain, saving the woman’s life, bringing the war to an end. That there is some such descriptive distinction to be drawn here seems clear. The ‘doctrine of double effect’ makes two claims about cases in which that difference is given. First, it claims that it is morally permissible to cause death as a ‘mere side effect’ in the pairings just described. Second, it offers a set of conditions whose satisfaction it claims is necessary and sufﬁcient for permissibility in these kinds of cases. The conditions can be thought of both as specifying what we should understand by ‘side effects’ and as adding further conditions under which the descriptive difference thus made renders the agent’s behaviour permissible. According to the New Catholic Encyclopedia, there are four such conditions that have to be met conjunctively: 1) “The act itself” is either morally good or indifferent. 2) The bad effect is not “positively willed”, but “merely permitted”. 3) The bad effect is not a “means” to the good or indifferent end, i.e. the good or indifferent effect is “produced directly” by “the action”, not by the bad effect. 4) There is a degree of proportionality, the good effect “compensating for” the bad one.2 Condition (4), according to which bad effects should only be considered allowable where they are compensated for by good effects, is uncontroversial. However, each of the other three conditions raises conceptual questions to which we need answers in order to isolate the claim at the heart of the doctrine that merits serious consideration. 1.1 On (1): “the act itself” and “double effect” If we take literally talk of “double effect”, then the two events that are up for comparative moral evaluation ought both to be caused by the agent’s action. At ﬁrst glance it might appear that it is this action that is designated by the phrase “the act itself”. In (i) that would appear to be the injection of morphine and in (iii) the dropping of the bombs. Example (ii) pairs two different actions on the part of the in the Second World War. See Anscombe 1961, 59 and Finnis 1991, 42f. The case also features in Powers 1995, 345–8; Quinn 1989, 336–47 and in Michael Bratman’s application of his analysis of intention to the doctrine in Bratman 1987, 139–64. 2 Connell 2003, 880. I have given the conditions a slight pruning. In particular, I have left out the distinction between “the order of causality” and “the order of time”, by means of which the author attempts to clarify the notion of “means” employed in condition (3).

94

Intentionality, Deliberation and Autonomy

doctor. In the ﬁrst of these it is the hysterectomy that has the two morally competing effects; in the second we seem constrained to pick out the mere bodily movements of the doctor – his ‘basic action’ – if we want to separate his action from the death of the foetus as well as from the woman’s survival. Construing things in this way would, however, make condition (1) redundant: actions under these descriptions are all morally neutral; ‘basic actions’ would seem to be necessarily so. Interpretative charity, therefore, demands a different construal. What the phrase “the act itself” is presumably meant to isolate is the action under the description provided by the content of the agent’s intention. In other words, it is the content of the agent’s intention that has to be either morally good or at least indifferent. This has two consequences: ﬁrst, the two ‘effects’ at issue here are not caused by the ‘act itself’. Rather, the description that bequeaths us ‘the act itself’ is just the description of the agent’s behaviour as the ‘cause’ of one of the ‘effects’, that effect that is morally good or indifferent. What, then, are the two ‘effects’ effects of? If they are indeed effects, then they will both be caused by some ‘basic action’ – an action characterized in a way that makes it clear that it was not performed by performing any other action. However, as a bit of rudimentary action theory makes clear, the relationship between ‘basic actions’ and the events that make broader action descriptions true need not be causal. For instance, doing something that alerts someone to a danger and thus saves her life may also count as an insult or as breaking a promise. In such cases, we do not have two effects, but rather two results or consequences, one of which is an effect, the other of which is not. For this reason, we are dealing, strictly, with a principle that concerns consequences or results, not necessarily effects. This is also a reason to avoid the everyday talk of ‘side effects’. Picking up the term beloved of certain military leaders, I shall instead talk here of collateral consequences. The claim that the moral status of an action depends signiﬁcantly on whether the content of the agent’s intention was morally good or indifferent may, secondly, seem to leave the advocate of the doctrine open to an objection that Pascal raised against its use by the Jesuits. According to Pascal, the permissibility of just about any action might appear obtainable by an agent’s merely redirecting her intention.3 This is indeed a problem in certain cases that have been seen as providing applications of the doctrine. Example (ii) is a case in point: what is supposed to be the criterion for claiming that it is the second doctor’s intention to kill the foetus, rather than simply to crush its skull? On the one hand, the causal connection between the realization of the latter intention and the former effect cannot be sufﬁcient, otherwise the doctrinaire would also have to lumber the ﬁrst doctor with the intention to kill the foetus. On the other hand, the problem is not solved by insisting on a conceptual connection, as there is little plausibility to the claim that skull crushing entails killing.4

3 Kenny 1973, 140f. A similar suspicion has been reiterated by John Harris, who has claimed that where permissibility or responsibility is at issue, “intention can be so narrowly deﬁned as to yield any answer that is wanted”. See Harris 1980, 50. 4 This is the problem of ‘closeness’. On this point see Bennett 1981, 107–13; Anscombe 1982, 22f.; Davis 1984, 111–13; Quinn 1989, 336–41.

The Double Failure of ‘Double Effect’

95

Nevertheless, other cases, for instance examples (i) and (iii), look to pose no real problem as to which consequence is intended and which is not. Rather, the obstetrics/ abortion case seems simply to be a dubious application of the doctrine. In order to be clear on whether this is correct, that is, on whether identifying the contents of a person’s intentions is indeed less criterially wobbly than Pascal assumed, we will need an analysis of the concept of intention. 1.2 On (2): “positive willing” and “permitting” Condition (2) offers an attitudinal explanation of the notion of collateral consequences, which the negatively evaluated event has to instantiate: collateral consequences are not “positively willed” but “merely permitted”. A word is in order about both expressions. To begin with, the concepts of intention or aiming are to be preferred to the former, not only because they are more ﬁrmly grounded in everyday usage, but because talk of ‘positive willing’ as opposed to ‘permitting’ may encourage a confusion of the doctrine with a further controversial normative principle often cited by ‘deontologists’. This is the claim that there is an important moral difference between acting to make some proposition the case and allowing it to become or continue to be the case. The important point for our purposes is that this latter difference is orthogonal to the distinction which the ‘doctrine of double effect’ claims is of moral signiﬁcance. A person’s non-action can, like her action, result from her intending to behave the way she does in order that some result may come about. Similarly, either an action or an omission can be performed in the face of the belief that that ‘performance’ will lead to some collateral consequences. Talk of ‘positive willing’ thus should be replaced by talk of ‘intending’ or ‘aiming’. However, although the term ‘permitting’ can be equally misleading, there is also something felicitous about it. If we keep in mind that talk of ‘permitting’ refers to a particular kind of attitude and not to a speciﬁcally non-active way of realizing an attitude, then it is certainly more appropriate than another characterization one frequently ﬁnds here, namely that of “merely foreseeing”. Contrasting the actionguiding concept of intention with the merely epistemic notion of foreseeing is an unhelpful move because it appears to situate the objects of the latter attitude outside ‘the action itself’. However, things are attitudinally more complicated. A minimally rational agent who intends to A in order to bring about p, whilst recognizing that doing so will (probably) also bring about q, cannot simply maintain a purely epistemic perspective towards her (basic) action’s having these consequences. If Gill knocks back a number of gins in order to get drunk, knowing that she is also risking an awful hangover, she must, in so far as she is minimally rational, be accepting that risk. Accepting a proposition is a species of opting for it, as is intending to bring it about. Both attitudes, one could say, are optative, being expressible by the locution “Let it be the case that p”. Of course, under other circumstances Gill would not opt for the splitting headache she knows she is risking. But under other circumstances she would not opt to get drunk either. Such counterfactuals only tell us something about counterfactual opting; they do not change what the agent has opted for in the circumstances given. Agents such as Gill, like the agents in examples

96

Intentionality, Deliberation and Autonomy

(i) to (iii), opt strictly for what Gilbert Harman has aptly termed a “total package” (Harman 1986, 98; cf. Bratman 1987, 143), that is for a conjunctive state of affairs (p and q). One of the conjoins is aimed at, the other only accepted. The ‘doctrine of double effect’ tells us that in spite of the fact that both propositions are conjunctively opted for, the fact that one is only accepted, whereas the other is aimed at or intended, makes a signiﬁcant normative difference. An adequate discussion of the doctrine thus needs to take account of two attitudinal facts: one, that both sets of consequences are opted for in the circumstances and, two, that we nevertheless distinguish between the proposition aimed at and the proposition merely accepted. In order to work out whether a signiﬁcant normative difference can really be built on this attitudinal distinction, we need an analysis that clariﬁes what it is about intending or aiming that distinguishes it from acceptance. This task is taken up in section 2. 1.3 On (3): “means” If the ﬁrst two conditions are given the suggested interpretations, then they provide a simple and coherent core doctrine, according to which there is some kind of signiﬁcant normative difference between, on the one hand, bringing about some state of affairs because one aims to bring it about and, on the other hand, bringing it about because one accepts it in the light of its consequential relation to a state of affairs one aims to bring about. I shall call a slightly speciﬁed version of this core claim the principle of collateral consequences (PCC). In the light of this characterization, it is not immediately clear what role the third condition is supposed to play. According to the most natural reading, it would appear to pick out a particularly salient case of the second condition: if (2) excludes the bad action being intended, (3) may simply seem to specify that it may not be intended as a means. We can say that performing some action A as a means to some other action B involves A–ing because one intends to A, where the intention to A results from one’s belief that A–ing is an antecedent causal condition of one’s B–ing. A great number of our intentions come into being because we believe that, as a result of the causal structure of our environment, we can only achieve something else we intend by ﬁrst A–ing. However, the reasons I gave for talking of ‘consequences’ rather than of ‘effects’ ought to make it clear that the causal relation cannot be what is normatively decisive. If breaking a promise is forbidden, but one can only help someone in need by deliberately doing something that counts as breaking a promise, then surely an advocate of the doctrine ought to ﬁnd the behaviour of the promise breaker morally problematic in the same way that she takes causing physical harm as a means to helping to be morally problematic. Taking the bringing about of some morally problematic state of affairs to be a means or a way of bringing about a morally good or indifferent state of affairs ought, if the PCC is indeed the core of the doctrine, both involve the same kind of norm contravention. Taken literally, however, condition (3) is not just a speciﬁcation of a particular kind of reason for intending to bring about some bad state of affairs. As it stands, the condition does not stipulate that the bad ‘consequence’ not be taken as a means, that is, intended because it is believed to be an antecedent causal condition of the agent’s

The Double Failure of ‘Double Effect’

97

aim. What it stipulates is that the bad effect not be a means. In other words, condition (3) is a purely causal matter, that is, a condition on the way the relevant events in the world relate to each other, not a requirement on the conception of those events on the part of the agent.5 Thus understood, a ‘means’ is an antecedent causal condition, irrespective of whether anyone knows this or not. Taken together with condition (2), we would now have the requirement that the bringing about of some negatively evaluated state of affairs neither be intended, for instance taken to be a means, nor be an antecedent causal condition of a positively evaluated end. Note that as a purely extensional condition it would apply even if the negatively evaluated state of affairs should, contrary to what the agent thought, turn out to be an antecedent causal condition of the positively evaluated state of affairs. Think of the bombing example: in bombing the munitions depot, the collateral bomber accepts the deaths of the children because he sees the destruction of the arms depot as a means to bring the war to a speedy end. Imagine, however, that the enemy ends up surrendering, but only because of the deaths of the children, in spite of the fact that this was not the collateral bomber’s plan. Extensionally understood, the deaths of the children turn out to be a means to the end of shortening the war. According to condition (3), that makes dropping the bomb impermissible, whatever happened to be going through the pilot’s mind. Notice that this construction has the bizarre conclusion that the good outcome, the end of the war, makes the action impermissible, whereas the bomb dropping and the children’s consequent deaths would have been permissible if the entire action had turned out to be completely ineffectual. As an extensional reading of condition (3) would render it an absurd requirement,6 it is therefore most reasonable to read it intensionally. But, as that is to see it as merely picking out the most salient type of case already covered by (2), the condition can simply be dropped. 1.4 Abstracting from absolutism There is one ﬁnal feature of the ‘doctrine of double effect’ which should be mentioned and which should be recognized as extraneous to the core claim of the PCC. It is not explicitly stated in the four conditions, but presupposed in the claim that their satisfaction is necessary in order that certain actions may be permitted. The presupposition is that morality includes absolute prohibitions, that is, that it forbids certain actions, whatever competing reasons may be adduced for performing the action in the circumstances. Expressed in the language of means and ends: “there are no ends which justify every means, and … there are some means which no end will justify” (Kenny 1995, 87). 5 On the ambiguity between extensional and intensional readings of the third condition, see Davis 1984, 114–16. 6 The article in the New Catholic Encyclopedia, on which I have based my own wording, suggests primarily an extensional interpretation. However, the formulation of Joannes P. Gury in his Compendium Theologiae Moralis (1850/1874), which, according to Mangan, provides the canonical version of all four conditions, is clearly meant to be intensional. Gury’s explanation of the requirement “The evil effect must not be the means to the good effect” is that, if it were, “then the good [would be] sought by willing the evil” (my emphasis). See Mangan 1949, 59–61.

98

Intentionality, Deliberation and Autonomy

The function that the doctrine is supposed to fulﬁl can be made particularly clear by examining its role in absolutism. The absolutist presupposition is that certain actions, pre-eminent among which is killing, can under no circumstances ever be justiﬁed. However, as there can frequently be situations in which these absolute prohibitions would lead to undesirable normative paralysis, a principle that makes certain of their contraventions non-reproachable, in spite of not being justiﬁed, can save the absolutist from the predicament of having frequently to condemn agents whatever they do.7 Although this is one explanation of the attraction of the doctrine, it clearly cannot constitute its rationale. For one thing, it makes the doctrine’s plausibility depend on the validity of absolutism. Most important, the doctrine’s capacity to loosen absolutism’s hold depends itself on the independent plausibility of the PCC, that is, on the claim that the distinction between intending and merely accepting makes a decisive normative difference where we are evaluating actions. But if this attitudinal distinction really is able to fulﬁl this function in the case of absolute prohibitions, then it should result in the same sort of normative relaxation relative to nonabsolute prohibitions, that is, with respect to prohibitions that can also, under other circumstances, be outweighed by stronger justifying reasons. It should thus make itself felt relative to norms that have no more than pro tanto character, for instance, the norm not to cause pain. Recent secular advocates of the doctrine, prominent among whom is Thomas Nagel, argue that precisely this is the case (cf. Nagel 1986, 176). An example pairing that illustrates the point can be provided by a couple of dentists, one of whom sees the pain he is inﬂicting as a side effect of his drilling, whereas his twin inﬂicts the same amount of pain in the course of the same operation on purpose. It seems clear that the attitudinal difference between the two dentists gives rise to exactly the same intuitions about the normative difference between the two cases as do the attitudinal differences in examples (i) and (iii). For this reason, we not only can but should detach the core of the doctrine, that is, the PCC, from absolutism. The PCC’s plausibility does not depend on absolutism; rather, any plausibility absolutism is thought to have may very well depend on the validity of the PCC, without which the lack of a mechanism for restricting the dilemmas to which absolutism gives rise is likely to appear intolerable.8 1.5 Two questions This preliminary analysis shows that the ‘doctrine of double effect’ depends on the core claim I have labelled the ‘principle of collateral consequences’. According to the PCC, accepting rather than aiming at some negatively evaluated p one is bringing about brings with it some kind of normative relaxation of the prohibition on the action of bringing about p. If this is correct, then an investigation of the 7 Cf. Anscombe 1982, 19–22; Mackie 1977, 161f. The reason for the particular concentration on killing in the Catholic tradition is that the causation of death can result from deeds with other aims, whereas the contravention of other absolute norms – Anscombe names the prohibition of “sodomy” (!) – are hardly possible without corresponding intentions. 8 The other candidate for a dilemma-restricting mechanism is the doctrine of a normative distinction between doing and allowing.

The Double Failure of ‘Double Effect’

99

doctrine has two questions to answer. The ﬁrst concerns the precise character of the descriptive difference between the two attitudes that is claimed to justify the normative relaxation. Only if we are clear what it is about intending that goes beyond accepting can we come to understand the idea that its absence might explain the peculiar normative relaxation of a prohibition that is supposed to result from that absence. The second question concerns the normative relaxation itself. How precisely are we to understand the idea of the normative relaxation of a prohibition where that prohibition is not cancelled by stronger counter-reasons? I shall argue that distinct intuitions feed into this idea. These, it seems, need to be taken apart and discussed separately in order for us to be able to appreciate their force. Note that there is nothing in the ‘doctrine of double effect’ that explains why its core principle is supposed to be valid. Anyone who accepts the principle ought to do so because she accepts its rationale or, at the very least, accepts an explanation of why no rationale of the principle can be provided. Let us begin with the ﬁrst question. 2. Intending, aiming and accepting How, then, is the attitudinal distinction that grounds the idea of collateral consequences to be analysed? 2.1 Commitment and irreducibility The ﬁrst step towards an answer consists in clarifying what it is that distinguishes intending from other optative attitudes, that is, wants in the general sense of proattitudes explicable by locutions of the form “Let it be the case that p”. In the literature on intending, the central feature invoked here is that of commitment to bringing about the attitude’s content (Harman 1986, 94f.; Bratman 1987, 4f., 16–20, 107–10; Mele 1992, 158–62). Commitment to some action goes beyond merely wanting to perform it, even where one believes oneself capable of performing it. The majority opinion on this component at the moment is that it is in some sense irreducible to beliefs and wants, even where the latter are supplemented by motivational speciﬁcations. Commitment to perform an action is something other than being most strongly motivated to do it, with or without supplementary doxastic conditions (Brand 1984, 123–7; Harman 1986, 78–95; Bratman 1987, 10, 100, 121; Mele 1992, 154–70; 2003, 28). Now, an irreducibility claim is a claim that there is, for some reason, a point beyond which analysis is impossible. Were this to be the case with the commitment component of intention, then it might appear that the descriptive mystery facing us here grounds the normative mystery that is the unjustiﬁed normative relaxation of a prohibition. Commitment, so it might be thought, is just something special and when it is absent from the causation of certain consequences, then we should relax those normative strictures that only come into play where the commitment component essential to intention is at work. Adapting a move central to Michael Bratman’s normative functionalist theory of intention, someone might argue that one of the functional roles deﬁnitive of commitment is precisely its requirement if moral norms relative to an agent’s action are to be strictly applicable. According to Bratman, commitment is partly deﬁned by

100

Intentionality, Deliberation and Autonomy

its normative role, although in his conception the relevant role is one within norms of prudential rationality, not morality: someone who intends to do something ought to deliberate so as to form subordinate intentions, refrain from forming conﬂicting intentions and endeavour to bring about the intention’s content (Bratman 1987, 16– 18; 140–41). What makes Bratman’s theory irreducibilist is the idea that intention, in particular its commitment component, is in part deﬁned as whatever attitudinal feature it is that makes it rational for its bearer to adhere to such practical norms. This strategy leaves open a question that one would certainly like a theory of intention to answer: namely, in virtue of what does being ‘committed’ to some action make it rational to adhere to those norms? Normative functionalism effectively tells us that this question cannot be answered. An adherent of the PCC might very well respond in a parallel manner to the question as to what it is about intending in virtue of which norms contravened in its absence suffer a kind of normative debilitation: the question just cannot be answered. Nevertheless, part of intending’s essence, so it might seem, is visible in the normative role it plays in morality, as expressed in the PCC. What is important here is not that an advocate of the doctrine might see the role of intention in the PCC as part-deﬁnitive of intention, but the fact that an irreducibilist conception of intending confers plausibility on the principle. Our inability to provide an analysis of intention in descriptive terms would explain why we are unable to say with any clarity what it is about intention that explains its particular normative role in morality. For this reason a discussion of the PCC requires a discussion of the irreducibility thesis. The obvious way to challenge an irreducibility thesis is to offer an analysis of the concept in question. The general line I think an analysis of intention should take involves focusing criterially not on the characteristic normative or motivational results, but on the genetic conditions of its instantiation. Schematically put: we should be looking not downstream, but upstream.9 Brieﬂy, the optative attitudes we primarily pick out by means of the term ‘intention’ are, quite simply, the products of decisions. Alongside these primary intentions there is a second set of wants picked out by the term, namely spontaneously generated wants that are qualiﬁed by sufﬁcient motivational strength for them to directly control unreﬂective action. The idea of commitment that gives rise to characteristic motivational consequences and grounds norms of prudential rationality is, according to such an analysis, bipartite.10 In our context, however, for reasons to be mentioned in a moment, we can focus entirely on the ﬁrst, paradigmatic sort of intention. If this is correct, then the criteria we require here are going to be criteria for the concept of decision. Decisions are optative stands that bring to a close episodes of what can be called ‘minimal practical deliberation’. Minimal practical deliberation is the mental process set in motion by the desire of the agent to resolve an optative uncertainty as to what to do. Deliberation that is more than minimal, as it usually is, involves the weighing of pro- and contra-considerations with the same end in view, but the resolution of optative uncertainty that is deﬁnitive of deciding is not 9 This symmetry in the accounts thus characterized was pointed out by Michael Bratman. The account that I offer is argued for in detail in Roughley (forthcoming). 10 In this point I agree with David Velleman (cf. Velleman 1989, 112).

The Double Failure of ‘Double Effect’

101

tied to the prior weighing of reasons. Sometimes we ‘just decide’. But whatever it is that leads an agent to take that particular stand, her paradigmatic intentions are those optative attitudes distinguished by having terminated optative uncertainty and the (minimal) practical deliberation she initiated in order to do so. According to this perspective, it is our having decided to A that tends to motivate us to make further attitudinal moves to ensure that we A and which makes it rational to do so. 2.2 Opting for a package Returning to the speciﬁc kinds of examples that are relevant for the PCC, it ought to be clear why only decisional intentions are relevant: opting for a conjunctive package, the conjoins of which have been weighed against each other, is obviously a postdeliberative matter. For this reason we can ignore secondary, spontaneously formed intentions. However, the fact that, in the cases we are concerned with, the agent decides to realize a complex proposition only one conjoin of which is aimed at would appear to throw doubts on the genetic analysis: if deciding to bring about p is deﬁnitive of intending to bring about p, then that seems to make an agent in such a case intend to bring about both consequences of his action. This would appear to deny the descriptive premise on which the PCC is built, a premise I have claimed is enshrined in everyday psychology. If the genetic analysis really does force us to this conclusion, then we should surely reject it. Bratman rejects such an analysis because of what he sees as the falsity of the “choice-intention principle”, according to which choosing to do A and B necessarily involves intending to A and to B (Bratman 1987, 145). The principle, Bratman argues, is false because choosing a package does not commit one to forming further intentions conducive to the realization of both conjoins and to endeavouring to realize both. Rather, these characteristic downstream roles of intending only need come into play with respect to one of the conjoins. If the other is only accepted, then nothing of the sort need be true of it (ibid., 154f.). The upstreamist reply to this is to insist on a clear distinction between the agent’s attitude to the package and his attitude to its components. It is true that choosing or deciding to (A and B) does not necessarily involve intending to A and intending to B. What it does necessarily involve, the upstreamist claims, is intending to (A and B). This is because we do not have to intend singularly to bring about every consequence we take it we will be bringing about when we realize a package we intend to realize. Bratman’s arguments against this claim beg the question. The collateral bomber, he argues, does not intend to kill the children even as part of a conjunctive intention because he is not disposed to reason about how to kill them or to endeavour to kill them should doing so be separable from the destruction of the munitions depot (ibid., 148).11 But were such separability to be possible, then the bomber could drop the conjunctive intention. He only intends the package because of what he takes to be the inseparability of the conjoins.12 Given that they cannot be taken apart, he intends both of them together. 11 This is his argument against what he calls “the principle of intention division”. 12 Cf. my remarks about counterfactual opting under 1.2.

102

Intentionality, Deliberation and Autonomy

2.3 Aiming This defence of the upstream analysis has only so far shown why downstreamist arguments do not falsify the claim that paradigmatic intentions concern what the agent has decided to do. It has done nothing to solve the problem we set out to solve, namely how to distinguish between what is individually intended and what is individually accepted within the content of a complex intention. How can the upstreamist make sense of this distinction? The answer involves saying something more about the agent’s view of the relationship between the proposition aimed at and the overall package. The point is simply this: an agent intends to bring about that singular state of affairs as part of a conjunctively intended package the prospect of which she takes to provide the reason for her decision to go for the whole package.13 Where she opts for (p and q), she intends or aims to bring about p and not q if she sees her overall stand as justiﬁed in the light of the prospect of bringing about p. It is important to note that this relationship between aiming and accepting within an intended package is by no means exclusive to intention. It is, on the contrary, a perfectly general phenomenon in the optative sphere and is, for instance, to be found in cases of wanting where no action on the part of the want’s bearer is at issue. Take the case of Polly, who is observing with interest the political developments in some far-off country. She considers the possible outcomes of approaching elections there. In doing so she acquires the want that the x-party win the election because she believes their economic policy is the one most likely to put an end to the country’s poverty. However, this want is not formed wholeheartedly, because Polly also believes that the x-party will implement certain forms of unpalatable social repression. Nevertheless, compared with the alternatives on offer, she certainly wants the x-party to win. This is an optative package deal scenario, in which the relevant attitudes are non-action wants. We would normally describe Polly’s optative state as one in which she hopes that the x-party wins. In doing so she accepts that this outcome will bring with it certain forms of repression to which she is averse. The decisive point of comparison is that Polly’s acceptance of the unpalatable consequences does not lead us to ascribe to her the hope that they take place. Note further that the optative terms of which this is true can be extended. If Polly happens to believe that a victory of the x-party is a certainty, then we might say that she is looking forward to their winning the election. But we would not say that she is looking forward to them carrying out the repressive measures she takes to be unavoidable. Note ﬁnally that we can also reverse the polarities: take someone who is, on balance, afraid of the package deal being offered by his rapidly approaching old age. Although he can also see certain beneﬁts that it may bring, he is not for that reason afraid of those prospective beneﬁts. Where someone desires singularly the consequences of the prospect that furnishes the reason for a package deal she opts for, we can say that she welcomes them. Welcoming is the positive variant of acceptance.

13 Cf. Scanlon’s proposal: “one’s intention … is an aspect of one’s action that is crucial to the reason one takes oneself to have to do it.” See Scanlon 2000, 306f.

The Double Failure of ‘Double Effect’

103

In all these cases, some proposition is the content of what one might call a ‘focal optative stand’: a singular want the content of which furnishes the reason for the overall optative stand on the package deal. In the case of effective intentions, it will also be the reason for which the agent acts. In the case of such intentions, we can say both metaphorically and literally, it furnishes the decisive reason: it is the consideration that makes the difference and the consideration on the basis of which the agent takes his decision. The relationship between aiming and accepting, then, involves a taking-as-justifying on the part of the agent. This will turn out to be important as we turn to the way in which the normative relaxation speciﬁed by the PCC is supposed to work. 3. Two versions of the principle of collateral consequences In order to clarify the form of normative relaxation that might plausibly supervene on the descriptive distinction, there is a distinction between types of normative principles that need to be made explicit. This is necessary because of an ambiguity in the intuitions that are adduced in the literature to support the doctrine. The ambiguity can, for instance, be seen in two claims made by Thomas Nagel within a short space. Although the two points are clearly distinct, Nagel obviously does not think it necessary to distinguish them, let alone to argue for them separately. The ﬁrst is this: “Intention”, he asserts, “appears to magnify the importance of evil aims by comparison with evil side effects”; intended evil is “lit up” by the “intensifying beam” of intention (Nagel 1986, 180f.). The claim behind these metaphors seems to be that the bringing about of a negatively evaluated state of affairs as a result of an intention constitutes a worse offence than doing so as a result of a mere acceptance. In other words, this is the de-absolutized variant of the original Catholic doctrine, according to which the latter may well be permitted where the former is forbidden. Now compare Nagel’s second claim: according to this, “we consider ourselves far more responsible for what we do (or permit) intentionally than for consequences of action that we foresee and decide to accept but that do not form part of our aims” (ibid., 180, my emphasis). This is a different matter and can be rephrased in the following way: where it is forbidden to bring about a certain state of affairs, the fact that someone brings it about as the result of an acceptance, rather than an intention, makes doing so in an important sense less fully the agent’s action and thus exculpates him, at least to some degree, from having contravened the relevant norm.14 Here, as in many other discussions of the doctrine, it appears that intuitions as to the principles of assignment of responsibility for the contravention of norms contribute to the plausibility of a claim about the contents of the norms contravened. But these are two different claims and we can only attain clarity on the strength of reasons that thus cluster around the cases in question if we take them apart.

14 Timothy Chappell sees this latter claim as the core of the doctrine of double effect (Chappell 2002, 223–5). David Chan also sees (his version of) the doctrine as a principle for the evaluation of the blameworthiness of agents (Chan 2000, 405–34).

104

Intentionality, Deliberation and Autonomy

The ambiguity at work here derives from the failure to distinguish between two different kinds of principles at work in systems of moral evaluation. The ﬁrst kinds are moral norms in the narrow sense of the word, norms whose primary function is to guide the prospective action of agents. Where these norms are employed in their secondary function, that is as standards for the retrospective evaluation of actions, they are complemented by principles of a second sort, principles that regulate the assignment of responsibility for the contravention of the norms. The fact that a person has violated some norm does not in itself decide the question as to how strongly that violation is to be attributed to her. That depends on whether there are further factors to be taken into account that should count as excusing her for what she has done. Considerations of both sorts combine to justify judgements as to the level of culpability of the agent: how guilty she is depends both on how important the norm is that she has violated and how strongly the violation is to be attributed to her. The two points are, however, clearly distinguishable and are distinguished by modern legal systems (Fletcher 2000, 454–9, 576–9). In our context, the important point is that the moral signiﬁcance of the intention/ acceptance distinction could be expressed in a principle of either sort, that is, its function could be thought to be either normative in the narrow sense or responsibilitytheoretic. As a result we can give the PCC two different formulations: PCC(N) Where an agent A brings about a state of affairs p the bringing about of which is morally problematic, p’s being merely accepted, rather than aimed at, necessarily counts in favour of the permissibility of A’s action token. PCC(R) Where an agent A brings about a state of affairs p the bringing about of which is morally prohibited, p’s being merely accepted, rather than aimed at, necessarily counts as reducing A’s responsibility for having contravened the relevant norm. We should, I shall be arguing, reject both variants. Dealing with them separately will, on the one hand, prevent independent intuitions from interfering with each other. That will make it easier to see that there are no arguments that give adequate support to either version of the principle. On the other hand, it will also enable us to see that there are indeed cogent arguments that support claims which are related to the two versions of the PCC in one way or another. There are, it turns out, important normative and responsibility-theoretic principles at work in the background. However, these neither derive from the intention/acceptance distinction nor justify judgements that are coextensive with those to which either version of the PCC would commit us.

The Double Failure of ‘Double Effect’

105

4. Acceptance and responsibility Let us begin with the claim that the distinction between intention and acceptance grounds a difference in the assignment of responsibility. 4.1 Responsibility and criminal homicide The role that the PCC(R) ascribes to the distinction can be characterized effectively by means of a comparison with legal practices. The US Model Penal Code, for instance, distinguishes what it calls four “kinds of culpability”: “purpose”, “knowledge”, “recklessness” and “negligence”. The Criminal Code for England and Wales establishes similar distinctions among “degrees of fault”, preferring to talk of “intention”, rather than “purpose” (American Law Institute 2001, sect. 2.02(2), 94f.; Law Commission 1989, sect. 18, 51f.). The most important threshold in both systems is the one that separates recklessness on the one hand from ‘knowledge’ and ‘intention’ or ‘purpose’ on the other. An agent is taken to be less responsible for breaking a law as a result of taking an unacceptable risk of producing some result than for either producing it purposely or for acting in the knowledge that what she is doing will have such a result. In Anglo-American law, the chief difference between murder and manslaughter is drawn by this means (American Law Institute 2001, sects 210.2–3, 305–71; Law Commission 1989, sects 54–5, 66–7). The PCC(R) is best understood as proposing that the difference between acceptance – which covers ‘knowledge’ – on the one hand and ‘purpose’ or ‘intention’ on the other should be seen as marking a responsibility threshold with a similar status. It may appear at ﬁrst glance that there are features of our legal systems that support this view. The deﬁnition of murder provided by the British Law Commission does in fact require an ‘intention’ either to cause death or to cause serious personal harm (Law Commission 1989, sects 54, 66–7). Certain philosophers, notably Anthony Kenny and John Finnis, have argued that the content of this requirement should be understood as the agent’s aiming at occurrences of either of those kinds, in the everyday sense of ‘aiming’ (Kenny 1977, 172f.; Finnis 1991, 49). However, that has not been the view of the British Law Lords. On the contrary, comments made by the Lords in various cases demonstrate that the Criminal Code’s distinction between ‘intention’ and ‘knowledge’ is not disjunctive, indeed that the notion of ‘intention’ employed in the deﬁnition of murder clearly extends beyond the everyday concept. The precise understanding of the ‘mental element’ necessary for murder has been the subject of considerable debate among judges and legal commentators. As has been repeatedly remarked (Kenny 1977, 161–74; Smith and Hogan 1978, 47–52, 285–91; Williams 1983, 249–54; Duff 1990, 1–37; Fletcher 2000, 269–74), the case of Mrs Hyam (1974) motivated the Lords to engage in extensive exercises of conceptual analysis. The defendant had poured petrol through a letter box and set light to it with the purpose of frightening her rival, but with the effect of killing the latter’s daughters. There was signiﬁcant disagreement amongst the Lords as to how the relevant notion of purpose is precisely to be understood and as to whether the mental state thus identiﬁed is indeed required for murder. However, there was no dissent in the question as to whether the conscious taking of a certain degree of risk

106

Intentionality, Deliberation and Autonomy

as to causing death could be sufﬁcient for murder, even if death had not been aimed at. Two of the Lords made this point in terms of the everyday notion of intention, adding that either such an everyday intention or a speciﬁc degree of subjective probability is required. Others insisted on the necessity of an ‘intention’ in a technical sense that incorporates the additional doxastic criteria into the concept.15 Either way, the consensus in both British and American law is that murder requires either an everyday intention or the acceptance of consequences believed to have a certain probability. This requirement can be terminologically clothed either in line with this everyday distinction, as the US Model Penal Code does (American Law Institute 2001, sect. 210.2(1), 305), or can be speciﬁed in terms of a misleading technical notion of intention, as incorporated into the Criminal Code for England and Wales. Here an ‘intention’ is said to be in play where the agent “acts either in order to bring about [a result] or being aware that it will occur in the ordinary course of events” (Law Commission 1989, sect. 18(b)(ii), 51). The important point for our purposes is that, where the law does see an ‘intention’ as necessary for murder, the operative concept is an expanded variant of the everyday notion. To avoid misunderstanding, we can refer to the technical notion of intention in use here as a ‘British legal intention’ or a blintention.16 Note that although in these cases British law builds the requirement of a blintention into the description of the offence – using it, for instance, to establish the primary distinction between murder and manslaughter – that does not entail that the object of the operative prohibition is killing-as-a-result-of-a-blintention, a norm that might be thought to exist alongside another, independent norm prohibiting killing-as-aresult-of-recklessness. Rather, the ‘offences’ of murder and manslaughter are both contraventions of the same norm, but are distinguished according to the extent to which the agent is responsible for those contraventions. In such cases it is the degree of attributability that is seen as justifying differences in sentencing – a difference that can be minimal in England where homicide is concerned, but is considerable for the distinct offences of ‘intended serious personal harm’ (life imprisonment) and ‘reckless serious personal harm’ (ﬁve years) (Law Commission 1989, 123). These latter two offences are again contraventions of the same norm (causing serious personal harm), but are distinguished by the extent to which they are attributed to their agents. Such offences are importantly different in structure from prohibitions whose contents are the bringing about of a certain level of probability of some undesirable event, as in the offence of ‘reckless driving’. The superﬁcial appearance that the concept of recklessness has the same function in the two offences ‘reckless serious personal harm’ and ‘reckless driving’ is illusory. In the former its fulﬁlment indicates that the 15 Even Lord Hailsham, whose ruling Kenny sees as pointing the way to a deﬁnition of murder that requires aiming on the part of the defendant, actually stipulated that what are taken to be “inseparable consequences of the end” are also ‘intended’. See Smith and Hogan 1978, 50. 16 Whether commentators approve or disapprove, there is little dispute that we are dealing here with a “term of art”. See Fletcher 2000, 443; Williams 1983, 250; Finnis 1991, 50. Blintentions are closely related to German “Vorsaetze”, which come in three grades, the lowest, ‘dolus eventualis’ involving being ‘reconciled’ to producing the relevant harm. See Lackner and Kuhl 1999, sects 15, 108–24; Fletcher 2000, 446–8.

The Double Failure of ‘Double Effect’

107

agent was less responsible for contravening the prohibition than he would have been had he done so as a result of a blintention. In the second case recklessness is part of the content of the prohibition itself: the norm forbids a certain level of risk-taking whilst driving.17 Where, as in such cases, recklessness is a genuine part of the norm’s content, no one thinks that the activity whose reckless performance is forbidden (driving) is itself morally problematic and that the agent would have committed a worse crime if he had performed it as a result of a blintention. 4.2 Control and subjective probabilities Unlike what ﬁrst appearances may suggest, then, our legal practices do not work with anything comparable to a responsibility-theoretic distinction between intention and acceptance. Rather, the decisive threshold remains that drawn by the orthogonal distinction between blintention and recklessness. The distinction is orthogonal because it is primarily a matter of the subjective probabilities opted for.18 Of course, the fact that this is the way the US and British legal systems assign responsibility for homicide is no proof that this is the reasonable way to view things. However, we do not have to look far to see why it is. The decisive point is the contribution that an increase in subjective probabilities, unlike the move from acceptance to intention, makes to the factor that grounds responsibility, namely the control exercised by the agent. An agent’s control over the production of states of affairs is, as Aristotle argued, dependent on the two criteria of knowledge and lack of compulsion (Aristotle, NE 1109b35–1110a1; cf. Hart 1968, 121f.). On the one hand, law or morality can see responsibility as diminished where the agent is subject to duress, to pathologically compulsive inner states or where he is provoked so severely that any reasonable person would be expected to lose control over himself. On the other hand, an agent’s (nonculpable) lack of knowledge of the results he is bringing about exculpates him from any harm he causes as a result of that ignorance. Moreover, it is equally reasonable to see the gradability of doxastic states as grounding degrees of culpability. Now, there is clearly a highly stipulative element to decisions as to where exactly thresholds are to be set and, indeed, as to how many such thresholds are appropriate. The important point here is that it is some contrast of this kind that we need, because it is the kind that concerns our capacities for control. The greater the probability we ascribe to some result of a ‘basic action’ of ours, the more we see that result as dependent on us and correspondingly the more we take on responsibility for bringing it about. This only changes once our control over the performance of the ‘basic action’ is diminished as a result of one or another form of compulsion. In contrast, the fact that the prospect of some state of affairs is not the reason for the agent deciding to bring 17 On the relationship between norms requiring the bringing about of bare states of affairs and norms whose contents are the creations of probabilities of states of affairs, see Seebass 1994, 381–97. 18 This is the primary, although not the entire basis of the distinction because liability is also seen as dependent on the social acceptability of the risks taken. Cf. Fletcher 2000, 259–62.

108

Intentionality, Deliberation and Autonomy

it about – in spite of his having it in his optative purview – is quite simply irrelevant for the question of control. Although the legal distinction between blintention and recklessness is orthogonal to the everyday distinction between intention and acceptance, there is, nevertheless, a simple connection between the two that helps to explain the responsibilitytheoretic intuitions expressed in the PCC(R). This is that everyday intentions tend to be associated with a fairly high level of subjective probability. There is no necessity to this: if a person ﬁres a machine gun at a moving car in order to bring it to a halt, he may be well aware that doing so makes the death of one of the passengers highly likely. Nevertheless, that does not entail that he intends to kill anyone. However, there is clearly a strong tendency for aims and high subjective probabilities to correlate. There are two reasons why this is so. First, people only tend to aim at things they think there is a fair likelihood of them being able to attain, whereas there is no characteristic level of subjective probability a person assigns to a proposition he merely accepts as a possible consequence of his action. Second, there are structural reasons why those cases are going to be in the minority in which the proposition accepted is thought to have a higher probability than that of the proposition aimed at. We can see this if we distinguish two kinds of example, which can be labelled ‘branching cases’ and ‘knock-on cases’. In branching cases an agent brings about p because doing so tends to bring about consequence q and in spite of the fact that bringing about p tends, for entirely independent reasons, also to bring about r. In knock-on cases the agent brings about p because of p’s tendency to lead to q and in spite of q’s tendency to lead to r. In branching cases it is an open question whether q or r is more probable. In knock-on cases it is not: because the instantiation of the proposition accepted is structurally dependent on the instantiation of the proposition aimed at, the agent’s bringing about of the latter cannot, for obvious reasons, be more probable than his bringing about of the former. Put concisely: although there are no structural determinants of the relative probabilities of an end and the consequences of ways or means of achieving that end, the consequences of an end are necessarily no more probable than is the achievement of that end. For these reasons, we may frequently be justiﬁed in seeing the agent as less culpable for bringing about some merely accepted consequences of what he is aiming at than for achieving his aim. However, any lessening in culpability here depends entirely on the agent’s subjective probabilities, not on any intrinsic features of the two optative attitudes themselves. This is only brought out by considering cases in which the agent assigns exactly the same likelihood to both results, a condition that is most obviously fulﬁlled where the agent assigns the coming into being of r, if q comes about, a probability of one. For instance, only when the deaths of the schoolchildren are taken to be strictly unavoidable if the munitions depot is bombed does the contrast between the two bombers get to the attitudinal heart of the matter. As long as the example allows some probability of the children escaping death when the bomb is dropped, we have a difference in subjective probabilities that can mobilize justiﬁable responsibility-theoretic intuitions. Subtract these, however, and the case for the PCC(R) dissolves.

The Double Failure of ‘Double Effect’

109

5. Acceptance and justiﬁcation In spite of the fact that the ‘doctrine of double effect’ is ofﬁcially formulated in terms of permissibility, many of the intuitions that give it force are best understood in terms of the PCC(R). However, once the responsibility-theoretic variant of the principle has been rejected, we are left with a claim that really just is about permissibility. In order to evaluate this claim, shorn of considerations of attribution, we need to return to the structure characterized in 1.4: the idea of a prohibition’s normative relaxation independent of the justiﬁcation of that relaxation by other moral norms that override the prohibition. If this is indeed the structure for which we are seeking an explanation, the question we are faced with is whether there is plausibly any room for such non-justiﬁcatory normative relaxation within an acceptable morality. 5.1 Normative relaxation and justiﬁcation This question appears particularly acute once absolutism is rejected. If it is not postulated that there are certain actions that could under no circumstances be justiﬁed, then it seems a reasonable hypothesis that the judgements that appear to be explained by the PCC(N) might be reconstructed in terms of the justiﬁed overriding of strong prohibitions in the light of other moral norms. A non-absolutist morality plausibly contains a strong prohibition on killing. Nevertheless, an agent is both morally and legally permitted to cause another person’s death where doing so is required for self-defence. Similarly, for any morality that is not strictly paciﬁst, certain forms of killing are permitted in war. And medical practitioners are also permitted under certain circumstances to cause their patient’s deaths. In a non-absolutist morality, the reasons for these permissions are naturally seen as grounding in norms that justify the causation of death under another description: as defending oneself, waging war under certain circumstances and relieving a patient’s pain. In order to make clear the lines of the debate, it is important to note that a view of this kind does not necessarily depend on ‘consequentialism’, if consequentialism is the doctrine that valid norms are justiﬁed in terms of the consequences of their being obeyed. The rejection of absolutism is perfectly compatible with the justiﬁcation of norms taking other forms, for instance, Kantian universalization. Whether, for instance, there are valid norms of fairness that are not justiﬁable in terms of their consequences is a question that is completely independent of whether such norms are necessarily to be accorded lexical priority. This point is obscured by Anscombe’s original characterization of consequentialism as encompassing any moral theory that rejects absolutism (Anscombe 1958, 34–6).19 Put in the easily misleading terms of ‘ends’: the absolutist defender of the PCC(N) claims that there are some actions that no ends can justify, whereas the PCC(N)’s non-absolutist advocate claims that there are some actions that, in the contexts under scrutiny, no ends justify. In both claims the relevant ‘ends’ encompass the realization of competing moral norms, however these are themselves justiﬁed. Both defenders of the PCC(N) claim that the actions in question can nevertheless be morally acceptable 19 It is also tends to be obscured by Kant’s own absolutism.

110

Intentionality, Deliberation and Autonomy

under speciﬁc attitudinal conditions. If this is correct, then the challenge that has to be met by the defender of the PCC(N) therefore does not concern the acceptability of non-consequentialist reasons for moral norms, but the acceptability of a relaxation of normative stringency in spite of the lack of an appropriate justiﬁcation by reference to further moral norms. But once one subtracts the idea that the relevant relaxation could be a consequence of the reduced attributability of the action’s prohibited result, the challenge, when characterized in this way, looks to be unmeetable. Why should we relax a prohibition if there is no justiﬁcation for doing so? Moreover, taking this claim together with our analysis of what it is to aim at one component of a package leads to absurd results. Remember that, in such a case, intending to bring about p rather than q is a matter of taking the prospect of bringing about the former proposition as the reason for one’s overall stand, that is, it is a matter of seeing the prospect of p as justifying opting for (p and q). But if the PCC(N) does indeed spell out a principle of non-justiﬁed relaxation of a prohibition, then the following must be true: seeing the prospect of p as justifying also opting to bring about q must relax the prohibition on bringing about q without justifying bringing about q. That is, the normative relaxation would have to result from the falsity of the agent’s perspective, according to which the prospect of p justiﬁes also opting to bring about q. This would make the normative relaxation speciﬁed by the PCC(N) dependent on the agent’s false belief in the existence of a justiﬁcatory relation. This would be a bizarre construction indeed. Of course, it might be thought that this result is an artefact of a tendentious characterization of the central idea of the PCC(N). Perhaps what it aims to establish is incorrectly characterized as moral acceptability minus justiﬁability. The PCC(N) is, after all, a normative principle and ought perhaps therefore to be understood as itself providing a kind of justiﬁcation, if a somewhat unusual one. Understood in this way it would stipulate that a person’s opting for a compound proposition because she takes the prospect of one of its components to justify opting thus, functions as a justiﬁcation for bringing about a second, otherwise forbidden component of her optative attitude’s content. There are two reasons why this cannot be right. Firstly, it is implausible that an action’s justiﬁability might be determined by the agent’s taking it to be justiﬁed. Of course, where someone contravenes a norm in good faith, believing wrongly that there are considerations which justify that contravention, we may well hold the person less responsible for doing so than if the offence is a result of purely egoistic motives. But this would be an application of a mitigating principle of attribution, not of a justiﬁcatory principle. Secondly, the doctrine thus understood would appear to be systematically dependent on moral agents adopting the very perspective it rejects. Where an agent takes the prospect of bringing about p to justify his opting for (p and q), in spite of the problematic character of bringing about q, he must have some reason for seeing things this way. Where we are dealing with examples that fall under conﬂicting moral norms, it is natural to assume that that reason refers to what the agent sees as the stronger justiﬁcatory power of a norm prescribing the bringing about of p than one condemning the bringing about of q. For instance, it appears that the reason why the collateral bomber pilot opts to bring about the children’s deaths as part of the

The Double Failure of ‘Double Effect’

111

package would have to be his belief that the prescription to end the war outweighs the prohibition on killing the children. But it is beliefs of this kind that the PCC(N) is supposed to be rejecting. In other words, if the PCC(N) is a correct justiﬁcatory principle, the justiﬁability of the agent’s resolution of the moral dilemma looks like it would have to be dependent on him being wrong about the justiﬁability issue. A theory that claimed this would not be self-contradictory. However, it would demand contradictory normative stances from moral agents and the moral theorist. Ironically, this is the very dissociation of practical and theoretical perspectives that ‘deontologists’ have frequently, and not implausibly, objected to in two-level utilitarian theories. There is one last possibility open to the advocate of a justiﬁcatory understanding of the PCC(N). This would be to treat the justiﬁcatory path recommended by the principle as self-referential. In this version, the agent opting for (p and q), in spite of the prohibition of bringing about q, would do so not because he takes some norm recommending p to outweigh a norm proscribing q. Instead, he would opt as he does because he takes his opting for that reason to justify his action, including its various unpalatable consequences. The action would be justiﬁed by the attitudinal stance according to which it is justiﬁed by that same attitudinal stance. This suggestion would avoid the inconsistency of the previous proposal. But in doing so, it would be all the more obviously subject to the objection that taking some behaviour to be justiﬁed does nothing to guarantee its justiﬁcation. A moral norm which prescribed intentions that are about themselves would be directing agents to take what they take to be justiﬁcatory, rather than properties of their actions, as justiﬁcatory. It is difﬁcult to see how what is called ‘justiﬁcation’ here could be recognized as a form of justiﬁcation at all. 5.2 Attitude evaluation Throughout this discussion I have assumed that moral norms prescribe, prohibit or permit actions, that is, intentional bringings about of (static or dynamic) states of affairs.20 The investigation of the ‘doctrine of double effect’ concerns the question of whether we have reason to accept permissions of actions – or a principle of diminished responsibility for norm contraventions – on the basis of a speciﬁc attitudinal constellation on the part of the action’s agent. The result of this investigation is that we do not. However, this does not entail that we have no reason to accept moral norms for the evaluation of agent’s attitudes. We certainly do so and there is a simple reason why we should: our understanding of a person’s attitudinal constitution, in particular of her motivational constitution, is the decisive factor in our estimation of how she is likely to act in morally relevant contexts.21 20 “Intentional” is here, of course, not equivalent to “intended”. 21 There is also a logical reason why motive evaluation is necessarily secondary relative to action evaluation: a motive is necessarily a motive to perform a speciﬁc type of action, that is, we need to specify an action type in order to individuate a motive in the ﬁrst place. A fortiori, we need to specify a morally problematic action type in order to individuate a morally problematic motive.

112

Intentionality, Deliberation and Autonomy

What is of particular signiﬁcance for our topic is the possibility of a gap opening up between our normative stance towards an agent’s action and our stance towards the attitudinal constellation responsible for that action. There are cases in which, although someone has adhered faultlessly to the moral norms relevant for her action, we may nevertheless, in spite of condoning her action, also criticize the motives behind it. Under such circumstances, the positive evaluation of what the person does is tempered by the negative evaluation of an attitudinal set which disposes her, under conditions close to the one under scrutiny, to morally unacceptable behaviour.22 In such cases the PCC(N) picks out a particular kind of attitudinal constellation for condemnation and conﬂates this attitudinal evaluation with the evaluation of the action it causes. It registers, for instance, that the attitudes of the ‘teleological’ bomber are at least prima facie more dubious than those of his ‘collateralist’ counterpart. It then moves illicitly from this judgement about the agents’ attitudes to a judgement about their respective actions. However, not only should the conﬂation of the two levels of evaluation be avoided. Differentiation among the attitudinal features of agents that may merit moral criticism reveals that the distinction between intention and acceptance has no special status on this level either.23 In order to see this it is important to compare cases in which the difference between intention and acceptance is not manifested in – however minimal – differences in the agent’s actions. Particularly where we are dealing with complex, temporally extended actions, such cases may be difﬁcult to describe plausibly. Someone who intends to bring about p will, as is emphasized by downstream conceptions of intention, typically (if not necessarily) devise ways of bringing about p and endeavour to do so. If this leads to ﬁne adjustments in the action of the teleological bomber that are not present in the case of his collateral counterpart, then we are judging different actions and we do not need the PCC(N) to tell us what is wrong. If, on the other hand, we restrict the objects of moral evaluation to agents’ attitudinal constellations, there is no intrinsic reason for seeing intentions to contravene norms as morally worse than acceptances of one’s doing so. Certainly, if someone has thought things through and opted to go for a morally problematic action, that tells more strongly against him than if he is merely the bearer of a corresponding desire, which he may be able to resist. Moreover, if he opts for a package deal, only reluctantly accepting certain morally prohibited consequences, then we will generally ﬁnd his attitudinal conﬁguration less morally unpalatable than if he takes the prospect of those consequences as the reason for his action. Nevertheless, we need to bear in mind that the acceptance of collateral consequences need by no means be reluctant. The fact that someone does not see the prospect of certain morally problematic consequences as the decisive reason for his action in some situation s1 is perfectly compatible with his being disposed to take their prospect as such a decisive reason in closely related situations s2, s3 … sn. Where an acceptance is a welcoming of collateral consequences (cf. above, pp. 102 f.), or derives from indifference to them, it may express a more dubious overall attitudinal constellation than an intention. The fact that the collateral 22 David McCarthy calls this a “mismatch” between ﬁrst- and second-order morality. See McCarthy 2002, 629, 634–9. 23 This point is made forcefully by McCarthy in the aforementioned article.

The Double Failure of ‘Double Effect’

113

bomber acts as he does because he happens to believe that destroying the munitions depot is the best military option does not exclude his readiness to intend to kill as many children as might turn out to be necessary. He may not care about the children in the slightest. In fact, there is nothing in scenario (iii) that excludes him being a sadist who acts for military reasons, whilst positively revelling in the destruction of innocent lives he also happens to be causing. His teleological counterpart, on the other hand, may only have decided to kill the children after agonizing about how to reconcile his belief that this is the only way to end the war with his moral aversion to doing such a deed.24 Thus it may be the case that there are far more circumstances under which the collateral bomber would be prepared to kill the children than there are circumstances under which this is true of his teleological counterpart. Because taking some consideration as the decisive reason for action in one situation is frequently an indication of the propensity to do so in similar situations, intentions are important objects of the moral evaluation of agents’ attitudinal sets. However, as indicators of such counterfactual tendencies, they are not necessarily superior to acceptances. Clearly then, although the PCC(N) draws on valid intuitions concerning the moral importance of agents’ attitudinal sets, it not only conﬂates the primary norms of action evaluation with the secondary norms for the evaluation of attitudes, it also misrepresents the latter. 6. The double failure The principle of collateral consequences, the attitudinally based core principle of the so-called doctrine of double effect, should, I conclude, be rejected: no intrinsic moral signiﬁcance attaches to the distinction between intending and accepting. The reconstruction of the intuitions that invest it with moral import suggests that they may be given form by one of two versions of the principle, where the ﬁrst is responsibility-theoretic and the second, closer to the ofﬁcial wording of the ‘doctrine’, normative in the narrow sense. In the light of the upstream analysis of the distinction between intending and accepting the conjoins of a compound intention, the PCC fails to convince in either variant. The intuitions behind the PCC(R) are not difﬁcult to articulate: where an agent takes the prospect of bringing about p to justify his opting to bring about (p and q), this tends to correlate with a lower subjective probability of his bringing about the collateral consequence q. This fact, not its dependence on the intention/acceptance distinction, may justify seeing some norm contravention on the part of the agent as less attributable to him. In contrast, when the PCC is construed in terms of permissibility, its status and rationale become considerably more difﬁcult to fathom. In an attempt to do so, two interpretations of the normative version of the principle were discussed. According to the ﬁrst, the principle stipulates the normative relaxation of a prohibition in spite of the lack of a decisive justiﬁcation of the action thus allowed; according to the

24 Indeed, acquiring the intention to kill the children is compatible with failing to carry it out as a result of weakness of will.

114

Intentionality, Deliberation and Autonomy

second, the PCC(N) is a normative principle that justiﬁes the action in the standard manner. The ﬁrst reading makes the relaxation of the prohibition dependent on a false justiﬁcatory perspective on the part of the agent. The second either makes the exception to the prohibition dependent on the validity of justiﬁcatory mechanisms the PCC(N) is supposed to supplant, or else makes the agent’s justiﬁcatory perspective self-referential. In the former variant the PCC(N) becomes redundant; in the latter it loses all appearance of being an account of justiﬁcation. The most plausible explanation of the PCC(N) seems to lie in a conﬂation of action evaluations and attitude evaluations. Even here, however, the distinction between intention and acceptance need not be morally decisive.25 References American Law Institute (2001), Model Penal Code, Uniform Laws Annotated, vol. 10A, West Group. Anscombe, G.E.M. (1958), ‘Modern Moral Philosophy’, in G.E.M. Anscombe, Collected Philosophical Papers, Minneapolis, MN: University of Minnesota Press, 1981, vol. 3, pp. 26–42. —— (1961), ‘War and Murder’, in G.E.M. Anscombe, Collected Philosophical Papers, Minneapolis, MN: University of Minnesota Press, 1981, vol. 3, pp. 51–61. —— (1982) ‘Medalist’s Address: Action, Intention and Double Effect,’ Proceedings of the American Catholic Philosophical Association, 12–25. Bennett, Jonathon (1981), ‘Morality and Consequences’, in S. McMurrin (ed.), The Tanner Lectures on Human Values, vol. II, Salt Lake City, UT and Cambridge: University of Utah Press and Cambridge University Press, pp. 45–116. Brand, Myles (1984), Intending and Acting: Towards a Naturalized Action Theory, Cambridge, MA: MIT Press. Bratman, Michael E. (1987), Intention, Plans and Practical Reason, Cambridge, MA: Harvard University Press. Chan, D.K. (2000), ‘Intention and Responsibility in Double Effect Cases’, Ethical Theory and Moral Practice, 3, 405–34. Chappell, Timothy (2002), ‘Two Distinctions that Do Make a Difference: The Action/Omission Distinction and the Principle of Double Effect’, Philosophy, 77, 211–33. Connell, F.J. (2003), ‘Double Effect, Principle of’, in New Catholic Encyclopedia, Detroit, MI: Thomson Gale, pp. 880–81. Davis, Nancy (1984), ‘The Doctrine of Double Effect: Problems of Interpretation’, Paciﬁc Philosophical Quarterly, 65, 107–23. 25 My thanks to the participants in the conference documented in this volume for their comments on one version of this chapter. I am particularly grateful to Michael Bratman, Al Mele and Ralf Stoecker for their intensive discussion. Special thanks also to Christoph Lumer, Julius Schaelike and Stephan Schlothfeldt, all of whom read and commented in detail on various written drafts. I am also grateful to the Swiss National Science Foundation for funding the research of which this paper is a product.

The Double Failure of ‘Double Effect’

115

Duff, R.A. (1990), Intention, Agency and Criminal Liability: Philosophy of Action and the Criminal Law, Oxford: Blackwell. Finnis, John (1991), ‘Intention and Side-Effects’, in R.G. Frey and C.W. Morris (eds), Liability and Responsibility, Cambridge: Cambridge University Press, pp. 32–64. Fletcher, George (2000), Rethinking Criminal Law, New York: Oxford University Press. Frey, R.G. (2003), ‘The Doctrine of Double Effect’, in R. G. Frey and C.H. Wellman (eds), A Companion to Applied Ethics, Oxford: Blackwell, pp. 464–74. Harman, Gilbert (1986), Change in View: Principles of Reasoning, Cambridge, MA: MIT Press. Harris, John (1980), Violence and Responsibility, London: Routledge. Hart, Herbert (1968), ‘Intention and Punishment’, in Herbert Hart, Punishment and Responsibility: Essays in the Philosophy of Law, Oxford: Clarendon, pp. 112–35. Kenny, Anthony (1973), Anatomy of the Soul: Historical Essays in the Philosophy of Mind, Oxford: Blackwell. —— (1977), ‘Intention and “Mens Rea” in Murder’, in P.M.S. Hacker and J. Raz (eds), Law, Morality, and Society: Essays in Honour of H.L.A. Hart, Oxford: Clarendon, pp. 161–74. —— (1995), ‘Philippa Foot on Double Effect’, in R. Hursthouse, G. Lawrence and W. Quinn (eds), Virtues and Reasons: Philippa Foot and Moral Theory, Oxford: Clarendon, pp. 77–87. Lackner, Karl and Kristian Kuhl (1999), Strafgesetzbuch mit Erlaeuterungen, Munich: Beck’sche Verlagsbuchhandlung. Law Commission (1989), A Criminal Code for England and Wales, Vol. 1, London: HMSO. Mackie, J.L. (1977), Ethics: Inventing Right and Wrong, London: Penguin. Mangan, Joseph T. (1949), ‘An Historical Analysis of the Principle of Double Effect’, Theological Studies, 10, 41–61. McCarthy, David (2002), ‘Intending Harm, Foreseeing Harm and Failures of the Will’, Noûs, 36, 622–42. Mele, Alfred R. (1992), Springs of Action: Understanding Intentional Behavior, New York: Oxford University Press. —— (2003), Motivation and Agency, New York: Oxford University Press. Nagel, Thomas (1986), The View from Nowhere, New York: Oxford University Press. Powers, M. (1995), ‘Contemporary Defenses of the Doctrine of Double Effect’, Revue Internationale de Philosophie, 3, 341–56. Quinn, Warren (1989), ‘Actions, Intentions and Consequences: The Doctrine of Double Effect’, Philosophy and Public Affairs, 18, 194–211. Roughley, Neil (forthcoming), Wanting and Intending: Elements of a Philosophy of Practical Mind, Dordrecht: Springer. Scanlon, T.M. (2000), ‘Intention and Permissibility’, Proceedings of the Aristotelian Society, supp. vol. LXXIV, 301–17.

116

Intentionality, Deliberation and Autonomy

Seebass, Gottfried (1994), ‘Handlungstheoretische Aspekte der Fahrlaessigkeit’, Jahrbuch fuer Recht und Ethik, 2, 375–411. Smith, J.C. and Brian Hogan (1978), Criminal Law, London: Butterworths. Velleman, J. David (1989), Practical Reﬂection, Princeton, NJ: Princeton University Press. Williams, Glanville (1983), A Textbook of Criminal Law, London: Stevens and Sons.

PART II Action-Theoretical Conceptions of Practical Deliberation

This page intentionally left blank

Chapter 6

The Will and The Good Hugh J. McCann

In common life the will and the good are closely related, in that we pursue the good through operations of the will. In moral epistemology, by contrast, the two are usually widely separated, in that the will is not considered a very useful guide to what is objectively good. This chapter explores the possibility of a closer relationship, defending the position that objects of desire are apprehended as good, not just in the sense that it seems so to the subject, but in the much stronger sense that, ceteris paribus, the fact that a thing is desired may be taken as good reason for counting it as a prima facie good. There are sound metaethical reasons for holding such a view, and practical rationality demands it. It is widely recognized that when human beings aim at getting things done – that is, when they form intentions, or engage in volitional activity to execute them – the will and the good are by nature related. Forming and executing intentions is a matter of active willing, of decision and volition, over which we exercise immediate control. Active willing is oriented towards the good in that when we decide and act, we do so in pursuit of some end we take to be valuable. In order to engage in active willing however, we must ﬁrst achieve some grasp of what is at least putatively good, and the role of the will in this process is less often explored and appreciated. The aim in this chapter is to help ameliorate this situation. The argument here is that passive willing – the sort of willing paradigmatically represented by the phenomenon of desire, over which we do not exert immediate control – ought to be viewed as a means, perhaps the chief means, by which we apprehend what is good. This argument involves examining two theses we might have in mind were we to claim that the will, conceived as a faculty of desire, is naturally oriented towards the good. The ﬁrst is relatively uncontroversial: that what is desired is always desired ‘under the aspect of good’. Many if not most philosophers would agree with this claim – in part because it does not invest any special trust in the will. It says only that to desire something is, in a certain way, to conceive of it as good. The second thesis is much stronger: that desire counts as a mode of objective awareness, that it is attuned to the good in such a way that what we desire has a prima facie claim to being actually good. This is a controversial claim, because if it is true then our inclinations count as a suitable starting point for an account of what is valuable and what is not; what we want becomes a reliable prima facie indicator of what is worth wanting, of what a person ought to want. There are good metaethical reasons for

120

Intentionality, Deliberation and Autonomy

holding this second thesis, and a few of them will be pointed out in what follows. But the main interest of this chapter has to do with practical rationality and it will be argued that unless this second thesis is true, it is impossible for intention formation to be, in the fullest sense of the term, a rational procedure. 1. Sub specie boni Let us begin by taking a quick look at the ﬁrst, relatively uncontroversial thesis. It was common among medieval philosophers to claim that what is desired is desired sub specie boni, or under the aspect of good.1 Indeed, this sort of thesis is as old as Aristotle, and on the usual understanding it does not come to very much. It says only that to desire X is to apprehend X, or some salient aspect of X, as good. Note that “apprehend” here does not mean judge, in the sense of taking some proposition to be true. Even in the strictly cognitive realm it is a mistake to confuse judgment and apprehension (to be aware of something as green is not necessarily to judge that it is green), and to confuse them in conative contexts is worse. That our desires and our judgments of value are very close cannot be denied, but to claim that desire taken in itself is or includes a kind of judgment is, as I understand things, to confuse the conative and cognitive functions of the mind. Judgment is a cognitive function, and its content is always a proposition: the sort of thing it would be appropriate to evaluate as either true or false. So if I judge X to be good, I afﬁrm the proposition that X is good, and my judgment will be correct or incorrect depending on whether this proposition is true or false. The content of a desire is not an appropriate candidate for such evaluation, for it is not a proposition at all. It is something optative: we might express it as, “Would that I have (or do) X”, or “Would that X be (or come to pass)”. What the claim that we always desire things under the aspect of good comes to is simply that to have this sort of thought, to think in the optative mode, is to experience or apprehend what is desired as worthy of pursuit, as worth desiring.2 Thus, although to have this experience is not in itself to make a cognitive judgment, we should hasten to add that it would be a natural next thought to judge X to be good – perhaps not überhaupt, but at least insofar as it displays the salient aspect, the aspect that arouses desire. The thesis that what is desired seems good to us has much to be said for it – so much, in fact, that it may even count as a kind of necessary truth. For how could it be that I desire something, and yet ﬁnd nothing in it? To be sure, the things we desire are often complex, and therefore frequently present themselves as having both good and bad aspects. And we often desire things we take to be bad overall. Yet it seems that to the extent we desire such things, they must appear good to us. If the opposite were possible, we would expect to ﬁnd people saying things like, “I would dearly love to have (or do) X, but there is nothing worthwhile about it, nothing at all valuable in 1 See, for example, Thomas Aquinas, Summa Theologica, I: q. 82, a. 2, ad 1. 2 Aversion can be treated simply as a desire that something not be. So something like: “Would that I avoid X”, or “Would that X not be”. I shall not discuss aversion in this paper, but most of what is said here would apply, mutatis mutandis, to aversion and the apprehension of evil.

The Will and The Good

121

having X (or in X-ing)”. And that seems a decidedly peculiar stance to take about anything. Not that no one has claimed it may occur. But the examples that might be proposed are either unpersuasive or have a pathological ring. One suggestion is that one might enjoy a philosophical activity like asking a humiliating question of a speaker simply for the sense of power thereby obtained, even though one does not consider such pleasures worthwhile (Setiya 2003, 353).3 But this is unconvincing. No one denies that the sense of power one gains from learning to walk, or to drive an automobile, is a good. Why, then, is it not a good in this case? To be sure, one may disapprove of this means of gaining a sense of power, but it does not follow that the sense of power – the sense of being able to exert inﬂuence, to make a difference – is not in itself a good. Still less does it follow that it is not desired as good by the agent. Indeed, if we think of enjoyment as desire fulﬁlled, what else would it mean to enjoy it? In the pathological vein is a second suggestion: that someone might have a brute desire to turn on radios, yet claim to see nothing of value in having them on (ibid., 354). The trouble with this kind of case is that precisely because no good is held to be in the ofﬁng, we cannot understand what is going on. Thus, rather than trust the agent’s claim to see no value in his act, we are likely to urge him to seek psychiatric help, so that he can come to recognize the value he saw. Finally, consider what purports to be a case of clear-eyed akrasia: I smoke an entire pack of cigarettes tonight knowing I will not enjoy them, because I plan to quit smoking as of tomorrow morning (ibid., 353). Here again, the example is implausible. If I get nothing from smoking, what supports the habit? And if I am quitting precisely because I get nothing out of it, then why should the prospect of quitting prompt me to smoke the whole pack? A case of waste not, want not? But then there is an apparent good, after all. Our weaker thesis, the sub specie boni claim, appears, therefore, to count as a necessary truth about desire. Indeed, the very weirdness of the cases offered as counter evidence seems to tell in its favour. It is an important thesis because if it is true, there is an implicitly normative dimension to desire. We cannot apprehend something as good yet remain neutral about its worth. Rather, to desire something is to ﬁnd it worthy of our desire, a ﬁtting object of pursuit for the will. That is why our next inclination, when we desire something, is to pronounce it good. To be sure, any such judgment is only prima facie, and therefore able to be defeated by other aspects of the case at hand. We may eventually conclude that what we desire, or desire in a certain respect, is not in the end good, not good overall. To the extent, however, that we desire it, we apprehend it as good. 2. Desire as a guide to the good Though opposed by some, the claim that things are desired as good does not in itself provoke deep or widespread controversy. The reason has already been hinted at: 3 Setiya is concerned to argue that it is possible to act for reasons one does not see as good. One could equally claim, however, that the desires or inclinations involved in his examples are devoid of any apprehension of goodness.

122

Intentionality, Deliberation and Autonomy

the import of the thesis is entirely subjective. To say a person who desires X ﬁnds X desirable is not to say the ﬁnding is in any way to be relied upon. To commit ourselves to this thesis is not, therefore, to adopt a position that has great bearing on questions of moral epistemology. And it may be thought that we should not wish to go any further, that a faculty as capricious as desire, as prone to wide variation from case to case and from individual to individual, could not possibly serve as a guide to anything objectively real. Indeed, the very idea that we might be able to learn about the world by consulting our desires may appear misguided. Learning, some might argue, is not even a conative function. It is a cognitive one, in which the world, which is what it is, somehow manages to imprint itself on our faculties, so that some of the things we think – namely, our beliefs – are brought into conformity with it. With the conative operations of the mind, as is so often said, the direction of ﬁt is the other way: we start from our thoughts, from what we want the world to be, and seek to make the world conform. That does not sound like the apprehension of anything real. How, then, can we attend to our desires and extract any kind of fact beyond that of the desire itself? How can we take the content of our states of inclination as telling us how things are? The force of this argument notwithstanding, I want to defend the stronger of our two theses: that we not only desire things as good, but that we may take our desires, in themselves, as a prima facie indicator – indeed, a kind of apprehension – of what is good, as a basis on which we could make at least a fair beginning at forming a list of good things. On such an account, to desire something counts as ﬁnding it good not just subjectively, but in the sense of taking hold of its true valuational character – so that when the will is functioning as it should, to desire a thing is to gain a sort of conative awareness of its valuational features, just as to see a thing is to become cognitively aware of its visual features. Usually we take the visual perceptions of ordinary folk as a pretty reliable indicator of the colours and shapes of things – to the point that we are apt to accept their judgments about what objects are around them. So if someone says, “I see a red cylindrical structure – a silo, I guess – standing next to that barn”, our disposition will be to take him at his word. In the same way, on the present suggestion, if someone were to say, “I want to go for a walk”, we could take the very fact that the individual has this desire as a pretty reliable indicator that walking is good, or at least that there is something good about it. In short, just as we treat what people see and do not see as giving a fair indication of what does or does not exist in their environment, so, on this understanding, we would treat what they want and do not want as a fair indication of what is good and bad. It should be emphasized that this is not a version of noncognitivism. That is, it is not being suggested that what appear to be propositional contents of cognitive value judgments – such as ‘Walking is good’ – are in fact not propositions, but rather effusions of emotion, disguised optations, endorsements, commendations, or something like that. Claims of this kind are completely misguided; indeed, if it is not self-evident that ‘Walking is good’ is a proposition, then nothing is self-evident – or so, at least, it seems to me. But the experience of desire, the conative apprehension of something as good, is worthy of a certain trust, and so can serve as a ground for cognitive judgments of value. One consideration that justiﬁes this trust has to do with what J.L. Mackie once referred to as the “queerness” of value properties – that

The Will and The Good

123

is, their implicitly teleological character (1990, 38–9). To call something good is not, by most accounts, to give a straightforward empirical description; rather, it is to present the thing in question as a prima facie legitimate objective, something worth having or doing, other things being equal. In other words, value properties have not so much to do with what exists, as with what deserves to exist. And of course it is this fact that makes moral epistemology so problematic, for it is not at all easy to see how such properties might be detected by the usual cognitive means. Indeed, for Mackie, even to accept value properties as objectively real commits us to postulating a special cognitive faculty by which they can be intuited (ibid.). To move in the direction of value intuitionism is, however, to court accusations of arbitrariness. The very postulation of a special faculty of value awareness may appear gratuitous; and certainly the deliverances of such a faculty may appear as attempts to settle things by ﬁat. If we are to claim walking is in some way good, it will not do to say we just see that this is so. For such a claim to be trustworthy we need some idea of how the goodness is seen, of how thinking about the so-called ‘natural’ features of walking – its merely descriptive characteristics – leads to the intimation of goodness (ibid., 41). The view being suggested here promises considerable help with these problems. Indeed, it is not all that distant from the intuitionist idea Mackie criticizes. The difference is that on this view our sense of good is not in the ﬁrst instance a cognitive matter, and so does not demand the postulation of an unfamiliar faculty to account for it, a faculty that is bound to appear ad hoc to the moral sceptic. Rather, this view lodges our primary awareness of the good in the familiar faculty of desire. Nor should it be thought implausible to treat desire as a source of such awareness. If value properties are teleological in nature, yet objectively real, it really does not make much sense that they should simply be ‘seen’. It is, however, both reasonable and ﬁtting that they should become manifest to us through a kind of tug – that is, through our being drawn towards things that have those properties, in an experience whereby they are presented to us precisely in the guise of the good. This sort of tug is exactly what occurs when we experience desire, if just the weaker, sub specie boni thesis is correct. Accepting the stronger thesis requires only that we take such representations seriously, as a basis for prima facie judgments of desirability or goodness. If we are willing to do so, then we can avoid introducing an ad hoc faculty of value intuition, and can ﬁnd a recognizable foundation for judgments of value in the phenomenology of moral experience.4 Treating intuitions of goodness as founded in desire also brings moral epistemology into accord with the practices of common life. The experience of desire is not the experience of a blind or arbitrary impulse. It is, as we have seen, the experience of being drawn towards an end that is seen as valuable – that is, as worthy of desire – in the very experience of desiring it. And there is good reason to think this experience 4 The phenomenology of value experience is less appreciated than it should be. When it is pursued, the central focus tends to be not on desire and the good but on our sense of moral obligation and the grounding of judgments of right and wrong. The seminal work in this area is Köhler 1938; see also Mandelbaum 1969. For a more recent treatment see Horgan and Timmons 2005.

124

Intentionality, Deliberation and Autonomy

is foundational to the values we actually hold. It is a simple fact of life that when it comes to determining desirability, desire is our most important guide. No amount of metaphysics or value theory could convince that something was good – that is, desirable – if on all sides people found it repulsive, or were simply indifferent to it. And when we disagree with others about the goodness of a thing, we seldom appeal to theory. Rather, we take a case-oriented approach: we rely, if not on experience itself, then on a detailed description of what we take to be the good-making features of the thing in question, to engender agreement. The reason is not that these tactics enable people simply to ‘see’ what is good, but that they evoke the appropriate conative appreciation, on the basis of which a value judgment can be made. Finally, understanding the experience of desire as a basis for prima facie judgments of objective value helps us make sense of the idea of value supervenience. It is a commonplace that value properties do not stand on their own, but rather supervene upon so-called ‘natural’ or non-value properties, in such a way that value properties will not vary unless the natural properties vary as well. Exactly why this is so is, however, somewhat mysterious on many accounts. For supervenience to be an objectively useful notion, it has to come to more than mere covariance: there has to be a way in which the descriptive properties of things genuinely give rise to their valuational features. But neither will supervenience be a useful idea if it turns out to be an occult relation, something of which we may insist we have a cognitive intuition, but an intuition whose origin we cannot explain. On the view suggested here supervenience is not an occult relation. Rather, goodness supervenes on those descriptive features of a thing or action that make it an appropriate object of desire, something that will be appreciated by a properly functioning will as worth having or pursuing. That is, goodness turns out to be objective desirability, and supervenes on those features of things that make them ﬁt objects of desire.5 Our stronger thesis is supported, then, by some weighty metaethical considerations, and also by the practices of common life. Of course, there are also considerations that may appear to militate against it. One is simply the way the will is treated on this view: that is, as a faculty by means of which we are able to lay hold of an aspect of objective reality, in experiences that then can serve as a basis for judgments of value. We are not used to thinking of our inclinations as anything but subjective in import, and some, no doubt, will balk at investing them with objective signiﬁcance. In fact, however, we do exactly that in daily living. And this move seems natural and appropriate when what is to be apprehended is something teleological, something having to do with what would be ﬁtting rather than what simply is. Moreover it is not at all misguided to treat the operations of the will as a basis for gaining this kind of information. Rather, what seems to be inappropriate is to believe that anything real must be apprehended through our cognitive faculties alone. It is preposterous to suppose that whereas our cognitive faculties are attuned to the way the world 5 It is worth noting that this characterization of goodness is itself implicitly evaluative. That is, what it holds to be valuable is not just what is desired, but what ought to be desired, what a properly functioning will should appreciate. This, however, is as it should be. Unless we are strict naturalists, we must expect that any deﬁnition of a value property such as goodness will have to be couched in evaluative terms.

The Will and The Good

125

truly is, our conative states track nothing at all that is objective. Our desires are not, after all, things we make up. Like sensation, desire is essentially passive: a kind of response we have when presented with various ends or courses of action we might pursue. And when we desire something it seems desirable to us in virtue of its nature, not ours (Mandelbaum 1969, 60). A faculty with these features ought not to be dismissed as merely subjective in import; to think some real dimension of the world is presented to us by means of it is much more reasonable. 3. The variability of desire Another consideration that may be thought to tell against the second thesis is the seeming capriciousness of desire. Desire is notoriously variable: what I desire you may not, and what I desire on one occasion I may not desire on another. This fact occasions no difﬁculty as long as we take desire to be strictly subjective in import. But if we wish to understand desire as a form of objective awareness, it will be argued, then we must address its apparent unruliness. Why is it, if desire is a vehicle of value awareness, that on so many occasions different subjects seem to be aware of different goods? Moreover desire is unsystematic: our wants seem to exhibit no special order among themselves, other than to reﬂect the exigencies of our practical situation at the moment, and desires are in frequent contradiction. By contrast, our cognitive beliefs about value tend to be structured, to be hierarchically ordered in terms of importance, or at least capable of being thus ordered; and when we detect contradictions in our value systems, we are at pains to remove them. How, it might be asked, are we to be guided in the transition from the supposed raw awareness of value postulated by our second thesis to a genuine system of values, something that might be useful for the satisfactory ordering of our lives? Part of the solution to this difﬁculty is to recognize that it is to a certain extent overblown. Desire may appear to vary greatly from individual to individual if we attend only to things like our preferences in food, clothing or literary genres, but they vary much less when it comes to important things. With rare exception, things like nourishment, shelter, love and a reasonable level of comfort are consistently desired by all. All of us want to be treated with justice and generosity, to be properly educated, to live fulﬁlling lives. In fact, part of the reason we are able to get along with one another is that our desires do not differ all that much; it is because we can safely anticipate that what others want will be rather like what we want that we are able to plan our actions so as to accommodate needs besides our own. This is not to deny that preferences may vary from person to person, and in the same person from time to time. In this, however, there are two things to remember. First, preference is often at least as much a matter of cognitive judgment as of desire. That my preferences should differ from yours might be due in part to my having different beliefs as to what is genuinely valuable, or most valuable all things considered. If so, a difference in preference need not necessarily bespeak a difference in desire. Second, even where desire is the main ingredient, it is most often preferences that are at issue, not desires. That one person should have a fondness for mystery novels while another enjoys military history does not imply that neither would ﬁnd the

126

Intentionality, Deliberation and Autonomy

other’s interest at all enjoyable. Still less does it imply – and this is what counts – that the mystery fan would not, if he thought about the matter, desire that military history even be read and understood, or that the armchair general would be repelled by the very idea of a mystery novel and abhor the suggestion that anybody might read one. What I apprehend as good is not, on the view we are discussing, to be identiﬁed with what I desire for myself, but with what I desire. An awful lot of the variability of desire concerns the former, not the latter, and so is not all that relevant to our second thesis. Still, it must be admitted that individual desires are at times seriously diverse. People do differ in the lifestyles they enjoy, in their standards of etiquette, in their preferred ways of dealing with others, and in a host of other things. In part, no doubt, this is due to differences in factual understanding. Value does supervene on nonvalue properties, and if we are not agreed on the facts about, say, what it is like to live extravagantly, we may have very different desires about doing so. Our desires may also differ depending on our life experience, what hurts or successes we may have endured, what tasks we think we are capable of, and so forth. Such differences need not, however, be considered damaging to the view being defended in this chapter. After all, the same applies to ordinary cognitive perception. Judgments of colour and shape, especially when made by a single individual, are often unreliable. Perhaps the perceiver is viewing the object from a poor angle, or in bad light, so that the shape is distorted or the colour mistaken. Or the fault may lie entirely within: the individual may be hallucinating, or colour blind. More bizarre things can go wrong, too. Sometimes things are misperceived simply because they violate our expectations – as when one hunts high and low for the keys one thinks one has lost, all the while failing to see them on the table in plain sight. We do not, however, allow such problems to overturn our trust in visual perception, largely because they are fairly overcome in the general case – that is, when we consider the collective judgments of percipients about common observables. Particular perceivers may have problems in particular cases, and we laymen may all have doubts when we leave the realm of the ordinary – doubts, for example, about whether certain objects postulated by sophisticated science are real. When it comes to objects of ordinary experience, however, we have no doubts. We know how our environment is populated, because we know that in concert, our faculties of perception are reliable. On the view being defended here, much the same holds of desire. There is no question that an individual can err in what she desires, apprehending things as good that – overall, at least – are not, or disliking things that are not in the end bad. There are bound to be vagaries of perception here, just as in the case of vision. Sometimes, in individual instances, the perspective or expectations of the person experiencing a desire will lead her to get things wrong; perhaps, too, as with vision, there are sophisticated or subtle cases where only the claims of experts may be trusted. Above all, there is the fact that our desires are ultimately the basis on which we form our intentions, and so are close to the objectives that deﬁne our practical lives. It is well known that we can become overly preoccupied with our personal projects, and therefore fail to appreciate anything that might stand in their way. Thus, almost inevitably, the wants of any individual will at times distort what is truly valuable. Here too, however, we can take refuge in the general case. Analogously to the case of

The Will and The Good

127

visual perception, we may view the collective desires of humankind as representing a concerted apprehension of what is truly good, that can serve as a corrective to individual error. It is important to realize that being mindful of the conative experience of others will not necessarily change our own. But that is not really all that important. The experience of a single individual is rarely to be taken as deﬁnitive of reality, and my desires need no more change when I come to realize yours are different than the experience of a colour-blind person must change when he comes to realize others see things differently. What is important, rather, is that my cognitive judgments of value be responsive to experience other than my own. That is one of the ways in which we are able to move from the relative disorder of our own conative experience to the construction of a system of values – that is, a structured and orderly set of beliefs about what is truly good. Indeed, many of our cognitive judgments about value are derived not from our own conative experience, but from that of others, imparted to us through education and training. Nor is this the only means by which order may be introduced into our value beliefs. There is also theoretical reﬂection, which may disclose common patterns and themes in the things we desire, and so assist us in organizing our value systems.6 It is worth noticing that there is a certain risk here. That our cognitive beliefs about value are founded on more than our own motivations can lead to cases in which our beliefs about what we should be doing are not matched by corresponding motives – the situation that is at the heart of the problem of weakness of will.7 Still, there is no reason to think the relative disorder of our conative experience taken in itself is incommensurate with the more structured character of our beliefs concerning value. 4. Rational decision-making From a metaethical point of view, then, there is much to be said for our stronger thesis, that the experience of desire ought to be viewed as an apprehension of objective value. It assists with the problem of ﬁnding an objective foundation for cognitive value judgments, which might otherwise be left high and dry; it makes credible our methods of persuading others about matters of value; and it lends substance to the notion of value supervenience without invoking special cognitive powers. There are, however, independent reasons for treating conative states as offering objective insight into matters of value, reasons having to do with rational decision-making. The fundamental problem here is well known: if we make the having of values an exclusively cognitive matter, we run the risk that no decision can properly be motivated by them. Let us for a moment assume the opposite of what has been argued in this chapter; that is, let us assume that our only sources of information about what is valuable are strictly cognitive in nature. And now suppose that by consulting those sources, whatever they are, you have discerned that it would be 6 An interesting illustration of how theory can help us organize value intuitions may be found in Audi 2004, ch. 3, where the moral principles of Rossian intuitionism are evaluated in terms of Kantian moral theory. 7 For a treatment of this issue, see McCann 1998, ch. 11.

128

Intentionality, Deliberation and Autonomy

good for me to get some exercise. The problem is that if you convey this information to me, you appear to give me no reason to act, in the sense of a motive that would justify my acting on your information. I might accept the truth of what you say, yet respond with complete indifference, loathing exercise as I do. I might even tell you your information is no good to me, since I am not motivated to exercise – that is, I do not like it – so you might as well go and try to help someone else. This kind of example seems to me quite reasonable and possible. Absent a motive, action cannot be rational, and judgments, even value judgments, do not appear to be motives. There is, however, a view, usually called motivational internalism, that would gainsay this.8 Essential to motivational internalism is the idea that the judgment that something is good necessarily implies motivation to act. On the strongest understanding this would mean that value judgments are in fact motives, in the sense that they are sufﬁcient in themselves to motivate action, and thus are able to provide rational justiﬁcation for action without the aid of any independent conative state.9 Alternatively, one might claim that the making of a value judgment entails the presence of a motive, though not necessarily one to be identiﬁed with the judgment, or the holding of it. Either way, our present example would be ruled out a priori. For on either version of motivational internalism, once I am presented with and accept the fact that exercise is good for me, I do have a reason for acting, in the sense of a motive. It may not, of course, prevail: my habitual laziness may very well win out – perhaps, we would say if we were determinists, it is the stronger motive. But presented with the fact that exercise is good, I at least have a competing motive, a basis on which it would have been rational to take some exercise. Motivational internalism is a debatable view on any formulation, and unconvincing to boot. Given the diversity of sources from which our beliefs about value are derived, it seems entirely possible that some might ﬁnd no resonance with our motives – that is, our conative states – and no amount of protestation or wishful thinking is going to make belief a form of conation.10 But we need not settle those issues here, because there are independent grounds for thinking motivational internalism is a misguided view. This applies especially to the stronger version articulated above – what I shall call strong internalism – according to which a value judgment counts in itself as a kind of motive. What this implies is that intention formation need not be conatively based: that is, that a decision to act can be reached in a rational way without any reliance on conative motivation. This is mistaken: unless it is grounded in some conative premise no act of intention formation can be rational in any very useful sense of the term. Rational decision-making is not just a matter of deciding to do what a reasonable person would say I should, even when that person is me. Just consider: I might be fully convinced that it would be good for me to take some exercise – and thus, 8 Motivational internalism has been defended in various forms by a number of authors, usually as a thesis about moral judgment. See, for example, Williams 1981, ch. 8; Nagel 1970, chs 8 and 11; Korsgaard 1996, ch. 11; and McDowell 1998, ch. 4. Critics of the view include Audi 1997, ch. 10; and Svavarsdóttir 1999. 9 For this sort of view see Nagel 1970, 110–11. My formulation of the strong version of intuitionism is owing to Svavarsdóttir 1999, 163; see also Audi 1997, 221. 10 Both Audi and Svavarsdóttir offer criticisms along these lines.

The Will and The Good

129

according to strong internalism, motivated to do so. And I might believe I will get some exercise if I take a walk. Yet I might decide to take a walk for an entirely different reason – say, to avoid a student I know is looking for me. If so, my decision may be rational, but it will not be made rational by my belief that some exercise would be good, for that belief will not, as we say, have ‘entered into’ my decisionmaking. What is it, then, for my belief to enter into my decision-making, and how does its doing so make my deciding to take a walk a rational act? We might be tempted to say that for my belief to rationalize my act of deciding is simply for it to cause that act. But that would be a poor answer. Suppose my belief that some exercise would be good is, as we say, ‘realized’ in some brain state of mine, and the same for my belief that walking would be a way of exercising. And suppose that these brain states together cause another brain state which realizes the event of my deciding to take a walk – so that I simply ﬁnd myself deciding to take a walk. I would most probably ﬁnd this a rather startling phenomenon, and if you asked me the reason for my decision, I might very well tell you there is none, that although I did have a reason for taking a walk, I had not decided for that reason, or indeed for any other, and that in fact I had no idea why I decided as I did. Why should this be so? Not because the causal story is false (I have no idea whether it is or not), but because that story is logically consistent with my having engaged in no process of actually reasoning, no process in which my decision was reached out of consideration of my reason. If we wish to claim cognitive value judgments can serve as a basis for rational decision-making, we need to explain how, starting from such a judgment, I can reason my way to a decision to act. Otherwise, strong internalism must fail. Is there, then, a way to portray my belief about the value of exercise as providing a basis on which I might reason my way to a decision to take a walk? Not as far as I can see. There are these two premises for me to work with: (1)

It would be good for me to get some exercise.

(2)

If I take a walk, I will get some exercise.

And these have to provide some kind of logical basis for my decision, which has the form: (3)

I shall take a walk.

But there seems to be no such basis. The content of my decision has to be identical with the intention thereby formed; and there appears no way, logically, to get that content out of a pair of beliefs. We might, perhaps, want to try for the related value judgment: (4)

It would be good for me to take a walk.

But of course this does not follow. Accepted principles of deontic logic would permit us to infer it if we had the converse of (2): (2a) If I get some exercise, I will take a walk.

130

Intentionality, Deliberation and Autonomy

But (2a) is false. There are dozens of other ways for me to get exercise – and, by the way, many of them would be as good or better than taking a walk. What is more, even if we were able to infer (4), it is not (3). In other words, there is still a gap between my coming to believe a walk would be good for me and my deciding to take the walk. And if we try to bridge it by having the belief simply cause the decision, we will wind up with a situation logically just as puzzling as that with which we began – not to mention the problem of now having a theory on which every judgment that some course of action would be good is liable to cause a decision to pursue that course. It does not look like we can get very far, then, in making decisions out to be exercises of actual reasoning by treating value judgments, that is, beliefs about value, as motivating. Suppose, however, that we allow the content of our desires to enter into practical decision-making. We can then have: (1a) Would that I get some exercise. (2)

If I take a walk, I will get some exercise.

(3)

Therefore, I shall take a walk.

Here again there is no question of obtaining the conclusion from the premises by valid deductive inference. What we must realize, however, is that a movement of that sort is out of place when, as in this case, the major premise is a desire, because the direction of ﬁt for practical content is not from mind to world, but from world to mind. Our business here is not to begin with a premise taken as true of the world and then travel from it to further information. Rather, the business of reasoning that is truly practical – practical not just in its subject matter but in its essential function, in what it actually does – is to progress towards bringing the world into conformity with the mind. To evaluate this reasoning, therefore, we must begin with the step it takes towards changing the world: we must begin with the conclusion. And what we then immediately notice is that if the conclusion is satisﬁed – that is, if my decision is carried out – then given the truth of the minor premise, the major premise will, by modus ponens, be satisﬁed also. The world will indeed be brought into conformity with my mind, in that my desire for exercise will be fulﬁlled. And now we are where we need to be. The thought process involved in my deciding to take a walk, which was utterly puzzling in logical terms as long as we treated my belief about the value of exercise as motivating, now becomes transparently rational – a veritable model of sensible practical thinking.11 Strong motivational internalism does not, as far as I can see, have any hope of matching this result. 5. What kind of rationality? But there is rationality and rationality. What the above discussion shows is that strong internalism fails not merely by misunderstanding the relation between cognitive 11 For this approach to evaluating validity for practical syllogisms, see Kenny 1975, ch. 5.

The Will and The Good

131

judgments of value and the phenomenon of motivation, but also in that it is unable to make sense of decision-making as a genuinely rational process. Decision-making is rational ﬁrst in the strictly descriptive sense that it is ratiocinative – that is, a process in which we reason our way to a conclusion – and second in the evaluative sense that the conclusion makes sense on the basis of its premises. What this requires is not a ‘motivating belief’ but a plain old motive – that is, in the present case, a desire. Only then do the premises of ‘practical reasoning’ have truly practical content, thereby enabling decision-making to be truly an exercise in reasoning, and truly reasonable, given the function it performs, and the way in which that function is carried out. Trying to get beliefs to do the same job only results in a mess. A weaker formulation of motivational internalism would fare better in this regard, since presumably it would allow the motivation allegedly entailed by a sincere value judgment to consist in some appropriate conative state accompanying the judgment. But even if as I doubt, the weaker entailment holds, internalism would point us in the wrong direction as far as the apprehension of value is concerned. For practical rationality is not just a matter of the logic of decision-making. The practical syllogism based on desire makes sense of decision-making because it presents intention formation as a process in which I frame as my project an action I take as a suitable means to a good end. But a piece of practical reasoning that is evaluatively rational only in this way may still fall short. That is because this kind of evaluative rationality is only the analogue of logical validity in deductive inference (we might call it practical validity), which limits it to merely subjective signiﬁcance: all that matters is that I have the desire to get some exercise, and the belief that taking a walk is a way to do so. Provided only that this is so, and that I decide on the basis of my desire and belief, I will have behaved as a rational being from a subjective point of view: I will have reached my decision in a way that is logically appropriate, given my premises. This sense of rationality is, however, limited, in essence because it requires only that the sub specie boni thesis be true; that is, it requires only that in desiring to get some exercise, I take it that doing so is a desirable thing, a good thing. Suppose, however, that in fact it was not ﬁnally a good thing – for example, that it would in my present state of health subject my heart to undue stress. Then we might say that in a more objective sense my reasoning here was not fully rational, because it was founded upon a desire that misrepresented what was truly good for me. An ideal agent, we might say, would not have had such a desire, because ideal agents do not desire things that are not good, anymore than ideal perceivers perceive things that are not there. For this situation to be rectiﬁed, it must ﬁrst be the case that my reasoning display the practical analogue of deductive soundness: that is, the desire on which my reasoning is founded must be well directed, in that the objective it embodies be truly good. The exercise I desire must be something that actually would serve me well. If this is so, then my conclusion – the intention formed through the decision I make on the basis of my desire – will be objectively worthwhile, an intention it is ﬁtting for me to proceed to enact. But even with this practical soundness achieved, the entire endeavour can turn out to be hollow. For suppose the thesis being urged here – that desire truly is a way, even our primary way, of grasping what is good – is false. Suppose desire is not a

132

Intentionality, Deliberation and Autonomy

way of apprehending the actual goodness of the things and courses of action we contemplate, but rather a mere projection of subjective preferences with no bearing on reality. Then my practical reasoning could not be an endeavour in which I behave as a fully rational being, for it would be completely blind. There could, of course, still be genuine goods – perceived, perhaps, by some unique cognitive intuition, or somehow arrived at by pure reason. But no genuine awareness of those goods would operate in the actual thought process of decision-making, even if our desires (perhaps owing to the causal inﬂuence of our cognitive judgments) reliably tracked them. The situation would be rather like one in which there truly was a physical world – a world, indeed, that was accurately described in our beliefs about it – but where those beliefs were not formed in the way we think they are, through legitimate cognitive contact with the world, but rather through extraneous intervention. Perhaps they would be formed by some Malebranchian God, who rigs our beliefs so they will match the world, but without, in their formation, ever allowing us to see the world. In the same way, a situation in which we formed intentions based on our desires, even if those desires were always for what is really good, would be a sham unless those desires themselves represented a form of insight into the world. In short, while the model described does, unlike other models, allow us to make sense of rational decision-making, it will do so in the fullest sense – a sense in which decisions may be said to be not only practically sound, but also legitimately founded – only if we take it that in experiencing desires, we really are apprehending what amounts to the teleology of the world, apprehending ways in which, prima facie at least, it is ﬁtting that things should be. I would urge, then, that the second of the two theses defended is not just valuable for its metaethical implications: its truth is indispensable to our being able fully to respect ourselves as practical thinkers.12 References Aquinas, Thomas (1945), Summa Theologica, excerpted in Basic Writings of Saint Thomas Aquinas, ed. A.C. Pegis, New York: Random House. Audi, Robert (1997), ‘Moral Judgment and Reasons for Action’, in Robert Audi, Moral Knowledge and Ethical Character, New York: Oxford University Press. —— (2004), The Good in the Right, Princeton, NJ: Princeton University Press. Horgan, Terry and Mark Timmons (2005), ‘Moral Phenomenology and Moral Theory’, Philosophical Issues, 15, 56–77. Kenny, Anthony (1975), Will, Freedom and Power, Oxford: Basil Blackwell. Köhler, Wolfgang (1938), The Place of Value in a World of Facts, New York: Liveright.

12 I am indebted to the participants in the International Conference on Intentionality, Deliberation and Autonomy: The Action-Theoretic Basis of Practical Philosophy, held at the Università Degli Studi di Siena in 2005, for comments on an earlier version of this paper. I have also beneﬁted from comments by Christoph Lumer, and from conversations with Robert Audi and Robert Kane on these matters.

The Will and The Good

133

Korsgaard, Christine M. (1996), ‘Skepticism about Practical Reason’, in Christine M. Korsgaard, Creating the Kingdom of Ends, New York: Cambridge University Press. Mackie, J.L. (1990), Ethics: Inventing Right and Wrong, New York: Penguin. Mandelbaum, Maurice (1969), The Phenomenology of Moral Experience, Baltimore, MD: Johns Hopkins Press. McCann, Hugh J. (1998), The Works of Agency, Ithaca, NY: Cornell University Press. McDowell, John (1998), ‘Are Moral Requirements Hypothetical Imperatives?’, in John McDowell, Mind, Value, and Reality, Cambridge, MA: Harvard University Press. Nagel, Thomas (1970), The Possibility of Altruism, Oxford: Clarendon. Setiya, Keiran (2003), ‘Explaining Action’, The Philosophical Review, 112, 339–93. Svavarsdóttir, Sigrún (1999), ‘Moral Cognition and Motivation’, The Philosophical Review, 108, 161–219. Williams, Bernard (1981), ‘Internal and External Reasons’, in Bernard Williams, Moral Luck, New York: Cambridge University Press.

This page intentionally left blank

Chapter 7

The Grounds and Structure of Reasons for Action Robert Audi

This chapter provides an account of the structure and grounding of reasons for action. Experience is represented as providing such grounding, both causally and normatively. The chapter also develops parallels between practical and theoretical reasons, both external and internal. Practical reasons are, however, seen to differ from theoretical reasons in the kind of content they have, and this difference is shown to be connected with the different explanatory and normative roles of the practical attitudes, such as intention and desire, and the theoretical ones, such as belief. The theory of reasons defended in the chapter provides an account of the relation between practical reasons and facts. It also describes a kind of dependence of practical reason on theoretical reason, but it afﬁrms their mutual irreducibility and the indispensability of both in the theory of rational action. We enter the world with a multitude of needs, entirely dependent on others, and utterly ignorant. But needs give rise to desires, experience remedies ignorance, and we learn to act on our own. Desire is essential for full-blooded action: nothing wanted, nothing done. Belief is also essential for full-blooded action,1 but desire is a ground of action, in a way belief is not. For one thing, we act in order to fulﬁl desire, but not in order to fulﬁl belief. More about this will follow, but let us begin by showing how even basic desires – those not grounded on other desires – can be grounded, and by comparing this kind of basis with that of their counterpart beliefs.2

1 It is true that if an action is basic for the agent (S), it may not be necessary for S to have a belief to guide it (though arguably S must believe at least something to the effect that to A one just does it – or does this). But for non-basic (intentional) action, some kind of instrumental belief seems required. This matter is discussed in some detail in Audi 1986a, 517–19. 2 My terminology may suggest to some readers that ‘desire’ and ‘want’ are normally used interchangeably. They are not; but the subtle differences between their uses will not cause trouble in my sometimes shifting from one to the other for stylistic reasons.

136

Intentionality, Deliberation and Autonomy

1. Experience, desire and belief Experience is the fabric of our lives. Life as lived, by contrast with mere biological animation, is constituted by experiences. This is largely why a good life is one in which good experiences, those having intrinsic value, predominate.3 As we are naturally constituted, good experiences are central among those we want, though there is no particular description under which we must want them. In the content of our desires for things that are good, plurality reigns, and the good things we want need not bear that generic name. It is clear that desires arise from experience, even if not solely from that source. Take an elemental case. The experience of eating a normal, satisfying kind of food (particularly where one needs food) commonly produces a desire for the sort of experience in question. To be sure, before children can develop concepts adequate for desire – which has a content that is in some sense conceivable to the subject – they must go through predecessor states of discrimination. In these states, although children respond differentially to the different properties that their experience reveals to them, they do not apprehend the corresponding concepts. There is no sharp distinction between these developmental stages. Consider a desire for, say, milk. This is distinct from a tendency to ingest it when it is offered; the desire requires at least a minimal concept of milk as opposed to other consumables. This kind of discriminative consummatory tendency may be a natural precursor of desire; but in either case, experience seems to be the most elemental (psychological) determinant of desire. Given this picture, we can say that experience is a source and sustainer of desire and hence a ground of it in at least that genetic sense. But it is important to see how experience grounds desire. For it is possible simply to want to drink milk and have no reason for this, nor even any sense of why one wants it. Even a desire that is (at least ultimately) genetically grounded in experience need not be one for which the person has a reason.4 But in normal cases of desire, we want the objects of our desire for something: either for qualities that are intrinsic to them (or that we take to be so) or for their contribution (or believed contribution) to something else we want. My interest here is in the basic conative case: in what may be called intrinsic desire. This is desire for the wanted object for its own sake. Intrinsic desire is, roughly, wanting the object on account of its intrinsic qualities (or qualities one takes to be intrinsic to it – a qualiﬁcation usually omitted hereafter). Thus, we might want to converse for the stimulation we ﬁnd in the exchange, to drink wine for its ﬂavour, and to feel a soft breeze for its cool gentleness. There are two kinds of grounds important here: causal and normative. The experience of enjoying wine can produce desire to sip it, and in that (genetic and causal) sense can ground the desire. The experience apparently does this causal work through our consciousness of certain of its qualities. Those gustatory qualities 3 The distinction between inherent and intrinsic value is developed at some length in Audi 2003b. This paper also makes a case for experientialism regarding the basic bearers of intrinsic value. 4 This is argued in Audi 2002.

The Grounds and Structure of Reasons for Action

137

are enjoyable (or perhaps in some other way rewarding). In virtue of our enjoying them, we tend to want to drink the wine. The appeal to enjoyment, then, can explain why we want to drink it. But surely experiencing these enjoyable qualities can also normatively ground our desire. Roughly, to ground it normatively is to support its rationality. This is easier to see where desire seems irrational than in a case of natural, perfectly rational desire like wanting to sip wine for its enjoyable ﬂavour. Imagine a man combing his hair with a brass clothes brush. Thinking he may be causing himself pain, we ask why on earth he wants to brush his hair with that. If he replies, “It feels good – it relieves a terrible itch”, we can see his desire as rational in the light of the qualities for which the action is (intrinsically) wanted. Since belief as well as desire is essential for action, it is also important to consider the grounding of belief. The parallel between belief and desire is striking here.5 It is plain that experience is a source of beliefs. “Seeing is believing”, we often hear, and comparable points might be made for the other senses.6 The parallel holds not only for causal grounding but also for normative grounding. Seeing lightning is a sensory experience, and it tends to produce (in those with concepts adequate for it) belief to the effect that lightning is occurring. Such beliefs, moreover, are rational on the basis of (and justiﬁed by) the sensory experience in question. We might generalize. In virtue of certain evidential or, we might say, truthindicative qualities of experience, beliefs are rational. There is, for instance, the ﬂashing, jagged brightness of lightning, the roaring sound of a motor, the furry feel of a cat. Similarly, we might say that in virtue of certain value-indicative qualities of experience, desires are rational. There is, for example, the pleasantly tangy quality of sipped wine, the relief from heat that comes with a cool breeze, the sense of free movement and high muscle tone we can have on a mountain hike. As the examples suggest, the value-indicative properties need not indicate value as such; at least in typical cases, they indicate something having or conferring value. They do not directly point to the property of being valuable (if they point to it at all), but rather to something that has or indeed underlies that property. Believability and desirability, and not just belief and desire, are (in the basic cases) grounded in experience. 2. The mutual irreducibility of practical and theoretical reason For philosophers seeking understanding of a complex kind of thing, it is natural and proper to want the most economical account that does justice to the data to be explained. The more extensive the analogy between any two kinds of phenomena under investigation, the stronger the pull towards economy of analysis. In the case of normativity, must we countenance two ultimately different dimensions, the practical and the theoretical? Could rational belief, for instance, be fully explicable in terms of rational desire or conversely?

5 I have discussed parallels between desire and belief in great detail in Audi 2001, esp. chs 3–6. 6 This adage is critically discussed in some detail in Audi 2003a, ch 1.

138

Intentionality, Deliberation and Autonomy

Few if any plausible attempts have been made to reduce theoretical to practical rationality, and this possibility will not be addressed here.7 But philosophers at least since Plato have been attracted to the idea that the will should be subordinate to the intellect. This idea may generate reductive efforts of two kinds, one causal, the other normative. First, one might argue that beliefs (or other theoretical attitudes) can adequately motivate action; second, one might argue that beliefs – or at least those that are justiﬁed – can render action rational. If these points hold, one might go on to argue for two reductive claims: that desire is not essential either for the performance of action or for its rationality, and that our practical rationality is a matter of the content and rationality of our beliefs. The ﬁrst, broadly causal claim has been associated with motivational internalism, which, in its most general form, I take to be the view that some degree of motivation to do a deed is internal to the judgment that one ought (on balance) to do it. That view as usually understood concerns moral beliefs rather than normative beliefs of all kinds, but the difﬁculties confronting it suggest why normative beliefs alone cannot account for all human motivation. I have elsewhere argued at length that they cannot, and I will not pursue the matter now.8 But it is central to our purposes here to consider whether, normatively, belief is the basis of desire: speciﬁcally, whether the rationality of the former is the ground of the rationality of the latter. If desire is essential to motivating at least some human actions (and the concern here is human action), then the theory of action must countenance desire, and there is less incentive to eliminate it in our normative theory. But the further question whether the rationality of desire is analysable in terms of, or at least depends on, that of belief is important in its own right. Some of what I have to say about this will be implicit in the case to be made shortly for a difference in the kinds of objects of the practical and theoretical attitudes. But two points are appropriate immediately. First, the analysability claim is surely too strong. A desire to drink milk can be rational on the part of a child without the child’s believing anything to the effect that drinking milk is good. Indeed, it seems likely that if, when we are conceptually capable of (rationally) believing drinking something to be intrinsically good, say as pleasantly refreshing, we do rationally believe that, this is very likely on the basis of enjoying it; and if one (rationally) believes it to be instrumentally good, this is very likely on the basis of enjoying (or suffering from the absence of) something to which one believes it is (in some way) a means. The point is also apparent in relation to aesthetic experience. I may enter a gallery not expecting to enjoy, and even thinking ill of, the unconventional works on display. Suppose I am surprised and I ﬁnd viewing them intriguing and, after a time, enjoyable. This response can render rational both my wanting to continue viewing them and my believing they are good. I need no antecedent or concurrent belief that they are good to render my desire rational. The second point concerns the cognitive dependency claim: the thesis that the rationality of one’s desire depends on that of one’s beliefs. There are two relevant 7 I have addressed the possibility in Audi 2001, ch. 5. 8 Motivational internalism is assessed in detail, and much literature on the topic is considered, in Audi 1997b, which contains other papers pertinent to the subject of this one.

The Grounds and Structure of Reasons for Action

139

kinds of normative dependency that are not always distinguished. One is negative, a kind of defeasiblity, a vulnerability to defeat. The other is positive, a kind of grounding. Suppose I want to do something, say to sky-dive, for what I take to be its enjoyable qualities. This may be a rational desire, but its rationality (and certainly the rationality of wanting on balance to sky-dive, or of intending to do it) can be defeated by my discovery that the experience would only be unpleasantly dizzying. The normative ground of my desire is the anticipation of enjoyable qualities;9 but its rationality negatively depends on what I do or should believe. In this way, practical reason has a negative dependence on theoretical reason. A necessary condition for rational desire, then, is the absence of certain beliefs or of other “theoretical” elements. But it does not follow and is not true that, for desires, practical rationality is constituted by possession of, or grounding of desire in, such elements. Defeasiblity is a kind of necessary condition dependency; it is not a constitutive dependency. This defence of the irreducibility and (qualiﬁed) autonomy of practical reason should be balanced by two further points, one negative, one positive. First, theoretical reason does not have a like dependence on practical reason. The rationality of belief, for instance, is not comparably defeasible by the presence of even rational desires (not comparably, but possible in one way, since a belief that I do not want x could have its rationality defeated by the perceptible presence of a desire for x). Even rationally wanting to believe that p is false – for instance, that it is false that a loved one has died – cannot undermine the rationality of believing it; and rationally wanting p to be true, though it can lead to excusable wishful thinking that it is true, does not make that belief rational (I here conceive wishful thinking that p as believing that p because of wanting that it be true). Second, on the positive side, even though belief is not required as a ground of rational intrinsic desire, belief can render even intrinsic desire rational. Believing that I would enjoy viewing the unconventional paintings can render rational my intrinsically wanting to view them. A belief can do this, however, only if rational, and its rationality in such a matter (where one has not experienced the type of thing in question) will normatively depend in some appropriate way on some experience that is relevantly similar to the kind of experience in prospect as an object of desire. There is another source of reasons for resisting a reduction of practical to theoretical rationality and for taking the former to have a signiﬁcant kind and degree of independence of the latter. If desire and belief, as paradigms of practical and theoretical attitudes, have different kinds of objects that, in turn, differ in the criteria for their appraisal, this would support (though it does not entail) the mutual irreducibility of practical and theoretical rationality. How this is so will become clear as we examine their objects. We begin with intuitive data and, in the next section, proceed to an outline of a theory of reasons for action.

9 Perhaps the relevant anticipation of an experience can serve as a normative ground for wanting it, only if I have actually experienced something I take to be appropriately similar, but I want to allow that anticipation, like memory impressions, can be among experiential elements that qualify as normative grounds.

140

Intentionality, Deliberation and Autonomy

Beliefs, but not desires, may be said to be true or false (I do not mean to rule out a third value here, for even if a belief is neither true nor false, there is no semantic impropriety, as there is with desires, in calling it true or calling it false). Moreover, where the objects of desires are one’s own actions, then, as in the case of intentions, these objects are most naturally identiﬁed inﬁnitivally: schematically, to A. Even where a desire is not for action, in English, at least, the subjunctive is preferable to the indicative for speciﬁcation of its object. I have a desire that you visit me (not that you will do so), a desire that my students prosper (not that they will prosper) a desire that my concerns be (not are) heard by my congressional representative. These phrases are not used in a way that calls for assessment of what they express as true or false. That fact goes with the point that the attitudes themselves are never properly called true or false in the way beliefs are (a true intention is a genuine one, to be sure, but there ‘true’ contrasts not with ‘false’ but with ‘apparent’). Intention should also be mentioned as a central practical attitude. If (in English, at least) the inﬁnitive is not used to express the object of an intention, the subjunctive is the norm: I intend that there be plenty of food at my party, not that there will be. Granted, there is the stilted ‘I intend that my children shall run my company after my death’, but the sense of this is something like ‘I intend to see to it that my children run my company’. None of this is to deny that all of the attitudes in question are propositional in the wide sense that they are individuated by content in a broadly propositional way. For instance, intending to A differs from intending to B in the way the proposition that one A’s differs from the proposition that one B’s. As this suggests, the practical attitudes are intentional and their objects correspond to propositions in a way soon to be indicated. What sorts of things, then, do constitute the objects of desire, where ‘object’ has the content sense appropriate to propositional attitudes? Wanting the ﬂower to be replanted to the left may be conceived as having the state of affairs, the ﬂower’s being replanted to the left, as object. I have long thought that the most natural generic category – though not the only category – to use in describing these attitudes is that of states of affairs, understood roughly as what is designated by certain kinds of descriptions of acts, situations or occurrences that the desire represents and that may or may not ever be actual. Roughly, I am speaking of states of affairs in the sense in which (unless impossible or necessary) they may or may not obtain, just as (normally) the object of a desire or intention may or may not be realized. If I want to view paintings, the relevant state of affairs is (my) viewing paintings; this may or may not occur. The act-type in question may or may not be tokened. The mention of an act-type here suggests a plausible direction in which we might go to achieve greater speciﬁcity. We might plausibly hold that the basic cases of intention and all cases of action-desires (desires to do) have act-types as their objects. These act-types are abstract enough to do justice to the many different tokens that can realize the objects, but are narrower than states of affairs conceived generically. We would say, then, that intending that there be plenty of food at my party is equivalent to some intention to do, say to supply plenty of food. That is plausible; but wanting may be wider: wanting to possess a painting, for instance, is

The Grounds and Structure of Reasons for Action

141

not equivalent to wanting to do something. We need states of affairs or something similar to do justice to the object of that kind of desire. One might think that since, in any of these cases of practical attitudes, the proposition that the relevant state of affairs does occur corresponds to the state of affairs and is true if and only if that occurs, we might simply speak of propositions and avoid complicating our terminology (and possibility our ontology). But whereas the proposition has truth value – presumably even before I act – act-types and states of affairs, when they ﬁgure as objects of intentions and certain desires, are instead viewed, at least by the agent, as to be brought about. The signiﬁcance of this contrast will emerge in the next section, which concerns the structure of reasons, especially practical reasons: normative reasons to act. 3. Normative reasons During the years when the controversy over whether reasons are causes raged, reasons for action were taken to be intentional attitudes or, not infrequently, desire– belief combinations of the kind Donald Davidson called primary reasons.10 More recently, reasons have been taken to correspond with the contents of such attitudes. This seems a more natural view. The most basic speciﬁcation – though not the only kind of speciﬁcation – of a reason to act seems to be of the form of ‘to x’, where the inﬁnitive clause is of the kind that follows ‘in order’ when an action is explained as performed in order to bring something about. The most basic speciﬁcation of a reason for belief, by contrast, is of the form of ‘that p’, where p is a proposition. It is true, however, that we can properly say, in answering ‘What is your reason?’ as applied to action, things with the forms of ‘I want to …’, ‘I intend to …’ and ‘I believe …’. Recognizing this, I have distinguished, in earlier work, between reason states and reasons proper.11 The former are intentional attitudes, the latter their contents (also often called their objects). Thus, asked for a reason for driving rather than taking a fast, comfortable train, I can say ‘I wanted to take a great deal of luggage’ or ‘to take a great deal of luggage’ or even ‘I believed I had too much luggage for the train’, or simply ‘I had too much luggage’. Often, citing the reason state is felt to be equivalent to citing the reason proper; but the former ascription is often preferable where one is uncertain, for instance unsure whether one was correct in thinking there was too much luggage for the train. If, as I suggest, we distinguish between reasons proper and reason states, we can leave open whether reasons in either sense are causes, though few would argue that the abstract contents of intentional attitudes are the right sorts of things to be causes. Normative reasons may be practical or theoretical. The practical kind of normative reasons are broadly speaking reasons for action, roughly, reasons that favour performing some action and support the rationality of doing so. The theoretical kind are broadly speaking reasons to believe, roughly, reasons that favour believing some 10 In Davidson 1963. 11 Acting for Reasons (Audi 1986a) is probably my earliest published work developing the distinction between reasons and reason states in detail.

142

Intentionality, Deliberation and Autonomy

proposition and support the rationality of believing it. The remainder of this chapter will suggest an overall conception of these two kinds of normative reason and to connect it both with the sorts of normative grounding sketched above and with the different structures of the objects of the practical and theoretical attitudes. One way to conceive practical reasons connects them with the theory of value. Speaking in quite general terms, we might say that a practical reason is (in good part) a projection of something of value (something good, if we simply treat reduction or elimination of pain or suffering as good): something of intrinsic value if the reason is basic, something of instrumental value if the reason is non-basic. Thus, someone might say that a good reason to go into medicine is to reduce suffering. I suggest we think of the phrase ‘to reduce suffering’ projectively and conceive afﬁrming the reason, as in saying it is a good one (or, for example, is one’s reason for wanting to A), as making the corresponding projection. Such (sincere) afﬁrmation does not entail having the intention to A, though it would naturally give rise to it. Practical reasons (roughly, normative reasons for action) are commonly expressed by rational desires, since these commonly project something of value. But not all desires express reasons; a sufﬁciently ill-grounded desire, say a hypnotically induced (non-instrumental) desire to be slowly burned to death, would not (though it might provide a motivational reason why someone lights a match in the absence of anything to be ignited).12 The notion of desirability (as applied to states of affairs) may be taken to be approximately equivalent to that of goodness, and it may help to describe the suggested conception of a practical reason as an invitation to view it as projecting something that is in some way desirable. Why speak of projection rather than simply of representation? A practical reason may be conceived as a projection because the agent (actual or hypothetical) sees the object as realizable in the future by A-ing. We could regiment and say that the projection must be correct (factual). But it is more natural to hold that if the projection is (sufﬁciently) plausible, it yields a good reason, and if not, a bad one. Where it is both plausible and correct, we might speak of a sound reason. The reason is provided (expressed, carried and so on) by a desire or intention; it is grounded – strictly, its normative force is grounded – ultimately in the good-making properties of the relevant experience, for example pleasurableness. (For convenience I construe “badness-reducing” properties as good-making.) The projection is fulﬁlled, or realized, if the relevant experience occurs. To be sure, the presence of one or more good-making properties of an experience allows that it have bad-making properties too. Hence, desire grounded in experience 12 Indeed, even if an ill-grounded desire has a content that describes something of value – as where a hypnotically induced desire is for good conversation – it may not provide the agent with a normative reason for action. There is (normative) reason to pursue good conversation (on the plausible assumption that it is an intrinsically good thing), but not just any desire for it provides the agent with normative reason to pursue it. The agent may (say, because of narrow experience and poor education) lack an adequate conception of good conversation. A merely non-instrumental desire, moreover – one that is neither intrinsic nor instrumental – projects neither kind of value. When I have forgotten why I want to go to the backyard, but still want to and ﬁnd myself doing it, my desire at the time points neither to any intrinsic good in the action nor to any good it leads to. This is supported by the arguments of Audi 2002.

The Grounds and Structure of Reasons for Action

143

of one good-making property need not be rational on balance, and its object need not be good overall. (This projectivist notion of a reason does not depend, however, on the experientialist view that fulﬁlment of the relevant projection is constituted by the occurrence of an experience; the notion would apply to non-experiential states of affairs as well.) By contrast, a theoretical reason, conceived as a normative reason for belief, is (in good part) a representation of something as true. Someone might say, for instance, ‘That she always sneezes around my cat is a reason to believe she has an allergy to cats.’ The kind of representation in question is apparently truth-valued and admits of justiﬁcation. We might say that if it constitutes a basic (normative) reason for the subject, it is non-inferentially (prima facie) justiﬁed for that person; otherwise it is in some sense inferentially justiﬁed for the person.13 Again, we could regiment and say that the representation must be or express a fact, but we could also do as above. Here, however, the representation need not concern the future and, a fortiori, is not constituted by a projection into the future. It concerns how things are (or were); and it is grounded (strictly, its normative force is grounded) in a truth-indicative property such as the sense of a hard, cool smooth surface beneath one’s ﬁngers (as indicating glass). If the reason is factual (hence externally sound), and its relation to what it is a reason for is entailment, then the factuality transmits and the entailed proposition is true. If the relation is inductive, then probability, a “measure” of truth, transmits. It is natural in speaking of propositional attitudes to talk as if they were relations to objects, and I will continue to do so. But my account of reasons can leave open the possibility of taking those attitudes to be properties of persons, projective or representational as the case may be. The attitudes would still be individuated in a ﬁne-grained way, and the practical ones would still differ from the theoretical ones in the kind of content they have. On either the relational or non-relational construal of the propositional attitudes, having such an attitude puts the possessor in different functional relations to the world. Roughly, in virtue of beliefs we are disposed to make corrections as a result of certain sensory inputs and relevant inferences; and in virtue of intentions we are disposed to perform actions that, on our instrumental beliefs, will bring about the state of affairs represented in our projections. My account does presuppose some role for intentional psychological properties, then, but does not strictly require that these be relational. 4. Intellect and will and the functioning of practical and theoretical reasons The proposed conceptions of the structure of reasons are highly consonant with a venerable idea that still has much to recommend it: the idea that the function of the intellect is to pursue truth, and the function of the will – the practical “faculty” – is 13 Two points will be clarifying here. First, I take it that a merely non-inferential belief – one that is neither experientially nor inferentially grounded – does not supply any theoretical reason. The case for this is similar to the case regarding merely non-instrumental desires and is outlined in Audi 2002. Second, I include, under “inferential” construed broadly, any justiﬁcation coherence might supply.

144

Intentionality, Deliberation and Autonomy

to pursue the good.14 Theoretical reasons at their best reﬂect how the world is – or even how all possible worlds are. Practical reasons at their best project how the world may be made better. There have been many versions of this idea, and there are numerous views about how intellect and will are related, including the famous Socratic doctrine that (baldly stated) knowledge of the good implies willing (or at least wanting) its realization.15 In contemporary philosophy this conception of intellect and will seems to have reappeared in the view that belief and desire – and more generally the theoretical and practical attitudes – have different directions of ﬁt to the world.16 Roughly, beliefs and other theoretical attitudes (those with truth-valued objects) succeed (in an external objective sense) when they ﬁt the world; desires and other practical attitudes (those with the kinds of objects intentions have) succeed when the world ﬁts them (at least typically, the future world, but otherwise the world contemporaneous with the desire). The normatively central point here is that different standards of appraisal are appropriate. The point does not commit us to attributing any particular sort of content to either kind of attitude. To be sure, true but unjustiﬁed beliefs, and similarly, realized but irrational desires, are not unqualiﬁedly successful. We might call them internally unsuccessful insofar as we think that despite their positive status in one respect – in virtue of the agent’s being correct and getting what is wanted – they violate standards that the person accepts or, if sufﬁciently rational and adequately reﬂective, would accept. Still, we may contrast beliefs and desires at least in this: being false is clearly a kind of defect in beliefs, whereas being, even at all times, unrealized (and thereby arguably instantiating a counterpart of falsehood) is not a defect in desires. What has generally not been noticed is that there is apparently a connection between the kinds of contents appropriate to the theoretical and practical attitudes and, on the other hand, their different directions of ﬁt and, more broadly, their different functions as basic elements of intellect and will respectively. In spite of these differences between beliefs and desires in function and directions of ﬁt, a number of philosophers have taken them to have the same kinds of contents – truth-valued ones – and many other philosophers have been at least casual about distinguishing these kinds of contents.17 Perhaps it need not be argued that (de dicto) beliefs have propositional contents, in the sense of items that are true or false (we can leave open

14 As Aquinas puts it at one point, “The object of the reason is the true … the object of the appetitive power is the appetitive good.” Summa Theologiae, q. 60, a. 1. 15 This kind of view survives in ethical theory under the name “motivational internalism”. For a detailed appraisal and many references to relevant literature, see Audi 1997a (reprinted in Cullity and Gaut 1997, which contains several other papers bearing on motivational internalism). 16 This version of the contrast is apparently due to Anscombe (1956). 17 Davidson, for example, has taken intending to be a kind of belief (see Davidson 1980). Others have supposed there are attitudes (sometimes called besires) with the function of both belief and desire.

The Grounds and Structure of Reasons for Action

145

whether some kind of linguistic entity could serve here). But something must be said about contents of desires, intentions and other practical attitudes.18 As stressed above, truth value may always (without semantic impropriety) be ascribed to beliefs, but never to desires. We express the content of desires using (primarily) inﬁnitives or subjunctives, not indicatives: I can want to talk with you, or that we talk, but I cannot want that I am talking with you, or that I will talk with you, or even that it is true that I am talking with you. There is a parallel set of data regarding theoretical and practical reasons, roughly, reasons for believing and reasons for action (or at least for desire). As suggested earlier, a reason for action is always expressible in a phrase of the form of ‘to A’ (or ‘in order to A’ where the action has occurred), where ‘A’ ranges over actiontypes, though a reason for action may also be expressed in other ways.19 A reason for believing is never so expressible. A reason why one believes might be claimed to be expressible in that way. A doxastic voluntarist might say, for instance, that it was in order to save his life that James believed he could jump across a precipice (where he brought himself to believe this to increase his conﬁdence of success). But this apparently makes good sense only where we think of the ‘in order to’ as expressing a reason to cause or sustain the belief. That is a reason for action.20 The positive side of reasons for believing is more complex. If a reason for believing is conceived strictly, as expressible in a sentence of the form of ‘My reason for believing that p is that q’, then such a reason is always a proposition (or similar entity, such as, on one conception, a fact conceived as a true proposition). This is because q must express something believed by the speaker (assuming the overall statement is true). Moreover, if what is cited as a reason for believing p really is a reason for believing it, one can at least sensibly ask whether it supports (for example, entails, explains or probabilistically implies) p. There are, however, grounds for believing a proposition that are not reasons in the strict sense for believing it (though I do not mean to overstress this distinction, since the proposition that a ground obtains is a reason for believing a proposition the ground supports). Simply seeing a ﬂash of lightning is a ground, not (I think) strictly a reason, for believing there really is a ﬂash. We can call seeing it a reason, particularly if we are thinking of the point that once we believe we see it, we have a reason, in the strict sense, for believing there is lightning: namely, the proposition 18 In (Castañeda 1975) Hector-Neri Castañeda construed the objects of the practical attitudes as “practitions”, which he contrasted with propositions as objects of the theoretical attitudes. Although I do not adopt his terminology (or his speciﬁc views on this issue), I have beneﬁted from his far-reaching work on the ontology of the intentional attitude. 19 As noted in the text, in answering “What was your reason for doing that?”, one may answer “I believed …” or (especially) “I wanted …”. I call these attitudes reason states and distinguish them from reasons proper, which are their contents. A partial account of this distinction is given in Audi 1986a. That paper also deals with the special case of things we do for their own sake, for which ‘in order to’ is perhaps not entirely appropriate in giving an explanation. In any event, I suggest that where a belief provides a reason for acting, it does so at least in part because there is a suitably related want or a reason for having such a want, as where one’s reason for telephoning is given by citing a belief that one has too little time for e-mail. 20 Cf. Harman 1999, esp. chs 1 and 4; and Foley 1993 on this point.

146

Intentionality, Deliberation and Autonomy

that we see it.21 But we would not normally cite just our seeing a ﬂash of lightning as a reason for believing there is one, but rather the proposition that we see it. Nonpropositional grounds do not even seem to constitute reasons for believing (except in the sense of reasons why it is that one believes) until they are believed to obtain. They are, to be sure, evidences; and reasons for believing, on my view, are propositional expressions of evidence (or at least apparent evidence). In any event, for both reasons proper and grounds, we may say that when they ﬁgure as reasons or grounds for believing, they support believing as truth-valued. A premise, as well as an experience or memory impression, may be a ground, and a belief of a premise may express a reason (an inferential one). In the case of nonpropositional grounds, such as a visual experience of lightning, the counterpart reason for belief is also a proposition: one that expresses the state of affairs experienced. Reasons for believing (and for other theoretical attitudes) are in this way unlike practical reasons, which are not truth-valued though they express a content that may be said to be realized or unrealized. A desire that one’s friend receive an honour, for instance, expresses (and projects) a content to be realized. With this much theory in view, we can see something important. Suppose we think of the will – or, in my terminology here, of practical reason – as properly functioning to change the world in the direction of either some good (whether it is objectively or subjectively conceived does not matter at this level of generality) or at least some desired state of affairs. And suppose we also conceive the will as guided in exercising this function by practical reasons and as successful when its functioning – by its acts, we might say – produces the changes. If the guiding reasons are practical, they must point to acts. This is exactly what practical reasons, in the central cases of their operation, do. For intention, and equally for action-desires, the content of practical reasons is expressible by an inﬁnitive clause designating an act-type, say ‘to signal my vote’. This clause expresses my reason for raising my hand; my will produces that action; and the action changes the world (or so one may hope). By contrast, think of the intellect as properly functioning to provide a true representation of the world – or the part of it relevant to the subject, or at least to the subject’s survival. The intellect is properly guided by theoretical reasons. Insofar as it represents the world correctly, it is externally and objectively successful. If the guiding reasons are theoretical, they must point to truths. (Whether the pointing must be conceived in terms of epistemically internalist or externalist standards may be left open here.) If the content of the reasons is propositional, they can do this; if it is not (at least in the indirect way a ground is) – if, for example, the content is practical, as where ‘to make life bearable’ expresses a presumptive reason for believing one’s friend is honest – they cannot. Thus, suppose my reason for believing there is lightning nearby is that a ﬂash and loud cracking sound have occurred. This clause expresses a proposition that constitutes a reason for my belief; my coming to believe that evidential proposition produces, and my believing it causally grounds,

21 One epistemologist who uses “reason” to encompass grounds is Swain (1981). A sorting out of different uses of “reason” is given in Audi 1998.

The Grounds and Structure of Reasons for Action

147

my believing that there is lightning; and my intellectual state changes to reﬂect the world. The ﬁrst thing that may come to mind in the comparison is the different directions of causation: from the will to the world in the ﬁrst case, and from the world to the intellect in the second. This is as it should be: if the will succeeds when it changes the world, it must produce a causal condition sufﬁcient for the change; and if the intellect succeeds when it truly represents the world, it must achieve such a representation through a causal connection running from the world to it.22 If, however, the broad perspective on intellect and will so far presented is correct, there is something more. It bears on the causal contrast just drawn, but, as the next two sections will show, it has more to do with the ontology of the objects of intentional attitudes. 5. The relation between reasons and facts I have described practical reasons as non-truth-valued and best represented by contents of inﬁnitive clauses, and I have characterized theoretical reasons as truthvalued and best represented by contents of truth-valued propositional clauses. If I am correct, then I am bound to show how my view can take account of two points that may suggest a univocal conception – speciﬁcally, a facticity view – of the nature of reasons.23 Let us begin with the notions of a person’s reasons for believing and for doing something. The ﬁrst thing that may strike one here is that there is at least an oddity in saying that p is a person’s reason for believing or doing something if the person takes p to be false. Second, it appears that reasons for action can always be expressed in a propositional mode and, when they are, also seem subject to this same kind of constraint: if p is false, we ﬁnd it at best odd to call it the person’s reason for doing it. These worries require considerable analysis. Let me start with the theoretical case. Why is it odd to say (at least in English) that someone’s reason for believing p is q, where one does not think q is true, and especially where one thinks q is false? Is this because, if q is false, we would be referring merely to a case of believing for a (cognitively) motivating, as opposed to normative, reason? Surely the reason need not be merely motivating. One can have excellent evidence for q, say joint testimonial and inductive evidence, even if q is 22 We cannot say ‘causing it’; beliefs about the future are not caused by future events. But there is still a causal connection: from causes of the relevant events, such as my decision to do something tomorrow, to the belief representing those events. 23 In epistemological literature, something at least close to this view has been afﬁrmed as early as 1975. Unger said that “[I]f someone’s reason for something is that p, then it follows that it is true that p.” (Unger 1975, 208). A similar view regarding evidence has been recently defended by Williamson: “Although we may treat false propositions as evidence, it does not follow that they are [evidence] … If e is evidence for h, then e is true” (Williamson 2000, 186 and 201). In the domain of practical reason, Stampe has held, “Reasons are what we mean to reason from, and reasons are facts … Thus the sailors who, believing that the earth is ﬂat, declined to sail with Columbus had in that belief no reason to decline: since the earth is not ﬂat, its being ﬂat was no reason.” (Stampe 1987, 337; cf. Parﬁt 1997 and Dancy 2000, esp. ch 5.)

148

Intentionality, Deliberation and Autonomy

false. I think that when, speaking sincerely, we give our own reasons for believing p, we believe the proposition(s) we cite; and the most common case with others – we normally assume – is their having truths as their reasons. We so regard our own reasons; and we tend to give others credit for getting such things right unless we have reason to think otherwise.24 This pragmatic point is sufﬁcient to explain why it is odd to say such things as that his reason for believing that (for example) there has been life on Mars is that there was once water there (where we think there was not, though there is evidence of water). We want to say something like ‘His reason is that – as he sees it – there was once water there.’ Notice, though, that – speaking as we commonly do in giving our reasons – he might say his reason is that there was once water there. This explanation is equivalent to ‘The reason for which he believes that there has been life on Mars is that there was water there.’ This is a similarly odd thing to say if we think there has never been water there. But consider the following equivalences. The proposition that someone believes that p (wholly) on the basis of (the proposition that) q, seems equivalent to the proposition that the person believes that p (wholly) on the basis of believing that q and to the proposition that the reason for which the person believes p is that q.25 But the second of these, employing the basis locution, clearly does not imply that q is true. If so, neither do the others. I think that the best explanation of the data here is again that pragmatic considerations heavily inﬂuence what reason-ascribing locutions we use and that, as is not uncommonly the case with equivalent locutions in English, some ascriptions are odd where their equivalents are not. Equivalents may also differ in what their common uses imply, as distinct from what their content entails. This pragmatic difference is consistent with the view I am presenting, on which we can believe for reasons in the same sense of the term whether the propositions in question are true or false. It should also be noted that on the facticity view, if I am asked my reason for believing p (or for what reason I believe p), and I sincerely say that my reason for believing it (or the reason for which I believe it) is q, I can be corrected – by being told that q is not my reason – (1) without being thought insincere or mistaken about why I believe p and (2) on the ground that q is false. But we do not correct selfascriptions of (actuating) reasons for believing in this way. We normally presuppose that others (apart from self-deception and other special cases) know for what reasons they believe what they do (when they do believe for a reason rather than simply on the basis of, say, perception or memory), and we normally do not take the falsehood of p to falsify this presupposition of self-knowledge. Suppose, for instance, that I am told by a credible person that q is false. Assuming that I now doubt that q but do not immediately cease to believe q, I will not now doubt that q is the reason for which I 24 We might speak here of true propositions rather than of truths. But suppose a truth is not just a true proposition but one that is undergirded by a corresponding fact. This might be so and may be why some people think of true propositions as facts. Nothing in this paper will turn on my conceiving truths as true propositions and leaving open how they are related to facts. 25 This equivalence is argued for in Audi 1986b.

The Grounds and Structure of Reasons for Action

149

believe p, but rather will think that perhaps I ought not to believe p on the basis of q (and may also come to doubt that p is true). A further point is this. Suppose that, as is common, I formed my belief that p on the basis of reasoning from q, I will ﬁnd it especially strange to think that I believe p for no reason. My reasoning might even be valid, hence formally good. In one terminology, the trouble with my reasoning (hence my inferential ground for p) is external; it is with the inputs to my reasoning. This defect does not preclude my conclusion’s being reason-based. To be sure, we might perhaps say that in second-and third-person ascriptions of reasons for belief, whether of (a) reasons for anyone at all to believe p, or (b) reasons for a speciﬁc person to believe it, or (c) reasons for which a particular person does believe it, a (defeasible) presupposition of objectivity predominates. Hence, if we think q false, we normally do not cite q as someone’s reason for (holding the belief that) p, unless we cancel the predominance presupposition, for example with ‘that – as he claims – q’. Suppose the question is ‘What reason does Sally have for her belief that p?’ We usually would not say ‘that q’ if we disbelieve q – though we might cite the corresponding reason state, saying that she believes q. If we are thinking of q as false and only as a subjective or a motivating reason or both, we are likely to say ‘She believes that q’, or use ‘as she sees it’, or in some other way distance ourselves from the objectivity presupposition. It may help in understanding such cases to notice that in the kind of context in question, where the inquirer presumably seeks an objective reason, the notion of a reason there is to believe p seems to be operating. And it is no surprise that such an external reason must be true (factive). Suppose that one could have a reason to believe p only if one had an external one – a reason that is factive. Now consider a person with excellent basis for believing p, say q, for which the person has evidence in the light of which any rational person would make the (objective) mistake of believing q (which is false). We would have to conclude that the person has no reason to believe p. This holds even if good evidence, or indeed any genuine evidence, must be true (on the plausible assumption that some evidence is inductive and hence nontruth-entailing). This consequence of the facticity view not only forces us to draw what is at best an implausible conclusion about beliefs based on excellent though ultimately misleading evidence. It also drives a wedge between, on the one hand the notion of rational belief and, on the other, the notion of reasons for believing a proposition as constituted by something that, at least given the evidential resources of the subject, supports that proposition and renders believing it rational. Similarly, the view would sever the connection between reasons for belief and reasoning as a process normally understood to supply them (even if not always good ones). (Parallel points hold for practical reasons.) It is neither plausible nor theoretically desirable to hold that the kinds of considerations on the basis of which it is rational to hold a belief cannot be reasons for it unless they are facts (or true propositions). This view seems to lead to an avoidable paradox. Granted, on my view we must acknowledge that a reason one has, a possessed reason, need not be external or objective – a reason there is to believe p (or to A) – and that reason-ascriptions for beliefs tend to presuppose truth and so, apart from

150

Intentionality, Deliberation and Autonomy

qualiﬁers like ‘as she sees it’ which select subjective (or at least internal) reasons for p, are odd and misleading given the ascriber’s disbelief of the proposition in question. But granting this point sacriﬁces nothing major if anything at all. For many truths, there are some contexts in which any unqualiﬁed afﬁrmation of them is odd or misleading. Let us now see how nearly parallel the practical case is and then proceed to further considerations favouring the distinction earlier suggested between the kinds of objects of the practical and theoretical attitudes. Consider the most general kind of ascription of a reason for action: the kind employing the purposive locution ‘in order to’. Every intentional action for a further end – and arguably every intentional action – admits of an in-order-to explanation in which the agent’s reason is expressed by the inﬁnitive clause.26 We raise our hands in order to greet others, open books in order to ﬁnd information, make appointments in order to meet people, and so forth. These locutions imply both a desire, in a very wide sense encompassing (though not entailing) intending, and, except possibly where the action is basic for the agent, an instrumental belief. But they do not presuppose that the instrumental belief is true. Thus, even if the same explanation can normally be given by citing the belief, for instance by saying that she opened the book because she believed it would contain a map of Siena, this equivalence would not hold if that mode of explanation presupposed a true belief. One might reply that if the instrumental belief is false, then both explanations cite only a motive, not a reason. But an instrumental belief does not by itself constitute a motive: the agent may have no desire for the state of affairs to which the belief represents the action as a means. One might go on to claim that the explanation cites only a motivating reason. But what would this come to, if it implies any more than that the instrumental belief is false? Not that the action or agent lacks rationality; not that the goal sought is not desirable; not that there is no sense of ‘reason’ in which the agent had a reason to act and acted for that reason; not – so far as I can tell – anything implying that practical reasons must be true or in any other sense factive. Connected with these points are other untoward consequences of taking reasons for action – not just objective or sound or “external” reasons for action – to be factive. Let us focus on the practical case (the theoretical one is quite parallel). If my reason for action must be factive, say, constituted by some truth connecting my action with its goal (what I want in performing it), three untoward consequences follow. First, I cannot, in a way that seems plainly possible,27 know what my reasons are without knowing the relevant propositions to be true, say cannot know that my reason for signing a check is that it will discharge a debt (where the check will be lost in the mail). Second, I can do something in order to achieve some end, yet either do it for no reason or do it for a reason though I have no reason for doing it. Third, others who see that my relevant belief is false can tell me that I acted for no reason 26 This is argued in Audi 1986a, which admits a possible exception for explanations of basic actions. 27 There is a subtlety here that I cannot discuss: if I do not believe that reasons are factive, I might know that R is my reason for an act but not believe, hence not know, that R is true. (I cannot believe it false; but I might, for example, believe only that it is probably true.)

The Grounds and Structure of Reasons for Action

151

or acted for a reason though I had no reason. But all of these implications seem quite mistaken. If, for instance, you take a medicine in order to relieve a headache, I may not say to you, simply on the ground that the medicine will have no effect, that you had no reason for taking it, or that you were not acting for a reason. I can perhaps say to you that despite appearances, there was no good reason to take it, where now (if this paper is correct in the matter) I invoke the notion of an objective reason, an external reason there is, for an action. But this is neither the only notion of a reason for action nor – even more important – the only kind that bears on the rationality of action.28 The contrast with appearances prepares us to take ‘good reason’ to entail truth, but ordinarily the notion of a good reason is not this narrow. I think, then, that the external, objective notion of a reason does not even reﬂect the only notion of a good reason for action. Regarding the medicine that I have excellent but ultimately misleading reason to think effective, I could reply that since I had excellent evidence that the medicine would work, I had a good reason to take it: one that was evidentially good even if factually mistaken. To suggest that my reason was only motivating fails to do justice to its role in rational assessment of the action. 6. Facticity versus factual groundability Nothing written here requires denying that there is an important connection between objective reasons for action (and belief) and facts. In my view, which is an axiological objectivism, unless it is true that there is something good about a state of affairs – in which case it is a fact that there is – there is no (objective) reason to bring it about. Moreover, if there is (objective) reason to believe p, say that q, then the proposition that q is a truth. We must grant, then, that if someone believes that p on the basis of a false proposition, this false proposition, though it is the person’s reason, is not an objective (external) one. Thus, we can say such things as that, though his reason for believing that there is life elsewhere in our galaxy is convincing to him and backed by some reﬂection, it is mistaken (not a reason for the rest of us, not a good reason, not objectively sound, not in line with the facts, and so on). Indeed, on the kind of experiential groundedness view sketched here, there is a still deeper connection between reasons and facts. Apart from the reality, hence the factual character, of the qualities of experience (including abstract thinking as a kind of intellectual experience) in virtue of which experience renders desires and beliefs rational, there are no rational propositional attitudes to provide good reasons for action or belief, objective or other. Neither true beliefs nor desires for something objectively good provide normative reasons for action if these attitudes are ill-grounded. Wishful thinking may by good fortune lead to my truly believing 28 Another problem with the facticity of reasons view arises on the plausible epistemically internalist view that we have internal access (roughly, access by reﬂection or introspection) to our justifying reasons, hence to their content: we do not have internal access to all the relevant facts. I may have such access to my believing that p and hence to p as an object of thought. But apart from facts about my own inner life (and perhaps certain a priori facts) I do not have such access to the fact that p (assuming p is true).

152

Intentionality, Deliberation and Autonomy

that a medicine will cure me, and a lucky post-hypnotic suggestion may lead me to want to do something that is desirable but could not be known by me to be so given my limited experience. In neither case do I have good reason for acting. We can now see that just as normative reasons for action, in the wide sense of reasons for the agent in question to act, need not themselves be true propositions (or facts, conceived as entailing the relevant truths even if different kinds of entities), not all factive propositions that are a person’s de facto reasons for action are genuinely normative. It is not only factive propositions that can support the rationality of action, nor is it supported by every factive proposition that constitutes a reason for which a person acts. I believe, however, that in a certain way practical reasons depend on facts. Reasons there are to do something (practical reasons in the external, objective sense of ‘reason’) depend on something’s in fact having value; and even a person’s reason for doing something, if it is a good reason for the person to do it, must be well-grounded in certain facts of experience. We can retain a facticity view regarding ultimate (normative) grounds, then, even if we reject the overall facticity of reasons view. To see more clearly how normative reasons depend on facts, recall the case in which one takes a medicine in order to reduce pain but is mistaken, though well justiﬁed, in thinking that it will reduce pain. It has already been noted that if reducing pain is an objective reason, then reducing pain – to reduce pain, to put it in projective language – is in fact good. But now consider the false belief. It provides a normative reason for the act (supports the rationality of performing it) only if there are facts appropriately accessible to the agent in the light of which there is objective reason for the act. One such fact would be that there is evidence making it probable that the medicine will reduce pain. If there is no evidence, then the agent is not well justiﬁed in the instrumental belief. On my view, the justiﬁcatory bearing of the evidence in such a case of mistaken belief is supported by an evidential principle that the agent can in principle see to be true: that when there is information of the relevant kind (say, a high proportion of positive instances in a variety of circumstances, and no negative instances) regarding a proposition, it is likely to be true. I would myself argue that epistemic principles of the relevant kind are a priori, but that need not be assumed here. It also need not be assumed that the agent believes (as opposed to being disposed to believe) such a supporting principle or can even formulate one, at least apart from Socratic prodding. The kind of questioning that may lead to formulation may also lead to belief formation. The general point here is that, although the agent’s (normative) reason for action, propositionally expressed, need not be true, it is such as to render the action truthgroundable – factually groundable, if you like. (The counterpart point holds for normative reasons for belief, but the analogy will not be pursued here.) A proponent of the facticity view could insist on regimentation and claim that the agent’s “real reason” is the true proposition (or the fact) that there is evidence that the medicine will reduce pain. But this seems inadequate both to our actual thinking about what our reasons are and to our discourse in which they ﬁgure. I believe, then, that practical reasons as most perspicuously expressed – inﬁnitivally – are not truthvalued, and that normative reasons for action that are propositionally expressed

The Grounds and Structure of Reasons for Action

153

need not be factive. But basic practical reasons correspond to axiological truths, and propositional expressions of practical reasons, if not themselves true, are so connected with facts as to render the actions in question truth-groundable. It should now be clear that denying that practical reasons must be construed as facts does not commit one to denying the plausible idea that every practical reason has a propositional expression and that, if this proposition is true, the reason is indeed factive (either because a true proposition constitutes a fact in the relevant sense or because there must be a fact in virtue of which it is true). Suppose the reason for Aing is to bring about G. This reason can be expressed by an instrumental proposition. The “simplest” kind expresses a sufﬁcient condition relation between the act and the goal: that A-ing will bring about G. 7. Moral reasons So far, moral reasons have not been explicitly brought into the present account. Moral reasons can be theoretical or practical. Both are of concern here, particularly the latter. But even when a moral reason is truth-valued and in that generic way theoretical, as where it is a judgment that one is morally obligated to do something for a friend, it supports one or another type of action. As I view morality, its basic principles are epistemically autonomous: they can be seen to be true on the basis of adequate reﬂection on their content, even if they can also be known on a non-moral basis, such as an appeal to a sound general theory of practical reason. There is, then, a cognitive basis, a grounding in theoretical reason, on which they can be known.29 I have maintained that theoretical reason has a special authority in the practical domain. The epistemic autonomy of moral principles seems to me to constitute one case in point. If I know that I ought to keep my promises, I thereby have reason to do so. As a result of knowing this, I may also want to do so. Even a justiﬁed false belief of this kind could provide a reason to act accordingly that can both motivate doing so and is good enough to render such action rational. But quite apart from whether motivation is implied by self-addressed moral knowledge, the existence of objective reasons for action is implied by the truth of moral principles. The moral principles that underlie one’s knowledge and belief ascribing obligation to oneself, then, have considerable normative authority in the practical domain. These points do not entail that moral facts, as opposed to normatively authoritative facts of experience, are required as a ground for the existence of moral reasons (though this may well be true, depending on what constitutes a moral fact). The argument here has been that there can even be instrumental reasons good enough to sustain rational action without the reason-constituting propositions being true, as opposed to believed in a sufﬁciently rational way. But even if moral facts (in some special, speciﬁc sense) are not necessary in grounding good moral reasons, there are morally signiﬁcant non-moral facts that are sufﬁcient. Just as an action’s reducing pain (a non-moral fact about it) makes the action desirable and yields a reason for performing it (where the reason is to reduce pain), an action’s fulﬁlling 29 I defend the moral epistemology suggested here in Audi 2004, esp. chs 1 and 2.

154

Intentionality, Deliberation and Autonomy

a promise or avoiding an injury (a non-moral fact about it) makes it morally (prima facie) obligatory and yields a reason for it: to keep the promise or to avoid injuring someone. Maintaining the epistemic autonomy of moral principles does not, however, require maintaining their isolation. Indeed, I think, as Plato and Aristotle and Kant (among others) have, that there is a route to seeing their truth from considerations about practical reason. But this is not the place to argue for that.30 My main concern has been to indicate how actions are grounded in desires and beliefs, which in turn are both causally grounded in experience and, when rational, also normatively grounded therein. Experience has qualities in virtue of which desires can be rational and can thereby provide (which is not to say desires can constitute) reasons for action. The reasons that desires can provide are objectively good when the object of the desire is also objectively good, whether intrinsically or instrumentally. We can speak of reasons for action as grounds for it – speciﬁcally, of reason states as in some sense causally grounding it and of reasons proper as normatively grounding it. This point holds for moral reasons as well as for other kinds of reasons for action. ***** The overall picture offered in this chapter represents experience as grounding reasons, both causally and normatively, and whether they are practical or theoretical, or external or internal. This foundational commonality implies many parallels, but, even on the assumption that non-factive normative reasons are truth-groundable, it does not imply that either of our rational capacities is reducible to the other. I have granted one kind of dependence of practical reason on theoretical reason – a kind of defeasibility of practical considerations by theoretical ones – but have also afﬁrmed their mutual irreducibility. I have maintained not only that neither dimension of rationality can be reduced to the other, but also that, connected with this difference, the two sorts of attitudes have different kinds of objects. Only the objects of theoretical attitudes are truth-valued. This does not imply, however, that truth plays no role in the notion of practical reasons. The sound ones, the reasons there are for any of us to do certain kinds of things, correspond with truths. These objective, external reasons for action may be seen as projections of something good, rather as sound theoretical reasons may be seen as representations of something true. Experience supplies grounds for reasons of both kinds. When it supplies grounds for normative beliefs, and particularly for moral judgments – all of which have a certain future-oriented element – it may, in that way, provide grounds for action that accords with those judgments. But this does not entail that theoretical grounds exhaust the basis of practical reasons. Experiential grounds are not reasons, but I do not deny that the category of grounds includes reasons, actual or potential, for action or belief. The ultimate grounds, which on my view underlie reasons, are experiential. The enjoyable and painful qualities of experience that are the readiest paradigms of practical grounds, 30 In The Architecture of Reason (Audi 2001, ch. 6) I have made a case for the view that considerations of practical reason support the rationality of being moral.

The Grounds and Structure of Reasons for Action

155

like the sensory qualities of experiences that are the readiest paradigms of theoretical grounds, confer rationality but do not admit of it. Experiences and their qualities are not as such rational or irrational. The two kinds of grounds, most prominently sensory grounds for belief and hedonic grounds for desire, go with the different roles of belief and the other theoretical attitudes and, on the other hand, of desires, intentions and the other practical attitudes. The theoretical attitudes give us our map of the world, and the normative beliefs among them can tell us what destinations are worth visiting, as well as, in the moral case, which ones we must, or must not, visit. The practical attitudes, when they are well-grounded in good experiences, will at once help us to discover worthwhile destinations and motivate us to reach them.31 References Anscombe, Elizabeth (1956), Intention, Oxford: Basil Blackwell. Audi, Robert (1986a), ‘Acting for Reasons’, Philosophical Review, 95, 511–46. —— (1986b), ‘Belief, Reason, and Inference’, Philosophical Topics, 14, 27–65. —— (1997a), ‘Moral Judgment and Reasons for Action’, in Audi 1997b, 217–47. —— (1997b), Moral Knowledge and Ethical Character, Oxford: Oxford University Press. —— (1998), ‘Reasons for Believing’, in Edward Craig, (ed.), Routledge Encyclopedia of Philosophy, London: Routledge, 127–30. —— (2001), The Architecture of Reason, Oxford: Oxford University Press. —— (2002), ‘Prospects for a Naturalization of Practical Reason: Humean Instrumentalism and the Normative Authority of Desire’, International Journal of Philosophical Studies, 10, 235–63. —— (2003a), ‘Intrinsic Value and Reasons for Action’, Southern Journal of Philosophy, 41, 30–56. —— (2003b), Epistemology, London: Routledge. —— (2004), The Good in the Right, Princeton: Princeton University Press. Castañeda, Hector-Neri (1975), Thinking and Doing, Dordrecht: Reidel. Craig, Edward (ed.) (1998), Routledge Encyclopedia of Philosophy, London: Routledge. Cullity, Garrett and Berys Gaut (1997), Ethics and Practical Reason, Oxford: Oxford University Press. Dancy, Jonathan (2000), Practical Reality, Oxford: Oxford University Press. Davidson, Donald (1963), ‘Actions, Reasons and Causes’, Journal of Philosophy, 60. —— (1980), ‘Intending’, in Donald Davidson, Essays on Actions and Events, Oxford: Oxford University Press.

31 For helpful comments on this chapter I am grateful to audiences at Northwestern University, the Oxford Moral Philosophy Society, the State University of New York at Albany, the Siena Conference on Intentionality, Deliberation, and Autonomy, and to Christoph Lumer, Hugh McCann and Derek Parﬁt.

156

Intentionality, Deliberation and Autonomy

Foley, Richard (1993), Working without a Net, Oxford: Oxford University Press. Harman, Gilbert (1999), Reasoning, Meaning, and Mind, Oxford: Clarendon. Parﬁt, Derek (1997) ‘Reasons and Motivation’, Proceedings of the Aristotelian Society. Stampe, Dennis (1987), ‘The Authority of Desire’, Philosophical Review, 96. Swain, Marshall (1981), Reasons and Knowledge, Ithaca, NY: Cornell University Press. Unger, Peter (1975), Ignorance, Oxford: Oxford University Press. Williamson, Timothy (2000), Knowledge and Its Limits, Oxford: Oxford University Press.

Chapter 8

An Empirical Theory of Practical Reasons and its Use for Practical Philosophy Christoph Lumer

In the ﬁrst part of this chapter (sections 2–5) an empirical theory of practical reasons is sketched and defended. It consists of: hypotheses about what intentions are, namely optimality beliefs, (2), hypotheses about how intentions are formed on the basis of probabilistic beliefs and intrinsic desires (3), a pluralist theory about intrinsic desires (4) and a theory about motives for moral action (5). In the second part (sections 6–8) it is argued that normative practical philosophy must rely on empirical theories of practical reasons (6). Then it is shown that theories of prudential practical rationality and of morals that are not based on such empirical theories run into problems. And it is sketched how the empirical theory outlined in the ﬁrst part can be used as a basis for a reconstructive approach to practical rationality and of an internalist approach to morals (7–8). 1. The aims of this chapter and the concept of ‘practical reason’ Theoretical reasons are reasons for believing, practical reasons are reasons for acting. Two main types of practical reasons have to be distinguished. 1. Normative reasons – which have to be differentiated again into 1.1. rational or prudential and 1.2. moral reasons – are good reasons, which serve to choose between actions (and to justify them later on) in a rational, prudential or moral way. Ethics and the theory of practical rationality imply theories about normative reasons. 2. Explanatory reasons are reasons for which someone acted; they explain why the person acted as he did. They are dealt with in philosophy of action.1 In practical philosophy nowadays, at the beginning of the twenty-ﬁrst century, there is much talk of “practical reasons”, their ontology, their content, their rationality, their intelligibility, if and how they motivate and so on. One problem of these debates 1 This is a usual distinction (see for example Nagel 1970, 4, 18; Smith 1987, 38f.; Dancy 2000, 1–4), though the explanatory reasons sometimes are called “motivating reasons”. Holding that normative reasons should be motivating, too, I prefer to speak of “explanatory” versus “normative reasons”.

158

Intentionality, Deliberation and Autonomy

is that the normative theories often make or presuppose strong assumptions about the empirical features of our explanatory practical reasons but that these assumptions are almost never proved and developed into a coherent theory, so that the normative theories remain unfounded in this respect.2 This chapter has two aims and parts. First it will sketch a theory of explanatory reasons, that is, an empirical decision theory about how people deliberate, decide and form intentions, culminating in several empirical hypotheses (sections 2–5). (This part to a great extent summarizes pieces of my research. References will be provided in the single sections.) Subsequently it will be explained why and how normative theories of practical reasons depend on this type of empirical information (section 6); and some exemplary consequences of the empirical theory for normative theories of practical reasons will be drawn, namely for the theory of practical rationality and for ethics (sections 7–8). The mentioned boom of talking about “practical reasons” has a further problematic aspect. This talk is often very unclear and vague; sometimes it is even intended to be so. But there is also a cause to the point of this vagueness: ‘Explanatory reason’ is only a comprehensive notion, a variable for all kinds of considerations that have inﬂuenced the ﬁnal intention. (In ordinary language we can call, for example, any of the following instances that contributed to the decision an “(explanatory) reason” for the action: desires, criteria underlying such desires, (probabilistic) beliefs about means-ends-relations, about other implications of the action, about the set of alternatives, comparative valuations of actions, beliefs about events that have caused certain desires, about circumstances that constitute particular occasions for actions, beliefs about facts that may justify the beliefs listed so far, beliefs about authority arguments for the beliefs listed so far. And if the speaker thinks the agent’s beliefs to be true he may also call the respective states of affairs “reasons”. (Lumer 2004a, 709f.)) So ‘explanatory reason for an action a’ may best be deﬁned as: (selfcontained) part of the relevant considerations and mental attitudes in the deliberation leading to the intention to a, where this intention has caused a in the right way (cf. ibid., 710).3 The empirical theory to be sketched in the following (sections 2–5) is 2 Examples are Dancy’s attempts to prove cognitivism and to prove that all explanatory reasons are objective facts (Dancy 2000, 85, 98–116, 126–33). These endeavours are criticized in: Lumer 2004a, sect. 4. Another example is Searle’s “proof” that there are desireindependent motivating reasons. At the crucial point he simply assumes that recognizing one’s social obligation is already motivating (Searle 2001, 170f., 181) – without providing even a trace of an argument or explanation for this assumption. 3 Some words about the ontology of reasons. Whereas normative reasons have an objective form – ‘that the shares of IBM will increase in price tomorrow is a normative reason to buy them today’ – explanatory reasons are presented in a subjective as well as in an objective form: ‘She bought IBM because she thought these shares would increase the following day’ versus ‘She buys IBM because these shares will rise tomorrow.’ The classical explanation of this discrepancy is: the correct and primary form of explanatory reasons is subjective; but we can and should use the objective form when we have a justiﬁed belief that this reason is true; the speaker thereby gives additional information. Explanatory reasons are subjective because they report the considerations and propositional attitudes from which the agent decided and acted and by which the action shall be explained – though the doxastic

An Empirical Theory of Practical Reasons

159

also an effort to explain for what kinds of mental attitudes the variable ‘explanatory reason’ empirically stands. 2. The decisive last mental step before action: intentions and what they are 4 The desire–belief theory of action holds that action arises from pairs of desires for some end and beliefs that a certain action will realize that end. Qualiﬁcations of this idea have been, for example, that actions arise from the strongest desire or the desire–belief pair with the highest product of desirability and probability and so on (cf. for example Davidson 1963, in particular 3–12; 1978, 87; Goldman 1970, 49–57, 72). The desire–belief theory is problematic because it tries to do without intentions, thus underrating their role which cannot be fulﬁlled by sets of desire–belief pairs (Bratman 1987): desire–belief pairs do not have the necessary resoluteness; the action “chosen” by them cannot be executed later on; they do not explain difﬁcult decision processes and so on. Therefore, the immediate and decisive mental cause of actions must be intentions. There is a host of theories about what propositional attitudes intentions are. The most important ones are: sui-generis theory, which says intentions are propositional attitudes in their own right, not reducible to other types of propositional attitudes (Bratman 1987, 10, 20, 110; Donagan 1987, 41, 81; Mele 1992, 127, 162); prevision theory, which maintains that intentions are – self-fulﬁlling – previsions of one’s own actions (Harman 1976; Velleman 1989a, 109 and 109–42; 1989b); optimality belief theory, according to which intentions are beliefs that among the considered set of open alternatives, one speciﬁc course of action is optimum (Lumer 2005);5 models of practical inference (Aristotle, NE 1111b–1113b, 1139a, 1147a; von Wright 1963; 1971, 96–107; 1972); psychological normativism, which holds that an action is caused by the belief that this action is socially required in the particular situation (Mead 1934, 152–64; a partial psychological normativism is included in Habermas’ concepts of ‘communicative action’ and ‘normatively regulated action’, cf.: Habermas 1981, 127f., 132–4, 143, 148–51, 385–7, 412, 418; Habermas 1975, 280–82); models of needs, presuming that we act out of our strongest or most developed need and a belief that a certain action will fulﬁl that need (Maslow 1954, chs 4–7). explanatory reasons aim at truth, that is they try to capture objective facts, which are the normative reasons. Explanatory reasons cannot be primarily objective just because the states of affairs in question are in part causal relations (and not events), happen in the future or are not facts and therefore cannot explain the action. (This ontology is neutral with respect to the question of what is foundationally primary, normative reasons, as externalism and in particular value objectivism holds, or explanatory reasons, as internalism maintains.) 4 This and the following section are based on: Lumer 2005. 5 Somewhat more precisely, the optimality belief theory holds the following. Forming an intention consists in an optimality judgement that a certain action is optimum among the considered set of open alternatives. And the intention is the resulting optimality belief. But they are intentions only up to the moment when the action is ﬁnished or, if it is never ﬁnished, when the prospected end of its execution has been reached or, if the person believes in deictic form, something with respect to these two dates (end of action, projected end of action), when these assumed moments have been reached. (Lumer 2005, sect. 4.)

160

Intentionality, Deliberation and Autonomy

The majority of these theories are doxastic theories of intention, holding that intentions are certain types of beliefs that the intended act token (I will perform an action of type A at time t) has a certain property F. They differ in their respective assumptions about that F. Prevision theory holds that F is empty (or equal to ‘is true’ – the intention then is the belief ‘I will do A at t’ or ‘It is true that I will do A at t’); psychological normativism holds that F is equal to ‘is socially required or prescribed’, and so on. Further doxastic theories about intention are the optimalitybelief theory, models of practical inference and perhaps (depending on the exact interpretation) models of need. Intentions have many features; in the following only important empirical features apt to falsify some theory of intention are considered. Doxastic theories of intention as such have some very important advantages in the respect that they can easily explain the following features. I1. Present and distal intentions: There are present and distal intentions. Doxastic theories explain this difference easily with the action description’s differing time indices. I2. Various logical forms of intentions: Intentions come in various logical forms; there are, for example, general, conditional or disjunctive intentions. Doxastic theories explain this by assuming that such intentions are of the form ‘all a (of that and that type) are F’, ‘if c then a is F’, ‘a is F or b is F’ and that we may arrive at elementary intentions via logical inferences. I3. Transition deliberation to intention: Usually there is an easy transition from the deliberation, which may precede the intention, to intending. Doxastic theories easily explain this by assuming that the deliberation in such cases consists in a cognitive process during which a true proposition of the intention’s form is sought and that the intention formation consists in, ﬁnally, after having found sufﬁcient reasons, to come to believe in the respective proposition. I4. Settling effect of intentions: Intentions, to a certain degree, settle what will be done and create a commitment to the execution,6 which, though, may eventually be withdrawn later. Doxastic theories explain this so: forming an intention, at least temporarily, stops the consideration of alternatives because after pondering various pros and cons the person ﬁnally comes to believe that a is F. And withdrawal of the commitment is due to ﬁnding new reasons that lead to a change in one’s opinion. I5. Small and big actions: Human intentions aim at a very wide spectrum of intended actions, for example from scratching one’s knee up to doing what one can to diminish hunger in the world. And often small actions are intended as parts of bigger actions. Doxastic theories obviously allow “small” as well as “big” actions to be the object to which the predicate ‘F’ may be applied. And they explain the transition from intending a big action to intending its

6 Bratman has emphasized this settling effect as a prerequisite for intrapersonal and interpersonal coordination (Bratman 1987, 4, 16f.).

An Empirical Theory of Practical Reasons

161

parts this way. There are several important candidates for ‘F’ for which holds: if F applies to a big action then F applies to all its (relevant) parts, too, so that the coming into existence of the small intentions can follow the paths of logic. However, apart from optimality belief theory, none of the other doxastic theories (prevision theory, psychological normativism, models of practical inference and models of needs) can explain a further highly important feature of intentions, namely this: I6. The weighing content of deliberation: The most important content of deliberation, which might precede the formation of intention, is to ponder the possible advantages and disadvantages of the considered options. Normal prevision requires an inference from beliefs about circumstances and general (statistical) laws; self-fulﬁlling predictive judgements, however, could be completely arbitrary, thus expressing freedom of the will and making our body follow this liberum arbitrium. Judgements about the social requirement of an action mostly are deontic conclusions from beliefs about some social standard and the given situation. And practical inferences (of the standard form) as well as beliefs about need satisfaction, ﬁnally, require deduction from the aim or need and a proposition that a certain action would fulﬁl them. In none of these cases is pondering pros and cons of importance. In addition, psychological normativism, models of practical inference and models of needs (but not prevision theory and again not the optimality belief theory) fail to explain the uniqueness of intended actions. I7. Deﬁniteness of intentions: In executive intentions we choose exactly one action as the one to be executed (or a well circumscribed set of actions about which the agent is undecided and between which he lets the executive system decide as a kind of random device). This deﬁniteness is necessary for the executive system to “know” what to do. All the restricting mechanisms appealed to by psychological normativism, models of practical inference and models of needs, that is, social norms, aims and needs, leave much room for various alternatives. On the other hand optimality belief theory can easily explain the latter two features of intention: I6. Pondering pros and cons of an isolated action means considering what makes this action better or worse, respectively. Considering the pros and cons of one action as compared to those of another action means trying to ﬁnd out which is the better one. I7. Uniqueness of optimality beliefs results from restricting the set of candidates of possible actions via the relation ‘is better than’. (‘a1 is optimum among the set of options {a1, a2, …, an}’ means: a1 is better than all the other options, that is a2, a3, …, an.) In case of options with utilities close to each other deﬁniteness may be reached by ﬁne-tuning the options’ valuation. The limiting cases then are Buridan examples where we end up with two (or more) options held to be exactly equally desirable. The relation ‘is better than’ does not help any longer in this situation. But instead of falsifying optimality belief theory, such cases even conﬁrm it because optimality belief theory explains the difﬁculties we actually have in real life with such situations; we still search for

Intentionality, Deliberation and Autonomy

162

the better action. (A way out for cases where we ultimately do not ﬁnd any value difference is to, ﬁnally, form a disjunctive intention and thus let the executive system decide between the weakly best options. From the intention system’s viewpoint this comes up to a random decision.) The currently most famous non-doxastic theory of intention is the sui-generis theory, which says that intentions are irreducible propositional attitudes. Apart from the fact that introducing a new propositional attitude seems to be somewhat ad hoc, such theories have some difﬁculties in explaining the above mentioned features I1, I2, I3, I5 and I6 of intentions. These difﬁculties are not insurmountable in all cases but every condition requires an extra explanation and an additional, complicating part of the theory – only Mele has offered such an explanation and only for features I3 and I6 (Mele 1992, 228–34). The sui-generis theory for example must hold that conditional intentions (cf. I2) are attitudes of type Φ with the propositional content ‘if p then I will do A at t’; and it must explain why, after coming to believe that p, the person develops an attitude of type Φ with the propositional content ‘I will do A at t’. There may be an explanation for this but it is one that has to be explicitly developed. Analogous difﬁculties arise with the features I1 (present and distal intentions) and I5 (small and big actions). But above all, the sui-generis model of intention does extremely badly with feature I3 (transition from deliberation to intention) and, consequently, with feature I6 (the weighing content of intentions). According to the sui-generis model, the intention is quite isolated from the deliberation. Mele has tried to resolve this problem by complementing the sui-generis model with a theory about the acquisition of intentions (Mele 1992, 228–41). According to this theory, the most important cause of intentions are optimality judgements about actions, which by default, that is in the absence of preemption, lead to the respective intention (ibid., 228–34). However, if this were so then the optimality belief itself could function as an intention, as is assumed by the optimality belief theory, and the sui-generis intention would be superﬂuous. Summarizing, we may conclude that if there are no further very important features of intentions or further promising explanations of intentions not yet considered, the optimality belief theory is the best explanatory theory of intention. H1: Intentions as optimality beliefs: Intentions are optimality beliefs that a certain action is optimum (in a sense to be speciﬁed) among the considered set of options. However, there are some well-known objections to the optimality belief theory of intention. Here I can deal only with the most important of them: the phenomenological objection.7 It says, considering the content of our deliberation and decisions introspectively we only rarely ﬁnd more or less explicit thoughts of the content ‘This (action, option) would be best.’ In particular tiny children, when they are just barely able to intend, probably do not dispose of the concepts ﬁguring essentially in optimality propositions (Robert Audi, personal communication). This problem of the interpretation of optimality propositions will be dealt with in the next section. 7

Some further objections are discussed in: Lumer 2005, sect. 4.

An Empirical Theory of Practical Reasons

163

3. Forming intentions: aggregating intrinsic valuations to optimality judgements The optimality belief theory introduced so far leaves open what kind of desirability (including desirability prospects) people try to maximize in deliberation and, accordingly, what concept of a ‘best’ or ‘optimum’ action people use in their intentions. The second part of the empirical theory of practical reasons to be outlined here deals with how the ﬁnal optimality judgement is aggregated from more basic valuations, thereby taking into account various consequences of the actions considered and the different probabilities of these consequences. But ﬁrst some terminology has to be explained. 1. I speak of ‘desirabilities’ instead of ‘utilities’ because the latter notion is bound to a very speciﬁc theory of empirical desirabilities, namely subjective expected utility theory. 2. Empirical desirabilities are the desirabilities people have in mind during deliberation. Ideally such desirabilities are quantitative. But usually people have only vague ideas about the respective quantity; they may perhaps use comparisons to other quantities but not numbers. These ideas about quantities, nonetheless, go beyond mere comparative judgements, that is, subjects can say not only that a1 is better than a2 but also how much better it is. 3. Apart from empirical desirabilities there are rational desirabilities of objects, which follow from empirical qualities of that object and the deﬁnition of ‘desirability’ given in rational or prudential desirability theories. The present part of this chapter deals with empirical desirabilities only. 4. The total desirability of an object a is the desirability of a under the condition that all its relevant consequences are known with certainty. The total desirability is equal to a’s intrinsic desirability plus the intrinsic desirabilities of a’s consequences. 5. The prospect desirability of an object a, on the other hand, can also be determined when not all relevant consequences of a are known with certainty. Apart from the intrinsic desirability of a itself, a’s prospect desirability includes weighted desirabilities of a’s (more or less probable) consequences. There are various ways of weighting possible consequences; the most prominent is to weight them according to their expected desirability. So expected desirability and total desirability are special cases of prospect desirability. The idea behind prospect desirabilities is to summarize everything that has some more basic desirability and probably accompanies a into one judgement about a. The rationale behind this is that in this way some sort of index is generated, saying how good the world accompanying a will be. The concept of ‘total desirability’ does the same for certain prospects. According to the meaning of “Action a1 is optimum (that is, has the highest prospect desirability) among the options {a1, a2, …, an} taken into consideration”, such judgements should be veriﬁed by, ﬁrst, assessing the single options, attributing to each of them a prospect desirability, and, second, by comparing the prospect desirabilities of all options to ﬁnd out which is highest. In practice, however, we ﬁnd little of this ideal procedure. There is no straightforward chronological order of the single steps, numbers are almost never used, and the procedure is greatly abridged. Empirical decision theory, which is mainly advanced by psychologists and economists, deals with the problem of aggregating more basic desirabilities to

164

Intentionality, Deliberation and Autonomy

decisions. There are three main approaches. They all say that decision consists in considering the various pros and cons of options, that is their expected positive or negative consequences, and in aggregating the respective probability assumptions and desirabilities of the consequences. But these theories differ in their hypotheses about how this is done exactly. The ﬁrst one is subjective expected utility theory, which says that we arrive at decisions as if they had been taken according to the advices of rational decision theory, that is, by calculating the expected utility of the options taken into consideration and choosing one of the options with the highest expected utility. Because rational decision theory is a normative theory it would be very astonishing if all persons complied with that theory. Actually psychologists have falsiﬁed subjective expected utility theory in many respects, especially since the late 1960s (for example Kahneman and Tversky 1979; overview: Camerer 1995, 622–6, 644–9, 652–70). It has been shown, for example, that the great majority of subjects does not weight risky prospects linearly, that is multiplying the desirability of the respective outcome with its probability. Highly probable but not certain prospects are undervalued, whereas prospects with little probabilities are overvalued. The second type of empirical decision theories are axiomatic theories of empirical desirabilities somewhat in the fashion of subjective expected utility theory but generalizing this approach in allowing for interpersonally differing ways of nonlinear weighting of not certain (that is, risky or uncertain) prospects. This means these theories provide only a general formula of risk weighting, and the exact weighting function for risky prospects, in general a non-linear function, has to be established for each person separately. The most prominent model of this kind is Kahneman’s and Tversky’s prospect theory (Kahneman and Tversky 1979; Tversky and Kahneman 1992) but there are many others (overview: Camerer 1995, 626–44). However, all the axiomatic or general theories of empirical desirabilities developed so far have two big problems. First, even though some of these theories are elegant and many of them mathematically rather ambitious, several studies have found that none of them is prognostically satisfying (cf. for example Harless and Camerer 1994; Currim and Sarin 1989). In normal cases they are not prognostically better than subjective expected utility theory. This holds even for prospect theory. Second, those axiomatic theories as well as subjective expected utility theory are hydraulic (and not cognitive) in the sense of not pretending to model what people actually think during deliberation but only to model their ﬁnal decision based on their assumptions about desirabilities and probabilities, which in some way and beyond conscious awareness then leads to the predicted decision. Most people would not even understand the formulas used for description. This means that these models do not reﬂect what goes on in conscious deliberation and which concept of desirability is used in the resulting optimality belief, which in turn is identical to the intention. The third approach in empirical decision theory is process tracing research, which with several techniques tries to ﬁnd out what people actually think, consider or calculate and which decision strategies they follow during deliberation.8 Thus far the one theory of the decision process has not come within reach, but the mentioned 8 Some important contributions of this type are, for example: Gigerenzer et al. 1999; Payne et al. 1993; Shugan 1980; Svenson 1996. Overview: Crozier and Ranyard 1997.

An Empirical Theory of Practical Reasons

165

type of research has provided a huge wealth of empirical ﬁndings, about very diverging decision strategies. One general idea crystallizes from these ﬁndings. The general aim of deliberation is to ﬁnd the best option – but with some qualiﬁcations. First, the aim of ﬁnding the best option does not require calculating in the ideal form described above, that is calculating the prospect desirability of each option and then comparing these values. There are many possible abbreviations, which do not even change the result. If, for example, option an+1 is equal to an in all but one respects and worse in that particular respect, then an+1 is worse than an; and if an+1 is found to be worse than an it cannot be optimum and its prospect desirability need not be calculated. Second, estimating prospect desirabilities may be more or less precise, depending on the inclusion of more or less important aspects of the object. A very rough deliberation in the end may regard only one important aspect of the option but still provide an approximation of the option’s prospect desirability. Third, options are more or less important. The cost of possible errors due to less precise assessments (that is, the costs of not having chosen the really best option but only the, let us say, third best option) diminish with the importance of the options. Fourth, deliberation is costly, and its costs have to be added to those of the action itself. On the other hand, deliberation may have (perhaps big) advantages, namely to lead to ﬁnding a better option. So we have a second problem of optimization here, that of optimizing deliberation. Although this kind of optimizing cannot be very precise (the attempt to follow it through would lead to more and more costly higher-order optimizations up to inﬁnite levels) it is possible to use rules of thumb that say how much one should deliberate in case of a certain sort of decision problem. Now at least adults are implicitly aware of all this and somehow take it into account in that they “decide” how to decide, that is, they change their decision strategies intrapersonally with the type of decision problem, in the end aiming at optimizing deliberation, thereby optimizing the executive action as well (cf. for example Payne et al. 1993; Shugan 1980; Svenson 1996). The observable (via process tracing methodologies) result of this underlying tendency is the huge and at ﬁrst sight confusing multitude of decision strategies used. The simplest strategy is to choose an option immediately when it comes to mind if it has at least one positive aspect and no (grossly) negative aspect comes to mind. This decision strategy is often used in real life, it is open even to tiny children, and in philosophy it has been formalized as practical syllogism. This implicit qualiﬁcation of the optimality belief theory should sufﬁce for rejecting the phenomenological objection that in deciding we usually do not think in terms of ‘optimality’ (cf. above, section 2). Now if behind the great variety of decision strategies used there is a tendency of optimizing deliberation, then the following must hold. During deliberation about executive actions people use many decision strategies implying many secondary criteria of the desirability and optimality of the options. These secondary criteria are not anthropologically ﬁxed; they are not even intrapersonally ﬁxed but have been cognitively developed. However, on a deeper level there seems to be one primary, that is, fundamental and most exact, criterion of the desirability of options with which the quality of the secondary criteria can be measured. The primary desirability may be a form of prospect desirability or it may be a form of total desirability. Of

166

Intentionality, Deliberation and Autonomy

course, people in most cases will not be able to formulate this primary criterion; they need to know it only implicitly. H2: Primary and secondary desirabilities: Humans use a wide variety of decision strategies implying different criteria of the desirability and optimality of alternatives. These strategies are not inborn but are developed in cognitive processes, and their use in different situations is evaluated and chosen according to optimizing deliberations that imply a fundamental criterion of desirability and optimality. If secondary desirability criteria are not anthropologically ﬁxed, at least primary criteria may be so. Thus far there has been no empirical research about this question. But the fact that even decision theorists, who should be able to formulate their most fundamental desirability criterion, strongly disagree about many aspects of such a criterion makes it quite unlikely that there is such an anthropologically ﬁxed primary desirability criterion. On the other hand the discussion among them is not simply intuitionistic and does not only consist of diverging statements but includes arguments. And it leads to new deﬁnitions of ‘desirability’ in a process of technical invention for solving certain problems. A weaker but still anthropologically universal hypothesis explaining these developments then is this: there is no universally accepted primary desirability criterion; but there are universally accepted adequacy conditions for deciding between proposals for fundamental desirability criteria. People do not explicitly dispose of such adequacy conditions; these conditions only show up implicitly when people decide between different proposals for the fundamental desirability criterion. Observing the proposals made during the discussion, one may formulate two adequacy conditions underlying them: H3: Adequacy conditions for fundamental desirability concepts: Humans use two fundamental or primary desirability concepts at the same time, a concept of total desirability for decisions under certainty and a concept of prospect desirability for decisions under risk or uncertainty. If humans choose between concepts (D1, D2, …, Dn) of fundamental desirability and they believe that among them D1 comes closest to fulﬁlling the following adequacy conditions whereas the others do not, they adopt D1 for their fundamental decisions. H3.1:Condition for ‘total desirability’ for decisions under certainty: For all events x and y with respect to which the person can decide under certainty holds: x (according to the person’s information) is totally better (in the sense of Di) than y iff the sum of the intrinsic desirabilities of all intrinsically relevant events accompanying event x is higher than the respective sum for y. H3.2:Condition for ‘prospect desirability’ for decisions without certainty: The desirability criterion Di in question is materially equivalent (that is, on the basis of the same information it leads to the same preferences) to that desirability criterion Dx (from the set of prospect desirability criteria, which deﬁne the prospect desirability of events only on the basis of the person’s empirical information and his intrinsic desirability function) for which holds:

An Empirical Theory of Practical Reasons

167

if one disregards decision costs, the constant use of Dx as the criterion for decisions without certainty is totally optimum (that is, optimum according to the fundamental desirability criterion for decisions under certainty). (Cf. Lumer 2005, 252.) This adequacy condition hypothesis can only rarely be used for prognostic purposes. But it nonetheless can be used for explaining some fundamental decisions people take. And, what is much more important, it can be used in theories of practical rationality for constructing such desirability criteria which are stable in view of criticism and of getting new information. Such stability then may be considered as a guarantee of their rationality. 4. Intrinsic valuations 4.1 Conceptual distinctions – what are originally intrinsic desires? The second part of the empirical theory of decision sketched here (section 3) dealt with the aggregation of more basic desirabilities of actions themselves and their consequences as well as information about the probabilities of such consequences to optimality judgements about actions in terms of ‘total’ or ‘prospect desirabilities’. The more basic desirabilities of consequences may again be prospect desirabilities or total desirabilities. But in order not to lead to an inﬁnite regress there must be some more basic desirabilities that are not prospect or total desirabilities (cf. Aristotle, NE 1094a18–22). These most basic desirabilities are intrinsic desirabilities, that is, desirabilities that we do not attribute to the object because of its consequences but because we are interested in some feature of the object itself.9 The more basic desirabilities for actions themselves are always immediately intrinsic – otherwise the more basic desirability would have to include again some of the action’s consequences, which already are considered separately. Intrinsic desirabilities express what we are interested in properly. Total and prospect desirabilities only summarize such intrinsic desirabilities for facilitating decisions. Even some prospect desirabilities are more basic than others in that the ﬁrst are the desirabilities of the latter’s consequences; so, for example the desirability of having food will usually be more basic than the desirability of having the money to buy it. However, even the desirability of having food is still a prospect desirability, and having food itself is probably intrinsically neutral. Assessing intrinsic desirabilities sometimes seems to be rather trivial; at least adults aggregate them (nearly) automatically to some rather basic prospect desirabilities. Therefore, most desirabilities, even the most basic ones, we are explicitly considering during deliberation are already prospect desirabilities. So most people do not even differentiate explicitly between intrinsic and prospect desirabilities. Nonetheless, we 9 In the last section I distinguished between primary and secondary desirability criteria. The desirability criteria meant there were always criteria for total or prospect desirabilities. Intrinsic desirability has only now been introduced. So, for example, the primary criteria for ‘total desirability’ should not be confused with those for ‘intrinsic desirability’.

168

Intentionality, Deliberation and Autonomy

can ﬁnd out people’s intrinsic desirabilities by asking them whether they do not think that they desire some object because of its consequences. And we may invite them to reﬂect if their valuation perhaps might change if they imagined for a moment that the object in question would not have the assumed consequences. Another way of ﬁnding out people’s intrinsic desirabilities is deep psychology, which historically reconstructs how people come to value certain objects in a speciﬁc way. Because these two methods may lead to different results we should distinguish between originally intrinsic desires / desirabilities, which should be the aim of deep psychological research, and (at time t) conscious intrinsic desires / desirabilities, which are those intrinsic desires that could be discovered by means of the ﬁrst method. We may ﬁnd conscious intrinsic desires that originally were prospect desires only. And in case of repression we may even ﬁnd originally intrinsic desires that presently have no conscious counterpart at all. The difference between intrinsic desirabilities and prospect desirabilities is not only mostly neglected in usual deliberation, even standard utility theory and rational decision theory do not make this distinction as they only look for coherence of unqualiﬁed utilities. But in this way the justiﬁcatory relations, that is the reasons behind our valuations, are disregarded, which is a serious omission (cf. Lumer 1998, 33–7). From the standpoint of a stronger rationality theory we should be very interested in an empirical theory about the originally intrinsic desires behind our prospect desires, which thus would clarify more deeply the possible ways of deciding. Unfortunately, there is a severe disproportion between the extensive psychological and economic research about our dealing with not certain prospects and the scarce psychological research about the content of (originally) intrinsic desires.10 The following theses are based on the latter material.11 4.2 Hedonic desires and weak psychological hedonism H4: Originally intrinsic desires: There are three types of originally intrinsic desires: 1. hedonic desires, 2. feeling-induced desires and 3. corrected hedonic desires. This hypothesis will now be explained and defended. Psychological hedonism is the thesis that all sorts of feelings of the agent, namely pleasant or unpleasant bodily feelings, emotions and moods, are the objects of his originally intrinsic desires and that they are valued according to the integral of their 10 An immediate problem with most of the respective psychological literature is that the expression “intrinsic (motive)” is not used in the philosophical sense. Some psychological results and empirical theories about intrinsic desires are presented in: Batson et al. 1983; Cialdini et al. 1987; Deci and Ryan 1985; Heckhausen 1989, 455–66 (as well as chs 8–12 and 14); Maslow 1954; McClelland et al. 1989; Morillo 1992; Sansone and Harackiewicz 2000; Waterman 1990. 11 The following theory of originally intrinsic desires has been developed in: Lumer 2000, 428–521. In particular about feeling-induced intrinsic desires see: Lumer 1997; Lumer 2000, 477–93; about corrected hedonic desires see: Lumer 2000, 493–521.

An Empirical Theory of Practical Reasons

169

intensity over time, where this intensity has a positive sign for pleasant feelings and a negative sign for unpleasant feelings. Strong psychological hedonism claims that the agent’s feelings are exactly the objects of his originally intrinsic desires; weak psychological hedonism holds that beyond the agent’s feelings there may be other objects in his originally intrinsic desirability function. Hypothesis H4 is a version of weak psychological hedonism, the ﬁrst part making it a form of hedonism and the second and third part, which allow for non-hedonic originally intrinsic desires, making it a form of weak hedonism. Weak psychological hedonism seems to be a truism. Its explanatory power has been proved quite often. The real point of dispute is only if also strong hedonism is true or if there are originally intrinsic desires beyond hedonic ones and, if so, what their nature is. Therefore, only the second and third part of that thesis shall be substantiated here. 4.3 Feeling-induced intrinsic desires All originally intrinsic desires different from hedonic desires, in particular feelinginduced desires, are always in danger of being explained away by some hedonistic reduction. So making a case for the former desires a – rather neutral – paradigm case is helpful where a hedonistic explanation is implausible. Think of a three- or four-year-old little girl who has been provoked by her elder brother (ten years old), and being really furious with him she grabs his hand and bites his forearm, right above his wristwatch with all her strength. Surely our girl knows what she is doing – biting her brother’s forearm – and she will have known in advance that this action will “damage” or “destroy” the aggressor, probably even that it will injure him and that he will suffer. And she has chosen well the point where to bite him: in a place where her mouth has enough grip, and not on the wristwatch, where she would hurt herself. Damaging or destroying the aggressor seems to be the (desired) aim of the action, and there seems to be no other ulterior aim. But that the aggressor is damaged or destroyed is no feeling of the agent so that we have a nonhedonic originally intrinsic desire. Afterwards the girl will be satisﬁed in a crude moral way. But it seems to be too farfetched to suppose that such a little girl already knows about such hedonic consequences of her acting out of rage; at least when she acts out of rage the ﬁrst time she cannot have the empirical knowledge about these hedonic consequences, but must acquire this knowledge and perhaps some time, perhaps some years, later may even have the intrinsic hedonic aim of being morally satisﬁed. The general explanation for this type of intrinsic desire runs as follows. Emotions like rage are caused by some classiﬁcation, for example: ‘This was an (unjustiﬁed) aggression by x (somebody else) against me (or somebody else one sympathizes with)’, where the aggressor is neither too strong nor too insigniﬁcant. This classiﬁcation implicitly is a valuation, that is, the classiﬁcatory attributes are meant to be positive or negative. Such classiﬁcations and valuations with respect to their content are not hedonic; but they are not yet motivational; they are only affective in the sense that they cause certain emotions. So they are not the intrinsic value judgements with motivational force we are looking for. However, every type of emotion is linked with another, ‘satisfying’ type of emotion in the way that tokens of the ﬁrst emotion seem

170

Intentionality, Deliberation and Autonomy

to aim at tokens of the second, satisfying emotion. Rage has (moral) satisfaction as its satisfying match, happiness has attachment as its satisfying match, fear has relief as its satisfying match, and so on. These matches again have their own affective evaluative classiﬁcations as their causes. Let us call the original type of emotion “E”, the predicate used in its underlying affective valuation “VE”, the type of matching, satisfying emotion “M” and the predicate used in the underlying valuation “VM”. The mechanism behind emotion-induced intrinsic desires then is this: H5: Emotion-induced intrinsic desires: If somebody is in the emotional state E and during her emotional arousal classiﬁes some consequence or outcome o of a potential action a of herself as fulﬁlling the value criterion VM of the matching emotion M (that is, she believes: o is VM) then she (originally) intrinsically desires outcome o in proportion to the strength of the inducing emotion of type E. In our example the girl is in a rage (E); she considers the option of biting her brother (a) with the consequence that he will suffer (o); the matching satisfying emotion is (moral) satisfaction (M); the affective valuation criterion for causing this (moral) satisfaction is ‘x is a just punishment or, a bit more primitive, a damaging or destruction of some villain or aggressor’ (VM); and the girl judges ‘that my brother will suffer terribly is a just punishment of this villain’ (VM(o)). This judgement in a certain sense already coincides with the (positive) motivational intrinsic desire to make the brother suffer; and the strength of that desire corresponds to the intensity of our girl’s rage (E). Psychologists have tried to ﬁnd out if acts of compassion might be explained roughly along these same lines (Manucia et al. 1984; replication and extension by: Cialdini et al. 1987; cf. also: Batson et al. 1983). They tried to exclude that their subjects could aim at positive hedonic effects of their helping. The results were, ﬁrst, that emotions were necessary for helping and, second, that subjects helped independently of the fact that they could or could not expect pleasant feelings from their help. These results do not conﬁrm the exact hypothesis given above. But they conﬁrm at least the more general hypothesis that in case of acts of compassion there may be non-hedonic intrinsic desires to help other people. Up to now we have considered only emotionally induced desires. There seems to be a similar phenomenon in bodily feelings and in moods as well, that is, that these feeling states, provided they are strong enough, induce originally intrinsic desires as well. Here I can only mention this generalization of the theory of feeling-induced desires. Feeling-induced desires seem to be an evolutionary older motivational system than the normal hedonic motivational system. And it is a much more primitive and less rational motivational system than the latter. First, emotionally induced intrinsic desires aim at rather immediate changes of the exterior situation. This change often will be beneﬁcial to the person; otherwise this motivational system would not have survived evolution. But there is no room for changing these aims if they are not beneﬁcial or if there are possibilities of advancing the (long-term) well-being of the person much more. Second, feeling-induced desires are bound to current feelings

An Empirical Theory of Practical Reasons

171

so that they change rapidly over time. This implies that they are not apt as a basis for long-term planning and long-term decisions. This distinguishes them sharply from hedonic desires, which are stable over time: if I know that I might suffer from hunger or anxiety one year from now I intrinsically disapprove of such feelings now, one year later or whenever. And the stability of such valuation is the basis for the fact that I can currently plan to avoid such feelings which otherwise will arise only very much later. So feeling-induced desires are not suitable as a basis for a rational desirability function. Third, feeling-induced desires cause important forms of weakness of will. We have formed a rational (in the sense just explained) judgement that a is (not) optimum, but when it comes to (not) doing a feelinginduced desires lead to a divergent emotional optimality judgment, which causes the emotionally best action. 4.4 Corrected hedonic desires The other type of not classically hedonic but nonetheless originally intrinsic desires is corrected hedonic desires. These desires may come into play in cases of Nozick’s experience machine12 or in cases of improving one’s feeling states by drugs (and not going the “normal” way, that is, doing something fulﬁlling or rewarding) or in other cases of manipulation or alteration of one’s feelings. Nozick introduced his thought experiment in order to defeat hedonism. To a certain degree he was successful. Most persons discount positive feelings that have been induced by some manipulation, but not negative feelings. This means they correct their hedonic valuations. They continue to give intrinsic value (in a cool mood) only to feelings; in this respect they remain hedonists. But they correct the attributed desirabilities, thus becoming corrected hedonists. People I have asked about their reasons for this correction could not really give an answer; the corrected valuation seemed to be a question of taste. And this indicates that the valuation remains intrinsic. It is prompted or triggered by the reﬂection about the possibility of different sorts of manipulation; but the new way of valuing is not justiﬁed by this reﬂection. H6: Corrected hedonism: People who reﬂect on the manipulation of their feelings tend to become corrected hedonists. In their intrinsic hedonic desires they discount manipulated positive feelings according to the degree of manipulation, whereas they continue to desire negative feelings according to the assumed negative integral of the feeling’s intensity over time.

12 This machine, provided by some friendly extraterrestrials, can be programmed to make you experience whatever you like if you enter the machine for ever; having entered the machine you will not be aware that everything that you now experience is ﬁction, so to speak. The machine is programmed in such a way that the experiences you will have in it will be signiﬁcantly, though not tremendously better in hedonic terms than what you would experience in real life. Would you enter the machine? (Nozick 1989, ch. 10.)

172

Intentionality, Deliberation and Autonomy

Some philosophers have suggested further originally intrinsic desires, like the desires for autonomy, virtue and love.13 But these desires may be explained rather easily in a hedonistic way, and the authors of these proposals have not provided evidence that would refute such an explanation. In addition their proposals remain ad hoc and isolated in the sense of not being integrated in a complete theory of intrinsic desires or motives and of desires in general – as has been sketched here. 4.5 Dynamics of intrinsic desires Originally intrinsic valuations rely on general but usually unconscious ‘criteria’, that is, they are not executed arbitrarily and differing from case to case but do show some pattern. So people value as if applying a criterion, which normally is not conscious. Such criteria for originally intrinsic valuation are inborn; people do not construct them through reﬂection – though they may be triggered by reﬂection, as in the case of corrected hedonic desires; and they are rather ﬁxed. This does not imply that the criteria for intrinsic valuations are static. In what has been said so far some dynamics of this kind has already been hinted at. There are (at least) ﬁve kinds of changes of criteria for intrinsic valuations.14 1. Cognitive development: Criteria for originally intrinsic valuations may be changed due to some cognitive reﬂection. This reﬂection does not really provide a reason for the new criterion, it only stimulates or triggers the change – otherwise the new criterion probably would be only a criterion for prospect desirabilities. The only example for such a change found so far is the development from simple to corrected hedonism. Such changes represent real cognitive development in that the new criterion is ‘based’ on a more extensive cognitive background, though they are not justiﬁed but only triggered by these new cognitions. For a rational desirability theory it would be interesting to discover the perhaps ﬁnal level of such cognitive developments. Corrected hedonism probably is such a ﬁnal level. 2. Autonomization of prospect desirabilities: People use many particular secondary instrumentally justiﬁed criteria for prospect desirabilities. Sometimes they forget this genesis and justiﬁcation and take the originally prospect desirabilities to be intrinsic desirabilities in the sense that they no longer can give a reason for their respective desires. Whereas cognitive development of criteria for intrinsic valuation is a cognitive progress, such autonomization of originally prospect desirabilities to consciously intrinsic desirabilities is a sort of step back because it is based on forgetting.

13 Among others, intrinsic desires for the following objects have been proposed in the literature: autonomy (Young 1982), personal expressiveness, that is, living in accordance with one’s true self (Waterman 1990), virtue and love (Davis 1977), intellectual consistency, psychological maturity and moral rectitude (Nelson and Wilker 1975), perfection of character (Tatarkiewicz 1947/1962, ch. 24), among other things having the right contact with reality, wisdom (Nozick 1989, chs 10, 11 and 23). 14 For the dynamics of criteria for intrinsic valuation see: Lumer 2000, 205–18.

An Empirical Theory of Practical Reasons

173

3. Dissolution of autonomized desirabilities: However this regression may be caught up with. If people learn about the genesis of their consciously intrinsic desirability criterion they tend no longer to use it as such. Either they keep on holding the old (or a better) justiﬁcation for this criterion to be valid; then they continue to use it, but once again (as originally) as a criterion for prospect desirabilities. Or they ﬁnd the old justiﬁcation to be fallacious and see no better substitute for it; then they tend to stop using the criterion. 4. Repression of originally intrinsic desirability criteria: If some sorts of originally intrinsically good objects (for example sexual pleasure or pleasure resulting from aggressive behaviour, which leads to punishment) over a longer period generally have strong negative consequences, people may repress their respective intrinsic desires: they value the respective objects as prospectively negative and ‘forget’ their positive intrinsic component. Such repression is a cognitive step back, too, because it is based on forgetting and because it tends to overrate the negativity of the respective objects (because their positive component is no longer available). 5. Recovering the repressed: However this kind of regression may be caught up with, too, by restoring the repressed intrinsic valuation to people’s consciousness. H7: Dynamics of intrinsic desires: Intrinsic desirability functions can change as a consequence of the following processes: cognitive development, autonomization of prospect desirabilities, dissolution of autonomized desirabilities, repression of originally intrinsic desirability criteria and recovering the repressed. An important consequence of these mechanisms of changing intrinsic desirability function is that more information tends to lead to a stable state with maximally cognitively developed intrinsic desirability functions, where all perhaps at some time autonomized desirability criteria are dissolved and all perhaps at some time repressed originally intrinsic desirability criteria are reinstalled. 5. Motives for moral action Completing all the details of the picture outlined so far is not possible here. But some further details of particular interest for practical philosophy shall be sketched in this section. They regard motives for moral action. Actions in accordance with the requirements of morals may be motivated in rather arbitrary ways. But there are some motives substantially connected to acting morally, in particular the following.15 H8: Moral motives and motives for moral action: The following motives are substantially connected to acting morally. H8.1:Self-interested cooperation: Informed prudential self-interest in reciprocal cooperation can lead to respecting another person’s interests.

15 For a more extensive analysis of motives for moral action see: Lumer 2002.

174

Intentionality, Deliberation and Autonomy

H8.2:Sympathy: Sympathy is an emotion and not a motive. But it can lead to moral action in two different ways. 1. Being an emotion, sympathy can induce an intrinsic desire to improve another person’s situation. However, such desires are not temporally stable (cf. above, section 4.3). 2. Pity is a negative, unpleasant emotion; it has a positive, pleasant counterpart, a sort of satisfaction about someone else’s positive well-being. Being pleasant or unpleasant, sympathy falls in the domain of hedonism, and one can try to optimize one’s sympathy: one can change the world in such a way as to have more occasions for feeling positive sympathy and fewer occasions for feeling negative sympathy. H8.3:In-group solidarity: A certain form of solidarity, which may be called “ingroup solidarity”, is based on one’s identiﬁcation with a certain group. This leads to augmenting the salience of and the interest in the fate of in-group members. This in turn can focus sympathy, which in principle is universal, as to be dedicated to a stronger degree to the proper in-group. The most ambitious and universal identiﬁcation is the identiﬁcation with mankind; less ambitious identiﬁcations can, for example, be nationalistic or have the proletariat or all oppressed people as their scope. H8.4:Respect: One form of respect is emotions out of a group of emotions like admiration, reverence, fascination, being deeply moved, which are caused by an appraising classiﬁcation of a certain object, a person, animal, nature, artefact and so forth, as highly developed, elaborated or ingenious and so on, and fragile. Like sympathy, respect can lead to moral action in two ways. 1. These emotions can induce a motivation out of respect, namely a positive intrinsic desire of the respective object’s existence, which in turn may lead to appropriate action like protecting this object. 2. On the other hand one may seek to optimize respect emotions by creating occasions for positive respect emotions by means of protecting and supporting the object. H8.5:Self-transcendent motives: There is a group of self-transcendent emotions like love, affect, liking, creative self-expansion and pride in the proper community or culture. All these self-transcendent emotions arise from a deep connection to a rather particular object. And they can induce motivation to protect the respective object and to make it ﬂourish. H8.6:Moral motives: Finally, there are moral motives in the narrow sense, that is motives originating from some moral judgement and translating it, so to speak, into moral practice. 1. There are, once again, emotion-induced moral motives, originating from feelings of guilt or indignation or outrage and which motivate to punishing or, more rarely, to repairing. 2. On the other hand there is a hedonist version of moral motives, namely to raise one’s moral selfesteem (and to prevent its being lowered) by acting morally and then positively appraising one’s deeds or character. Positive self-esteem after all is (or, in the mere cognitive meaning of “self-esteem”, results in) a pleasant feeling, and negative self-esteem is an unpleasant one. Therefore, we can be hedonistically motivated to optimize our moral self-esteem by acting morally.

An Empirical Theory of Practical Reasons

175

6. Why normative theories of practical reason presuppose empirical theories of practical reasons Having sketched the empirical theory of practical reasons, we can now draw some normative consequences. The central thesis underlying the second part of this chapter is: normative theories of practical reason, in particular the theory of practical rationality and ethics, presuppose empirical theories of practical reason in the sense that the former must be based on rather strong empirical information about the psychological laws governing our deliberations and decisions. This thesis holds for the following reasons. The theory of practical rationality is normative in the broad sense; more precisely it makes recommendations for deciding and acting rationally. Ethics on the other hand is (in part) normative in the strict sense, but it is also normative in this broad sense; it makes recommendations for deciding and acting morally. Now reasonable recommendations must fulﬁl the following adequacy conditions. A1: Realizability: The recommendation must be realizable. A2: Selectivity: The recommended course may not be such as to be followed necessarily anyway (for example due to some law of nature); that is, the recommended course has to be selected from a possible range of open courses. Otherwise the recommendation would be void. A3: Goodness: The recommended course must use the open spectrum of different ways of deciding and courses of action in a good manner so that the recommendation is appealing to the addressee. Otherwise there would be no reason to follow the recommendation. Expressed slightly differently: the proposals must be designed in such a way as to make people, with the help of some argument, inclined to follow these proposals. These are only necessary adequacy conditions implying something about the empirical underpinnings of normative theories of practical reasons. Here is not the place for developing sufﬁcient adequacy conditions. Now, fulﬁlling these adequacy conditions requires precise empirical information about our ways of deciding. The argument for this claim is straightforward. Ad A1: The recommendations do not regard (executive) action only but also our way of deciding. Therefore, we must not only be able to act in the required way (where “to be able to do a” is meant in the conditional sense, that is as: if I intend to a I will do a); we must also be able to follow the suggested decision rules (this holds for the theory of practical rationality) and to arrive at the required intention (this holds for normative ethics in the narrow sense). And for checking if the addressees of our theories are able to do so, and for bringing it about that they really do so we must know how people decide and, in particular, how their ways of deciding react on which input. Ad A2: For fulﬁlling the selectivity condition we must know which features of people’s ways of deciding are necessary and where there are options for deciding differently. Ad A3: For making good recommendations for our ways of deciding and for the content of our intentions we must at least know which options

176

Intentionality, Deliberation and Autonomy

people have for deciding differently and what the consequences of these options are. – Note that these arguments do not depend on an internalist or externalist conception of normative practical reasons. The ﬁrst parts of each of the following two sections give some examples of how normative theories of practical reason fail because they are not at all or not sufﬁciently based on empirical theories of practical reasons and therefore violate these adequacy conditions. The second parts of the following two sections, on the other hand, show how, on the basis of the sketched empirical decision theory, a satisfying ethics and theory of practical rationality that satisﬁes the adequacy conditions can be developed. 7. Consequences for the theory of practical rationality A ﬁrst implication of the above sketched empirical theory for non-moral practical philosophy is that it tends to exclude libertarian views of free will in that it shows how decisions are subjected to empirical laws (H1–H8).16 On the other hand this theory opens the way to a form of freedom consisting in deciding rationally and maximizing the individual good. Another interesting implication is that, according to H1, optimality beliefs simultaneously have two directions of ﬁt: as beliefs they try to ﬁt to the world, that is, to be true and to identify the really best action (in subjective terms, though); as intentions they aim at making the world ﬁt to the intended course of action, that is, to the action held to be optimum. A third important implication regards the question of instrumentalism: can actions be rationally criticized only with respect to the means used or also with respect to their (ultimate) ends? On the one hand the originally intrinsic desirability functions described in subsections 4.2–4.4 and thus the possible originally ultimate ends are ﬁxed. On the other hand the above sketched dynamics of intrinsic desires (subsection 4.5) leaves some room for rationally criticizing actual intrinsic desires: it is possible that actual intrinsic desires are not originally intrinsic, not cognitively maximally developed or temporally unstable; in all these cases they are unstable with respect to acquiring new information. A rational theory of ultimate ends that tries to fulﬁl adequacy condition A1 (realizability) can use this room but is bound to its limits. In the rest of this section some consequences for the theory of practical rationality will be developed more systematically. 7.1 Criticisms of rational decision theory and full information accounts of rationality Rational decision theory on the lines of a von Neumann/Morgenstern-like axiomatization does not fulﬁl the adequacy conditions A1–A3. In particular its recommendations are not realizable (cf. A1). Nobody fulﬁls the axioms of these theories (Kahneman and Tversky 1979, 263–73; Camerer 1995, 622–6, 644–70; cf.

16 Of course, the hypotheses sketched above, as usually in psychology, at best report very high frequencies and thus cannot exclude undetermined decisions altogether.

An Empirical Theory of Practical Reasons

177

above, H2). But fulﬁlling the axioms is a necessary condition for subjective utilities to be deﬁned. Therefore, the subjective (expected) utilities for the respective person are not deﬁned, and consequently it is impossible to calculate expected utilities of the options at hand. In addition, even if one could calculate them in principle, in most cases it would be practically impossible or irrational to calculate them because this would be too time-consuming (cf. A3). This means rational decision theory as such does not tackle the problem of deciding how to decide (cf. H2 and H3) and does not take the problem of scarce decision resources behind H2 seriously enough. Finally, the theory is pragmatically useless. It states only what a rational system of preferences and decisions looks like. But it does not give us procedural rules on how to arrive at that ideal ﬁnal state. Even in this pragmatic sense the recommendations of that theory are not realizable (cf. A1). But many philosophers as well as some economists too have criticized rational decision theory from a quite different point of view that aims at completing rather than replacing it. They say that rational decision theory’s standards of rationality are too low. In particular, they say that it is irrational to accept people’s unﬁltered preferences uncritically, even if they are coherent, as a basis for deﬁning prudential desirabilities. After all these preferences may be based on false information (Harsanyi 1982, 55f.). This criticism implies that rational decision theory does not recommend the best way of deciding, thereby violating again A3. The full information approach to practical rationality seems to be the most obvious and insurmountably rational reply to the latter criticism.17 Roughly, this approach holds that a decision is rational and represents the choice of the best option if it would be the result of a decision made with complete (relevant) and vividly represented information. Full information approaches to practical rationality and desirability face severe problems too. First, we never have full information and even if we had we could not represent it all vividly at the same time. Therefore we can never check empirically which of two options is prudentially better (according to full information theory). And the approach does not provide the means for ﬁnding out which alternative would be preferred. Thus A1 is violated.18 Second, in a certain sense full information approaches are rationally insurmountable because the person, at least hypothetically, always knows everything that is relevant. However this is exactly the problem. If the person really had all the empirical information she still would not know what to do with it, which way of deciding to follow. So one has to add the information about the best way of deciding to the full empirical information. But 17 Richard Brandt (1979, 10–13, 70–88, 110–29) has developed the most important proposal of this kind. Another full information approach is, for example: Grifﬁn 1986. 18 To be fair it must be recognized that Brandt has undertaken particular efforts, more than any other theoretician of practical rationality in the second half of the twentieth century, to insert psychological information into his rational model for ﬁnding out what the result of a fully informed decision would be (Brandt 1979, 24–148). However, ﬁrst, due to its very nature, the late behaviouristic psychology he used is more than imprecise and cannot capture and explain what cognitive processes lead to our decisions. And second, these psychological laws can tell us only what the effect of various stimuli would be but nothing about the response on full and vividly imagined information. So Brandt’s assumptions about the results of rational decisions are rather speculative.

178

Intentionality, Deliberation and Autonomy

this would be precisely the information that the theory of practical rationality should provide. So the full information approach presupposes that the person hypothetically has solved the general problems of rationality theory, which even the theoretician himself has not been able to solve so far. This is asking a bit too much. Therefore the realizability condition (A1) is violated again. So even the full information approach (like rational decision theory) is pragmatically useless (cf. A3) since it does not give procedural recommendations on how to arrive at rational valuations. And one reason for this is that the proposals of this approach refer to the empirical outcome of certain decisions but do not make sufﬁcient use of the empirical information about our ways of deciding. 7.2 Reconstructive prudentialism An approach that avoids these problems and nonetheless reaches far beyond the rationality level of rational decision theory is a reconstructive desirability theory, which relies heavily on empirical information about our decision-making. The basic idea of this approach is to discover the empirically possible ways of deciding and of establishing desirabilities, and to recommend that way as the way for calculating prudential desirabilities that remains stable in view of new knowledge (and perhaps fulﬁls some further conditions). More precisely, this approach gets along with the following steps. First, it has to be reconstructed how people really decide, or more exactly, which possible ways of deliberation and deciding exist and under what conditions they are taken. Second, it has to be found out which of these ways is stable in view of new knowledge and which fulﬁls further requirements like being stable in other respects. Exactly this way then deﬁnes prudential ‘desirability’. (Thus all three adequacy conditions, A1–A3, are fulﬁlled.) So this deﬁnition has taken over from the agents only the way of deciding and only the wisest way of deciding (wisest in the sense of relying on all relevant structural information so that stability is reached) and thus the purely conative part of their deliberation. The empirical information necessary for ﬁnally establishing the (real) desirability of a given object must be objectively true and can be provided by whoever has it; so the empirical information is inserted from the outside. In this way among others the overﬂow of information in full information approaches is avoided (cf. A1). Third, based on this prudential deﬁnition of ‘desirability’, more convenient (that is, easier to apply or requiring less information) secondary criteria for ‘desirability’ can be developed, which then should be used in rational decision (cf. A3). Obviously, this prudential and reconstructive approach requires very strong empirical information about our ways of deliberation, that is, an empirical theory of practical reasons. The respective theory presented above (sections 2–5) roughly leads to a desirability theory of the following type (detailed exposition: Lumer 2000, 241–427, 521–48). 1. The gross structure is anthropologically ﬁxed and consists of three parts: intrinsic desirabilities (H4–H7), prospect desirabilities, deﬁned in terms of intrinsic desirabilities (H2–H3), and optimization (H1). 2. If we apply the criterion of stability in view of further information, and if we consider what has been said about the dynamics of intrinsic desirability criteria (H7), only original criteria of ‘intrinsic desirability’ are eligible as prudential criteria of ‘intrinsic desirability’

An Empirical Theory of Practical Reasons

179

and among these only the cognitively most developed ones (if there is any cognitive development at all). The only empirical criteria fulﬁlling these conditions are corrected hedonism and the criteria inherent in feeling-induced intrinsic desires. The latter desires strongly depend on the underlying feelings (H5), however, and thus are completely unstable over time and do not permit any long-term planning. Therefore, for formal reasons of rationality the latter group should also be excluded as the basis of prudential intrinsic desirabilities. So we end up with corrected hedonism as the prudential criterion of intrinsic desirabilities (cf. H6). 3. The fundamental deﬁnitions of ‘total desirability’ and ‘prospect desirability’ must follow the lines set out in the adequacy condition hypothesis about fundamental prospect desirability concepts (H3). 4. An empirical theory of deliberation, ﬁnally, is heuristically very rich in providing a huge wealth of secondary criteria of prospect desirability, which then should be scrutinized as to if and under which conditions their application is optimum according to the fundamental criteria. This theory of prudential desirability fulﬁls the three adequacy conditions A1–A3. One might suspect that it is impossible to practically neglect the criteria of ‘intrinsic desirability’ that have been discarded as irrational, so that A1, realizability, would be violated. But this is not true. Autonomized prospect desirabilities, which have become conscious intrinsic desirabilities, can be dissolved by confronting people with the history of these intrinsic desirabilities. Simple hedonism may be overcome by confronting people with thought experiments of the experience machine type. And feeling-induced desirabilities can be made ineffective by delaying one’s decision until the respective feelings have cooled off a bit. (This actually is what folk wisdom advises: do not decide in the heat of passion because you may regret your decision later!) It is rather obvious that the recommendations of reconstructive desirability theory are not followed anyway (A2) because it proposes to follow precisely one of the many possible ways of deciding. It may be difﬁcult to prove that the recommendations of this theory are good (A3) and as good as possible among the recommendations of competing theories that also fulﬁl conditions A1 and A2 because such theories just deﬁne ‘good’ and so on. Nonetheless the suggested theory has some clear advantages. First, ﬁltering out ways of deciding that are unstable in view of further knowledge or during the course of time prevents false investments, that is investments in projects whose success later is regarded to be vain or even harmful. Second, stability in view of further information guarantees something like wisdom and seeks to satisfy ideals of enlightenment. 8. Consequences for ethics What are the consequences of the above sketched empirical theory of practical reasons for ethics? A straightforward example for critical consequences of this theory is utilitarianism. The standard interpretation of utilitarianism obliges us always to maximize moral desirability. It does not take into account that, aside from their moral motives, people have much stronger self-interested motives (cf. H4– H6), which make it impossible for them to obey that obligation. So condition A1 (realizability) is violated. In order to fulﬁl adequacy condition A1, moral obligation

180

Intentionality, Deliberation and Autonomy

has to be designed on the basis of information about the strength of our moral motivation and about other motives that might be apt to support the fulﬁlment of moral obligations. Another consequence of the empirical theory outlined above may be that the axiological character of practical reasons assumed by it (cf. H1– H3) strongly favours an axiological, for example welfarist, ethics as opposed to deontological or virtue ethics. Deontologism, for example, seems to presuppose a decision psychology of “acting according to principles / subjective maxims”, where the will is always acting according to the representation of a law (cf. Kant, GMS BA 15, 36, 63). Of course, humans can do this – but always on the basis of respective desires (cf. H2–H4). So, given humans’ axiological character, deontologism will always have difﬁculties in making us motivationally accept deontologically justiﬁed duties (cf. A1).19 These criticisms are rather simple. Consider a more complex example, foundational externalism. 8.1 Critique of foundational externalism Most ethicists accept what, introducing some distinction here, may be called “effective internalism”, that is the thesis that a sound moral justiﬁcation must motivate rational or prudent persons to follow that moral to at least some degree – which does not exclude that in the end there may be much stronger motives to the contrary. (Some ethicists, like Brink (1989) or Schaber (1997), do not accept effective internalism. But if their ethics do not fulﬁl the requirements of effective internalism, that is, if they are not motivating, their ethics have little practical sense; even as theorists we can ignore them because people in their practical decisions will do so as well.) Effective internalism is an implication of adequacy condition A3 (goodness): the ethics proposed must be practically appealing. Foundational internalism is a stronger thesis: it accepts effective internalism and claims that for fulﬁlling the motivational requirement moral justiﬁcations have to rely on motives that precede these justiﬁcations (Williams 1979). (Foundational externalism on the other hand claims that the motivational requirements of effective internalism can be fulﬁlled independently of such pre-existing motives. In a certain sense the justiﬁcation should create the right motivation.) Kant, according to these deﬁnitions, is an effective internalist because he not only wants people to follow the morality he justiﬁes but he also wants them to follow it for the right motives (cf. for example Kant, GMS BA X, 26). At the same time he is a foundational externalist. His basic idea is that pure reason has to establish what is morally right independently of any motivational inclinations of men and, therefore, independently of empirical theories as introduced above. Only in a second step the insights of pure reason shall motivate to action, thus fulﬁlling the requirements of effective internalism. (Kant, GMS BA 33, BA 36f., BA 63f.; KpV A 56–8, A 126– 8.) So Kant, like today’s foundationally externalist ethicist, denies the relevance of empirical theories of practical reasons for the foundational tasks of ethics altogether. 19 This argument and several other respects in which axiological ethics ﬁt better to human decision psychology are elaborated in: Lumer 2004b.

An Empirical Theory of Practical Reasons

181

However, for fulﬁlling the requirements of effective internalism, empirical information like the one introduced above is indispensable because motivating for action is a causal and, according to a Humean type of causality, contingent empirical relation. In other words: Kant’s original project, to ﬁnd out moral laws by pure reason and then to let the respective insights alone determine action, is even analytically impossible (cf. A1). This holds because to determine action or to motivate to act is a causal relation, and causal relations depend on respective empirical laws. According to the idea of pure practical reason, very speciﬁc cognitions about an action, say that this action is F, should cause the execution of that action, whereas different cognitions, that the action is G, H and so on, should not have this effect. But if this is to happen then there have to be psychic mechanisms, describable by means of psychological laws, bringing about that exactly cognitions of this kind, namely that an action is F, lead to the action’s execution. Now, which of the many possible qualities F, G, H and so on of an action recognizable by pure reason exactly is relevant for executing the respective action is ﬁxed by those mechanisms – and not by pure reason. Pure reason cannot decree the practical relevance of its cognitions. (Lumer 2002/03, 269–71.)20 What has been said so far does not exclude a more down to earth version of foundational externalism in ethics, though. There may be motivational mechanisms that translate moral judgements into some respective motivation rather independently of the speciﬁc content of the moral judgement. According to this hypothesis, a moral conception may be justiﬁed independently of motivating reasons; and if someone is convinced of this conception and of the truth of certain moral judgements about an action then that mechanism induces some motivation to act accordingly. Differences with respect to what Kant intended would be, ﬁrst, that the whole construction would depend on the ‘contingent’ existence of such an empirical mechanism in humans, which also would determine the kind of cognition to which we would be motivationally responsive, and, second, that this mechanism also would limit to a certain degree the spectrum of moral conceptions that might be motivationally supported. But such a mechanism coincides exactly with what above (H8.6) has been called “moral motives” (in the narrow sense). The prudentially (and also in moral practice) most important moral motive is raising one’s moral self-esteem (H8.6.2). However, whether seeking positive self-esteem, ﬁnally, is a sufﬁcient basis for foundational externalism in ethics still depends on the moral neutrality of this mechanism. Does this mechanism support (within certain limits) whatever moral conceptions so that pure reason would be free to ﬁnd out the right morals? The answer to this question now depends on what moral conceptions we can rationally adopt in such a manner that the adoption leads to inﬂuencing our self-esteem in the right way. This is a question regarding the psychology of moral emotions, which is beyond the scope of this paper. But the fact that emotions, too, are caused by very speciﬁc and anthropologically ﬁxed judgements makes it unlikely that our 20 Similar criticisms can be advanced against all versions of value objectivism and moral realism whose values are not bound to psychic mechanisms and which nonetheless hold that recognition of the value object generates motivation to act in favour of these objects.

182

Intentionality, Deliberation and Autonomy

moral self-esteem should be open to whatever moral conceptions, as foundational externalism requires (cf. A1).21 8.2. The basis of foundational internalism Foundationally internalist approaches in ethics assume that, for providing understandable reasons and for being able to motivate, justiﬁcations of morals must be based on (in a certain way) pre-existing rational motives, which later on shall also be the reasons for following the justiﬁed morals. The above sketched theory of motives for moral action (H8) has presented the candidates that may serve as such a basis. But moral justiﬁcation and motivation should be practically rational, which includes stability in view of new knowledge. Not all motives for moral action listed in H8 fulﬁl this requirement and thus are not suitable as the basis of an internalist justiﬁcation of morals.22 Sympathy-induced motives (H8.2.1) and motives induced by respect (H8.4.1) as well as emotion-induced moral motives (H8.6.1) are feelinginduced motives, hence unstable over time. Self-transcendent motives (H8.5) may function as the basis of an individual morality, but they are too idiosyncratic for being the basis of a social and universalistic morality. All moral motives in the narrow sense (H8.6), including those not induced by emotions, ﬁnally, are not apt as such a basis either because they already originate from moral judgements; therefore, a justiﬁcation strategy based on these motives would lead to a vicious circle. All the motives mentioned so far play more or less important roles as moral motivation ampliﬁers, but they are not apt to provide the initial moral “signal”. The remaining motives for moral action of the list in H8.6, on the other hand, can serve as justifying motives in an internalist foundation of morals, and have mostly been used as such in the history of philosophy. Based on a prudential and informed self-interest in cooperation (H8.1) a weak business morality, without altruism and redistribution of utilities, can be justiﬁed – as it has been, for example, by Gauthier (1986). Sympathy optimization (H8.2.2) may be the basis of a universal morality of benevolence as British moral sentimentalists intended it, whereas the locus classicus for a sympathybased morality is Schopenhauer. In-group solidarity (H8.3) is only a mechanism for altering the extent of sympathy, in particular increasing sympathetic motivation in favour of the members of one’s in-group. This feature is apt to justify hierarchical pathocentrism for example. Optimizing respect emotions (H8.4.2), ﬁnally, may be the basis for justifying a prohibitive morality of respect, such as the principle ‘Do not harm anybody!’ or an ethics of conservation of nature and culture. 21 The ontogenetic beginning of morality is heteronomous in that morality is taken over from the socialization agents and thus justiﬁed by authority. But later, in particular beginning with adolescence, people may not only systematize their morals and question the parts that they ﬁnd unjustiﬁed, they also bring in new autonomous sources of morals, which can then become and remain stable bases of their morals. These autonomous sources of morals seem to be exactly those motives, however, that internalist justiﬁcations of morals are based upon: rational self-interest in cooperation, sympathy and respect. (Lumer 2002, sect. 7; Lumer 2002/03, sect. 8.1.) 22 For a more extensive justiﬁcation of the following assessments of the motives for moral action see: Lumer 2002, 171–82.

An Empirical Theory of Practical Reasons

183

Because all the latter motives can be used as a basis for justifying morals, the question of how to integrate all these motives into one morality or how to exclude some of them arises. But this question is beyond the scope of this chapter. However, the empirical theory of practical reasons sketched here could already be shown to be useful for an internalist approach to ethics in individuating the motives that are suitable as a basis for such a foundational task. Working out this ethics up to precise moral criteria then requires much more scrutiny of the exact features of the respective motives. For example the exact limits of rational self-interest in cooperation will also deﬁne the limits of the business morality, or the exact features of our sympathy will also deﬁne the degree and distribution of benevolence in a sympathy based ethics of benevolence and so on.23 This means, in practical philosophy we still need much more of an empirical theory of practical reasons than could be outlined here.24 References Batson, C. Daniel; Karen O’Quinn; Jim Fultz; Mary Vanderplas and Alice M. Isen (1983), ‘Inﬂuence of Self-Reported Distress and Empathy on Egoistic Versus Altruistic Motivation to Help’, Journal of Personality and Social Psychology, 45, 706–18. Brandt, Richard B. (1979), A Theory of the Good and the Right, Oxford: Clarendon. Bratman, Michael E. (1987), Intention, Plans, and Practical Reason, Cambridge, MA and London: Harvard University Press. Brink, David Owen (1989), Moral Realism and the Foundation of Ethics, Cambridge: Cambridge University Press. Camerer, Colin (1995), ‘Individual Decision Making’, in John H. Kagel and Alvin E. Roth (eds), The Handbook of Experimental Economics, Princeton, NJ: Princeton University Press, pp. 587–703. Cialdini, Robert B.; Mark Schaller; Donald Houlihan; Kevin Arps; Jim Fultz and Arthur L. Beaman (1987), ‘Empathy-Based Helping: Is It Selﬂessly or Selﬁshly Motivated?’, Journal of Personality and Social Psychology, 52, 749–58. Crozier, Ray and Rob Ranyard (1997), ‘Cognitive Process Models and Explanations of Decision Making’ in Rob Ranyard, W. Ray Crozier and Ola Svenson (eds), Decision Making: Cognitive Models and Explanations, Oxford: Routledge, pp. 3–20. Currim, Imran S. and Rakesh K. Sarin (1989), ‘Prospect versus Utility’, Management Science, 35, 22–41. Dancy, Jonathan (2000), Practical Reality, Oxford: Oxford University Press. Davidson, Donald (1963), ‘Actions, Reasons, and Causes’, in Donald Davidson, Essays on Actions and Events, Oxford: Oxford University Press, 1980, pp. 3–19.

23 An attempt to establish the exact features of our sympathy and to draw (prioritarian) ethical consequences from them is: Lumer 2000, 589–632. 24 I would like to thank several people for discussions of this chapter, in particular: Robert Audi, Michael Bratman, Carl Ginet, Hugh McCann and Alfred Mele.

184

Intentionality, Deliberation and Autonomy

—— (1978), ‘Intending’, in Donald Davidson, Essays on Actions and Events, Oxford: Oxford University Press, 1980, pp. 83–102. Davis, William H. (1977), ‘Transcendent Needs’, Philosophy Today, 21, 184–93. Deci, Edward L. and Richard M. Ryan (1985), Intrinsic Motivation and SelfDetermination in Human Behavior, New York and London: Plenum. Donagan, Alan (1987), Choice: The Essential Element in Human Action, London and New York: Routledge & Kegan Paul. Gauthier, David (1986), Morals by Agreement, Oxford: Clarendon. Gigerenzer, Gerd; Peter M. Todd and the ABC Research Group (1999), Simple Heuristics That Make Us Smart, Oxford [etc.]: Oxford University Press. Goldman, Alvin I. (1970), A Theory of Human Action, Englewood Cliffs, NJ: Prentice-Hall. Grifﬁn, James (1986), Well-being: Its Meaning, Measurement, and Moral Importance, Oxford: Clarendon. Habermas, Jürgen (1975), ‘Handlungen, Operationen, körperliche Bewegungen’, in Jürgen Habermas, Vorstudien und Ergänzungen zur Theorie des kommunikativen Handelns, Frankfurt am Main: Suhrkamp, 1984, 273–306. —— (1981), Theorie des kommunikativen Handelns. Vol. 1: Handlungsrationalität und gesellschaftliche Rationalisierung, Frankfurt am Main: Suhrkamp. Harless, David W. and Colin F. Camerer (1994), ‘The Predictive Utility of Generalized Expected Utility Theories’, Econometrica, 62, 1251–89. Harman, Gilbert (1976), ‘Practical Reasoning’, Review of Metaphysics, 29, 431–63. Harsanyi, John C. (1982/<1977>), ‘Morality and the Theory of Rational Behaviour’, in Amartya Sen and Bernard Williams (eds), Utilitarianism and Beyond, Cambridge: Cambridge University Press, pp. 39–62. Heckhausen, Heinz (1989), Motivation und Handeln, 2nd edn, Berlin [etc.]: Springer. Kahneman, Daniel and Amos Tversky (1979), ‘Prospect Theory: An Analysis of Decision under Risk’, Econometrica, 47, 263–91. Lumer, Christoph (1997), ‘The Content of Originally Intrinsic Desires and of Intrinsic Motivation’, Acta analytica – philosophy and psychology, 18, 107–21. —— (1998), ‘Which Preferences Shall Be the Basis of Rational Decision?’, in Christoph Fehige and Ulla Wessels (eds), Preferences, Berlin and New York: de Gruyter, pp. 33–56. —— (2000), Rationaler Altruismus. Eine prudentielle Theorie der Rationalität und des Altruismus, Osnabrück: Universitätsverlag Rasch. —— (2002), ‘Motive zu moralischem Handeln’, Analyse & Kritik, 24, 163–88. —— (2002/03), ‘Kantischer Externalismus und Motive zu moralischem Handeln’ Conceptus, 35, 263–86. Abridged version at <www.gap-im-netz.de/gap4Konf/ Proceedings4/Proc.htm>. —— (2004a), ‘Desires and Explanatory Reasons’, in Roland Bluhm and Christian Nimtz (eds), Selected Papers Contributed to the Sections of GAP.5, Fifth International Congress of the Society for Analytical Philosophy, Bielefeld, 22–26 September 2003, (CD-ROM), Paderborn: Mentis, pp. 704–14.

An Empirical Theory of Practical Reasons

185

—— (2004b), ‘Vom Primat der Werte – Wertethik versus Pﬂicht- und Tugendethik’, in Reinhold Mokrosch and Elk Franke (eds), Wertethik und Werterziehung, Göttingen: Vandenhoeck & Ruprecht unipress, pp. 39–61. —— (2005): ‘Intentions Are Optimality Beliefs – But Optimizing What?’, Erkenntnis, 62, 235–62. Manucia, G.K.; D.J. Baumann and R.B. Cialdini (1984), ‘Mood Inﬂuences on Helping: Direct Effects or Side Effects?’, Journal of Personality and Social Psychology, 46, 357–64. Maslow, Abraham H. (1954), Motivation and Personality, 2nd edn, New York: Harper and Row, 2nd edn 1970. McClelland, David; Richard Koestner and Joel Weinberger (1989), ‘How Do SelfAttributed and Implicit Motives Differ?’, Psychological Review, 96, 690–702. Mead, George Herbert (1934), Mind, Self and Society: From the Standpoint of a Social Behaviorist, ed. Charles W. Morris, Chicago and London: University of Chicago Press, 18th edn 1972. Mele, Alfred R. (1992), Springs of Action: Understanding Intentional Behavior, New York and Oxford: Oxford University Press. Morillo, Carolyn R. (1992), ‘Reward Event Systems: Reconceptualizing the Explanatory Roles of Motivation Desire and Pleasure’, Philosophical Psychology, 5, 7–32. Nagel, Thomas (1970), The Possibility of Altruism, Oxford: Clarendon. Nelson, Jack and David Wilker (1975), ‘Pleasure and the Intrinsically Desired’, Analysis, 35, 152–9. Nozick, Robert (1989), The Examined Life: Philosophical Meditations, New York [etc.]: Simon & Schuster. Payne, John W.; James R. Bettman and Eric J. Johnson (1993), The Adaptive Decision Maker, Cambridge: Cambridge University Press. Sansone, Carol and Judith M. Harackiewicz (eds) (2000), Intrinsic and Extrinsic Motivation: The Search for Optimal Motivation and Performance, San Diego, CA: Academic Press. Schaber, Peter (1997), Moralischer Realismus, Freiburg and München: Alber. Searle, John R. (2001), Rationality in Action, Cambridge, MA and London: MIT Press. Shugan, S.M. (1980), ‘The Cost of Thinking’, Journal of Consumer Research, 7, 99–111. Smith, Michael (1987), ‘The Humean Theory of Motivation’, Mind, 96, 36–61. Svenson, Ola (1996), ‘Decision Making and the Search for Fundamental Psychological Regularities: What Can Be Learned from a Process Perspective?’, Organizational Behavior and Human Decision Processes, 65, 252–67. Tatarkiewicz, Wladyslaw (1947/1962), Analysis of Happiness, transl. from the Polish by Edward Rothert and Danuta Zieliuskn, The Hague: Nijhoff [etc.], 1976. Tversky, Amos and Daniel Kahneman (1992), ‘Advances in Prospect Theory: Cumulative Representation of Uncertainty’, Journal of Risk and Uncertainty, 5, 297–323. Velleman, J. David (1989a), Practical Reﬂection, Princeton, NJ: Princeton University Press.

186

Intentionality, Deliberation and Autonomy

—— (1989b), ‘Epistemic Freedom’, Paciﬁc Philosophical Quarterly, 70, 73–97. Von Wright, Georg Henrik (1963), ‘Practical Inference’, The Philosophical Review, 72, 159–79. —— (1971), Explanation and Understanding, London: Routledge & Kegan Paul. —— (1972), ‘On So-Called Practical Inference’, Acta Sociologica, 15, 39–53. Waterman, Alan S. (1990), ‘Personal Expressiveness: Philosophical and Psychological Foundations’, The Journal of Mind and Behavior, 11, 47–74. Williams, Bernard (1979), ‘Internal and External Reasons’, in Ross Harrsion (ed.), Rational Action: Studies in Philosophy and Social Science, Cambridge and London: Cambridge University Press, pp. 17–28. Reprinted in Bernard Williams, Moral Luck: Philosophical Papers 1973–1980, Cambridge: Cambridge University Press, 1981, pp. 101–13. Young, Robert (1982), ‘The Value of Autonomy’, Philosophical Quarterly, 32, 35–44.

Chapter 9

Anchors for Deliberation Michael E. Bratman

This chapter sketches a model of deliberation that is anchored in planlike commitments of the agent, commitments that that constitute a form of valuing. These anchors need not be inescapable, they can sensibly vary from person to person, they can stand in complex relations to judgments about the good, and they play basic roles in the crosstemporal organization of practical thought and action. And deliberation so understood is, I conjecture, central to autonomy and self-government. The model sketched here is located in the space between Frankfurtian appeals to volitional necessity and Platonic appeals to judgments about the best. It sees anchors for deliberation as valuings that have agential authority, are potentially revocable, can diverge across rational agents, and bear complex relations to judgments about the best. Their reasonable stability involves defeasible presumptions both against reconsideration and against revision, presumptions characteristic of planlike commitments and supported by considerations of cross-temporal instrumentality, integrity and autonomy. And this support for stability by connections to cross-temporal organization goes in part by way of the connection between cross-temporal organization and agential authority. 1. Deliberation, agential authority and the problem of stability The aim of this chapter is to sketch a model of deliberation that is anchored in certain plan-like commitments of the agent. These anchors need not be inescapable or necessary, they can sensibly vary from person to person, they can stand in complex relations to judgments about the good, and they play basic roles in the cross-temporal organization of practical thought and action. And such deliberation is, I conjecture, central to autonomy and self-government. Let us begin with a problem about deliberation. We normally think of deliberation as a form of reasoning that is directed by the agent; and we normally think that such agential direction is central to self-government. But not all reasoning processes are directed by the agent in the strong sense that is relevant here. A way to see this is to reﬂect on an idea, introduced by Allan Gibbard, that there are cases in which one’s thinking is “in the grip of a norm”. One of Gibbard’s examples involves

188

Intentionality, Deliberation and Autonomy

the well-known Milgram experiments.1 As I understand this example, some of the participants act as a result of attenuated pieces of practical reasoning, reasoning that involves thoughts like “I guess I should do what the person in charge tells me to do …”. There is such reasoning even though the participants are not fully behind it but are, rather, in its grip. Again, I might be in the grip of normative thoughts about revenge, along the lines of “That guy deserves to pay!”; or I might be in the grip of wistful thoughts in favour of procrastination along the lines of “It would be nice to put off to tomorrow what I could do today.” In such cases action may be the issue of a process of practical reasoning, but this is reasoning that holds the agent in its grip, rather than reasoning that the agent herself is fully behind and directs. The conclusion I draw from this is not that we need to appeal irreducibly to the agent as cause. The conclusion I draw is, rather, that if deliberation is to be directed by the agent it needs to be guided by commitments that speak for that agent, that constitute, as Harry Frankfurt puts it, “where … the person himself stands” (Frankfurt 1988d, 166; and see Velleman 2000c). So we need to ask: what makes it the case that certain attitudes have authority to speak for the agent, have “agential authority”?2 A model of strong forms of deliberation needs an account of the agential authority of basic attitudes that guide the deliberation. In several earlier essays I sketched three main ideas in an effort to develop an account of such agential authority.3 First, normal human agents are temporally persisting agents whose agency is temporally extended. Their temporally extended agency normally involves overlapping, semantically interlocking, causally connected, coordinated strands of practical thought and action. Practical reasoning and deliberation involve references to and continuities with earlier and later thought and action; and intentional action manifests and involves continuities with and referential connections to earlier and later practical thought and action. These various overlapping, causally connected and semantically interlocking webs of thought and action normally realize a signiﬁcant degree of coordination: earlier thought and action prepares the way for later thought and action, which continues and completes earlier plans and activities. And these cross-temporal continuities and connections are among the cross-temporal ties that a broadly Lockean approach sees as partly constituting the identity of that very person over time.4 Second, once we have this picture of temporally extended agency it seems plausible to say that what gives certain attitudes agential authority – what makes them not merely another “wiggle in the psychic stew”5 – is, in large part, their role in constituting and supporting such Lockean, identity-constituting ties in such overlapping, causally connected and interlocking quilts of temporally extended agency. And to this we can add two qualiﬁcations. First, for an attitude to have 1 Gibbard 1990, 58–60. I also discuss this idea of Gibbard’s in Bratman 2007c, 79–80. 2 For this terminology see Bratman 2007d. 3 See esp. Bratman 2007b; 2007d; 2007h; 2007i; 2007j. 4 Talk of continuities and connections comes from Parﬁt (1984). I note differences in my understanding of these ideas in Bratman 2007b, 29–30. My thinking about these Lockean ideas has beneﬁted greatly from discussions with Gideon Yaffe. 5 As I say in Bratman 2007b, 24.

Anchors for Deliberation

189

authority to speak for the agent in the deliberative contexts at issue here – contexts of, in particular, self-government – the agent will need to know, at least in broad outlines, of its role in structuring her agency (see Bratman 2007h). Second (and drawing on an idea from Frankfurt), agential authority of a sort characteristic of selfgovernment requires as well that one is “satisﬁed” with such attitudes in the sense, roughly, that there are not in one’s psychic economy certain kinds of pressures that disrupt their normal functioning, perhaps by way of leading to a change in those attitudes (Frankfurt 1999b, 103–5; and Bratman 2007b, 34–5, 44). Call this the Lockean model of agential authority. The third main idea is that certain policies about what to treat as having justifying signiﬁcance in one’s motivationally effective practical reasoning will normally satisfy these conditions of agential authority. This is because such policies will normally help constitute and support the cited Lockean dimensions of the crosstemporal organization of practical thought and action. Further, such policies will normally be in part higher-order: they will be, at least in part, policies about whether to treat certain desires, and/or what they are for, as having (or lacking) justiﬁcatory force in deliberation. Such (as I call them) self-governing policies will have (assuming satisfaction and self-knowledge) a signiﬁcant claim to agential authority.6 Further, given their role in shaping what is given justifying weight in deliberation, such policies constitute a kind of valuing. Valuing of this kind is different from, and related in complex ways to, inter-subjectively accountable value judgment. We will return to this point below; but for now let us just say that such valuings will frequently go beyond value judgment in settling on speciﬁc contours of the agent’s life in the face of signiﬁcant under-determination by value judgment (see Bratman 2007g, 165–7; 2007f). So we have our three ideas: a conception of our temporally extended agency, a Lockean model of agential authority, and an account of self-governing policies as valuings that have a claim to agential authority. Taken together, these ideas allow us to model agentially directed deliberation as deliberation that is guided by authoritative valuings, authoritative self-governing policies.7 These valuings have agential authority in large part because of their role in constituting and supporting Lockean aspects of the cross-temporal organization of practical thought and action. And these valuings help specify what is to be given justifying signiﬁcance in the agent’s deliberation. This is how I would try to provide for agential direction of deliberation. But now we need to address the worry that, for all that has stated above, such valuings may be too unstable to serve as anchors for deliberation that is characteristic of 6 I explain why I think the self-governing policies central to autonomy will normally be higher-order in Bratman 2007g. I also think that the self-governing policies that are central to autonomy will involve a reﬂexive element: they will, in part, support their own role in deliberation. But here I put this complexity to one side. Concerning such reﬂexivity see Bratman 2007h. Robert Nozick emphasizes a related idea of a “self-subsuming decision that bestows weights” (Nozick 1981, 294–306; quote from 300). I discuss this and other related ideas of Nozick in Bratman 2007e. 7 Or quasi-policies. See Bratman 2007b, 42–4. In my discussion in this essay I bracket this complexity.

190

Intentionality, Deliberation and Autonomy

self-government. Self-government is not a mere time-slice phenomenon. At least in its clearest cases it involves an agent who persists over time and governs practical thought and action in ways that cohere and make sense over time. The present proposal is that the deliberation characteristic of self-government is anchored in valuings that are kinds of policies, kinds of general intentions. And the worry is that an agent might simply change such policies at will: for all that has been said so far, an agent’s valuings – understood as his relevant self-governing policies – could themselves change over time abruptly and willy-nilly. But self-government involves deliberation that is grounded in elements that are more stable over time than this. So it is not clear that we have yet an adequate model of a kind of deliberation that is central to autonomy and self-government. 2. Volitional necessity and stability Indeed, a closely related concern about instability has led Harry Frankfurt to the idea that the deliberation central to autonomy must involve “volitional necessities” (Frankfurt 1999c, 110; see also Frankfurt 1988c). Frankfurt develops the idea of volitional necessity in several places. This is one representative remark: Unless a person makes choices within restrictions from which he cannot escape by merely choosing to do so, the notion of self-direction, of autonomy, cannot ﬁnd a grip. Someone free of all such restrictions is so vacant of identiﬁable and stable volitional tendencies and constraints that he cannot deliberate or make decisions in any conscientious way. (Frankfurt 1999c, 110)8

In this passage Frankfurt’s concern is with what is needed for “the notion of selfdirection, of autonomy” to “ﬁnd a grip”. But there is a distinction between self-direction and autonomy: autonomy involves self-direction, but it also involves, further, rational guidance – guidance by what the agent treats as reasons (see Bratman 2007g, 177).9 So let us focus speciﬁcally on autonomy – what I will also call self-government. Frankfurt’s picture is that autonomy requires “conscientious” deliberation, and that such deliberation needs to be anchored in volitional necessities. Let us take it for granted that Frankfurt is right in connecting autonomy and “conscientious” deliberation. As for what makes deliberation “conscientious”: I will 8 Frankfurt may have other reasons for thinking forms of volitional necessity are important in our practical lives. Here my concern is only with the idea that such necessity is needed for “the notion of … autonomy … [to] ﬁnd a grip”. I also address this concern in Bratman 2006. Let me also note that some of Frankfurt’s earlier work suggests that what is needed for autonomy is, rather, identiﬁcation, where identiﬁcation does not entail volitional necessity (see for example Frankfurt 1999b). The criticism I go on to make in the text is only offered of the claim that volitional necessity is needed for autonomy. Indeed, the approach I go on to defend here is to a signiﬁcant extent in the spirit of the earlier Frankfurtian view. 9 As I see it, Frankfurt’s early appeals to higher-order volitions in favour of what is to motivate one’s action – where motivation need not involve normative content – are more plausible as models of self-direction than of self-government (see Frankfurt 1988b).

Anchors for Deliberation

191

return brieﬂy to this below. For now, let us drop the qualiﬁer and simply ask what is needed for deliberation to be appropriately anchored. Frankfurt’s answer is volitional necessity. One of his examples is a parent’s love for his children, a love that it is not in the parent’s direct voluntary control simply to abandon (Frankfurt 1999d, 136). Frankfurt’s discussion of exactly what is involved in such volitional necessity is complex. He sees this phenomenon as tied to certain issues about personal identity (ibid., 139, n. 8; and see Velleman 2002); and he insists that volitional necessity is not solely an incapacity of the will, but involves as well an unwillingness to change this incapacity (Frankfurt 1988c, at 87; Frankfurt 1999c, 111–12; and see Watson 2002). But let us here bracket these complexities and work directly with the unanalysed, core idea of a volitional incapacity: this is the idea of an incapacity to escape certain volitional “restrictions” – for example certain demands on one’s thought and action imposed by one’s love for one’s children – “by merely choosing to do so”. How does this Frankfurtian view compare to a model that sees the relevant deliberation as anchored in valuings that have Lockean agential authority? I think there is here a signiﬁcant disagreement, though it can be characterized in two different ways. To express these two alternatives let us take the liberty of assuming that some appeal to Lockean ties is common ground, though Frankfurt might not agree. The important point here, however, is that even given this common ground there are two different Frankfurtian objections that might be made. A Frankfurtian might claim that agential authority requires not merely the cited connections with Lockean ties, in a context of satisfaction and self-knowledge, but also volitional necessity. Or a Frankfurtian might claim (more plausibly, I think) that deliberation characteristic of autonomy requires not only that the central, guiding attitudes have agential authority, but also that they be volitionally necessary. The important point is that on either view volitional necessity is needed for relevant deliberation – deliberation characteristic of autonomy – to be appropriately anchored. And that seems to be an overly strong claim about the conditions for such deliberation. When I take my commitment to scholarship as given for my deliberation I do not see myself as incapable of abandoning my commitment. And, indeed, I do seem capable of abandoning it. Nevertheless, I do not think that I should abandon it; and this thought seems quite reasonable. Such a commitment on my part normally seems enough to anchor my deliberation, even in cases of self-government. So it seems that there are common and important cases of deliberation, of a sort characteristic of self-government, whose anchors are not volitionally necessary. That said, it is not obvious what to put in the place of volitional necessity in a model of self-government. According to our alternative model, what is needed are policy-like valuings that have Lockean agential authority in a context of selfknowledge, where such authority does not require volitional necessity. But the current concern about this alternative model is that an attitude can have such agential authority and still not be sufﬁciently stable. The worry is that there is – at least within the theory as so far developed – nothing to ensure that an agent does not keep shifting, over time, from one system of attitudes with agential authority to another quite different system that newly has such authority. But such an agent would, to return to Frankfurt’s remark, be “so vacant of identiﬁable and stable volitional tendencies and

192

Intentionality, Deliberation and Autonomy

constraints that he cannot deliberate or make decisions in any conscientious way”. Or so someone might object. Is that right? A preliminary point is that this statement of the worry does not do full justice to connections between Lockean agential authority and the de facto tendency towards stability. One aspect of constituting and supporting Lockean, cross-temporal continuities and connections is normally a tendency to be stable over time. Further, agential authority involves satisfaction, and thus the absence of certain psychic pressures for change even given a capacity for change. So the objection is to some extent overstated. Nevertheless, it might still be suggested that the de facto tendency towards stability of valuings that is needed for Lockean agential authority is not yet enough to block the Frankfurtian worry about the need for “stable volitional tendencies and constraints”. As an initial response let me propose an end-run: we simply add to our model that there is in fact the kind of tendency towards stability that is purportedly needed, within this framework, for appropriately anchored deliberation (and so for autonomy). It can remain an open question whether the kind of de facto tendency towards stability that is needed to anchor the relevant kind of deliberation is ensured by agential authority. What the current proposal must say, however, is that this tendency towards stability of authoritative attitudes is not an incapacity, and so not an instance of volitional necessity. So we have an alternative to the Frankfurtian view under consideration: anchored deliberation, of a sort characteristic of self-government, involves valuings with Lockean agential authority that have an appropriate tendency towards stability. And an attitude’s relevant tendency towards stability is to be distinguished from an incapacity to abandon that attitude. What should we say about this alternative view? 3. Reﬂection and stability We need to clarify the problem we are trying to solve. To begin we can think of the idea of adequately anchored deliberation as a kind of design speciﬁcation (cf. Velleman 2000b, 11). We want to characterize a psychic economy whose functioning would satisfy this design speciﬁcation: the system would in fact be capable of, and systematically engage in, adequately anchored deliberation. We want to know whether a psychic economy that satisﬁed the cited conditions of authority and stability, as well as other conditions we are implicitly taking for granted here, would satisfy the design speciﬁcation of adequately anchored deliberation. Now, at ﬁrst glance it seems plausible that we can specify the relevant tendency towards stability so that the answer to this question is “yes”. It seems plausible that we can specify the relevant tendency towards stability so that the dynamics of a psychic economy that satisﬁes our conditions of agential authority and stability would in fact exhibit sufﬁciently stable “volitional tendencies” for relevant deliberation to proceed apace. And it seems plausible that we can do this without introducing Frankfurtian incapacities. After all, stable tendencies are one thing, incapacities another.

Anchors for Deliberation

193

But this does not yet address a problem about the agent’s reﬂective deliberation. Perhaps a system that moved along over time in accordance with the conditions we have described would count as a deliberating system. Indeed, we might build an AI (artiﬁcial intelligence) system along these lines, and plausibly describe it as a deliberating system (see Bratman, Israel and Pollack 1991). But we can still go on to ask whether, if the agent is to understand herself as deliberating in a way that is adequately anchored, it sufﬁces for her to see her relevant guiding commitments as having both agential authority and a relevant tendency towards stability. And here the worry is that authority plus a mere de facto tendency towards stability seems not to be enough. The agent might recognize that certain practical attitudes of hers in fact have both agential authority and such a tendency towards stability, but still, upon reﬂection, wonder whether or not to stick with them or, rather, abandon them in favour of alternative practical attitudes that would then newly come to have agential authority.10 If she thought, along Frankfurtian lines, that the relevant attitudes were volitionally necessary for her, this question would not have practical force, at least with respect to the present status of these attitudes.11 But if she were instead to suppose that it really is up to her whether or not to stick with these attitudes, the absence of a persuasive afﬁrmative answer to her question would threaten to undermine the tendency towards stability. This suggests that we have not yet provided a model of adequately anchored reﬂective deliberation. But the deliberation characteristic of self-government will normally be reﬂective. Further, it seems clear that Frankfurt’s concern is with reﬂective deliberation – indeed, I propose that we interpret his talk of “conscientious” deliberation as building in this reﬂective character. So we do not yet have an adequate alternative to Frankfurt’s account. If we are not at this point to follow Frankfurt in appealing to volitional necessity or incapacity, we need to say that not only is there a de facto tendency towards stability, but there is also rational pressure in favour of stability of relevant authoritative attitudes. We need to appeal not only to de facto stability but also to norms of stability.12 This would allow us to say that the reﬂective agent could, if pressed, appeal – at least on some occasions – to this rational pressure to stick with certain authoritative valuings and thereby sensibly continue to deliberate in reﬂective ways that are guided by those valuings. But why think there is such rational pressure towards stability? And, if there is, what is its nature? 10 This is a special case of a general theme emphasized by Christine M. Korsgaard (1996). 11 Though Frankfurt does allow that even if an attitude is volitionally necessary now, one may be able to take steps now that would change that attitude at some later time. So the appeal to volitional necessity may not block all versions of the cited practical question. 12 My thinking about stability owes much to Gilbert Harman’s appeal to the inertia of intention in his 1976 paper, “Practical Reasoning” (reprinted as Harman 1999b, in particular 62–3). In this essay, though, Harman’s notion of inertia seems to be primarily a descriptive notion, whereas I want to emphasize norms of stability. Harman’s later appeal to a principle of conservatism as applied to intentions does, however, begin to address this normative concern (see Harman 1986, 46). Richard Holton also emphasizes the need for such norms in Holton 1999.

Intentionality, Deliberation and Autonomy

194

4. Why not the best? A broadly Platonic answer is that rational pressure in favour of stability of one’s valuing of X (if such there be) comes from a judgment that X is strictly better than its alternatives.13 But I think that this proposal would not adequately explain the rational pressure towards stability: if this were the story of rational stability then Frankfurt’s worry about “identiﬁable and stable volitional tendencies and constraints” would remain in force in a wide range of cases of reﬂective deliberation. The basic problem is that in many cases the relevant alternatives are not located by the agent within an evaluative ordering in which one is judged by that agent to be strictly better than the others. A life of scholarship, a life of working for the American Friends Service Committee, a life of teaching junior high school mathematics, a life of academic administration – we may suppose that these are all judged by the agent to be potentially good lives. One might perhaps try to weave at least several of them into a single life; but to keep discussion manageable let us assume that each precludes the other. Suppose, in particular, that the agent is now committed to a life of scholarship. In the sense of valuing noted earlier, she highly values such a life and does not similarly value competing ways of life. If there is rational pressure against her giving up this commitment and switching instead to a different sort of life, it is not plausible that that pressure is always grounded in a judgment that a life of scholarship is strictly better than these other alternatives. There are too many cases of roughly equal value, or of signiﬁcant incommensurability,14 or of an agent’s self-conscious inability or unwillingness (especially given our many cognitive limits and the enormous complexity of such matters) conﬁdently to reach such strict comparative evaluative judgments. Whatever metaethical view we have of such comparative evaluative judgments, we will want to allow for common and important cases of signiﬁcant under-determination of important contours of an agent’s life by the agent’s judgments about what is good or best. So it seems that reasonable stability of basic commitments is not simply a matter of responsiveness to these judgments about the best. This point can be extended by reﬂecting on some ideas from Joshua Cohen and John Skorupski.15 Cohen, in interpreting the work of John Rawls, develops the “idea of reasonable pluralism”, the idea that “there are distinct understandings of value, each of which is fully reasonable”. And this, says Cohen, is suggested by the absence of convergence in reﬂection on issues of value, which leaves disagreements, for example, about the value of choice, welfare, and self-actualization; about the value of contemplative and practical lives; about the value of devotions to

13 I also discuss a modiﬁed version of this Platonic answer below, in section 8. 14 Concerning incommensurability see Raz 1999. Ruth Chang argues that in addition to better than, worse than, and equally good as, there is a fourth comparative value relation of being “on a par” with each other. If we agree we can interpret talk of roughly equal value as including both Chang’s third and her fourth case. See Chang 1997, 4, 25–7. 15 I also discuss the ideas to follow in Bratman 2007j.

Anchors for Deliberation

195

friends and lovers as distinct from more diffuse concerns about abstract others; and about the values of poetic expression and political engagement.16

And, as Cohen sees it, in recognizing such reasonable pluralism, “we are acknowledging the scope of practical reason” (Cohen 1993, 272). The second idea is that judgments about the good, like judgments quite generally, involve a commitment to a kind of rational convergence. This is what John Skorupski calls “the convergence thesis”: When I judge that p, I enter a commitment that inquirers who scrutinized the relevant evidence and argument available to them carefully enough would agree that p – unless I could fault their rationality or their evidence (Skorupski 1999, 73).17

What should we say about the convergence thesis when we consider, in particular, judgments about value? Well, it is perhaps possible for a person to judge that something is good or better than certain alternatives, even while recognizing that there would not be such convergence. Such a person might so judge but not – pace Skorupski – “enter a commitment” to convergence. Nevertheless, it does seem, at the least, to be a reasonable ideal that one’s judgments of value be constrained by expectations of or commitments to such convergence – an ideal that one not settle on a value judgment unless one thinks there would be relevant convergence. Borrowing a term from Cohen, who borrows it from Quine, we can call this an ideal of nonsectarianism in value judgment.18 And now the point to make is that – though there are complexities that would need to be sorted out in a fuller discussion – it seems reasonable to suppose that a non-sectarian evaluator who thought that she was in a world of reasonable pluralism, would be signiﬁcantly modest in her judgments of value. Nevertheless, in order to give coherent shape to her life she would probably still need to settle on articulated forms of valuing that go beyond these limited, constrained, modest judgments of value. But then, if there is rational pressure in favour of the stability of these commitments it is unlikely to come solely from her judgments of value. This is not to say that the reasonable stability of authoritative valuings has no connection at all to the agent’s own judgments about what is good. For example, we will want some minimal constraint that one’s authoritative valuing of X not be held in the face of one’s view that there is nothing at all good about either X or being committed to X, when that view itself has agential authority and it is in one’s power to make an appropriate change in one’s valuing. But a switch from commitments to

16 Cohen 1993, 281–2. And see Rawls 1993. I put to one side the question of whether Cohen has accurately interpreted Rawls. 17 Note that the cited commitment to convergence is seen as coming from the nature of the attitude – judgment – not from the speciﬁc content of the attitude. This should be distinguished from Michael Smith’s view that the content of evaluative judgment is about rational convergence (Smith 1994, ch. 5). 18 Cohen talks of “the sectarian route of afﬁrming one’s own view, that is, believing it as a matter of faith” in the face of “irresoluble rivalry” (Cohen 1993, 282). Cohen takes the phrases “irresoluble rivalry” and “sectarian” from Quine (1990, 98–101).

196

Intentionality, Deliberation and Autonomy

scholarship to commitments to academic administration – if we suppose, say, that one has been asked to be a college president – will not be a switch that violates such a minimal constraint. And the decision between these commitments may well not be settled by the agent’s relevant comparative evaluative judgments. So our problem now is this: once we see that an agent’s judgments of the good may well underdetermine many of her basic practical commitments, can we adequately explain the rational pressure for stability of authoritative commitments in reﬂective deliberation, or must we retreat, in our effort to articulate conditions of reﬂective deliberation, to volitional necessity? Can we ﬁnd a satisfactory way between a Platonic appeal to the best and a Frankfurtian appeal to volitional necessities? I want to sketch a tentative conjecture about how we might steer a middle path. The key is the dual, inter-related roles of cross-temporal organization in supporting claims both to agential authority and, in part by way of those claims to agential authority, reasonable stability. 5. Reasonable stability, cross-temporal organization, agential authority Suppose I stop and reﬂect on my commitment to scholarship, a commitment that has agential authority for me. This commitment consists primarily in a structure of selfgoverning policies concerning what to treat as justifying in relevant deliberation. These policies are not volitionally necessary for me: while I am conﬁdent that these policies ﬁt well into my life, I have, and am glad that I have, the capacity to give them up. Suppose now that I see that a commitment to academic administration, rather than to scholarship, might also be made, with some tinkering, to ﬁt well into my life. And suppose I do not think – perhaps because of my non-sectarianism – either that a life of scholarship is strictly better than a life of academic administration, or that a life of scholarship is strictly worse than a life of academic administration. Is there nevertheless a ground of reasonable stability of my commitment to scholarship? I think that there is, and that it involves general concerns with the cross-temporal organization of one’s life. We have already appealed to the role of such commitments in structuring our temporally extended agency as a ground of their agential authority. But we can also appeal to this role as a ground of reasonable stability, one that to some extent draws on the fact that it is, as well, a ground of authority. An initial point is that the roles of authoritative valuings in organizing our lives over time are broadly useful: they help us to achieve many of the complex, temporally extended ends that we authoritatively value.19 Instability in authoritative valuing tends to undermine not only the achievement of what is thereby speciﬁcally valued but, more generally, the cross-temporal organization and coordination central to our pursuit of a very wide range of temporally extended ends that we authoritatively value. So there is a broad instrumental pressure in the direction of the stability of such valuings: such stability is a more-or-less universal means. 19 Indeed, as Elijah Millgram has emphasized, the contribution of authoritative valuings to Lockean personal identity means that they contribute to the cross-temporal identity of the agent that is presupposed in many of these ends.

Anchors for Deliberation

197

The next point is that the agential authority of such valuings – an authority grounded in a cross-temporal Lockean role – feeds back into a pair of further grounds for reasonable stability. The ﬁrst is that instability of authoritative valuings tends to undermine a kind of temporally extended integrity that we normally authoritatively value; the second is that instability of authoritative valuings tends to undermine a kind of temporally extended autonomy that we normally authoritatively value. To explain this particular idea of temporally extended integrity let us reﬂect on a poignant remark of Adlai Stevenson. Stevenson once concluded his commencement address at Harvard University with the advice: “And don’t forget when you leave why you came.”20 After wiping tears from their eyes, the graduating seniors might have asked: Why not? Why keep track of your original reasons for coming rather than focus on the reasons you have embraced, after years of study, for getting an undergraduate education? Perhaps Stevenson wanted to exalt the idealism of 18 year olds, in contrast with 22 year olds. But a different idea is that we normally value an integrity over time that involves reminding ourselves of, and – other things equal – sticking with, our initial, authoritative reasons for engaging in a major life activity. Instability of authoritative valuings tends to undermine such cross-temporal integrity. Turn now to temporally-extended autonomy. Here the idea is that stability of authoritative valuings is normally needed for the cross-temporal deliberation that is an element of cross-temporal autonomy.21 Autonomous agency over time normally involves deliberative guidance by authoritative valuings that stably persist over that time. And we do normally value, and authoritatively so, such cross-temporal autonomy. We value it both generally and as embedded in speciﬁc aims. What I value, for example, is not just that my book be completed but, further, that I complete it, and that I complete it in a way that constitutes a form of temporally extended selfgovernment. On the one hand, then, the cross-temporal stability of authoritative valuings is a universal means, useful in pursuit of an enormously wide range of ends we authoritatively value. And, on the other hand, such cross-temporal stability of authoritative valuings is partly constitutive of forms of cross-temporal integrity and autonomy that we normally authoritatively value. In these ways the Lockean agential authority of valuings feeds back into a ground for their stability. In all three cases – appeals to universal means, to cross-temporal integrity and to cross-temporal autonomy – we identify rational pressure that is robust in the face of wide-ranging differences in speciﬁc valuings and ends. The idea of a universal means is explicitly an idea of this sort; and the ends of cross-temporal integrity and autonomy are compatible with wide differences in particular substantive ends. Granted, we also frequently value novelty and experimentation; and we recognize that rigidity in one’s valuings over time can sometimes be stultifying. Nevertheless,

20 As quoted by Stanford University President Donald Kennedy at many Stanford graduations. 21 My talk of cross-temporal autonomy derives from J. David Velleman’s talk of “diachronic autonomy” (Velleman 2000d, 240–41). There are related themes in Rubenfeld 2001.

198

Intentionality, Deliberation and Autonomy

the cited trio of considerations22 normally provides some rational support for a transition in thought from “this valuing now speaks for me” to “let this valuing continue to speak for me” – a support that is robust in the face of divergent particular ends and valuings, and a support that does not depend on the absence of the capacity for change. 6. Instrumentality, integrity, autonomy Does this support for stability itself depend on the presence of ends that might not be accepted by the agent? We might try to side-step this worry by seeing the bases for this support for stability as goods, whether or not the agent herself cares about them. Effective agency, integrity, autonomy are each, we might try to say, good whether or not the agent cares about them; and that is enough for them to play the cited roles in grounding the reasonable stability of authoritative valuings. But here we need to recall that our target is deliberation that is characteristic of self-governance. In the case of self-governance it seems that we should be able to locate grounds for stability of basic commitments within the practical framework of the agent who is doing the governing. So we need to address more directly the worry that the agent might not accept the ends of effective agency, integrity and/or autonomy. The ﬁrst point to make here is that the instrumental pressure in favour of stability is broadly immune to this worry. So long as one has ends that have a temporally extended structure, and so long as one has self-governing policies at all, one will be subject to these instrumental pressures in favour of stability of those policies. That is the point of the appeal to the idea of a universal means. But what about concerns with temporally extended integrity and autonomy? Could one have self-governing policies but not authoritatively care about one’s own temporally extended integrity and autonomy? And if so, does this possibility weaken the argument for stability of authoritative valuings? These are difﬁcult questions. Kantian theories of agency see a concern for autonomy as in some sense inescapable for an agent.23 And even Frankfurt thinks that “[i]t is a necessary truth about us … that we wholeheartedly desire to be wholehearted” (Frankfurt 1999b, 106). If we think of wholeheartedness as having a cross-temporal aspect – since it is the wholeheartedness of a temporally persisting agent whose agency is temporally extended – this is close to a claim about the necessity of a wholehearted concern with cross-temporal integrity. Of course, even if a general concern with cross-temporal integrity were volitionally necessary, it would not follow that the more speciﬁc authoritative commitments that do much of the work in deliberation, and that have been our primary concern here, share in this necessity. So even if we were to appeal to the necessity of a general concern with integrity, we would not be back to the Frankfurtian strategy of appealing to volitional 22 Stability of authoritative valuings also makes us more reliable participants in social cooperation that we value. A fuller discussion would also consider such social pressures. 23 They also see autonomy as the unique ground for basic structures of practical rationality; but we need not accept this uniqueness claim in order to give an important role to autonomy.

Anchors for Deliberation

199

necessity to explain the status of speciﬁc, substantive anchors of deliberation, such as my love of philosophy or of my children. Nevertheless, I am inclined to favour an approach that does not require, though it is compatible with, such necessities. Many of us do in fact care a great deal about cross-temporal integrity and self-government; and these concerns help support, in the way described here, general pressures for stability, including the stability of these very concerns with integrity and autonomy. Once these concerns are in place they reﬂexively support their own stability, as well as the stability of more speciﬁc authoritative valuings. And even in cases – if such there be – in which there are not these agential concerns with cross-temporal integrity and autonomy, there would remain the broad instrumental pressures in favour of stability of authoritative valuings. So even an agent who does not care about her own cross-temporal integrity and autonomy can be, to a certain extent, a self-governing agent with integrity. Even given only the cited instrumental pressures, an agent’s self-governing policies may exhibit a thoughtful and reasonable stability, and be in a position to anchor deliberation of a sort characteristic of self-governance. However, the presence of a concern on the part of the agent with her own cross-temporal integrity and autonomy – a concern that is common in agents like us – would itself further support these forms of stability, and thereby further support cross-temporal integrity and selfgovernance. And this does not depend on the idea that these concerns with integrity and autonomy are themselves necessary. 7. Planning agency, reasonable stability and default structures So we have a trio of routes from a role in cross-temporal organization to the reasonable stability of authoritative valuings. These routes may be explicitly cited by a reﬂective agent as justiﬁcations for sticking with particular, prior authoritative valuings. But there is a second way in which concerns with cross-temporal organization help shape the very structure of our practical thinking. To explain this let us look more carefully at the general case of planning (Bratman 1987). Let us consider the idea that the reasonable stability of authoritative valuings is a special case of the reasonable stability of planning structures characteristic of our planning agency. We can learn something about the structures of reasonable stability of valuings by looking at the more general case of planning – including, in particular, planning what to do over an extended period of time. In pursuit of this idea let us now focus brieﬂy on planning in general, returning later to the special case of valuing. The basic point to make here is that plan-like commitments normally involve a kind of settledness or stability. This does not, of course, mean that these commitments are irrevocable. But prior plan-like commitments, in contrast with ordinary desires, do normally involve a characteristic stability in the agent’s practical thinking. How should we understand this stability of planning structures? One idea concerns reconsideration, and has two aspects (Bratman 1987, ch. 5). The ﬁrst aspect is that, normally, once one has arrived at a plan-like commitment the matter is settled in the absence of reconsideration. The second aspect is that in many cases it is reasonable

200

Intentionality, Deliberation and Autonomy

not to reconsider. Reconsideration, after all, has its costs. On the assumption that one was a fairly reliable planner in the ﬁrst place (and was not, say, drunk when one made the plan) (Hinchman 2003), it will normally be reasonable not to reconsider unless there is some special reason for re-thinking the practical issue. Now, one might think that these points about reconsideration exhaust what we should say about the stability of prior plan-like commitments. But it seems that even if one does reconsider a prior plan-like commitment, there is a way in which that commitment reasonably functions in one’s thinking as a kind of default: it is normally in force unless it is defeated (Bratman 2007k).24 There is here a minimal burden of proof on a challenge to the prior plan-like commitment.25 This is not to say that the prior plan-like commitment directly provides a further positive reason in its own favour. To say that would introduce a problematic kind of bootstrapping (Bratman 1987, 24–7). In the face of a determinative and authoritative case against the prior plan-like commitment, that commitment does not add a further reason to its own defence, one that allows it to bootstrap itself into acceptability. Nevertheless, it is not enough to defeat that commitment that the agent does not judge that it would be strictly better not to change. Why exactly should we say that our planning agency involves not only the idea that a plan-like commitment remains in force if not reconsidered, but also the idea of a minimal default? The argument is an inference to the best explanation. We need to explain what seems, prior to theory, to be a minimally presumptive status in our practical thinking of prior plan-like commitments, even once we re-visit the relevant practical issue; but we need to do this without sanctioning unacceptable bootstrapping. The idea that these plan-like commitments provide a minimal default helps us do this. This is to see these prior commitments in ways that to some extent parallel the doctrine of stare decisis in law.26 And this allows us to steer a path between ignoring this apparent presumptive status in the thinking precipitated by reconsideration, and sanctioning illegitimate bootstrapping. This dual structure concerning reconsideration and minimal burden of proof is partly constitutive of our structures of planning agency. These planning structures are important to us in large part because of their support of cross-temporal organization. And the basis for our interest in that cross-temporal organization involves the three ideas highlighted earlier: it is a universal means and it can help constitute forms of 24 In going beyond appeal to norms of non-reconsideration, to such a default structure I diverge from Richard Holton’s treatment of these issues (in Holton 2004). 25 This is in the spirit of a broad conservatism that Gilbert Harman has defended. Harman endorses the following principle: “Principle of Conservatism: One is justiﬁed in continuing fully to accept something in the absence of a special reason not to” (Harman 1986, 46). (See also Harman’s remarks about “general conservatism” in Harman 1999a, 27.) Harman sees both beliefs and intentions as cases of full acceptance. I am trying to ground a kind of practical conservatism in part in considerations of cross-temporal organization; but I do not try to argue here that we should see conservatism of belief in the same way. For a related appeal to the idea of a “presumption or burden of persuasion” see Schauer 1991, 205. 26 Shelly Kagan made this point in correspondence. For a related idea see Nozick 1981, 297.

Anchors for Deliberation

201

cross-temporal integrity and autonomy that we commonly value. To be a planning agent is, in part, to plan in accord with this dual structure; and we have strong reasons of cross-temporal organization to be such planning agents.27 We can now return to authoritative valuings, and to our question of how concerns with cross-temporal organization shape the structure of their reasonable stability. It was noted earlier that an agent might appeal explicitly to concerns with crosstemporal organization in her reﬂection on whether to stick with her prior authoritative valuings. But a second way was also alluded to in which concerns with cross-temporal organization shape the structure of reasonable stability of authoritative valuings. We can now say what that is. The basic point is that authoritative valuings are special kinds of plan-like commitments – in particular, they are plan-like commitments to signiﬁcance in deliberation. As plan-like commitments they are embedded in the general structures of practical thinking characteristic of such commitments. In particular, they are in force in the absence of reconsideration, and upon reconsideration they serve as a minimal default. And this minimal default structure is supported by rational pressures towards stability that derive from needs for cross-temporal organization of thought and action. Indeed, these pressures are especially signiﬁcant in the case of basic, authoritative valuings, since the stability of authoritative valuings is partly constitutive of the cross-temporal integrity and autonomy that helps support the rational pressure in favour of stability. So, authoritative valuings are embedded in this minimal default structure, a structure that is grounded in inter-related connections to cross-temporal organization, agential authority and stability. And while this default can be defeated in reﬂective deliberation, the burden of proof is on the challenger.28 8. A modiﬁed Platonic model? Consider now a modiﬁed Platonic model: what explains the rational pressure (if such there be) for stability of a prior valuing of X are authoritative judgments that it would be strictly better not to make a change from this prior valuing to a valuing, instead, of Y, where such judgments can themselves appeal, inter alia, to the cited trio of 27 I do not mean that for this reason we choose to be planning agents. Though we are planning agents, this is not because we have chosen to be. Nevertheless, we can appeal to these reasons for being a planning agent in order to make sense of ourselves to ourselves, and to fend off challenges to structures of planning agency. 28 Drawing on work of David Wiggins, Jeffrey Seidman has argued that certain practical “concerns” reasonably function in a way I would describe as a kind of default (though he does not put his point in terms of the idea of a “default”). He writes that “we can have reasons to opt out of our concerns, but in the absence of such reasons, we are reasonable to harbour them, and to let them inform our practical deliberation.” Seidman sees this as a feature of our actual practical thinking which it is reasonable to endorse. I agree; but I also seek a further underlying rationale for some such default practical structures, and claim to ﬁnd it in considerations of cross-temporal organization, considerations that provide characteristic forms of support for our planning agency. (See Seidman 2003, 210; Seidman refers to Wiggins 1998.)

202

Intentionality, Deliberation and Autonomy

general considerations about cross-temporal organization. Could such a modiﬁed Platonic model work as a general account of reasonable stability of authoritative valuings? I doubt it. It was earlier remarked that when faced with a comparison of X with alternative Y, there are cases of roughly equal value, of incommensurabilities, and of an agent’s reasonable inability or unwillingness to reach relevant, strict comparative evaluative judgments. And this last point is especially forceful for a non-sectarian evaluator who believes she is in a world of reasonable pluralism. And now the point to make is that versions of all three kinds of cases will also arise when the comparison is between holding onto a prior valuing of X and changing to a new valuing of Y, even if we take into account the cited trio of considerations concerning cross-temporal organization. This is, again, true in particular, for a non-sectarian evaluator who believes she is in a world of reasonable pluralism. But the rational pressure towards stability of one’s prior valuing of X persists even in cases in which, though one does not judge it would be strictly better to change to a new valuing of Y, one also does not judge that it would be strictly better not to change to a new valuing of Y. Prior authoritative valuings provide a default, one supported by the cited considerations of cross-temporal organization; and this creates a burden of proof in their favour. The modiﬁed Platonic view of reasonable stability does not sufﬁciently provide for this burden of proof. 9. Conclusion We have, then, an outline of a non-Frankfurtian and non-Platonic model of reﬂective, anchored deliberation that is characteristic of self-government. The anchors for such deliberation are valuings that have agential authority, are potentially revocable, can diverge across rational agents, and bear complex relations to judgments about the best. Their reasonable stability involves defeasible presumptions both against reconsideration and against revision, presumptions characteristic of plan-like commitments and supported by considerations of cross-temporal instrumentality and, at least normally, cross-temporal integrity, and autonomy. And this support for stability by connections to cross-temporal organization goes in part by way of the connection between cross-temporal organization and agential authority. This is a welcome result, because for a wide range of candidates for anchored deliberation and autonomy neither the Frankfurtian nor the Platonic model accurately represent the relevant structures of agency. In many cases the deliberation characteristic of self-governance is anchored in authoritative valuings whose stability is thoughtful and reasonable, even though these valuings are neither volitionally necessary nor required by judgments about the best. It does not follow from this that there are no cases of anchored deliberation that conform to either the Frankfurtian or the Platonic model. For all that has been said here, there may be a plurality of modes of reﬂective, anchored deliberation and autonomy. If what has been argued here is correct, anchored deliberation and autonomy need not wait upon volitional necessity, or highly speciﬁc and determinative judgments about the best. But, for all that has been argued, it still may be that there are cases of

Anchors for Deliberation

203

anchored deliberation and autonomy that do involve one or both of these features. In theorizing about human autonomy we should not assume, without argument, that it may not come in different forms.29 References Bratman, Michael E. (1987), Intention, Plans, and Practical Reason, Cambridge, MA: Harvard University Press, reissued by CSLI Publications, 1999. —— (2006), ‘A Thoughtful and Reasonable Stability: Comments on Harry Frankfurt’s 2004 Tanner Lectures’, in Harry Frankfurt, Taking Ourselves Seriously and Getting it Right, (ed.) Debra Satz, Stanford, CA: Stanford University Press, pp. 77–90. —— (2007a), Structures of Agency: Essays, New York: Oxford University Press. —— (2007b), ‘Reﬂection, Planning, and Temporally Extended Agency’, in Bratman 2007a, pp. 21–46. —— (2007c), ‘Hierarchy, Circularity, and Double Reduction’, in Bratman, 2007a, pp. 68–88. —— (2007d), ‘Two Problems About Human Agency’, in Bratman 2007a, pp. 89–105. —— (2007e), ‘Nozick on Free Will’, in Bratman 2007a, pp. 106–26. —— (2007f), ‘A Desire of One’s Own’, in Bratman 2007a, pp. 137–61. —— (2007g), ‘Autonomy and Hierarchy’, in Bratman 2007a, pp. 162–86. —— (2007h), ‘Three Forms of Agential Commitment: Reply to Cullity and Gerrans’, in Bratman 2007a, pp. 187–94. —— (2007i), ‘Planning Agency, Autonomous Agency’, in Bratman 2007a, pp. 195–221. —— (2007j), ‘Three Theories of Self-Governance’, in Bratman 2007a, pp. 222–53. —— (2007k), ‘Temptation Revisited’, in Bratman 2007a, pp. 257–82. —— David J. Israel and Martha E. Pollack (1991), ‘Plans and Resource-Bounded Practical Reasoning’, in Robert Cummins and John Pollock (eds), Philosophy and AI: Essays at the Interface, Cambridge, MA: MIT Press, pp. 7–22. Chang, Ruth (1997), ‘Introduction’, in Ruth Chang (ed.), Incommensurability, Incomparability, and Practical Reason, Cambridge, MA: Harvard University Press, pp. 1–34.

29 This chapter is a somewhat distant descendant of papers I gave at the University of Illinois at Chicago June 2003 Conference on Personal Identity and Practical Reason, and at a meeting of the Bay Area Forum for Law and Ethics. A more recent version was presented at the 2005 Siena Conference on Intentionality, Deliberation and Autonomy. In each case I beneﬁted greatly from the discussions. I have beneﬁted in particular from discussions with and/or comments from Meier Dan-Cohen, John Fischer, Harry Frankfurt, Christoph Lumer, Elijah Millgram, Samuel Schefﬂer, Jeffrey Seidman, Gideon Yaffe and Manuel Vargas. Some of the work on this essay was done while I was a Fellow at the Center for Advanced Study in Behavioral Sciences. I am grateful for ﬁnancial support provided by The Andrew W. Mellon foundation.

204

Intentionality, Deliberation and Autonomy

Cohen, Joshua (1993), ‘Moral Pluralism and Political Consensus’, in David Copp, Jean Hampton and John E. Roemer (eds), The Idea of Democracy, Cambridge: Cambridge University Press, pp. 270–91. Frankfurt, Harry (1988a), The Importance of What We Care About, Cambridge: Cambridge University Press. —— (1988b), ‘Freedom of the Will and the Concept of a Person’, in Frankfurt 1988a, pp. 11–25. —— (1988c), ‘The Importance of What We Care About’, in Frankfurt 1988a, pp. 80–94. —— (1988d), ‘Identiﬁcation and Wholeheartedness’, in Frankfurt 1988a, pp. 159–76. —— (1999a), Necessity, Volition, and Love, Cambridge: Cambridge University Press. —— (1999b), ‘The Faintest Passion’, in Frankfurt 1999a, pp. 95–107. —— (1999c), ‘On the Necessity of Ideals’, in Frankfurt 1999a, pp. 108–16. —— (1999d), ‘Autonomy, Necessity, and Love’, in Frankfurt 1999a, pp. 129–41. Gibbard, Allan (1990), Wise Choices, Apt Feelings, Cambridge, MA: Harvard University Press. Harman, Gilbert (1986), Change in View, Cambridge, MA: MIT Press. —— (1999a), ‘Rationality’, in Gilbert Harman, Reasoning, Meaning, and Mind, Oxford: Oxford University Press, pp. 9–45. —— (1999b), ‘Practical Reasoning’, in Gilbert Harman, Reasoning, Meaning, and Mind, Oxford: Oxford University Press, pp. 46–74. Hinchman, Edward (2003), ‘Trust and Diachronic Agency’, Noûs, 37, 25–51. Holton, Richard (1999), ‘Intention and Weakness of Will’, Journal of Philosophy, 96, 241–62. —— (2004), ‘Rational Resolve’, The Philosophical Review, 113, 507–35. Korsgaard, Christine M. (1996), The Sources of Normativity, Cambridge: Cambridge University Press. Nozick, Robert (1981), Philosophical Explanations, Cambridge, MA: Harvard University Press. Parﬁt, Derek (1984), Reasons and Persons, Oxford: Oxford University Press. Quine, W.V.O. (1990), Pursuit of Truth, Cambridge, MA: Harvard University Press. Rawls, John (1993), ‘The Domain of the Political and Overlapping Consensus’, in David Copp, Jean Hampton and John E. Roemer (eds), The Idea of Democracy, Cambridge: Cambridge University Press, pp. 245–69. Raz, Joseph (1999), ‘Incommensurability and Agency’, as reprinted in Joseph Raz, Engaging Reason: On the Theory of Value and Action, Oxford: Oxford University Press, pp. 46–66. Rubenfeld, Jed (2001), Freedom and Time: A Theory of Constitutional SelfGovernment, New Haven, CT: Yale University Press. Schauer, Frederick (1991), Playing by the Rules: A Philosophical Examination of Rule-Based Decision-Making in Law and in Life, Oxford: Clarendon. Seidman, Jeffrey (2003), ‘Rationality and Reﬂection’, Ethical Theory and Moral Practice, 6, 201–14.

Anchors for Deliberation

205

Skorupski, John (1999), ‘Value Pluralism’, in John Skorupski, Ethical Explorations, Oxford: Oxford University Press, pp. 65–81. Smith, Michael (1994), The Moral Problem, Oxford: Blackwell. Velleman, J. David (2000a), The Possibility of Practical Reason, Oxford: Oxford University Press. —— (2000b), ‘Introduction’, in Velleman 2000a, pp. 1–31. —— (2000c), ‘What Happens When Someone Acts?’, in Velleman 2000a, pp. 123–43. —— (2000d), ‘Deciding How to Decide’, in Velleman 2000a, 221–43. —— (2002), ‘Identiﬁcation and Identity’, in Sarah Buss and Lee Overton (eds), Contours of Agency, Cambridge, MA: MIT Press, pp. 91–123. Watson, Gary (2002), ‘Volitional Necessities’, in Sarah Buss and Lee Overton (eds), Contours of Agency, Cambridge, MA: MIT Press, pp. 129–59. Wiggins, David (1998), ‘The Concern to Survive’, in David Wiggins, Needs, Values, Truth: Essays in the Philosophy of Value, 3rd edn, Oxford: Clarendon, pp. 303–11.

This page intentionally left blank

PART III Action-Theoretical Approaches to Freedom, Autonomy and Responsibility

This page intentionally left blank

Chapter 10

Autonomy for Real People Michael Quante

This chapter claims that we need a ﬁne-grained account of personal autonomy. In our everyday practice of attributing autonomy and taking responsibility we need a ﬁne-tuned notion of personal autonomy, especially in contexts like long-term care, psychiatry or in biomedical contexts like the assessment of neurocognitive enhancement or medical drugs which have effects on the patient’s personality. The philosophical debate concerning free will and determinism bypasses these hard questions of our everyday practice since it is focused on the compatibility–incompatibility question which can be regarded as the hard philosophical question concerning personal autonomy. At ﬁrst it is shown that both libertarians and compatibilists need a ﬁne-grained account of personal autonomy to be able to deal with the hard questions of our everyday practice. Furthermore it is shown that in compatibilists answers to the manipulation argument some conceptual distinctions can be found which should be integrated into such a ﬁne-grained account of personal autonomy. The overall conclusion is that the focus on the hard philosophical questions leads philosophers to concentrate on questions of causality, but that the answers to the hard questions of our everyday practice have to be found in the ethical realm. The critical problem with respect to our interest in freedom is not whether the events in our volitional lives are causally determined by conditions outsides ourselves. What really counts, so far as the issue of freedom goes, is not causal independence. It is autonomy. (Harry G. Frankfurt, The Reasons of Love)

1. Autonomy, incompatibilism and the manipulation argument I am an autonomous person of the average kind and I take each individual to be one too. We hold each other responsible for our actions and pay respect to each other in the many different forms which are essential for our overall form of life, at least for Western culture. Practically we all share the opinion that our practice of treating each other as autonomous and responsible agents is justiﬁed in principle. Without doubt it takes a long journey to become an autonomous person of the average kind – think of childhood, puberty and all the battles of education. Politically it has been an even longer and even more arduous journey to establish

210

Intentionality, Deliberation and Autonomy

social institutions which help people to become autonomous persons and which can be justiﬁed as manifestations of our mutual respect for autonomy. Without doubt this journey has not come to an end everywhere (to say the least). And a good deal of arguments can be made to claim that the societies we live in are somehow developing in ways which might endanger the autonomy of persons (think of information politics especially in times of war as one example of undermining our autonomy as political subjects; think of all the developments in neuroscience and biotechnology at the dawn of the twenty-ﬁrst century). Nevertheless most of us are involved in the practice of attributing and taking responsibility every day.1 Most of us act and lead our lives in a way that shows our commitment to being an autonomous person. We all accept that autonomy and responsibility come in degrees and that there are more or less autonomous persons. Besides this, we take into account the fact that for some persons it has been more difﬁcult to become an autonomous person of the average kind than for others. This comes to the fore in the way in which we try to balance praise and blame (take “extenuating circumstances” as the prime example). Since more and more of us grow very old in our societies, we all know that it is not only a long journey to become autonomous but that it is an even longer and probably even tougher ﬁght to stay an autonomous person. As a result of this our practice of realizing, attributing and respecting autonomy under these conditions faces many difﬁculties, becomes vague and is in need of nuance. But although all this is well known and we all feel that there are many problems in detail and much uncertainty in single cases (or regarding special kinds of cases like “sex offenders”), most of us are sure that basically our practice of attributing and taking responsibility is well justiﬁed: we take ourselves to be autonomous persons (at least of the average kind) and we value being autonomous and being treated as an autonomous person. From the philosophical standpoint things begin to look different. In real life the hard problems seem to be the many shades of grey we need for a good practice of attributing and taking responsibility while the basics are clear and steady. In philosophy the hard problem seems to be this very basis of autonomy itself. As for this problem there is a fundamental worry which runs so deep that all these more speciﬁc problems we have to deal with in everyday life seem to be pre-empted. This fundamental and hard philosophical problem lurking at the roots of autonomy and responsibility is the question of whether we are really free, whether freedom requires a special kind of causality, which is incompatible with determinism on the one hand and can be made intelligible on the other. This quarrel seems to be a matter of principle, not a matter of ﬁne-tuning and context- or case-sensitive judgement. There has been a stalemate of philosophical positions for more than 2000 years concerning the question of free will. During this time an amazingly rich arsenal of conceptual distinctions has been developed which is worth a lot since it can help us make our most basic intuitions clear and allow us to better understand the opposing lines of thought. Furthermore, an admirable panoply of philosophical 1 My use of the phrase “taking responsibility” is meant to be of the pre-philosophical, everyday kind. The more technical sense, in which it is introduced and developed in Fischer and Ravizza (1998, ch. 8) will be discussed below (see section 4.2).

Autonomy for Real People

211

positions has been developed, defended, defeated and re-developed again. This, too, is very helpful since philosophers have been able to distinguish several questions which should not be conﬂated when they develop their accounts. Nevertheless, for most philosophers the question which has to be answered most urgently is whether freedom, autonomy and responsibility on the one hand and determinism on the other hand are compatible or not. Both camps, compatibilists and incompatibilists alike, can appeal to our pre-philosophical intuitions. The reason for this is simply that in everyday life we all know cases in which being determined in certain ways is incompatible with being autonomous and responsible; and we all know that mere indeterminism alone is not enough for being an autonomous person and being able to act responsibly. If, as philosophers are used to do, the problem of determinism is posed generally, it seems that most of us are intuitively torn in different directions at once. The problem to deal with the many shades of grey which we face as the hard problems in our everyday practice has been replaced by the really hard problem of a fundamental and deep metaphysical tension which seems to tear apart the fundaments of our practice as such. Although I am a philosopher I am not convinced that the problem compatibilists and incompatibilists primarily discuss really is the hard problem regarding our practice of attributing and taking responsibility. More precisely: it is only indirectly so since the really hard problems of our practice are often misleadingly framed in terms of and along the main lines of the compatibilism–incompatibilism debate. Unfortunately even this statement puts me in one of the camps insofar as it is a version of soft compatibilism: as I see it autonomy and responsibility (and our evaluative practices related to our being autonomous persons and responsible agents) are compatible both with determinism (as such) and with indeterminism.2 In this chapter, however, I will try to discuss problems of autonomy and responsibility beyond (or besides) the classical free-will debate. The aim is to reach a position from where we as philosophers can begin to deal with the hard problems we face in our daily practice without having answered the hard problem we face in the philosophical free-will debate. I hope to show that the questions I am concerned with here can be asked and have to be answered in both compatibilist and libertarian accounts if reasonable answers to problems we face in our daily life shall be given by philosophy. Since the hard cases concerning autonomy and responsibility we are confronted with in everyday life make up a wide spectrum and require case-sensitive analysis my aim in this paper must be very modest: I will try to make clear how we can separate these questions from the classical philosophical one, how the former have been superimposed by the latter in such a way that plausible philosophical answers were bypassed or stayed out of reach. But although the topic I have in mind has been somehow superimposed, fortunately a lot of conceptual distinctions which

2 Here a cautionary note is called for: to say that our being autonomous and responsible is compatible with determinism is neither to say that it is compatible with every form of determinism, nor does it exclude that there can be circumstances in which autonomy and responsibility are impossible.

212

Intentionality, Deliberation and Autonomy

are important for my project have been made under the heading of compatibilist defences of personal autonomy against incompatibilist or libertarian attacks.3 In his paper “Autonomy and Manipulated Freedom” Tomis Kapitan states, “compatibilism has been the target of two powerful challenges”: the one is the well-known “consequence argument”, the other the probably just as well-known “manipulation argument” (Kapitan 2000, 81). Both intend to show that determinism as such rules out responsibility. As the consequence argument directly leads to the hard philosophical question I will analyse the manipulation argument only, because this will make visible the phenomena and problems I am interested in here. In Kapitan’s words the manipulation argument can be formulated this way: “By allowing agents to be fully determined, compatibilist accounts of practical freedom and responsibility are unable to preclude those who are subject to global manipulation from being free and responsible” (ibid., 81).4 According to Kapitan this argument works in a different way than the consequence argument does, and we will see that Kapitan’s formulation of the manipulation argument can be understood only as the result of a discussion caused by Robert Kane’s former libertarian objections against compatibilism.5 Therefore I will start by scrutinizing Kane’s arguments (section 2). Then I will assess the main compatibilist responses and Kapitan’s criticism of these compatibilists’ replies, which have prompted his way of phrasing the manipulation argument cited above (sections 2, 3 and 4). At the end, I hope, we will be in a position to see that the manipulation argument and its discussion can be very useful for developing an account of autonomy which is suitable for the hard problems real people have to face concerning autonomy.6 3 Since many of the arguments I will deal with in the following have been framed in the terminology of the free-will debate, I will use the terminology of freedom and autonomy interchangeably in this paper. Furthermore it should be kept in mind that my main target here is autonomy, not responsibility, so that my discussion of Fischer and Ravizza and Kapitan includes a shift of emphasis, since these authors primarily stress responsibility. For the purposes of my paper the differences lurking here can be ignored. 4 Probably Robert Kane has developed the manipulation argument for the ﬁrst time in his book Free Will and Values (cf. Kane 1985). 5 Some readers may object that arguing this way I evade incompatibilists who deny that freedom in the libertarian sense is possible, because freedom and determinism are not compatible and our world is determined. This is true: since I take for granted that personal autonomy and our practice of attributing and taking responsibility are ineliminable aspects of our Lebensform, I will ignore this tragic incompatibilism as I will ignore nihilism with respect to free will, that is, a position that denies the compatibility of free will with both determinism and indeterminism – both are no viable options for us, since the evidence for the proposition that we are autonomous and responsible agents will always overrule the evidence for both the proposition that our world is determined and the proposition that freedom and determinism are incompatible. 6 One short note to clarify two things. First, I do not want to deny that the ‘classical’ problem of free will and determinism which is in the main line of the consequence argument is of philosophical importance. Elsewhere I have defended the thesis that this metaphysical question is of no direct relevance for our practice of attributing autonomy and responsibility (cf. Quante 1998; 2003a; 2003b, ch. 10). But this does not commit me to the claim that the

Autonomy for Real People

213

2. Non-constraining control and personal autonomy: Kane’s libertarian attack In the third chapter of his Free Will and Values Robert Kane develops an argument which is meant to show two things: ﬁrst, he intends to demonstrate that there is a need for developing a libertarian conception of freedom, named ultimate responsibility. Since I am not interested in the compatibilism–incompatibilism quarrel concerning the ‘real’ meaning of freedom I will not discuss this aspect of Kane’s argument. Second, the aim of Kane’s argument is to show that compatibilism cannot show that S is not autonomous in cases where S is the subject of covert non-constraining control. Since Kane’s version of the manipulation argument is somewhat complicated I will analyse it in this section. At the beginning some terminology suggested by Kane has to be introduced. Kane distinguishes between constraining and non-constraining inﬂuences on the will of S.7 If S is subject to constraining inﬂuences S notices them as thwarting her will. In cases of non-constraining inﬂuences S does not notice these inﬂuences on her will. The constraining inﬂuences are subdivided into two classes: coercion, which is a threat of harm making some option less desirable, and compulsion, which is a threat frustrating a higher order desire.8 Furthermore Kane distinguishes between mere determination and control. The former is an inﬂuence on the will of S which is not intended by an agent, the latter is an inﬂuence intended by an agent other than S itself.9 Additionally Kane distinguishes between covert and non-covert inﬂuences on the will of S. It is reasonable to suppose that constraining inﬂuences cannot be covert, since S notices such an inﬂuence as thwarting her will. But we should leave open the possibility that non-constraining inﬂuences can be both covert and noncovert (so that noticing a non-constraining inﬂuence should not count as a constraint per se thereby turning the inﬂuence into a constraining inﬂuence). Starting from B.F. Skinner’s utopian community Walden Two Kane develops an argument which is intended to show that compatibilists cannot draw a distinction

metaphysical problem is of no philosophical interest in itself. Second, in this chapter I am only interested in autonomy. In so far as problems of autonomy, freedom and responsibility do not cover the same ground, I will ignore freedom and responsibility. Therefore my reading of the accounts discussed in the following departs in two respects from the main focus of their authors: on the one hand I pay attention solely to autonomy, and on the other hand I do not answer to incompatibilist worries. 7 Leaving magical inﬂuences aside we can say that inﬂuence implies causal inﬂuence in every case. And since I want to bypass the free-will debate, I will restrict myself to actual inﬂuences, so that counterfactual inﬂuences are ignored. In the context of the consequence argument, such counterfactual inﬂuences are important in analysing the meaning of “couldhave-done-otherwise”. Furthermore it is important to notice that inﬂuences are taken as tokens here, that is, as concrete dated events, and not as types of events. For simplicity I do not mention this aspect of time boundedness in the following. 8 Although this is not clear in Kane’s text, I take compulsions and coercions as distinct species of constraint (cf. Kane 1985, 51, n. 16). 9 Therefore self-control is not a species of control. For the purposes of this chapter we do not have to consider cases of self-control across time (in these cases complications come into play as regards the problem of S’s diachronic identity).

214

Intentionality, Deliberation and Autonomy

between mere determination and covert non-constraining control (CNC control). In CNC control S’s will is not frustrated at t, but she is controlled by another agent who has manipulated past circumstances so that S at t wants, desires and intends what the controller has planned before. Since there is no internal conﬂict, S does not notice her will as thwarted at t and since the control is covered, S has no motive to change her will to defeat the controller.10 In such a situation, Kane suggests, we would not judge S to be autonomous. Compatibilists who take the soft line therefore will try to show that mere determination and CNC control differ to the effect that we can justify a difference in our evaluation (compatibility of mere determinism and autonomy on the one hand, incompatibility of CNC control and autonomy on the other hand).11 But compatibilists, according to Kane, are not in a position to justify this claim. CNC control is a case in which S is manipulated and therefore intuitively cannot be regarded as being autonomous. But this is exactly what compatibilists have to do – according to Kane they do not have the conceptual resources to distinguish CNC control from mere determinism. They have to take the hard line claiming that CNC control at t is never incompatible with S being autonomous at t. The problem, insoluble for compatibilists, is “to say, what ‘signiﬁcant’ kind of freedom it is that CNC control takes away, but mere determination does not” (Kane 1985, 40).12 As a libertarian Kane thinks that only a kind of freedom which is incompatible with determinism can do this job. Kane admits that his argument is no logical refutation of compatibilism. But his arguments purport to show that compatibilists either have to take the hard line concerning CNC control or have to introduce occult powers – powers S is deprived of by CNC control but not by mere non-constraining determination – if they want to draw a line between CNC control and mere non-constraining determination (ibid., 45). Incompatibilists, according to Kane, are in a better position, since they can argue that no such line has to be drawn. Let us start our discussion with the last point that libertarians do not need to draw distinctions in the two pairs of cases (covert constraining control and covert constraining determination on the one hand, CNC control and mere non-constraining determination on the other hand). In all four cases the autonomy of S is threatened since freedom in the sense of ultimate responsibility is excluded in virtue of determination. So let us suppose that S lives in an indetermined world, in which ultimate responsibility is possible and in which ultimate responsibility actually occurs (at least sometimes). Even in such a world there might be circumstances in which S’s 10 As long as we do not take the controller to be an omnipotent agent, the covertness of her control might be a necessary condition for success. Additionally the covertness might be needed if among the wants of S is a want of not being subject to manipulation (or other forms of non-constraining control, if there are any). 11 This is not precise enough: if we do not want to close the option that some kinds of CNC control might be compatible with autonomy, the incompatibility-claim has to be speciﬁed. But I will ignore this for the moment. 12 If we take a more ﬁne-grained approach, allowing that some forms of CNC control are compatible with S being autonomous, the task for compatibilists must be to say which features of CNC control are incompatible with autonomy.

Autonomy for Real People

215

autonomy is threatened by constraint, be it constraining control or non-intentional inﬂuences. The same goes for covert non-constraining inﬂuences in such a world: some are threats to S’s autonomy, some are not. Probably even some forms of covert non-constraining control are no threat to S’s autonomy (for example education). Surely not all forms of covert non-constraining non-intentional inﬂuence will have to be counted as threats to S’s autonomy. Since we suppose that S lives in an indetermined world, the notions of control and inﬂuence should not be understood as implying determination. Nevertheless, even in an indetermined world in some situations control or inﬂuence are of a kind that threatens autonomy, in other situations not. Besides, even in an indetermined world intuitively we would count cases of control as more likely autonomy-threatening than cases of mere causal inﬂuence. Even if we agree that indeterminism is a necessary condition for autonomy we can ask why there are these distinctions in our evaluation of different (kinds of) cases. Now either the libertarian can take the hard line and claim that there are no differences to be made in attributing autonomy and taking responsibility if ultimate responsibility is possible in the world under consideration. Such a response would ignore all the hard cases of our real practice and would imply that there cannot be shades of grey that are important for our evaluation of autonomy or responsibility (this seems to be a notion of autonomy held by existentialist philosophers like Sartre). Or the libertarian takes the soft line and admits that there are more features which are relevant for our practice of attributing autonomy and taking responsibility. If he argues in this way the libertarian is in the same position as the compatibilist: he has to make explicit which criteria we use (or should use) in distinguishing the many shades of grey we de facto distinguish in our real evaluative praxes. The starting premise of the libertarian’s reasoning (at least in Kane’s argument) is that either determinism is the only threat to autonomy since it is incompatible with ultimate responsibility, in which case indeterminism becomes sufﬁcient for being autonomous and being responsible, and further qualiﬁcations or grades of autonomy are not necessary. Or (as Kane argues) determinism excluding ultimate responsibility is the dominant feature outweighing all the distinctions compatibilists can offer to make their soft line defence plausible. But this premise is not at all plausible: consider Walden Two, the Skinnerian utopian society, Kane’s arguments depart from. The non-constraining behavioural modiﬁcations and engineering in human societies, described in Walden Two, are already a reality in our society, as neurocognitive enhancement tells us (and so do the prospects of genetic enhancement hovering on the horizon).13 As Kane argues, most people are afraid of Skinner’s utopia since he “extols, and promotes as social policy, one of the things that libertarians in Western Culture have been most concerned to rule out when they asserted that humans have free will, namely, non-constraining control of the will by another agent” (ibid., 34). For sure, the chances of enhancement open to us today are far away from control in the strong sense Kane presupposes in his discussion of determinism. But it is evident that most arguments against enhancement of this kind are not of the libertarian or incompatibilist kind but rest on very different conceptual resources. This shows 13 See Farah et al. (2004) for an overview on neurocognitive enhancement; and compare Bayertz and Quante (2003) for a short discussion of genetic enhancement.

216

Intentionality, Deliberation and Autonomy

that in these contexts, more ethically relevant features are involved than only an incompatibilist notion of autonomy.14 To sum up: both the libertarian and the compatibilist are required to develop a ﬁne-grained theory of the features we rely on in our everyday practice of attributing autonomy and taking responsibility if they do not want to take their hard lines. Furthermore the difference between non-intentional inﬂuence and control emphasized by the compatibilists has also to be taken into account in a libertarian framework. Both compatibilists and incompatibilists need a philosophical theory of being autonomous and being responsible which takes into account the hard problems of our everyday practice of attributing autonomy and taking responsibility. Therefore libertarians cannot rely on the claim that determinism as such is a threat to autonomy. This might, if successfully defended, be enough to refute compatibilism, but it is not sufﬁcient as a theory of autonomy. Even developing an indeterministic account of freedom, like Kane’s conception of ultimate responsibility, is not enough, since indeterminism is no sufﬁcient basis for reconstructing our notion of being autonomous and our practice in detail. The reason for this is that the complexities of the conditio humana15 and the world we live in are under-described by being characterized as indeterministic or deterministic. Therefore we need a philosophical account which is rich enough to identify those features which guide our practices of attributing autonomy and taking responsibility. Since incompatibilists (and libertarians) often restrict themselves to refuting compatibilists’ arguments, not much of interest for my questions can be found in their accounts. On the contrary, compatibilists confronted with the manipulation argument are, if they do not want to take the hard line, obliged to offer an analysis of those features which justify our

14 Ironically many arguments against enhancement are framed in the language of indisposability or “Do not play God”, that is, arguments which plea against enhancement and for a restriction of our freedom to enhance our biological or psychological nature. Besides that, some defenders of enhancement are staunch liberals, who regard enhancement as a matter of fairness to compensate natural inequalities; (for this argument cf. Buchanan et al. 2000). We can conclude therefore that liberal reactions to Walden Two cannot be identiﬁed with Kane’s libertarian opposition to it simpliciter. And if Kane insists that part of the ethical reaction is that Walden Two threatens ultimate responsibility, my argument is sufﬁcient to demonstrate that we need a more precise theory concerning the other relevant features of enhancement if we want to develop a useful philosophical account which helps us to answer the question of how to deal with enhancement in our societies. 15 My arguments so far leave open two options. On the one hand it might be the case that theories answering the hard cases of our daily practice do not divide along the compatibilist– incompatibilist line. It might be that we can have an account dealing with these questions which is neutral with respect to the hard question of free-will in philosophy. On the other hand it might come out that compatibilists and libertarians have to develop divergent (and incompatible) theories dealing with the hard problems of our daily practice. If the latter turns out to be true, then my arguments above show that even libertarians are in need of such a ﬁne-grained theory if they want to give an adequate analysis of our evaluative practice of attributing autonomy and taking responsibility. Furthermore my strategy of developing such a theory starting from compatibilist answers to the manipulation argument is justiﬁed because of the overall dialectical situation.

Autonomy for Real People

217

evaluative practice (beyond the determinism–indeterminism clash). Therefore a lot more can be found in compatibilists’ responses to the manipulation argument. 3. Compatibilists replies and the problem of global manipulation The incompatibilist challenge that no (existing) compatibilist can answer the manipulation argument in a satisfying manner has prompted reactions in the compatibilists’ camp. This section presents what I take to be the two most promising lines of defence, that is, the accounts of Alfred Mele on the one hand and John Martin Fischer and Mark Ravizza on the other. 3.1 Mele’s autonomous agents In chapters nine and ten of his Autonomous Agents Alfred Mele addresses the CNC control cases and tries to show that compatibilists can take the soft line to answer it successfully. The overall premise of Mele’s argument is that “autonomy is possible for human beings like us” (Mele 1995, 144). His basic idea is that a history-sensitive account of psychological autonomy which is compatible with compatibilism can handle cases of CNC control adequately. In setting the stage this way Mele’s argument resembles my own strategy. In claiming that only a history-sensitive account of psychological autonomy is compatible with compatibilism he leaves open whether it is also compatible with libertarianism. If this were the case the history-sensitive account which can answer cases of CNC control adequately might be on neutral ground, so that both compatibilists and libertarians can rely on such an account to defend autonomy against tragic incompatibilists or free-will nihilists who hold that autonomy is impossible for human beings (at least in the actual world) (cf. ibid., 145). As we will see later, Mele’s account is open to criticism due to the fact that he develops his account as a reply to the libertarian objection Kane has made. Recall Kane’s CNC control scenario: S is subject to manipulation of an intentional agent (call him Big Brother) in such a way that Big Brother has manipulated S at some time prior to t in such a way that S at t decides as Big Brother has planned without S’s will being thwarted. At t there is no coercion since the overall psychological structure of S has been designed by Big Brother in such a way that at t S wants to do what Big Brother had planned her to do. S wholeheartedly decides without suffering internal conﬂict at t. Given that S has all capacities necessary for being autonomous at her disposal at t, and uses them in the right way at t, S has to be regarded as being autonomous at t. Compatibilists cannot claim that the fact that S at t is determined to decide as Big Brother has planned and arranged before is incompatible with S being autonomous at t. Therefore they either have to say that CNC control is no threat to autonomy (the hard line) or to explain why CNC control is a threat to autonomy, but mere determination is not (the soft line). At this point two philosophical discussions converge. Within the compatibilist camp a debate took place whether autonomy can be analysed in purely internal

218

Intentionality, Deliberation and Autonomy

ways or not.16 Internalists such as Harry G. Frankfurt develop accounts of autonomy which rely only on the actual psychological make-up of an agent considered at the moment of decision. The causal history of the agent’s actual psychological make-up is relevant only for developing the capacities necessary for autonomous decision. Once these capacities have evolved internalists take into account only the actual structure of an agent’s psychological make-up. Against this internalist account externalists with respect to autonomy have argued in a way similar to Kane’s CNC control objection that S might have been the victim of perverse socialization, manipulation like brain-washing, interventions by neuroscientists (or what have you) in the past. They do not address the problem of determination as such but try to show that we need the right kind of causal histories to guarantee that S is autonomous at t. From Kane’s perspective these externalists take the compatibilists’ soft line, from my perspective they address (some of) the hard problems of our everyday practice of attributing autonomy and taking responsibility. John Christman, for example, has demonstrated that Frankfurt’s account is open to counter examples due to its restriction to the actual psychological make-up of the agent under consideration. If an agent’s actual psychological make-up is the result of massive manipulation in her history, this will affect our judging her as autonomous (Christman 1991).17 While Christman’s objections to Frankfurt’s internalist account of autonomy suggest that the history-sensitive account is sufﬁcient for personal autonomy, Alfred Mele has shown that integrating such a history-sensitive condition is a necessary but not a sufﬁcient condition for S being autonomous in a special situation (cf. Christman 1991; Mele 1993). Let us take “special situation” to mean that S has a desire D at t and has to decide whether D should become effective, that is, result in an action intended to satisfy D. Furthermore D should cover all kinds of mental episodes, which are evaluative and volitional; D itself can be a complex desire, but need not be. It should only be excluded both that D is a mere belief without evaluative or volitional aspects, and that D is identical with the psychological structure of S as such, that is with S’s personality. The former guarantees that D is related somehow to the volitional aspect of S’s psychological make-up (therewith it is excluded only that mere beliefs (if there are any) are proper objects of decisions). The latter makes clear that we are scrutinizing local autonomy in the sense that only a part of S’s overall psychological make-up, that is, not her personality as such, is under consideration. According to Mele we have to distinguish three aspects of S’s autonomy relative to D: developmental autonomy (S is autonomous with respect to the development of D in her); possession autonomy (S is autonomous with respect to the continued possession of D at some time t under consideration); and inﬂuence autonomy (S is autonomous with respect to the inﬂuence of D on her behaviour at t). These three aspects belong to local autonomy, since we take into account only the relation 16 I take Frankfurt to be a compatibilist here. As far as I can see, the discussion concerning Frankfurt’s internalism with respect to autonomy has not been led under the heading of compatibilism and incompatibilism. 17 Cf. Quante (1997) for a discussion of these arguments.

Autonomy for Real People

219

between S and D. We can call the uniﬁed autonomy that comprises these three aspects the uniﬁed local autonomy. Now there are a number of constellations to be considered. On the one hand S may have developmental autonomy. This means that S has developed D in such a way that her overall local autonomy is not threatened in this dimension. But D may, to use Mele’s term, be practically unsheddable during an interval of time, so that S cannot resist D (except in extraordinary circumstances or after a longer period of time) – S does not have possession autonomy with respect to D.18 Additionally, suppose that S cannot resist D, but has the capacity to control in which way D inﬂuences her behaviour (take D to be the desire to eat something, which leaves open, what will be eaten) – S has inﬂuence autonomy regarding D. Many further constellations are possible, but they cannot all be discussed here in detail. The crucial point is that attributing uniﬁed local autonomy to S means taking into account several dimensions of local autonomy, which might have to be balanced against each other (if not all aspects of local autonomy are realized by S in a certain situation). As a reply to Kane’s objection this multi-dimensional analysis of local autonomy enables the compatibilist to demonstrate why cases of CNC control differ from cases of autonomous decision so that the aspect of CNC control which is incompatible with S’s autonomy must not be determination. If S has been subject to CNC control, then S has no authenticity with respect to D, which she is compelled to have at t. It is this “negative historical constraint on the autonomous possession of pro-attitudes” which explains why CNC control is incompatible with autonomy (Mele 1995, 173). Together with Christman’s arguments, this makes clear that a purely internalist account like Frankfurt’s cannot sufﬁce. But – beyond Christman’s account – it also demonstrates why a purely historical-sensitive account is not enough either. Autonomy can be endangered in the dimension of possession and inﬂuence autonomy too. And this is something which is related to the actual psychological make-up, so that the right kind of history with respect to D is not sufﬁcient to guarantee S’s autonomy with respect to D. Autonomy is a multi-layered concept including many criteria which have to be weighed against each other in cases where S is not autonomous in every dimension (we can take it to have a matrix structure).19 Since Mele is primarily interested in blocking the manipulation argument (as Kane has stated it in the form of CNC control), he does not try to plot a detailed picture of this matrix, but states sufﬁcient conditions for autonomy. In addition to the conditions for being an ideally selfcontrolled and mentally healthy agent we have a psychologically autonomous agent if we add the following three conditions (the compatibilist trio): (i)

The agent has no compelled*20 motivational states, or any coercively produced motivational states.

18 Mele (1995, 153) has introduced this term. 19 There are similarities here to Robert Nozick’s closest continuer theory in which a matrix of our concept of diachronic identity is developed (cf. Nozick 1981, ch. 1). 20 “Compelled*” means that the compulsion was not arranged by the agent herself.

Intentionality, Deliberation and Autonomy

220

(ii) The agent’s beliefs are conducive to informed deliberation about all matters that concern her. (iii) The agent is a reliable deliberator. (Ibid., 187)

Mele himself sees clearly that having answered Kane’s objection in the sense of defending compatibilism along the soft line is not sufﬁcient to answer the hard problems of our everyday practice, since the question remains: “How far can an agent fall short of satisfying these conditions and still be psychologically autonomous?” (ibid., 189) Answering this question would come down to ﬁlling in the details of the matrix of our concept of autonomy beyond the determinism-threat emphasized by incompatibilists and libertarians. 3.2 Fischer and Ravizza on ‘taking responsibility’ In chapter nine of their Responsibility and Control John Martin Fischer and Mark Ravizza add another feature to their compatibilist history-sensitive account of autonomy and responsibility developed up to that point.21 They suggest that “the past must contain a process of ‘taking responsibility’. Taking responsibility, we believe, is a necessary feature of moral responsibility” (Fischer and Ravizza 1998, 207). This feature involves, as they say, “three major ingredients” (ibid., 210–13): (i)

S must see herself as an agent, she must see that her choices and actions are efﬁcacious in the world.

(ii) S must accept that she is a fair target of the reactive attitudes as a result of how she exercises this agency in certain contexts. (iii) S’s view of herself as expressed in (i) and (ii) must be based, in an appropriate way, on evidence.

As Fischer and Ravizza say, it is an important component of the process of moral education that a child is coming to see herself as an agent. This is achieved by treating the child as if it were an agent and by already introducing it into the world of norms and values and by taking it as a proper target of reactive attitudes. In this process the child learns to take over the perspective on its own behaviour it is confronted with, thereby internalizing the self-understanding of being a responsible agent. At the end of this process S is able to take responsibility for her actions (and her way of leading a life as a person). Since deliberating consists in a complex reasonresponsive mechanism, embracing a lot of sub mechanisms which we do not have (detailed) knowledge about, Fischer and Ravizza suggest that S takes responsibility for the mechanisms her reason-responsive mechanism has consisted in (since taking responsibility is post festum, taking responsibility refers to the actual world only).22

21 Fischer and Ravizza are primarily concerned with responsibility but I will, in what follows, treat their account as an analysis of autonomy. 22 This is important to block certain interpretations of “could have done otherwise”, which I ignore here since I am not interested in the compatibilism–incompatibilism issue primarily.

Autonomy for Real People

221

Fischer and Ravizza now distinguish two ways in which S might hold (ii): in one case S is not a metaphysically reﬂecting subject worrying about determinism and freedom. Put this way, holding (ii) implies that S judges our social practices of praising and blaming someone for her actions as fair. This overall identiﬁcation of S with this practice does not exclude that S might disagree with being praised or blamed in a special situation (quite the contrary: quarrelling about singular judgements presupposes agreement concerning the practice in general). In the other case S is a metaphysically reﬂecting subject; here the only relevant constellation is a subject who decisively believes in incompatibilism (either freedom incompatible with determinism in a deterministic world, or freedom incompatible with indeterminism in an indeterministic world). In this case Fischer and Ravizza hold that “an individual must view himself as, prima facie at least, an apt candidate for the application of the reactive attitudes, and be willing to put aside his residual doubts, for all practical purposes” (ibid., 227). Having answered the incompatibilist objection on a general level up to this point, defending ‘taking responsibility’ at least for all practical purposes as an unavoidable stance (in Western culture, I would like to add), Fischer and Ravizza concede that there are special problem-cases which cannot be answered by applying the historysensitive criterion related to local autonomy but require the more “globally historical approach” already included in their notion of taking responsibility (ibid., 231). The problems they have in mind, which cannot be dealt with on the basis of a historysensitive criterion related to local autonomy, are, as we will see now, cases of manipulation. In the ﬁrst two cases S is the victim of covert manipulation by an agent who covertly implants in S an irresistible desire (case 1), consisting of a mechanism which is not reason-responsive at all, or a strong but not irresistible desire (case 2), which is reason-responsive only under very extreme circumstances (which are normally not given in our everyday practice). Since S has not taken responsibility for this implanted mechanism – reason-responsive or not – S should not be regarded as responsible with respect to D. The most problematic case Fischer and Ravizza discuss is a third one, in which the reason-responsive mechanism of taking responsibility herself is covertly implanted into S. With respect to this, Fischer and Ravizza say that S need not be taken to be responsible since the third condition for taking responsibility is not fulﬁlled: if the reason-responsive mechanism of taking responsibility is “electronically implanted”, S has not formed her self-image as a responsible agent in the appropriate way (ibid., 235ff.). Fischer and Ravizza have claimed at the outset that education is the normal way of implanting this self-image in young human beings. Imagine that our society improves in a way that this main task of education which is unfortunately not reached in some cases can be reached more effectively by implanting such a mechanism electronically. Then the crucial question is why the latter should be a threat to autonomy but the former should not. As Fischer and Ravizza admit in their account, “the relevant notion of appropriateness must remain unanalysed” (ibid., 236). And as our little thought experiment shows, reference to ‘normal’ conditions will not help if we are not willing to hold that technical intervention in these cases is generally incompatible with autonomy but education is not. As we all know, not every form of education is compatible with S

222

Intentionality, Deliberation and Autonomy

becoming autonomous, and it is hard to believe that technical intervention should be incompatible per se with S being autonomous later. Therefore either Fischer and Ravizza claim that their third condition is not relevant, in case the implantation of the reason-responsive mechanism of taking responsibility is under consideration, or they owe us further criteria to draw relevant distinctions here. For sure, all this does not show that the libertarian objection is correct. But it is evident that we need a more ﬁne-grained account of the matrix of autonomy embedded in our practice of attributing autonomy and taking responsibility. This is especially so if we take into consideration that our society is on its way to develop techniques (like cognitive or genetic enhancement) which will replace the normal course of education by engineering. It seems that the notion of appropriateness relevant here remains not only unanalysed in the account developed so far but refers us to conditions which are open to change by progress in science and technologies. For my purpose two features of the account of taking responsibility, developed by Fischer and Ravizza, are of special importance: ﬁrst, they emphasize that education is not a threat but a necessary condition for becoming an autonomous agent – at least in the case of human beings. Second, their account integrates internalist and externalist conditions in the epistemic sense: conditions (i) and (ii) of ‘taking responsibility’ refer to S’s self-understanding while (iii) refers to social standards of knowledge. If we take this seriously, the claim becomes plausible that S’s being responsible is a socially mediated affair. This will come as no surprise if it is accepted that attribution of autonomy is not backed by causal relations alone but inevitably includes evaluative judgements. 4. Kapitan and the victims of global manipulation In his paper entitled “Autonomy and Manipulated Freedom”, Tomis Kapitan’s aim is to show that the incompatibilists’ manipulation argument fails to refute compatibilism. Like Mele and Fischer and Ravizza, Kapitan starts from the premise that “in some sense the existence of responsible agents is beyond doubt” (Kapitan 2000, 82).23 But Kapitan thinks that neither Mele nor Fischer and Ravizza have answered the manipulation argument sufﬁciently. According to Kapitan, Mele’s reply to the problem of CNC control is confronted with a dilemma: either autonomy will become a condition which is realized in human beings so rarely that it cannot be a relevant aspect of our practice of attributing responsibility to human agents, or Mele’s account cannot show that CNC control makes S non-autonomous (cf. ibid., 88). On the one hand Mele has already accepted the ﬁrst horn of this dilemma since he admits that his account provides only sufﬁcient conditions for autonomy, but no conditions that are both necessary and sufﬁcient. To block Kane’s objection and to defend the soft line, this might be 23 Kapitan’s account is intended to be an analysis of responsibility and he explicitly argues against the claim that responsibility needs an independent condition guaranteeing autonomy to block the manipulation argument (cf. Kapitan 2000, 86). Therefore I will restrict my discussion of Kapitan’s arguments dealing only with his criticism of the compatibilist defences developed by Mele and Fischer and Ravizza.

Autonomy for Real People

223

enough. But, as Kapitan makes clear, the price is that human beings probably will not fulﬁl these sufﬁcient conditions. Although this is no objection in principle to Mele’s account – we can try to ﬁnd the missing conditions – it shows that we need a more ﬁne-grained approach to attain a concept of autonomy which is suitable for real human agents. On the other hand, the cases Kapitan describes to show that Mele is drawn back to the compatibilists’ hard line are socialization and education. In this process pro-attitudes are installed in S which are practically unsheddable at t and which S wholeheartedly identiﬁes with. This is the result of a process in which S’s pro-attitudes (think, for example, of some religious or moral beliefs) have been anchored so deeply in S’s personality that S cannot shed them without destroying her personality. Since we are all subject to “physical and social forces of which we are largely ignorant, over which we have no control, yet from which we acquire values, beliefs, motivations, and capacities for rational evaluation that subsequently guide our choices and actions”, this case of CNC control is ubiquitous (ibid., 89). Once S has become the person she is, these “forces ‘destroyed’ any capacity to become a different sort of person with self-control regarding any unsheddable pro-attitude that we happen to have” (ibid., 89). If we ignore the possibility of innate unsheddable proattitudes and take “unsheddable” in Mele’s sense to mean ‘practically unsheddable’, Kapitan seems to be right concerning the following: if in socialization or education capacities for critical reﬂection of certain pro-attitudes (like religious beliefs) are blocked, although S might have had the potential at some time in her life to develop such capacities later, S might not be able to critically evaluate these pro-attitudes. Since S has been subject to CNC control, we have either to conclude that CNC control is never incompatible with autonomy (the hard line) or to develop an account which makes clear which forms of CNC control are compatible with autonomy (and why they are, or are not). If we accept that education is a case of CNC control, we have to ask which forms of CNC control are compatible with autonomy. And since education will not go away in an indeterministic world, this problem concerns libertarians and compatibilists alike, except that libertarians claim that in such a world education cannot be a form of control. Such a reply would come down to the incompatibilists’ hard line, taking the absence of mere determination as sufﬁcient for S being autonomous in every case (which is as implausible as the hard line of the compatibilist). In a similar vein, Kapitan argues that Fischer and Ravizza cannot block the manipulation argument using their account of ‘taking responsibility’. His main focus is the third case of manipulation discussed by Fischer and Ravizza (the implantation of the reason-responsive mechanism relevant for ‘taking responsibility’ itself). The overall problem is that in cases of CNC control in which some weak reason-responsive mechanism is installed in S at some time prior to t, there are three probably relevant aspects: the fact that S has been manipulated, the implanted mechanism and all the other mechanisms S has acquired during her life-time. The crucial question is why S should not be taken to be autonomous with respect to her action (at t) if there are mechanisms operative at t for which S takes responsibility at t. Why should the fact that manipulation took place in the past or that the implanted mechanism is operative at t, too, be sufﬁcient to show that S is not autonomous at t? If we take into account that “the processes of character formation” are similar to covert manipulation in that

224

Intentionality, Deliberation and Autonomy

S does not take responsibility for these processes, we can ask why S is not regarded as not being autonomous at t because of these manipulations in her youth (ibid., 93). This is the same problem as stated above: if we take socialization and education as forms of CNC control, we have to show why they do not threaten autonomy in every case and where the difference lies between them and those cases of manipulation which make S non-autonomous at t. Besides this overall result which makes clear that a discussion of the manipulation argument beyond the quarrel of incompatibilists and compatibilists is needed, Kapitan’s objections are relevant inasmuch as this critique of the answer Fischer and Ravizza have given to the third case of manipulation (wherein ‘taking responsibility’ itself was implanted in S) enables us to distinguish different kinds of control. Fischer and Ravizza have argued that in this kind of case the evidence S has for her selfunderstanding is not based on evidence attained in the appropriate way. Now Kapitan distinguishes three kinds of CNC control S might be subject to: in case of global control S’s psychological make-up (desires, values, deliberations, choices, intentions and so on) within a given interval of time is what it is because of manipulation. Global control in this sense is opposed to local control which deals only with a part of S’s psychological make-up. Furthermore, two variants of global control can be distinguished: global control is interruptive if it begins some time after the beginning of S’s personal history (that is, after S has begun to develop a personality), and it is total, if this global control is executed right from the beginning of S’s history (in this sense).24 At the beginning of this chapter I quoted Kapitan’s formulation of the manipulation argument. There he spoke of global control, narrowing down the cases of CNC control which are intended to refute compatibilism. Now we are in a position to see that he has an even more speciﬁc version of the manipulation argument in mind, for he objects to Fischer’s and Ravizza’s reply to the third case of manipulation by reference to total manipulation. Since “total manipulation can duplicate much of the ordinary causal processes involved in conscious experience, belief acquisition, retention, and evaluation”, blocking the manipulation argument by referring to the evidence S can have for her self-understanding as an autonomous agent is no way out (ibid., 93). Since Fischer and Ravizza do not want to give up their epistemological internalism regarding autonomy, they cannot simply postulate the absence of total manipulation as a necessary element of being autonomous, since in case of total manipulation, S cannot be in the epistemic position to discover that her evidence for believing to be an autonomous agent is not of the appropriate kind. Fischer and Ravizza address this problem by claiming that no “coherent self” can develop in case of total global control (Fischer and Ravizza 1998, 234ff., n. 28). Like Kapitan, I cannot see that Fischer and Ravizza give any explicit reasons for this claim, but one argument can be given. First, total global control is not universal global control so that in the community of S there are other subjects who are not victims of total global 24 Total global control is permanent, but related only to S or other individuals, but not to all members of their community. We can deﬁne universal global control as permanent total global control every member of S’s community is subjected to; and absolute global control as permanent total control every human being is subjected to (cf. Kapitan 2000, 84).

Autonomy for Real People

225

control. Probably this is sufﬁcient to hinder those individuals subjected to total global control because of the reactions of the other members of their community. Besides that, we, as the evaluators of the case, draw a line. Against this one might take universal global control to restate the problem. In this case we, as the evaluators, are still not subject to control so that we can draw a line between ourselves and the members of the community under consideration (this is the situation given if we evaluate Walden Two). Against this one might try to shift to absolute global control. But this suggestion is inconsistent: we, the evaluators of the case, cannot detect whether we are subject to absolute global control since, as long as we are evaluating the cases, we have to do this under the idea of judging ourselves. Thereby we always consider a case of universal global control, since we do not conceive of ourselves as being subjected to control. This is important since it shows that the idea of absolute global control is irrelevant for our evaluating whether S or we ourselves are autonomous agents or not. And if we come to believe that we ourselves are subject to manipulative control, this clearly would infect our self-understanding. Taken this way, the response of Fischer and Ravizza becomes plausible and gets at an important feature of the manipulation argument: this argument can work only if we as the evaluators do not believe ourselves to be subject to the total or universal global control considered in the manipulation argument. The lesson to be learned here is that we cannot evaluate our practices of attributing autonomy from without, since we as the observers of the many cases presented in the line of the manipulation argument cannot but judge under the idea to be autonomously evaluating the case.25 In the end autonomy is not a question of powers or causality but of values, respect and self-respect.26 References Bayertz, K. and M. Quante (2003), ‘Genetic Enhancement’, in D.N. Cooper (ed.), Nature Encyclopedia of the Human Genome, vol. 2, London: Nature Publishing Group, 824–7. Buchanan, A. et al. (2000), From Chance to Choice, Cambridge: Cambridge University Press. Christman, J. (1991), ‘Autonomy and Personal History’, Canadian Journal of Philosophy, 21, 1–24. Farah, M.J. et al. (2004), ‘Neurocognitive Enhancement: What Can We Do and What Should We Do?’, Nature Reviews, 5, 421–5.

25 Does this not mean that by this we fell back into incompatibilism at the end? I do not think that this impression is correct as far as the incompatibilism of freedom, responsibility and determinism is concerned. As I see it there is indeed an incompatibility at stake here, but it is one of epistemic perspectives, not the classical one of the free-will debate. 26 Thanks to Simon Derpmann, Kristina Engelhard, Alexa Nossek and David Schweikard for suggestions and criticisms. Furthermore I wish to thank David Schweikard for the linguistic revision of the text.

226

Intentionality, Deliberation and Autonomy

Fischer, J.M. and M. Ravizza, S.J. (1998), Responsibility and Control, Cambridge: Cambridge University Press. Harry G. Frankfurt (2004), The Reasons of Love, Princeton, NJ and Oxford: Princeton University Press. Kane, R. (1985), Free Will and Values, Albany, NY: State University of New York Press. Kapitan, T. (2000), ‘Autonomy and Manipulated Freedom’, Philosophical Perspectives, 14, 81–103. Mele, A.R. (1993), ‘History and Personal Autonomy’, Canadian Journal of Philosophy, 23, 271–80. —— (1995), Autonomous Agents, New York: Oxford University Press. Nozick, R. (1981), Philosophical Explanations, Cambridge: The Belknap Press of Harvard University Press. Quante, M. (1997), ‘Personal Autonomy and the Structure of the Will’, in J. Kotkavirta (ed.), Right, Morality, Ethical Life, Jyväskylä: SoPhi, pp. 45–74. —— (1998), ‘Freiheit, Autonomie und Verantwortung in der neueren analytischen Philosophie’, Philosophischer Literaturanzeiger, 51, 281–309 (part I); 387–414 (part II). —— (2003a), ‘Philosophische Freiheiten’, in S. Mischer et al. (eds), Auf Freigang: metaphysische und ethische Annäherungen an die menschliche Freiheit, Münster: LIT-Verlag, pp. 11–37. —— (2003b), Einführung in die Allgemeine Ethik, Darmstadt: Wissenschaftliche Buchgesellschaft.

Chapter 11

Forming the Will Freely Gottfried Seebass

This chapter takes up the classical problem of freewill, arguing that its traditional understanding remains virulent up to the present and has been unduly declared obsolete by mainstream philosophy. The established division into “compatibilism” and “incompatibilism” is rejected as misleading. However, after a brief clariﬁcation of the relevant notions of ‘will’ and ‘freedom’, drawing on the author’s extended earlier research, it is argued in detail that it is not possible to meet the two central criteria of ‘willing freely’ if one subscribes to determinism. First, one cannot specify any relevant notion of practical possibility. Second, unless one accepts strong metaphysical premises, one cannot rely, alternatively, on the criterion of processes of forming the will being “essential” (in some relevant sense) to the person in question. Still it is not necessary to dismiss ‘freewill’, as the conception of volitional states or processes as indetermined in part is ruled out neither by an unprejudiced scientiﬁc view of the world nor by the bad old argument from “blind chance”. 1. Dismissing the freewill problem? ‘Freewill’ is one of the great, old and possibly “eternal” enigmas of mankind. In 1437 the Renaissance philosopher Lorenzo Valla diagnosed that there is “scarcely a question more in need of being answered and less close to an answer at the same time” (Valla 1987, 62f.). Since then things have not changed essentially. Yet many people think otherwise. Some are unconvinced of the topic’s oldness. In particular some scholars have claimed that ‘freewill’ is an invention of Judaeo-Christian theology unknown to the Greek tradition. This claim, however, proves false on both counts. On the one hand, it is simply not true that classical Greek philosophy and literature had no sense of the problem and its importance, although the ﬁrst explicit characterization of the “will” (“bulēsis”) as being “free” or “unfree” (“hekūsion” / “akūsion”) is not documented earlier (to my knowledge) than by the Aristotelian Magna Moralia.1 On the other hand, the mainstream of Christian theology did not afﬁrm but rather denied ‘freewill’ in human beings as its existence would scarcely

1

Aristotle, Magna moralia, 1188a30−35.

Intentionality, Deliberation and Autonomy

228

be compatible with the belief in God’s omniscience and omnipotence. So the idea is mistaken that ‘freewill’ was invented late or merely for theological reasons. Other scholars believe the “classical” problem of freewill to have been solved long ago or antiquated. Von Wright, for example, diagnosed that this problem “is now gradually receding into obsolescence” (von Wright 1985, 110). Schlick in 1930 and Schopenhauer almost a century earlier even expressed anger at all those “ignorants” who “still spill much ink” on a subject they believed had been long since settled by the arguments of philosophers such as Hobbes or Hume.2 And more recently Davidson dismissed it bluntly by citing a number of philosophical predecessors (including Hobbes, Hume and Schlick) who had done “what can be done, or ought ever to have been needed, to remove the confusions that can make determinism seem to frustrate freedom” (Davidson 1980, 63). Now, the history of philosophy is full of examples of alleged “deﬁnite solutions” that turned out to be over-hasty. ‘Freewill’ is a case in point. Despite the prolonged philosophical efforts, the very problem proclaimed to be obsolete appears to be virulent up to the present. This does not hold only with respect to everyday, prescientiﬁc thought and public discussion in the media. Even many psychologists and neuroscientists think that the traditional enigma of ‘freewill’ is unsolved and needs solving by experimental evidence. Some of them are even so bold as to claim that there is already sufﬁcient evidence to the effect that every act of conscious volition and thought is determined unconsciously on the physiological level. The ensuing controversial discussions show that the problem is still taken to be both unsolved and highly important. Accordingly, there are good reasons to be suspicious of straightforward claims to the contrary. To be sure, scientists who believe in physiological determinism are likely to ignore these facts and to dismiss ‘freewill’ readily. Also philosophers in the vein of Schopenhauer and Schlick will insist that the “classical” problem has been solved independently of whether the physiologistic claims are justiﬁed. To them the solution is conceptual rather than empirical and consists in the insight that the very idea of ‘freewill’ in question is ill-conceived and should be replaced by some other conception of willing and acting ‘freely’. Both parties agree, however, in the conviction that the traditional belief in indeterministic volition or choice is spurious and should be dropped as an illusion. So it is mainly this conviction which must be looked at critically if one suspects that the old enigma is still unsolved. 2. Two general misconceptions The widespread tendency to dismiss ‘freewill’ in its traditional sense stems from a double error. On the one hand, its signiﬁcance for human self-understanding tends to be underrated due for the most part to a misconception of, or resigned disengagement from, what it really means to be active. On the other hand, its importance is badly overrated in that it is assumed, explicitly or implicitly, that ‘willing freely’ means

2

Schlick <1930>/1984, 155; Schopenhauer <1840>/1977, sect. III, vol. VI, 83.

Forming the Will Freely

229

ipso facto ‘being entirely indetermined in what one wills’ or (in the words of Kant) being independent “of the mechanism of nature in its entirety”.3 The latter misconception is mirrored vividly by the established division into “compatibilism” and “incompatibilism”. This distinction suggests that it is one and the same thesis that is afﬁrmed by the one party and denied by the other. But this is misleading. First, the question of determination is only one of a number of different aspects of ‘willing freely’ and is not in itself a question of “all or nothing”. Second and most important, what is at stake among the two parties is not whether one should accept or reject the thesis that “freedom and determinism are compatible”. The real issue is how the concept of ‘freedom’ relevant to understanding ourselves as active beings, or as beings that are responsible for what they do, should be analysed or explicated. Consequently, the thesis afﬁrmed by the so-called “compatibilists” is not the same thesis that is denied by the “incompatibilists”. Also what kind, or kinds, of determination the latter want to deny and for what reasons and what kind, or kinds, of freedom are maintained by the former are open questions allowing for various answers. 3. Preliminary phenomenological clariﬁcations So let us forget from now on this unhappy categorization and go to work afresh. As elsewhere in philosophy, this work is for the most part conceptual and to a lesser degree phenomenological. Concerning the latter, two general points are obvious. First, human beings are far from independent “of the mechanism of nature in its entirety”. Quite to the contrary, they are inﬂuenced in what they will and do by a vast number of determining factors, social as well as natural. But this does not mean in the least that they are determined completely. Denying universal determinism is not equivalent to afﬁrming the extreme of universal indeterminism. And surely this extreme is not implied by the “classical” freewill problem. So let us not fool ourselves by quarrelling over spurious positions or ﬁghting for or against straw men. Second, the phenomenology of willing, intending and acting intentionally makes it clear that in discussing ‘freedom of will’ it is not possible for us to conﬁne ourselves to single choices, decisions or so-called ‘acts of volitions’ at one instant. We have to take into account the processes leading up to them. Accordingly, it would be inadequate simply to ask whether a given volitional state is ‘free’, let alone to deﬁne ‘freewill’ simply by ‘being indetermined in willing’. Rather, the adequate deﬁnition is: ‘freedom to form one’s will’. This covers many different processes most of which include deterministic parts and elements, though they are not normally thought to be ﬁxed throughout. Forming the will may consist in a spontaneous single act but need not. It may imply extended weighing of reasons and calculating means, ends and consequences. But it may also be conﬁned to a quick decision. A more extensively reﬂecting person may try to stir up (time permitting) unconscious attitudes in the endeavour to ﬁnd out what she “really wants” or “wants inmost” or to arrive at some “volitional equilibrium”. 3

Kant, Kritik der praktischen Vernunft, 1902ff., vol. V, 87.

230

Intentionality, Deliberation and Autonomy

Furthermore, reﬂection may or may not extend to volitions of a higher order, as has been countenanced by Augustine, Abelard and more recently Harry Frankfurt.4 All of these processes constitute relevant kinds of forming the will. And each of them raises the question of whether and to what extent they are ﬁxed in advance by inner and outer determinants, and if so, what this would mean for calling them “free” or “unfree”. 4. The concept of willing These questions, however, cannot be tackled unless we have some better idea of what we are talking about. So we must turn from mere phenomenology to conceptual clariﬁcation. Here we do not want to know, of course, what philosopher X or scientist Y have said that “freedom of will” is supposed to mean. There would be no end to listing different deﬁnitions and declarations. We want to know what we ordinarily mean when we say that active and responsible persons are capable of “forming volitional states freely”. And to know this we have to clarify ﬁrst the constitutive general notions: ‘will’, ‘volition’ and ‘willing’ on the one hand, ‘freedom’ and ‘free’ on the other. To shorten my argument I will restrict myself to the results of my earlier research on these matters. Concerning volition let us state ﬁrst what willing is not. Obviously it is not a brain state or brain process. Suppose I am giving a talk. I know that I want to ﬁnish my talk, and the audience may want to raise some objections in the discussion. But neither I nor the audience know what corresponds to this in our brains. Nor do expert neurologists up to now, despite some ill-considered claims to the contrary referring (for example) to the “readiness potentials” measured by Libet and his followers.5 Findings like these are much too unspeciﬁc to represent concrete volitional states such as ‘wanting’ or ‘willing that p’. Moreover, they cannot be the relevant neurological correlates, if (as is reported) they are prior to conscious volition. That there is something going on in the brains of people consciously wanting or willing something, most probably something very complex, is not in doubt. But for the purpose of identifying our volitions these facts are simply irrelevant. If willing is not a brain state or process, what is it? Behaviourists such as Hull or Berlyne6 proposed a dispositional analysis of intentional states like believing or willing, referring to interconnected internal states that are ascribed to human beings as “hypothetical constructs” (in the sense of MacCorquodale and Meehl [1948, 95–107]), viz. states or events thought to be realized physiologically. In a modiﬁed form this conception still survives under the name of “functionalism”. Now, though it is true that willing, in contrast to idle wishing, is qualiﬁed dispositionally, the 4 Augustine, De libero arbitrio, I: 12, 26; 13, 29; II: 19, 51; III: 3, 7−8; Confessiones, VIII: 5, 10; 8, 20−24; De civitate Dei, V: 9−10; De trinitate, X: 11, 18; Retractationes, I: 13, 5; Abelard: Ethica, cap. 1−3, see also Saarinen 1994, 46, 51−7, 66−71, 85; Frankfurt 1988, ch. 2. For a more detailed account and critical discussion see Seebass 2006, ch. 5, sect. 8. 5 Libet 1985, 529−39; Haggard and Eimer 2001, 47−63; Haggard, Clark and Kalogeras 2002, 382−5. 6 Hull 1943; Berlyne 1965; see also Seebass 1981/1982.

Forming the Will Freely

231

idea of its complete dispositional explication is hopeless.7 Surely willing, believing and other intentional states can not be deﬁned by the various “causal roles” they play in connecting relevant inner states or mediating between external stimuli and responses. The core of willing, stripped of its motivational and possible further qualiﬁcations, is a conscious, intentional optative attitude related reciprocally to an assertoric attitude as its conceptual counterpart. Both states are elucidated in a preliminary way by the metaphor of “directions of ﬁt” (in the vein of Anscombe, Searle and Tugendhat8). Much more can be said, however, about what this means and in what ways wanting, willing, deciding or intending differ from the mere optative.9 Still, as is the case with every basic notion, there are irreducible elements in it that cannot be analysed further, but must be grasped by the individual learning the relevant terms of language. Talk of unconscious or preconscious volition is derivative relative to its conscious form. Thus in the attempt to clarify ‘willing’ and processes of ‘forming one’s will’ we are referred to conscious, intentional and motivationally qualiﬁed states of willing. 5. The concept of freedom Turning next to the concept of ‘freedom’ I will also take a shortcut. In the most general sense, being ‘free’ means being ‘unhindered’. Hobbes and Schopenhauer have argued for this,10 and their diagnosis is conﬁrmed by ordinary language. More speciﬁc senses of “freedom” are subsumed easily. This refers in particular to the classical notions of ‘freedom of action’ and ‘freedom from constraint’ developed by Greek philosophy. Most forms of social, political and economic freedom are covered by these. Freedom of action means: being in a position to act as one wants or wills to act. Freedom from constraint is more complex, meaning in a broad sense almost the same as being unhindered but having several narrower meanings that overlap partly (already hinted at by Aristotle11) with what we would call “freedom of will”. Freedom of will is another special case branching again into various subcases. First, however, let us look more closely at the concept of ‘hindrance’ itself. Clearly this is a very general notion to be speciﬁed best by the questions as to what is hindered in what and by what. A river that is not hindered in its ﬂow by dams or embankments is said to “ﬂow freely”. A paralysed or tied man is “unfree” because he is hindered by abnormal (internal or external) impediments to move as he wants. The hindrance need not extend to every part or aspect and need not be absolute. Still it must be signiﬁcant. Common to all subcases is the idea that something is restrained severely from evolving, living or existing in ways which are 7 For those readers who are inclined to think that this is still a plausible option, see my detailed reasons in Seebass 1993, ch. 4, sect. 3. 8 Anscombe 1957, 4f., 56f.; Searle 1983, 7−9, 167ff.; Tugendhat 1976, 505ff. 9 See Seebass 1993, ch. 4, 4−6, and Roughley forthcoming. 10 Hobbes 1961, vol. III, 196; vol. IV, 273f., 275f.; vol. V, 367f.; Schopenhauer <1840>/1977, vol. VI, 43f. 11 Aristotle, Ethica Nicomachea, III: 1−2.

232

Intentionality, Deliberation and Autonomy

“natural” or “essential” to it and therefore should not be precluded. Hence whether, in what respects and to what degree something is free or unfree has to be judged with reference to two dimensions: the relevant standard of “naturalness” and the relevant realm of theoretical possibilities actually closed or open. Roughly, the greater the number of possibilities open, the lesser the hindrance and the greater the freedom. However, possibilities that do not touch on the relevant “nature” can be ruled out as inessential. Most of the inﬁnite number of actions I might envisage but actually cannot do are irrelevant to my freedom as I do not and will never care about them in the least. Having decided spontaneously and on my own, for example, to go to a concert tonight I do not feel restricted in the least by the fact that in doing so I am no longer able to watch to the eightieth episode of a trivial soap opera on television. Conversely, the fact that my TV-set offers 80 programmes to me today instead of the two programmes I used to have does not enhance my freedom 40-fold if I do not care about the 78 additional programmes. Thus freedom implies open possibilities but only possibilities “essential” to the person in question. Still what possibilities are “essential”? Starting with examples like the preceding one might suggest that this depends simply on what the relevant persons want or will. But even when conﬁned to freedom of action this criterion is badly insufﬁcient as has been shown even by ancient philosophers such as Plato, Aristotle and most lucidly Plotinus.12 It is not just that the volitional states of ordinary people can be, and all too often are, ﬂeeting, rash or capricious. Even our considered, reasoned will is not “essential” ipso facto and likely to be “inessential” if formed (for example) under pressure, in ignorance or without reﬂection on concurring wants and volitions. So we must rely on a stronger criterion of “essentialness” relevant to our freedom. However, what criterion is adequate here? One radical option would be to ignore the volitional states of people entirely and to state from an external position what is, or should be, considered “natural” or “essential” for them (similar to what we do when describing a river as “ﬂowing freely”). This would amount to the claim that there are objective ideal standards of what beﬁts people, taken individually or generically, in leading an “essential” form of life. Perhaps Plato’s critique of wilful democratic freedom is an early example of this.13 But even ignoring the strong metaphysical assumptions necessary to support this claim, it is quite implausible to deﬁne what belongs to the “essence” of a particular person independent of his or her particular mental states including volitions. Even the most experienced psychotherapists, closest personal friends, or most perceptive parents taking care of us during infancy cannot claim to “know better” what we “really want or will” independent of looking forward at least to our eventual reﬂective consent. Accordingly, individual volition cannot be given up as a necessary condition of “essentiality” although it needs to be strengthened substantially for it to become sufﬁcient.

12 Plotinus, Peri tū hekūsiū kai thelēmatos tū henos / Enneades, VI: 8, 1−7. 13 Plato, Res publica, 557a−564b.

Forming the Will Freely

233

6. The deterministic challenge Now, the most obvious way to do this is to add the requirement that the will in question is free, that is, has been formed freely on our own and unhindered (for example) by ignorance or severe physical, psychological or social pressure. However, in doing so we have to rely again on the double criteria of open possibilities and “essentialness”, with the only difference that both of these now do not refer to particular voluntary actions but to volitional states themselves. So it is all the more urgent to be clear about both conditions. Here at the latest we face the “classical” freewill problem claimed to be settled or obsolete by many writers (cf. above, section 1). For, at least the possibilities criterion presents an obvious difﬁculty for anyone sharing deterministic convictions. So let us not be distracted by the untoward charge of “obsolescence” and consider in an unbiased manner what determinism would mean for “forming the will freely”. As has been worked out lucidly by Aristotle, determinism is not conﬁned to causal or nomological types.14 Rather, determinism is the thesis that every event is ﬁxed without alternative – “determinatum ad unum” in the words of medieval writers15 – and therefore necessary in some sense, depending on the speciﬁc reasons for this ﬁxation. More precisely, for every pair of (elementary or complex) propositions p/¬p and corresponding states of affairs, the realm of possibilities open is reduced to one, the relevant alternatives being deﬁnitely excluded. Determinism is an ontological thesis, not an epistemic one and thus independent of predictability. It is also a general thesis covering mental as well as physical events. According to such a conception, all of what we think, believe and will, all of our practical and theoretical reasonings are ﬁxed without alternative. Whether this is due to the order of fate or god or to natural laws is immaterial. All that counts is the fact of ﬁxation and its relevant implications. 7. The Augustinian solution Obviously there are problems here, noticed at the latest by Aristotle, Epicurus and the Stoics. Accepting determinism for causal and theological reasons Augustine attempted to solve them.16 To save the ordinary assumption that freedom of action implies alternative options he invented the conditional analysis of the concept of practical possibility which allegedly ensures this formally. And to ensure the freedom of the conditioning will, he contrived a formal argument to the effect that volition is free ipso facto. Both of these theoretical devices have been adopted by myriads of theologians and philosophers up to now, at least in part and in a modiﬁed form. Unaware of their theological origin many of the so-called “compatibilists” 14 See in particular De interpretatione, cap. 9. For more details and a defence of an Aristotelian conception of determinism see Seebass 2006, ch. 6, sect. 1 (esp. notes 1−5). 15 See for example Thomas Aquinas, De veritate, q. 22 a. 6; Summa contra gentiles, I: 68; Summa Theologica, I: q. 103 a. 1; 1II: q. 10 a. 4; Duns Scotus, Ordinatio, II: d. 37 q. 2 n. 8, quoted in Sylwanowicz 1996, 199. 16 Augustine, De libero arbitrio, III: 2, 14−4, 41; De civitate Dei, V, 9−10.

234

Intentionality, Deliberation and Autonomy

rely on them in the belief that they are achievements of modern philosophers such as Hobbes, Hume and Moore. However, despite the respectable authority of St Augustine and all of his followers, both moves are ﬂawed. The argument for intrinsic volitional freedom is invalid.17 But without an argument to this effect problems are merely shifted back to volition, albeit perhaps to volitions of a higher order. Moreover, the conditional analysis is no analysis of practical possibility at all as it does not extend to its relevant modal sense. Of course we might say (following Duns Scotus and Leibniz18) that it is not logically impossible to form a different will. Also we might say that this possibility is not ruled out by the mere laws of nature or by the general abilities of ordinary people. However, abstract possibilities such as these are not relevant to the problems we are trying to solve. Certainly we do not care about alternatives existing in some imaginary possible world beyond, but are interested in concrete alternative ways of forming the will that are open to us in the actual world. And such alternatives are ruled out by assuming our world to be deterministic.19 So let us not waste time on the illusive endeavour to ﬁnd some notion of ‘practical’ or ‘volitional possibility’ that is relevant to freedom (in the usual sense explicated above) and ﬁts a deterministic universe. Rather the disillusioning fact is this: once we have subscribed to ontological determinism we have dismissed the possibilities criterion radically. Measured against this standard every kind of freedom will have a zero degree as there is no alternative to the result ﬁxed. Still we are left with the criterion of “naturalness” or “essentialness” which could be sufﬁcient in itself and therefore might be used by determinists to ﬁll the gap. But, leaving aside Augustine’s argument for intrinsic volitional freedom, how could this be provided? 8. Absence of external control First let us try the following line of argument. Obviously one necessary condition of volitional “essentialness” is the absence of a particular form of external control. Suppose the thoughts and volitions of a particular man are determined completely and at every instant by some god or neuropsychiatric “demi-god”. Certainly we would not call the resulting will “essential” as this man is but a marionette. Still one might argue that this is due solely to the fact that his mental processes are manipulated wilfully from outside and not due to the mere fact of determination. Accordingly, one might think that all we would need when taking some person’s will to be “essential” is the guarantee that it is not controlled completely by the will of another person. But this idea is wrong. To see this we only have to suppose that the will of our determining god or neuropsychiatric “demi-god” is determined in turn by some other god or “demi-god”. Surely the mere fact that the determining factors are of a volitional kind is no proof that it is this fact that renders the resulting will 17 See Rowe 1964, 356−63, and Seebass 1997, 239ff. 18 See Söder 1999, 92ff., 171ff., 203, and Leibniz 1967, 62ff.; 1948, 418ff.; 1999, vol. II, 654ff.; Théodicée, sections 36ff., Appendix III; Letter to Pierre Coste 19.12.1707. 19 For an extended argument to this effect and a detailed critique of the conditional analysis see Seebass 2006, ch. 7.

Forming the Will Freely

235

“inessential”. Only a will shown to be “essential” and free by a criterion different from the mere absence of external volitional control could be sufﬁcient to stop the regress into relevant antecedent conditions. Moreover, if the determining inﬂuence is complete and external, it does not matter at all whether or not it is volitional. Suppose some person’s brain gets (either by chance, natural determination, or wilful contrivance) under the inﬂuence of some electromagnetic super-ﬁeld determining in every detail its neuronal processes in every detail as well as, supposedly, the entire set of mental events attached to them in an epiphenomenalist way, including volitions. Could we say this person is able to form her will “freely and essentially”? Certainly not. But if so, why should we not come to exactly the same negative conclusion when we consider all such events to be the mere product of the indeﬁnite multiplicity of interacting inﬂuences emanating from the “super ﬁeld” of a deterministic universe? Obviously we are far away from being able to single out processes leading up to “essential” volitions and therefore badly in need of criteria qualiﬁed for this task. 9. General patterns of forming the will Trying to take up this problem within the conﬁnes of determinism, basically two solutions have been suggested. First it is said that the will is free if its formation satisﬁes certain general patterns that can be regarded as “natural” or “essential”. Some philosophers have been content even with the unspeciﬁc condition of being “amenable to reasons”. But this is wholly inadequate as it is not immune even to the classical objections raised against “inessential” volition since Plato. Others, along with jurists, have tried to specify lists of characteristic epistemic defects and restraints the absence of which deﬁnes, allegedly, what “willing freely” means. Alternatively, or in addition to this, it has been suggested that the degree of freedom and “essentialness” depends on the extent to which positive standards of reﬂectiveness, rationality, length and complexity are met when forming the will. All of this is quite plausible as far as it goes and adequate even for practical purposes, say in the courts. But it is surely inadequate theoretically. First, it is indifferent to the question of whether the optative attitudes that are fed into an informed, unrestrained and extended formation process are free and “essential” from the start. Second, it is conﬁned to the average, ignoring what is “essential” for a particular person at a particular time. Third and even more important, why are extended processes of reﬂective reasoning relevant to freedom at all? Normally we would say that engaging in these is the best, and most often only, way to ﬁnd out what alternative options we have, which is in turn a precondition for choosing freely which options we should realize. However, if determinism is true, belief in real alternatives is spurious, or can be no more than a subjectively unavoidable illusion. Objectively (as long since noted by Aristotle20) there would be no point in reasoning processes of that kind, as their entire courses and outcomes are ﬁxed. Some persons

20 Aristotle, De interpretatione, 18b31–3; Ethica Nicomachea, 1112a18−26, 1139b5−11, 1140a31−1140b1, 1141b10f.; Ethica Eudemica, 1226a20−31.

236

Intentionality, Deliberation and Autonomy

would be determined to need a long run-up before coming to a decisive volition, others to take a shortcut. But if so, why should the former course be more “essential” to the determined person than the latter? To make a substantial distinction here we would have to fall back on the strange idea that there exists some (Platonic) ideal standard of what beﬁts human beings to be applied from “beyond”. Moreover, are there not many quick, spontaneous decisions (such as to go to a concert tonight [cf. above, section 5]) that we take to be “essential” quite independent of the fact that they do not result from extended reasonings? Fourthly and ﬁnally, granting even the inherent “essentiality” of certain general patterns as a mere stipulation, we still get no answer to the most central point. Why should we say of some man luckily determined (by god, fate or natural causes) to satisfy them that he has formed an “essential” will in this case any more than in another case where he unluckily happens to be determined not to fulﬁl them? Here at the latest we are deﬁnitely at a loss. Clearly, this shallow-rooted notion of freedom, contrived for practical reasons at best, if not merely to ﬁt in with a determinist bias, is squarely beside the point. 10. Personal characteristics Still there is a second solution, proposed independently or in support of the ﬁrst. Here the idea is to relate the notion of an “essential will” to what we are as a character. Could we not simply say, independently of the speciﬁc patterns of their formation, that certain mental states and processes, including volitions, are “essential” for some person ipso facto if they are taken to be deﬁning? This would be a nice solution indeed. However, what volitions will qualify? Some people foster the idea that there might be some limited stable set of optative and assertoric attitudes to be discovered (by reﬂection or progressive self-experience) to be constitutive of what we are. However, if this were true, we should be able to list at least some of the most basic attitudes supposed to be deﬁnitional. But are you able to do this? And on what grounds could the deﬁning attitudes be selected? Merely because of their factual (genetic, educational or habitual) entrenchment? Hardly. Such attitudes can be taken as well, and are taken frequently, to be mere conditions under which we live or even hindrances to our thinking and forming the will in a way we regard as “essential”. To make them part of our personality, we have to accept them, that is, decisively to “appropriate” them (as Goethe put it) or to “identify ourselves” with them (as has been put more recently by Harry Frankfurt).21 If this is done, however, the resulting accommodation is not normally thought to be a passive reaction but to involve, contrary to determinist claims, an active decision among alternatives actually open. Moreover, although it is true that the possibilities of actively forming our character, or of dissociating ourselves by free decision from what has become habitual and entrenched inactively and without notice are limited, they surely are not inexistent altogether.

21 Goethe, Faust, verses 682f.; Frankfurt 1988, 21f., 166f.

Forming the Will Freely

237

Some authors claim to have found some general, stable volitions that qualify as deﬁnitive. Thus Hobbes thought that preserving biological life is the only unchanging object of our guiding will. Bentham, Freud and many others thought that it is gain and preservation of pleasure. But a short look at human history makes clear that even these very general and unspeciﬁc goals, which are unﬁt to single out individuals anyway, do not qualify as deﬁning personal characteristics. A suicidebomber surely does not care very much for his physical life and pleasure. Personal life is an ongoing dynamic process made up not only of passive, receptive experiences but also of instances of actively forming our attitudes towards facts and possibilities. This is the central element in setting the standards of what is “essential” for us and therefore of hindrances relevant to our freedom. Accordingly, they are subject to change. A volition we ﬁrmly “identify with” at present may well be dissociated later on. And even if some basic volitions were to be stable and deﬁnitive of what we are, it would certainly be absurd to claim that all of our concrete willings and doings are free and unhindered only to the extent they can safely be referred to those. Because he lacks the criterion of an objective (non-illusory) active choice among open alternatives, however, the only way out that a determinist can envisage is the heroic move taken by Kant and Schopenhauer.22 Both of them proposed that the entire set of our prevailing volitions deﬁnes the “empirical character”, indicating successively – to ourselves and to others as well – who we are as a person. Granting determinism, this is a consistent and respectable idea, even though it is highly counter-intuitive. Surely we do not think that the multifarious trivialities and contingencies of our lives are constitutive for us as persons. Moreover, we do not merely register passively whatever volitions happen to crop up in our minds. But let us waive these objections here as the most central problem of the proposal is this: if every prevailing volition is “essential” by deﬁnition, this predicate no longer refers to a distinguishing property which might be used as a criterion. Having lost our ﬁrst – possibilities – criterion of freedom anyway by accepting determinism (see above, section 7), we have lost now our second criterion too. Given a strictly deterministic universe, our ordinary concept of freedom, that is, freedom from relevant hindrances, simply does not apply. And this is true not only concerning freedom of will but also freedom of action as this concept depends (as has been argued above, section 8) on the criterion of “willing essentially” too. All of this leads to the conclusion that a consequent determinist must say, provided he is luckily determined to be consistent, that ‘freedom’ understood in its ordinary sense is spurious objectively and can be held to exist only by subjective illusion. 11. Counterintuitive consequences So we are confronted again with the great old enigma of freewill, which has turned out to be unsolved up to now and unsolvable in principle if one accepts determinism and does not want to give up our ordinary notion of ‘freedom’. Construed in this way

22 Kant, Kritik der reinen Vernunft, B 566ff.; Schopenhauer <1840>/1977, vol. VI, 87ff., 133ff.

238

Intentionality, Deliberation and Autonomy

we must say, after all, that these convictions are incompatible. Accordingly, one of them should be dropped. What about dropping freedom, then, as understood in its usual sense? Could one not argue, contrary to common belief, that this notion is not important to our lives and that this is the reason why the “classical” problem, while being still unsolved, should “recede into obsolescence” (cf. above, section 1)? In fact, many philosophers and theologians have argued that this sweeping move is the best, or at least the only advisable, way out of the “wand’ring mazes” in which we get lost (as Milton put it) when “reasoning high” about freewill, misguided by “vain wisdom” and “false philosophy”.23 Still it is ill-advised. As noted earlier (section 2), dismissing ‘freewill’ in its traditional and ordinary sense would mean underrating badly its signiﬁcance for human self-understanding due to a misconception of, or resigned disengagement from, what it really means to be active. If what we want and do, as well as the ways by which we come to it, are ﬁxed in advance and in every detail, we cannot – lest by illusion – understand ourselves as real actors, that is, as persons having an active inﬂuence by themselves on what actually happens. Much more even than in any mocking caricature drawn of the medieval “corporative state”, here every individual, every group or institution would have their “assigned places” within the structuring social system which in itself would be but a segment within the ﬁxed and all-inclusive order of being. Accordingly, such persons cannot understand themselves, or be taken by others, as rational and responsible actors. And as this condition is implied by the traditional system of normative social control, which differs from merely manipulatory forms of shaping behaviour, this system will become inapplicable, too. In fact the consequences are even more radical. They lead to a fatalist view of life. Of course this does not refer to the so-called fatalism of “lazy reason”, better called “foolish reason”, as it rests on the denial of obvious causal dependencies. This idea is ill-conceived and not consequently deterministic. Fatalism, rightly conceived, includes the mental realm and amounts to the thesis that the course of the world, being ﬁxed throughout, cannot be altered (theoretically) and therefore cannot be inﬂuenced actively (in the relevant practical sense of “can”). Within philosophy this should have been settled ever since Aristotle. More recently Richard Taylor (1962, 56−66; 1963, ch. 5) has renewed the same argument provoking dozens of articles which vainly tried to refute it. Clearly, a consistent determinist must be a fatalist – provided he is happily determined to be consistent. 12. Dismissing underlying prejudices In view of this, philosophers with a temper similar to that of Schlick and Schopenhauer (cf. above, section 1) might grow angry about the fact that there are still so many “ignorants” “spilling ink” in the hopeless endeavour to get around what is obvious by contriving, following the lead of St Augustine, some substitute notion of ‘freedom’ that ﬁts a ﬁxed universe and still preserves an understanding of ourselves as active and responsible persons. 23 Milton, Paradise Lost, II: 555−65.

Forming the Will Freely

239

In fact, these hopeless endeavours are motivated by a double belief, namely (1) that determinism is true, or might well be true, as the universe is conﬁned to what can be accounted for in strictly “naturalistic” terms and therefore must be seen as “causally” or “nomologically closed”, and (2) that the only way to deny determination is to afﬁrm “blind chance”. Now, I am not in a position to prove or refute either determinism or naturalism. But I think that it is possible to deny both of these and that to do so is (at the very least) not as desperate as are the continuing illusions or self-deceptions about the consequences that would result from their truth for rationality, activity, responsibility and personal freedom. Whether microphysical indeterminism, taken to be implied – pace David Bohm – by standard quantum theory, could be of any use to freewill is unclear to me. Some of the relevant proposals, for example those of Kane or Hameroff and Penrose,24 are certainly subject to the objection of “blind chance”. Others might well be better off. But I know of none which is satisfying. So I must leave it with this. Before ending this chapter, however, I would like to draw attention to one further important point. Even conﬁning ourselves to the physical world and natural science, it is amazing to see how many people still believe that determinism, though unproven, must be true. It is not just the existence of quantum physics. Any experimental physicist knows the insuperable dispersions in measurement. Most theories rely on probabilistic rather than deterministic laws if they formulate general laws at all and do not conﬁne themselves to regularities and correlations. All of this holds all the more strongly if we move up to psychological, social and cultural phenomena. Humankind has always lived in a world full of probabilities and contingencies not amenable to deterministic calculation. To believe that this is merely the result of epistemic defects that could be overcome is (at the very best) a useful methodological maxim and apart from that, as a general ontological thesis, simply an ideology. To me the only understandable reason for this is fear of “blind chance”. Surely our world is not full of this, though chance, apparently, cannot be excluded. But anyhow, to deny that something is determined is not in the least equivalent to afﬁrming that it happens “by mere chance”. The false belief to the contrary is one of the big blunders made repeatedly ever since the early critics of Epicurus. The existence of statistically constant marriage and suicide rates, for example, proves neither that individual marriages and suicides are matters of mere chance, nor that some people are determined somehow to marry or to kill themselves (as believed by some of the early nineteenth-century interpreters of “moral statistics”).25 Both kinds of events may well result from undetermined free decisions of individual persons. Indeterministic freedom may not exist in reality, but it is surely not inconceivable or conceivable only at the expense of being “chancy”. This argument against those who are still interested in making sense of our self-understanding as free, active and responsible persons, this bad old ploy at least should disappear completely from the discussion.

24 Hameroff and Penrose 1997, 177−95; Kane 1998, pt. II. 25 See Hacking 1983, 468ff.; 1990, chs 13−15.

240

Intentionality, Deliberation and Autonomy

References Anscombe, G.E.M. (1957), Intention, Oxford: Blackwell. Berlyne, D.E. (1965), Structure and Direction in Thinking, New York: Wiley. Davidson, D. (1980), Essays on Actions and Events, Oxford: Clarendon. Frankfurt, H. (1988), The Importance of What We Care About, Cambridge: Cambridge University Press. Hacking, I. (1983), ‘Nineteenth Century Cracks in the Concept of Determinism’, Journal of the History of Ideas, 44, 455−75. —— (1990), The Taming of Chance, Cambridge: Cambridge University Press. Haggard, P. and M. Eimer (2001), ‘On the Relation between Brain Potentials and the Awareness of Voluntary Movements’, Journal of Consciousness Studies, 8, 47−63. Haggard, P.; S. Clark and J. Kalogeras (2002), ‘Voluntary Action and Conscious Awareness’, Nature: Neuroscience, 5.4, 382−5. Hameroff, R. and R. Penrose (1997), ‘Conscious Events as Orchestrated Space-Time Selections’, in J. Shear (ed.), Explaining Consciousness: The ‘Hard Problem’, Cambridge, MA: MIT Press, pp. 177−95. Hobbes, Th. (1961), The English Works, ed. W. Molesworth, repr. Aalen: Scientia. Hull, C.L. (1943), Principles of Behaviour, New York: Appleton. Kane, R. (1998), The Signiﬁcance of Free Will, Oxford: Oxford University Press. Kant, I. (1902ff.), Gesammelte Schriften, Berlin: Akademie and de Gruyter. Leibniz, G.W. (1948), Textes inédits, ed. G. Grua, Paris: Presses Universitaires de France. —— (1967), Confessio philosophi, ed. O. Saame, Frankfurt am Main: Klostermann. —— (1999), Hauptschriften zur Grundlegung der Philosophie, ed. E. Cassirer, Hamburg: Meiner. Libet, B. (1985), ‘Unconscious Cerebral Initiative and the Role of Conscious Will in Voluntary Action’, The Behavioral and Brain Sciences, 8, 529−39. MacCorquodale, K. and E. Meehl (1948), ‘Hypothetical Constructs and Intervening Variables’, Psychological Review, 55, 95−107. Roughley, N. (forthcoming), Wanting and Intending: Elements of a Philosophy of Practical Mind, Dordrecht: Springer. Rowe, W. (1964), ‘Augustine on Foreknowledge and Free Will’, The Review of Metaphysics, 18, 356−63. Saarinen, R. (1994), Weakness of the Will in Medieval Thought, Leiden: Brill. Schlick, M. (<1930>/1984), Fragen der Ethik, Frankfurt am Main: Suhrkamp. Schopenhauer, A. (<1840>/1977), Preisschrift über die Freiheit des Willens, in Zürcher Ausgabe. Werke in zehn Bänden, vol. V, Zürich: Diogenes. Searle, J.R. (1983), Intentionality, Cambridge: Cambridge University Press. Seebass, G. (1981/1982), ‘Mediation Theory and the Problem of Psychological Discourse on “Inner” Events’, Ratio, 23, 81−95; 24, 29−42. —— (1993), Wollen, Frankfurt am Main: Suhrkamp. —— (1997), ‘When is an Action Free?’, in G. Holmström-Hintikka and R. Tuomela (eds), Contemporary Action Theory, vol. I, Dordrecht: Kluwer, pp. 233−50.

Forming the Will Freely

241

—— (2006), Handlung und Freiheit, Tübingen: Mohr-Siebeck. Söder, J.R. (1999), Kontingenz und Wissen. Die Lehre von den futura contingentia bei Johannes Duns Scotus, Münster: Aschendorff. Sylwanowicz, M. (1996), Contingent Causality and the Foundations of Duns Scotus’ Metaphysics, Leiden: Brill. Taylor, R. (1962), ‘Fatalism’, The Philosophical Review, 71, 56−66. —— (1963), Metaphysics, Englewood Cliffs, NJ: Prentice-Hall. Tugendhat, E. (1976), Vorlesungen zur Einführung in die sprachanalytische Philosophie, Frankfurt am Main: Suhrkamp. Valla, L. (1987), Über den freien Willen / De libero arbitrio , ed. E. Kessler, München: Fink. Von Wright, G.H. (1985), ‘Of Human Freedom’, in The Tanner Lectures on Human Values, Vol. VI, Cambridge: Cambridge University Press.

This page intentionally left blank

Chapter 12

An Action Can be Both Uncaused and Up to the Agent Carl Ginet

This chapter defends the claim that it is possible for there to be an action that was uncaused and also such that it was up to the agent at the time of the action whether it would occur. The paper criticizes various ways in which some thinkers have been led to deny the claim. In the end the chapter comes to the pessimistic conclusion that, not only is it possible for an uncaused action to have been up to the agent (at the time of the action), if the action was up to the agent then it must have been uncaused. I am an incompatibilist. I hold that it is not metaphysically possible for there to be free and morally responsible action in a deterministic world. I also believe that it is metaphysically possible for there to be free and responsible action in a world that is not deterministic – speciﬁcally, I believe it is possible for there to be free and responsible actions that are uncaused. Moreover I see no good reason to doubt that this is possible, but here I seem to be at odds with some other incompatibilists who think they see reason to deny that breaches of determinism can permit free and responsible action any better than determinism can.1 So they think there is reason to deny that it is metaphysically possible for there to be free and responsible action at all. Of course some compatibilists have also seen reason to deny that breaches of determinism would make possible free and responsible action; and some of these have even held that responsible action must be deterministically caused.2 I wish here to consider some attempts to persuade us that no sort of breach of determinism would permit the occurrence of free and responsible action and to explain why, with me at any rate, they do not succeed. Speciﬁcally, I want to respond to the claim that an uncaused action cannot be an action that was up to the agent, an action about which the agent had a choice.

1 2

For example: G. Strawson 1994; 2000; van Inwagen 2000; Smilansky 2000. For example: Hobart 1934; Nowell-Smith 1948.

244

Intentionality, Deliberation and Autonomy

1. Preliminaries First some preliminaries concerning my assumptions about action, about causation, deterministic and indeterministic, and about the notion of an event’s being up to someone. About the nature of action I will here assume without argument a number of things: (1) Actions can occur in a deterministic world: it is not in the very nature of an action that it must be free and responsible. (2) What makes an event of which a person is the subject an action of that person is something intrinsic to it and not how that event was caused. (3) The causally simplest events that count as actions are mental events, like decisions and volitions. (4) The causally simplest events that count as actions – or the events in the brain that realize such actions – may turn out to have internal causal structure but this is not required by the concept of an action. With respect to the causation of an event by other events I take there to be three mutually exclusive and exhaustive possibilities: an event is (1) deterministically caused, (2) indeterministically caused, or (3) uncaused. (1) An event e was deterministically caused just in case there was some antecedent (and perhaps very complex) event3 c such that c caused e, and c instantiates an event-type C and e instantiates an event-type E such that it is a causal law that an event of type C always causally produces an event of type E. (2) An event e was indeterministically caused just in case e was not deterministically caused but there was some antecedent event c such that c caused e. (3) An event was uncaused just in case it was neither deterministically nor indeterministically caused. To suppose it is possible for there to be indeterministic causation is to suppose that causation does not reduce, Humean fashion, to universal regularity but is rather a brute relation among particular events, a relation of production, a relation that may be impossible to specify in non-synonymous terms. This realist view of causation allows that there can be deterministic causal laws, laws that say that whenever an event of a certain kind occurs, it always causally produces a certain kind of event. It just denies that the very notion of causation requires that all causal laws must be exceptionless generalizations, must be deterministic. Of course, on the strictly Humean view, that there can be no such thing as real causation and that laws of nature cannot be about what causes what but only about what succeeds what, no 3 I take the notion of an event to cover not only changes but also states or conditions of things at times.

An Action Can be Both Uncaused and Up to the Agent

245

threat to free will and moral responsibility can arise from considerations about how extensive causation may be in the world.4 Suppose that an event of a certain sort, for example, a certain sort of action of mine, has occurred. What is it for it to have been up to me (in the sense that interests us here) whether that event would occur? More speciﬁcally what is it for it to have been up to me at a certain time whether that event would occur? I will not attempt a complete answer to this. For present purposes it sufﬁces to note the following. It was up to me at time T whether that event would occur only if I made it the case that it occurred and it was open to me at T to keep it from occurring; and it cannot have been up to me at T to keep it from occurring if whether it occurred depended entirely on facts in place by T; that is, it was up to me at T whether it would occur only if whether it would occur did not depend entirely on facts already in being at T. Of course, it can happen that at one time it was up to me whether a particular event would occur, but at some time before the event occurred it ceased to be up to me whether it would occur. But it seems clear that for any event, if it was up to me whether that event would occur, then that event must have been ultimately dependent on some action (or state of inaction) of mine such that it was up to me up until the time of its occurrence whether that action (or state of inaction) would occur. Hereafter, unless I specify otherwise, when I speak of its having been up to an agent whether a particular action of hers occurred – leaving out any mention of a time at which this was up to her – it should be taken to mean that it was up to her up until the time of the action. Finally, there is another unargued assumption I am making, having to do with the possibility of agent causation. Should it turn out that the causally simplest actions have internal causal structure and, moreover, the causal relation involved cannot be understood as a causal relation between events but must be understood as a brute causal relation between the agent qua substance as cause and some event as effect (and it is not clear to me that this is even a conceptual possibility), this would make no difference to the issue being addressed here. The question would still arise: can an event that is an action, whatever its internal structure may be, both be itself uncaused and be up to the agent? And the reasons some have suggested for answering no to this question would still be just as good or bad. 2. My claim and some bad reasons for denying it My claim, then, is this: (U) It is possible for there to be an action that was uncaused and also such that it was up to the agent at the time of the action whether it would occur. 4 I suppose that when experts in quantum mechanics speak of an indeterminate state “resolving itself” into, or becoming, one or another determinate state they mean to imply that the indeterminate state causally produced the ensuing determinate state, but I am very far from being an expert on quantum mechanics and I could be wrong about that. Such a statement is susceptible of a reading on which it implies no causal relation but means merely that the indeterminate state was followed by the determinate state of affairs.

Intentionality, Deliberation and Autonomy

246

This claim may strike some as implausible on its face. But I do not see that it should. In fact, I ﬁnd a prima facie case for accepting (U) in the fact that in my experience of acting it is typically the case that as I act I have the impression that it is up to me which way I act but I lack any impression of something causing me to act as I do. For example, as I gesture with my hand it seems to me to be up to me in what direction I move my hand and with what force, but it does not seem to me that something causes me to move it in the direction and with the force I move it. It would be odd if the content of this impression entailed something of which I have no impression. I admit, of course, that this is not impossible. This is a merely prima facie case against the entailment, but it is enough, I think, to make it reasonable for me to accept (U) until I become aware of good reason to doubt it. And as yet I have not. But I am aware of some bad reasons that do, or may, tempt some to doubt (U). Let me discuss four of these. 2.1 First bad reason Some may hastily generalize from what is true of events other than actions to what is true of all events, including actions. It does seem very plausible to suppose, and I am not inclined to dispute, that if an event that was not an action of mine was uncaused, then it could not have been up to me whether that event occurred, because it could not have been that I made it the case that the event occurred, given that it was uncaused. For me to make it the case that an event that is not an action of mine occurs is for me to cause that event by means of my action. But it does not follow that the same holds of all uncaused events, even those that are actions of mine. The reasoning I am claiming to be unsound can be spelled out this way: (1) For any event e that has actually occurred, it was up to an agent S whether e occurred only if S made it the case that e occurred. (2) For any event e, S made it the case that e occurred only if S caused e to occur. (3) If S caused e to occur, then e was not uncaused. (4) Therefore, an uncaused action cannot have been one such that it was up to the agent whether it would occur. The false premise here is (2);5 and my conjecture as to why it may nevertheless seem evident to some is that they fail to distinguish it from the claim that, for any event e that was not an action of S’s, S made it the case that e occurred only if S caused e to occur. This latter is extremely plausible. And it may well be that virtually all 5 It might be suggested that there is a weaker sense of “up to S at T” in which premise (1) is also false, a sense in which it is enough for it to have been up to S at T whether an event would occur that it had been open to S at T to keep it from occurring. If there is such a weaker sense then this argument is much worse against the weaker interpretation of (U) than it is against the stronger. But the same does not hold for the remaining three arguments against (U), to be considered below. They are as plausible for the weaker interpretation of (U) as they are for the stronger.

An Action Can be Both Uncaused and Up to the Agent

247

of the events concerning which we have occasion in ordinary life to ask whether a particular agent made them occur, or whether it was up to a particular agent whether they would occur, are events other than the actions of that agent. If so, that might explain a pre-reﬂective disposition to (mistakenly) infer from ‘S made it happen’ to ‘S caused it to happen’ even where the ‘it’ in question is an action of S’s. But, on reﬂection, we should see, I think, that we have no good reason to extend the extremely plausible generalization in the way that premise (2) does, to cover the agent’s own actions. For it seems evident to me that, given that an action was uncaused, all its agent had to do to make it the case that she performed that action was to perform it. If my deciding to vote for the motion, for example, was uncaused, then it follows that nothing other than me made it the case that I decided to vote for the motion, and it also follows that I made it the case that I so decided: I did so simply by so deciding. If my raising my hand was uncaused (that is, nothing other than me determined or made it the case that I raised my hand), then I made it the case that I raised my hand simply by raising my hand. (But the distinct event of my hand’s rising, which is a proper part of my raising my hand, is an event I did cause, by means of a mental activity of volition.6) 2.2 Second bad reason Another thing that appears to lead many philosophers to deny (U) is the thought that ‘uncaused’ implies ‘random’, or ‘by chance’, or ‘a matter of luck’, which in turn implies ‘not made the case by anyone’ and ‘not up to anyone’. Let me cite a few examples of this sort of thinking. Hume wrote that “… liberty, when opposed to necessity, is the same thing as chance …” (Hume 1902, sect. 8, pt 1). By “necessity” Hume here means causal determination, so he is saying that an event that was not causally determined would be one that happens by chance (and he does not restrict this generalization to events that are not actions). A.J. Ayer commits himself (Ayer 1959) to the proposition that if my acting as I do is not causally determined then it is an accident and a matter of chance that I act as I do.7 J.J.C. Smart suggests (Smart 1968, 300–301) that “acting from reasons is not merely random precisely because [and, presumably, only because] it is also acting from causes.”8 And Smart claims (Smart 1961, 291–306), without real argument, that 6 7

See Ginet 1990, ch. 2. The relevant passage reads as follows:

Either it is an accident that I choose to act as I do or it is not. If it is an accident, then it is merely a matter of chance that I did not choose otherwise; and if it is merely a matter of chance that I did not choose otherwise, it is surely irrational to hold me morally responsible for choosing as I did. But if it is not an accident that I choose to do one thing rather than another, then presumably there is some causal explanation of my choice: and in that case we are led back to determinism (Ayer 1959, 275).

8 The only argument Smart suggests for this is the observation that a computer programmed to select items from a set according to certain criteria can be said to have been “programmed to act in accordance with what we would call ‘good reasons’”. But the proposition that there can be phenomena acceptably described as “acting for reasons” in which

248

Intentionality, Deliberation and Autonomy

any acceptable deﬁnitions of “unbroken causal continuity” and “pure chance” must yield the result that the propositions ‘This event happened as a result of unbroken causal continuity’ and ‘This event happened by pure chance’ must be contradictories, that is, they cannot both be false. Bruce Waller (in Waller 1988 as quoted and paraphrased by Kane 1996, 171) says: Suppose two persons had exactly the same pasts and made exactly the same efforts of will, but one does the moral or prudential thing while the other does not. Given that their pasts were exactly the same up to the moment of choice, as indeterminism [or the action’s being uncaused] requires, wouldn’t that mean that the outcome was a matter of luck? … Would there then be any grounds for distinguishing between [them], for saying that one deserves censure for a selﬁsh decision and the other deserves praise for generosity? If they are really identical, and the difference in their acts results from chance, then it seems irrational to consider one more praiseworthy (or more blameworthy) than the other should be.

Galen Strawson (1994) repeatedly equates being undetermined with being random, as in, for example, his remark that “… it is absurd to suppose that indeterministic or random factors, for which one is ex hypothesi in no way responsible, can in themselves contribute in any way to one’s being truly morally responsible for how one is” (1994, 7). Alfred Mele (1999) claims that if nothing about an agent’s powers, capacities, states of mind, moral character and the like causally determines whether she will act in one way rather than another, then her acting as she does rather than in an alternative way is “just a matter of luck”.9 What does “random” or “by chance” or “a matter of luck” mean in claims like these? If it means only that the event was not predictable from antecedent causes, then it is evident that an uncaused action is a “random” event; but then it is very far from evident, and in need of argument these authors do not supply, that it cannot have been up to the agent whether a “random” action occurred and cannot have been the case that the agent made a “random” action occur. Let us grant, however, that there is a sense of “random” or “by chance” in which to say that an event was random or happened by chance is to imply not only that it was not predictable from causes, but also that no one made it happen and it was not up to anyone to make it happen, and that, therefore, for anyone who cared whether the event happened it was a matter of good or bad luck that it did. Then, using this sense of “random”, what the reasons cause the action does not entail that everything properly called acting for reasons must be a case in which the reasons caused the action, much less that an uncaused action must be a merely random event. 9 Mele considers an agent who fails to resist temptation in one possible world but who, in an alternative possible world where the laws of nature are the same and where she is in exactly the same situation, succeeds in resisting temptation. Referring to this agent in the ﬁrst world as “Ann” and in the second world as “Ann*”, he says: “If Ann’s effort to resist temptation fails where Ann*’s effort succeeds, and there is nothing about the agents’ powers, capacities, states of mind, moral character, and the like that explains [by which Mele means “causally determines”] this difference in outcome, then the difference really is just a matter of luck” (Mele 1999, 99).

An Action Can be Both Uncaused and Up to the Agent

249

is very far from evident and what these authors fail to argue for is that an uncaused action must be random. (Their thought here may be an instance of the ﬁrst error described above, generalizing from what is true of events that are not actions to all events, including actions; that is, they do not see that actions must be the unique exception to the principle that undetermined means random.) One author I know of who has tried to argue for the thesis that an uncaused action must be random is Peter van Inwagen. He writes: Suppose … that … Alice was faced with a choice between lying and telling the truth and that she freely chose to tell the truth … And let us assume that … Alice’s telling the truth … was … undetermined [that is, not deterministically caused]. Now suppose that immediately after Alice told the truth, God caused the universe to revert to precisely its state one minute before Alice told the truth … What would have happened the second time? … One can say only that she might have lied and she might have told the truth. … suppose that God a thousand times caused the universe to revert to exactly the state it was in [one minute before Alice told the truth]. Suppose that [after a thousand replays] Alice has told the truth four hundred and ninety-three times and has lied ﬁve hundred and eight times. Is it not true that as we watch the number of replays increase, we shall become convinced that what will happen in the next replay is a matter of chance? (Van Inwagen 2000, 14–15)

He clearly thinks this is a rhetorical question to which the answer is “yes”.10 But I do not see why he thinks this. If I contemplate just one Alice making an uncaused choice, I fail to see how the proposition that the choice was causally undetermined entails that it was random and not up to Alice which choice she made; and I quite fail to see how supposing there to be a great many duplicates of Alice in duplicate situations, sometimes making the same choice as Alice and sometimes making a different one, should make me any more inclined to think with regard to any one of these choices that its being undetermined entails that it is random and not up to its subject. Later in the same essay van Inwagen writes: You are a candidate for public ofﬁce, and I, your best friend, know some discreditable fact about your past that, if made public, would – and should – cost you the election. … You beg me not to tell. … I know that … the objective, “ground ﬂoor” probability of my “telling” is 0.43 and that the objective “ground ﬂoor” probability of my keeping silent is 0.57. Am I in a position to promise you that I will keep silent? − knowing, as I do, that if there were a million perfect duplicates of me, each placed in a perfect duplicate of my present situation, forty three percent of them would tell all and ﬁfty seven percent of them would hold their tongues? I do not see how, in good conscience, I could make this promise. I do not see how I could be in a position to make it. But if I believe that I am able to keep silent, I should, it would seem, regard myself as being in a position to make this promise. What more do I need to regard myself as being in a position to promise to do X than a belief that I am able to do X? Therefore, in this position, I should not regard myself as being able to keep silent. (Ibid., 17)

10 In van Inwagen 1983 (p. 128), however, he declared doubtful the assertion that if our acts are undetermined they are mere “random” or “chance” events.

250

Intentionality, Deliberation and Autonomy

It is not clear to me what van Inwagen means here by “objective ‘ground ﬂoor’ probability”. The phrase suggests that he is thinking of the situation as one in which the action in question will fail to be causally determined through being indeterministically caused rather than being just uncaused. If the action is uncaused then I do not see how a notion of objective probability could apply to it, if, as I suppose, objective probability implies more than just long run frequency. If van Inwagen’s action of keeping silent, supposing that is what he did, was uncaused then, in the frequency sense of probability, it would have meaning to say that the action had a 57 per cent chance of occurring, if and only if, for example, there were or will be many twin situations and in 57 per cent of them he or his counterpart keeps silent. To suppose that van Inwagen’s keeping silent will be uncaused, and it is only in the frequency sense that it is just 57 per cent probable that he will do so, is to suppose nothing that should make him think that it will not be up to him whether he does so, nothing that should give him any qualms about promising to do so. But it is another story if his keeping silent (or telling) is always indeterministically caused. An instance of indeterministic causation may be supposed to be governed by a law of nature which says that a certain kind of circumstance is such that in a certain percentage of cases of its occurrence it will cause one kind of effect, in another percentage of cases a different kind of effect, and so on (where all the percentages add up to 100 per cent). And it is plausible to think of such a law as attributing to the circumstance in question, to its very nature, a degree of propensity to produce each of the effects the law mentions, and the measure of that degree is an objective probability. If this is what indeterministic causation involves, then I think that there is a plausible argument from the premise that the promised action was indeterministically caused to the conclusion that it was not up to the agent whether she would so act if it was not up to her whether she would be in the circumstance that caused the action. Interestingly enough, the argument I have in mind, which will be explained later, seems to be suggested by van Inwagen in his discussion of “the third strand of the Mind argument” (van Inwagen 1983). So it is perhaps not too wild to conjecture, on the basis of van Inwagen’s talk of objective probability, that this argument was at work in his thinking in this later passage I have just quoted (from van Inwagen 2000). However that may be, it is important to note that the conclusion of that argument – that an indeterministically caused action cannot have been up to the agent – does not entail the claim van Inwagen makes in this later passage – that an undetermined action cannot have been up to the agent – because being undetermined does not entail being indeterministically caused: an undetermined event may be so because it is uncaused altogether. I should mention that there is one point that van Inwagen (in ibid., 2000) makes with which I strongly agree, namely, that the question of whether free and responsible action involves agent-causation is irrelevant to the issue of whether (U) holds. The supposition that an action was agent-caused, or consisted in an agent-causing, does not support either accepting or denying the possibility of the agent-causing event’s being uncaused and also up to the agent. Those whose intuitions lead them to deny (U) will of course say that if an agent-causing event is uncaused then it cannot have been up to the agent whether it occurred. And those, like me, whose intuitions lead

An Action Can be Both Uncaused and Up to the Agent

251

them to accept (U) will say that it can be up to the agent whether his uncaused agentcausings occur. Bringing agent-causation into the picture does not resolve the issue but just transfers it to agent-causing events. 2.3 Third bad reason Another argument for the denial of (U)11 takes as its premises that (a) an action could have been up to the agent only if it has an explanation in terms of the agent’s motives or reasons; and (b) its being uncaused would preclude such an explanation. I doubt (a) but will not pursue that issue here. It is near to being a moot point, because I certainly grant that most of our actions that we ordinarily regard (or would like to regard) as free and responsible actions that were up to the agent are ones that have explanations in terms of the agent’s reasons for performing them. I deny (b). I can here only brieﬂy explain why. I have argued,12 and continue to believe, that reasons explanations of actions need not be causal explanations; that, for example, the reasons explanation expressed in (1) below does not entail the causal claim expressed in (2). (1) My reason for doing A was that I wanted to obtain B and believed that by doing A I would obtain B. (2) My desire to obtain B, together with my belief that by doing A I would obtain B, caused my doing A. There is a condition that, as it seems to me, is clearly sufﬁcient for the truth of (1) but that does not entail the truth of (2) and that is compatible with its being the case that my doing A was uncaused, namely, the following: (3) Concurrently with my doing A, I intended of that doing of A that by it (and in virtue of its being a doing of A) I would satisfy (or contribute to satisfying) my desire to obtain B. It is compatible with the truth of this condition that my doing A was neither deterministically caused nor indeterministically caused. That is unlikely to be controversial. My more controversial claim is that this condition is sufﬁcient for the truth of the reasons explanation expressed in (1), namely, that my reason for doing A was that I wanted to obtain B and believed that by doing A I would obtain B. Some do not ﬁnd this claim intuitively compelling, but I have yet to see an argument against it that does not boil down to endorsing the contradictory intuition. In favour of my claim let me put forward the following consideration (which, I realize, falls short of being a compelling argument): I know with certainty that a condition of the form of 11 Found in Hobart 1934; Nowell-Smith 1948; Ayer 1959; Smart 1961. 12 See Ginet 1990, ch. 6 and Ginet 2002.

252

Intentionality, Deliberation and Autonomy

(3) holds for many of my actions and that the corresponding reasons explanation of the form of (1) is true. I think, for example, of the countless times when I know that my reason for pressing a certain button on my computer was in order to satisfy my desire to turn the computer on, and I also know that while I was pressing the button I intended of that action that it would satisfy that desire. I do not know with like certainty that those button-pressing actions were caused by a desire or intention of mine or even that they were caused at all (though there is no doubt some evidence suggesting that they were). Were we to discover somehow (perhaps to our surprise) that those actions were not caused, I would not feel in the least obliged to doubt the truth of the explanations that I performed those button-pressings for the reason that I wanted to turn on the computer. So I conclude that one should not take as a ground for denying that an uncaused action can be up to the agent the claim that an uncaused action cannot be one that the agent does for a reason. For there is no good reason to accept this latter claim. 2.4 Fourth bad reason The fourth and last way I will consider in which someone may be led to deny (U), is by a failure to appreciate adequately the difference between (U) and the following similar claim: (I) It is possible for there to be an action that was indeterministically caused by an antecedent event and also such that it was up to the agent at the time of the action whether that action would be caused by that antecedent event. They uncritically assume that (U) and (I) must stand or fall together. This may be because the longstanding focus in philosophical discussion of free will on whether determinism is compatible with free and responsible action has led them to overlook the possibility that, with respect to the issue of whether an undetermined action can be up to the agent, there may be a signiﬁcant difference between the two ways in which an action could be undetermined, that is, could fail to be deterministically caused: by being indeterministically caused and by being uncaused. In any case, some who think (U) and (I) stand or fall together also think there is reason to reject (I), and they therefore also reject (U). I have to confess that I too have been guilty of the same unthinking assumption. In my case, since I thought one should accept (U), I thought one should also accept (I). I committed myself to (I) in Ginet 1990 in commenting on an example discussed in van Inwagen 1983, sect. 4.4. Van Inwagen imagines a thief who, on the point of robbing the poor box in a church, decides not to rob it; and he imagines that this decision was caused, but not deterministically, by a certain desire and belief of the agent. About this example he says, … I wish I knew how it could be that … our thief had a choice about whether to repent, given that his repenting was caused, but not determined, by his prior inner states, and given that no other prior state “had anything to do with” – save negatively, in virtue of its noninterference with – his act. (Van Inwagen 1983, 149–50)

An Action Can be Both Uncaused and Up to the Agent

253

And he considers the suggestion that the thief may have had a choice about whether to repent owing to his having had a choice about whether, on the one hand, DB [a certain complex of desire and belief in the thief] caused R [the thief’s repenting], or, on the other, his desire for money and his belief that the poor-box contained money (DB*) jointly caused the event robbing the poorbox (R*). … DB and DB* both actually obtained. [According to the suggestion under consideration], what the thief had a choice about was which of these two potential causes became the actual cause of an effect appropriate to it. … If this account is correct then there are two events its coming to pass that DB causes R and its coming to pass that DB* causes R* such that, though one of them must happen, it’s causally undetermined which will happen, and it will have to be the case that the thief has a choice about which of them will happen. If this were so, I should ﬁnd it very puzzling and I should be at a loss to give an account of it. (Ibid., 239, n. 34)

I responded to this by saying that the proper account is that the thief “determines which of the antecedent motives he acts out of simply by acting in the way recommended by one of them while concurrently remembering the motive and intending his action to satisfy it” (Ginet 1990, 147, n. 22). But this, while true, did not really address van Inwagen’s question; for it is true only if “acting out of such-and-such a motive” has a noncausal sense, that is, it does not imply that the motive caused the action but only that the motive was the agent’s reason for the action. But van Inwagen’s question was: how can it be that the agent has a choice about which of the two competing motives causes his action? If we take him to be using “cause” strictly, then to point out, as I do, that the agent has a choice about which motive becomes his reason for acting as he does is not to say that he has a choice as to which motive causes his action. His acting out of one motive rather than the other – as I here understood this: the one motive rather than the other was his reason for acting as he did – is compatible with neither motive’s having caused his action, if I am right in my claim that there is a sufﬁcient condition for the truth of the statement that a certain motive was the agent’s reason for acting that does not entail that the motive caused the action. My supposing that I was addressing van Inwagen’s question can be explained, I think, only on the hypothesis that I was unthinkingly assuming that (U) and (I) stand or fall together.13 But I have recently come to think that this assumption is wrong. There now seems to me to be a reason to doubt (I) that does not apply to (U). For there certainly is intuitive appeal in the following proposition: it cannot be up to me at any given time to determine whether the state of the world already obtaining at that time will or will not causally produce a certain action of mine immediately after that time, that is, it cannot be open to me then to make it the case that it will and open to me then to 13 This assumption is at work again, much more recently, in Ginet 2003, where I say: “If it is plausible to think that [a] when an action is uncaused, nothing, not even the agent, determines or controls whether that action occurs, it should be plausible to think similarly that [b] when one of a conﬂicting pair of motives indeterministically causes an action, nothing, not even the agent, determines which motive causes the action” (Ginet 2003, 608). This remark was motivated by the thought that it cannot be plausible to afﬁrm or deny one of [a] or [b] without doing the same to the other.

254

Intentionality, Deliberation and Autonomy

make it the case that it will not. Though the laws of nature and the state of the world at T may leave it open at T whether or not the state of the world at T will cause the action in question, it is far from clear that it follows that they leave it up to me to determine whether or not it will. These thoughts bring me to a new understanding of the puzzlement expressed by van Inwagen. Though he does not put it this way, van Inwagen could have been puzzled as to how it could be up to the thief, at the time of his acting, to determine what a state of the world already in being at that time would causally produce. Whether or not this is what actually puzzled van Inwagen, it does now puzzle me. I have no difﬁculty in seeing how I can make it the case that I act in a certain way, rather in some alternative way, in the absence of anything causing me to act in that way − I do it simply by acting in that way. But I do fail to see how I can make it the case that the state I am in will cause me to perform one rather than another of the alternative actions that the laws of nature permit that state to cause. The problem with its having been up to me at the time of the action whether or not an indeterministically caused action would occur is not the chance involved in its coming about by indeterministic causation, but rather the causation involved. It is the same problem with deterministic causation of an action. If the state of the world up to T causes my action at T, whether deterministically or indeterministically, then it was (as it were) “up to” that state of the world, and not up to me, what action it caused at T. I did not make it the case that the state of the world at T caused that action; and it was not open to me to make it the case that it caused some other action, nor was it open to me to make it the case that it did not cause any action. If the causation was indeterministic then the state of the world “decided”, but either way at T it was entirely up to that state of the world – it was “in its hands” and not mine – what it would produce. The upshot of these reﬂections is this: the title of this chapter is too weak. Not only is it possible for an uncaused action to have been up to the agent (at the time of the action), if the action was up to the agent then it must have been uncaused. If this is right, then there is a disturbing consequence. The popularity of indeterministic theory in fundamental physics gives us some reason to think that typically our actions are not deterministically caused (though it by no means gives us justiﬁcation for claiming to know that this is so). But a reason to think that our actions are not deterministically caused is not an equally strong reason to think that they are not caused at all. For there is another way in which they may fail to be deterministically caused, namely, by being indeterministically caused. Our evidence does not rule out the possibility that our actions are indeterministically caused; and that means that, however strongly our evidence supports the proposition that they are either indeterministically caused or uncaused, it does not equally strongly support the proposition that they are uncaused. So, though it is possible, as far as we know, that typically our actions are up to us, we perhaps have less reason than we might have thought for believing that this is actually the case, and certainly less reason than we would like to have.14 14 For helpful comments on earlier versions of this chapter I would like to thank Sally McConnell-Ginet, David Widerker and the participants in the discussion of my paper at the Siena conference.

An Action Can be Both Uncaused and Up to the Agent

255

References Ayer, A.J. (1959), ‘Freedom and Necessity’, in A.J. Ayer, Philosophical Essays, London: Macmillan, pp. 271–84. Ginet, C. (1990), On Action, New York: Cambridge University Press. —— (2002), ‘Reasons Explanations of Action: Causalist Versus Noncausalist Accounts’, in R. Kane (ed.), The Oxford Handbook of Free Will, Oxford: Oxford University Press, pp. 386–405. —— (2003), ‘Libertarianism’, in M.J. Loux and D.W. Zimmerman (eds), The Oxford Handbook of Metaphysics, Oxford: Oxford University Press, pp. 587–612. Hobart, R.E. (1934), ‘Free-Will as Involving Determinism and Inconceivable Without It’, Mind, 43, 1–27. Hume, D. (1902), An Enquiry Concerning Human Understanding, ed. L.A. SelbyBigge, Oxford: Oxford University Press. Kane, R. (1996), The Signiﬁcance of Free Will, New York: Oxford University Press. Mele, A. (1999), ‘Kane, Luck, and the Signiﬁcance of Free Will’, Philosophical Explorations, 2, 96–104. Nowell-Smith, P. (1948), ‘Free Will and Moral Responsibility’, Mind, 55, 45–61. Smart, J.J.C. (1961), ‘Free-will, Praise and Blame’, Mind, 70, 291–306. —— (1968), Between Science and Philosophy, New York: Random House. Smilansky, S. (2000), Free Will and Illusion, New York: Oxford University Press. Strawson, G. (1994), ‘The Impossibility of Moral Responsibility’, Philosophical Studies, 75, 5–24. —— (2000), ‘The Unhelpfulness of Determinism’, Philosophy and Phenomenological Research, 60(1), 149–55. Van Inwagen, P. (1983), An Essay on Free Will, Oxford: Oxford University Press. —— (2000), ‘Free Will Remains a Mystery’, Philosophical Perspectives, 14, 1–19. Waller, B. (1988), ‘Free Will Gone Out of Control’, Behaviorism, 16, 149–67.

This page intentionally left blank

Chapter 13

Free Will: Action Theory Meets Neuroscience Alfred R. Mele

This chapter defends the thesis that Benjamin Libet’s data do not justify his claim that “the brain ‘decides’ to initiate or, at least, to prepare to initiate [certain actions] before there is any reportable subjective awareness that such a decision has taken place” and do not justify associated worries about free will. The data are examined in light of familiar distinctions in action theory: for example, the distinction between deciding and wanting and the distinction between intending and wanting. A 1983 article by physiologist Benjamin Libet and colleagues (Libet, Gleason et al. 1983) has been described as “one of the most philosophically challenging papers in modern scientiﬁc psychology” (Haggard et al. 1999, 291). A striking thesis of that 1983 article is that “the brain … ‘decides’ to initiate or, at the least, prepare to initiate [certain actions] at a time before there is any reportable subjective awareness that such a decision has taken place” (Libet, Gleason et al. 1983, 640; also see Libet 1985, 536).1 Libet regards this “decision” as the initiator of an “‘act now’ process” (2001, 61). He pointedly asserts: “If the ‘act now’ process is initiated unconsciously, then conscious free will is not doing it” (2001, 62; also see 2004, 136). Because Libet uses such terms as “decision”, “intention”, “wanting”, “wish” and “urge” interchangeably, some conceptual preliminaries are required in interpreting his work.2 That is the task of section 1. The experiments on which Libet bases his striking claims about decisions and free will are the topic of sections 2 and 3. I argue there that his data do not justify the thesis I quoted and do not justify associated worries about free will. Section 4 discusses the bearing of some reaction time data on Libet’s thesis. 1 In a later article Libet writes: “the brain has begun the speciﬁc preparatory processes for the voluntary act well before the subject is even aware of any wish or intention to act” (1992, 263). 2 Some passages in which two or more of these terms are used interchangeably are quoted in sections 2 and 3 below. Libet, Gleason et al. report that “the subject was asked to note and later report the time of appearance of his conscious awareness of ‘wanting’ to perform a given self-initiated movement. The experience was also described as an ‘urge’ or ‘intention’ or ‘decision’ to move, though subjects usually settled for the words ‘wanting’ or ‘urge’” (1983, 627).

258

Intentionality, Deliberation and Autonomy

1. Some conceptual distinctions Most people recognize that deciding to do something differs from having an urge or wanting to do something. For example, you might have an urge to shout at an irksome colleague but decide not to. And you might want to have a second helping of dessert but decide to stop at one. If, as I believe, to decide to A is to perform a momentary mental action of forming an intention to A (Mele 2003, ch. 9), then in deciding to stop at one dessert you form an intention to stop at one. Your having that intention also differs from your merely wanting to stop at one dessert. After all, you might want to have another dessert (it is very tempting) while also wanting to refrain from having it (you are concerned about your weight); but intending to have a second helping of dessert while intending to refrain from doing that, if it is possible at all, would be a sign of a serious disorder.3 Similarly, you might want to meet one friend at an 8:00 pm concert, want to meet another at an 8:00 pm lecture, and be unsettled about what to do. At this point, you want to do each of these things and lack an intention about which of them to do. In saying that deciding is momentary, I mean to distinguish it from, for example, a combination of deliberating and deciding. Someone who is speaking loosely may say, “I was up all night deciding to quit my job” when what he means is that he was up all night deliberating about whether to quit and eventually decided to quit. Deciding to A, on my view, is not a process but a momentary mental action of forming an intention to A, ‘form’ being understood as an action verb. Not all intentions are formed in acts of deciding. For example, “When I intentionally unlocked my ofﬁce door this morning, I intended to unlock it. But since I am in the habit of unlocking my door in the morning and conditions … were normal, nothing called for a decision to unlock it” (Mele 1992, 231). If I had heard a ﬁght in my ofﬁce, I might have paused to consider whether to unlock the door or walk away, and I might have decided to unlock it. But given the routine nature of my conduct, there is no need to posit an act of intention formation in this case. My intention to unlock the door may have been acquired without having been actively formed. If, as I believe, all decisions are prompted partly by uncertainty about what to do (Mele 2003, ch. 9), in situations in which there is no such uncertainty, no decisions will be made. Some of our decisions and intentions are for the nonimmediate future and others are not. I might decide on Monday to attend a party on Saturday, and I might decide now to phone my daughter now. The intention formed in the former decision is aimed at action ﬁve days in the future. The intention I form when I decide to phone my daughter now is about what to do now. I call intentions and decisions of these kinds, respectively, distal and proximal intentions and decisions (Mele 1992, 143–

3 Try to imagine that you intend to eat some cake now while also intending not to eat it now. How would you act? Would you reach for it with one hand and grab the reaching hand with your other hand? People who suffer from anarchic hand syndrome sometimes display behaviour of this kind (see Marcel 2003, 76–81). Sean Spence and Chris Frith suggest that these people “have conscious ‘intentions to act’ [that] are thwarted by … ‘intentions’ to which the patient does not experience conscious access” (1999, 24).

Free Will: Action Theory Meets Neuroscience

259

4, 158). Proximal decisions and intentions also include decisions and intentions to continue doing something that one is doing and decisions and intentions to start Aing (for example, start climbing a hill) straightaway. A distinction between relatively speciﬁc and relatively unspeciﬁc intentions also is in order. Bob now intends to attend next week’s departmental meeting. That is a more speciﬁc intention than the intention he had, at the beginning of the academic year, to attend at least a few departmental meetings during the year. He had the latter intention without being settled on any speciﬁc meetings to attend. In another illustration, Cathy has agreed to be a subject in an experiment in which subjects are instructed to salute whenever they feel like it on at least forty occasions during a two hour period. When Cathy begins her participation in the experiment she has a relatively unspeciﬁc intention to salute many times during the next two hours. At various times during the experiment she has speciﬁc proximal intentions to salute. 2. Libet’s work This section develops an interpretation of Libet’s work that is sensitive to the conceptual points just made. In some of his studies, subjects are instructed to ﬂex their right wrists or the ﬁngers of their right hands whenever they wish. Electrical readings from the scalp – averaged over at least 40 ﬂexings for each subject – show a “negative shift” in “readiness potentials” (RPs) beginning at about 550 milliseconds (ms) before the time at which an electromyogram shows relevant muscular motion to begin (Libet 1985, 529–30).4 Subjects are also instructed to “recall … the spatial clock position of a revolving spot at the time of [their] initial awareness” (ibid., 529) of something, x, that Libet variously describes as an “intention”, “urge”, “wanting”, “decision”, “will” or “wish” to move (see this chapter, note 2). On average, “RP onset” preceded what subjects reported to be the time of their initial awareness of x (time W) by 350 ms. Time W, then, preceded the beginning of muscle motion by about 200 ms (see Table. 13.1). Table 13.1

Libet’s ﬂexing study

-550 ms RP onset

-200 ms time W

0 ms muscle begins to move

(Libet ﬁnds independent evidence of a slight error in subjects’ recall of the times at which they ﬁrst become aware of sensations [ibid., 531, 534]. Correcting for that error, time W is -150 ms.) At what point, if any, does a speciﬁc intention to ﬂex arise in Libet’s subjects? Again, Libet writes: “the brain ... ‘decides’ to initiate or, at the least, to prepare to initiate the act before there is any reportable subjective awareness that such a decision has taken place” 4 For background on the generation, analysis and use of electroencephalograms (EEGs) and “event-related brain potentials”, including readiness potentials, see Coles and Rugg 1995.

260

Intentionality, Deliberation and Autonomy

(ibid., 536). If we ignore the second disjunct, this quotation (given its context) apparently offers the answer that a speciﬁc intention to ﬂex appears on the scene with “RP onset”, about 550 ms before relevant muscular motion and about 350 to 400 ms before the agent becomes aware of the intention (see ibid., 539); for to decide to initiate an act is to form an intention to initiate it.5 But are decision and intention the most suitable mental items to associate with RP onset? Again, Libet describes the relevant occurrence of which the agent later becomes aware not only as a “decision” and the onset of an “intention” to move, but also as the onset of an “urge”, “wanting” and a “wish” to move. This leaves it open that at -550 ms, rather than acquiring an intention or making a decision of which he is not conscious, the person instead acquires an urge or desire of which he is not conscious – and perhaps an urge or desire that is stronger than any competing urge or desire at the time, a preponderant urge or desire. It is also left open that what emerges around -550 ms is a pretty reliable causal contributor to an urge. I believe that if Libet himself were to distinguish between intending and wanting (including having an urge) along the lines I sketched, he might ﬁnd it more credible to associate the readiness potentials with the latter than with the former. To explain why, I turn to another experiment reported in Libet 1985 (and elsewhere). Libet proposes that “conscious volitional control may operate not to initiate the volitional process but to select and control it, either by permitting or triggering the ﬁnal motor outcome of the unconsciously initiated process or by vetoing the progression to actual motor activation” (1985, 529; also see 1999, 54; 2004, 139, 142–3, 149). “In a veto, the later phase of cerebral motor processing would be blocked, so that actual activation of the motoneurons to the muscles would not occur” (1985, 537). Libet offers two kinds of evidence to support the suggestion about vetoing. One kind is generated by an experiment in which subjects are instructed both to prepare to ﬂex their ﬁngers at a prearranged time (as indicated by a revolving spot on a clock face) and “to veto the developing intention/preparation to act … about 100 to 200 ms before the prearranged clock time” (ibid., 538). Subjects receive both instructions at the same time. Libet writes: a ramplike pre-event potential was still recorded … resembl[ing] the RP of self-initiated acts when preplanning is present. … The form of the ‘veto’ RP differed (in most but not all cases) from those ‘preset’ RPs that were followed by actual movements [in another experiment]; the main negative potential tended to alter in direction (ﬂattening or reversing) at about 150–250 ms before the preset time. … This difference suggests that the conscious veto interfered with the ﬁnal development of RP processes leading to action. … The preparatory cerebral processes associated with an RP can and do develop even when intended motor action is vetoed at approximately the time that conscious intention would normally appear before a voluntary act. (Ibid., 538)6

5 I say “apparently,” because an author may wish to distinguish an intention to ﬂex one’s wrist from an intention to initiate a ﬂexing of one’s wrist. Initiation is discussed in section 3. For completeness, I observe that if we instead ignore the quotation’s ﬁrst disjunct, it makes a claim about when an intention to prepare to ﬂex – or to prepare to initiate a ﬂexing of one’s wrist – arises. 6 For a more thorough discussion of the experiment, see Libet et al. 1983 or Libet, Gleason et al. 1983.

Free Will: Action Theory Meets Neuroscience

261

Keep in mind that the subjects were instructed in advance not to ﬂex their ﬁngers, but to prepare to ﬂex them at the prearranged time and to “veto” this. The subjects intentionally complied with the request. They intended from the beginning not to ﬂex their ﬁngers at the appointed time. So what is indicated by the RP? Presumably, not the acquisition or presence of an intention to ﬂex; for then, at some point in time, the subjects would have both an intention to ﬂex at the prearranged time and an intention not to ﬂex at that time. And how can a normal agent simultaneously be settled on A-ing at t and settled on not A-ing at t?7 That is, it is very plausible that Libet is mistaken in describing what is vetoed as “intended motor action” (ibid., 538; my emphasis). If the RP in the veto scenario is not associated with an intention to ﬂex at the appointed time, with what might it be associated? In the passage I quoted (from Libet 1985, 538), Libet compares “the ‘veto’ RP” with (a) “‘preset’ RPs that were followed by actual movements” and (b) “the RP of self-initiated acts when preplanning is present”. The RP referred to in a is produced in experiments in which subjects are instructed to watch the clock and ﬂex when the revolving spot reaches “a pre-set ‘clock time’” (Libet et al. 1982, 325). “The subject was encouraged to try to make his movement coincide as closely as possible with the arrival of the spot at the pre-set time.” The RP referred to in b is produced in two kinds of studies: studies in which subjects instructed to ﬂex spontaneously are not regularly encouraged to aim for spontaneity (ibid., 324–6), and studies in which subjects instructed to ﬂex spontaneously reported that they experienced “some ‘pre-planning’”, even if only in “a minority of the 40 self-initiated acts that occurred in the series for that averaged RP” (ibid., 328). “Even when some pre-plannings were recalled and reported, subjects insisted that the more speciﬁc urge or intention to actually move did not arise in that pre-planning stage” (ibid., 329). Reports of “pre-planning” seem to include reports of thoughts about when to ﬂex and reports of anticipations of ﬂexing (ibid., 328–9). Libet and his co-authors remark that “Subject S.B. described his advance feelings [of pre-planning] as ‘pre-tensions’ rather than pre-plannings to act” (ibid., 329). This subject may have meant that he experienced tension that he expected to result in ﬂexing. The RPs referred to in a and b have a very similar form (ibid., 330, 333–4; Libet 1985, 532). RPs with that form are called “type I RPs” (Libet et al. 1982, 326). They have signiﬁcantly earlier onsets than the RPs produced in studies in which subjects instructed to ﬂex spontaneously report that they experienced no “pre-planning” – “type II RPs”. “The form of the ‘veto’ RP” is the form of type I RPs until “about 150–250 ms before the preset time” (Libet 1985, 538). What does the veto group (group V) have in common until that time with the three kinds of subjects who produce type I RPs: those with a pre-set time for ﬂexing (group PS), those who are not regularly encouraged to aim for spontaneity (group N), and those who are regularly encouraged to aim for spontaneity but who report some “pre-planning” (group PP)? Presumably, subjects in group PS are watching the clock with the intention of ﬂexing at the pre-set time. But it certainly does not follow from that and the similar 7

On some abnormal agents, see n. 3.

262

Intentionality, Deliberation and Autonomy

RPs in groups N and PP – and V for a time – that members of each of these groups are watching the clock with a similar intention to ﬂex. For one thing, as I have explained, it is very likely that group V – subjects instructed in advance to prepare to ﬂex and then veto the preparation – are watching the clock without an intention to ﬂex at the targeted time. Given that the members of group V lack this intention, we should look for something that groups V and PS actually have in common that might be signiﬁed by the similarity in the RPs until “about 150–250 ms before the preset time”. One possibility is that members of both groups have urges to ﬂex (or to prepare to ﬂex) – or undergo brain events that are pretty reliable relatively proximal causal contributors to such urges – that are associated with an RP and regularly play a role in generating subsequent ﬂexings in the absence of “vetoing”.8 In the case of group V, perhaps a subject’s wanting to comply with the instructions – including the instruction to prepare to ﬂex at the appointed time – together with his recognition that the time is approaching produces a growing urge to (prepare to) ﬂex, or a pretty reliable causal contributor to such an urge, or a simulation of such an urge, or the motor preparedness typically associated with such an urge. A related possibility is suggested by the observation that “the pattern of brain activity associated with imagining making a movement is very similar to the pattern of activity associated with preparing to make a movement” (Spence and Frith 1999). The instructions given to group V would naturally elicit imagining ﬂexing very soon. Finally, the “ﬂattening or reversing” of the RP “at about 150–250 ms before the preset time” might indicate a consequence of the subject’s “vetoing” his preparation. What about groups N and PP? It is possible that they, along with the subjects in groups PS and V, begin acquiring urges to ﬂex at a greater temporal distance from 0 ms than do subjects instructed to ﬂex spontaneously who report no pre-planning. That difference may be indicated by type I RPs having signiﬁcantly earlier onsets than type II RPs. Another possibility is consistent with this. Earlier, I distinguished proximal from distal intentions, and Libet himself recognizes the distinction (see Libet et al. 1982, 329, 334; Libet 1989, 183–4). Presumably, subjects in group PS respond to the instruction to ﬂex at a pre-set time with an intention to ﬂex at that time. This is a distal intention. As the pre-set time for ﬂexing draws very near, that intention may become, help produce, or be replaced by a proximal intention to ﬂex, an intention to ﬂex now, as one naturally says (see Libet 1989, 183; 1999, 54; 2004, 148). That may happen around the time subjects in group V veto their urge to ﬂex or closer to 0 ms. And it may happen at or around the time subjects in groups N and PP acquire a proximal intention to ﬂex. They may acquire such an intention without having had a distal intention to ﬂex soon: recall that members of group V probably had no distal intention to ﬂex soon and that their RPs are very similar to those of groups N, PP and PS until “about 150–250 ms before the preset time”. All this is consistent with the similarities in RPs in the various groups of subjects, on the assumption that no segment of the RPs before about -150 to -250 ms for subjects in group PS speciﬁcally represents subjects’ distal intentions to ﬂex at the pre-set time – as opposed, for example, to something that such intentions have in common with 8 Another is that they have an intention to prepare to ﬂex, if preparing is understood in such a way that so intending does not entail intending to ﬂex.

Free Will: Action Theory Meets Neuroscience

263

distal urges to ﬂex (or to prepare to ﬂex) at the pre-set time – even though those intentions are present. The main difference between type I and type II RPs, in Patrick Haggard’s words, is that the former have “earlier onsets than” the latter (Haggard and Libet 2001, 49). The earlier onsets may be correlated with earlier acquisitions of urges to ﬂex soon – urges that may be brought on, variously, by the instruction to ﬂex at a pre-set time (group PS), the instruction to prepare to ﬂex at a pre-set time and to veto that later (group V), unsolicited conscious thoughts about when to ﬂex (groups N and PP), or unsolicited conscious anticipations of ﬂexing (groups N and PP). (Of course, it is possible that some such thoughts and anticipations are instead products, in part, of urges to ﬂex soon.) These urge inciters (or perhaps urge products, in the case of some experiences in groups N and PP) are absent in subjects instructed to ﬂex spontaneously who report no “pre-planning” – at least, if their reports are accurate. If type I RPs indicate urges, or urges together with proximal intentions that emerge later than the urges do, the same may be true of type II RPs. The difference in the two kinds of RP may mainly be a matter of when the urge emerges – that is, how long before 0 ms. Once again, Libet describes in a variety of ways the mental item that is indicated by RPs. Even if “intention” and “decision” (to ﬂex) are not apt choices, “urge” and “wanting” are still in the running. If “RP onset” in cases of “spontaneous” ﬂexing indicates the emergence of an urge to ﬂex soon, proximal intentions to ﬂex may emerge at some point between RP onset and time W, at time W, or after time W: at time W the agent may be aware only of an urge that has not yet issued in a proximal intention. Again, Libet asserts that “In a veto, the later phase of cerebral motor processing would be blocked, so that actual activation of the motoneurons to the muscles would not occur” (1985, 537). Perhaps, in non-veto cases, activation of these motoneurons is the direct result of the acquisition of a proximal intention (Gomes 1999, 68, 72; Mele 1997, 322–4). Libet suggests that this activation event occurs between 10 and 90 ms before the muscle begins moving and apparently favours an answer in the 10 to 50 ms range (Libet 1985, 537). Elsewhere, he asserts that the activation event can occur no later than 50 ms before the onset of muscle motion (2004, 137–8). Although I will not make much of the following point, it should be observed that urges that may be correlated with RP onset at -550 ms might not be proximal urges, strictly speaking. Possibly, they are urges to ﬂex very soon, as opposed to urges to ﬂex straightaway. And perhaps they evolve into, or produce, proximal urges. Another possibility is that urges to ﬂex very soon give rise to proximal intentions to ﬂex without ﬁrst evolving into or producing proximal urges to ﬂex. Some disambiguation is in order. A smoker who is rushing toward a smoking section in an airport with the intention of lighting up as soon as he enters it wants to smoke soon. That want or desire has a speciﬁc temporal target – the time at which he enters the smoking section. A smoker walking outside the airport may want to smoke soon without having a speciﬁc time in mind. Libet’s subjects, like the latter smoker, might at times have urges or desires to ﬂex that lack a speciﬁc temporal target. Desires to A very soon, or to A, beginning very soon, in this sense of “very soon”, are roughly proximal action-desires.

Intentionality, Deliberation and Autonomy

264

I have been using a (roughly) proximal urge to ﬂex as an example of something that might be indicated by type II RPs beginning around -550 ms. The alternatives I mentioned are (roughly) proximal urges to (prepare to) ﬂex, brain events that are pretty reliable relatively proximal causal contributors to such urges or to (roughly) proximal urges to ﬂex, relevant motor preparedness, and imagining ﬂexing very soon. It would be helpful to have a name for this collection of alternatives: I opt for the name (roughly) proximal urge*. What I dub the urge* hypothesis is the hypothesis that one or another of these things is indicated by type II RPs beginning around -550 ms. “Urge” sans asterisk continues to mean “urge” – not “urge*”. Libet’s experimental design promotes consciousness of urges and intentions to ﬂex, since his subjects are instructed in advance to be prepared to report on them – or something like them – later, using the clock to pinpoint the time they are ﬁrst noticed. For my purposes, what is of special interest are the relative times of the emergence of a (roughly) proximal urge* to ﬂex, the emergence of a proximal intention to ﬂex, and consciousness of the intention. If RP onset indicates the emergence of proximal, or roughly proximal, urges* to ﬂex, and if acquisitions of corresponding intentions directly activate the motoneurons to the relevant muscles, we have the following picture of subjects encouraged to ﬂex “spontaneously” who report no “pre-planning” – subjects who produce type II RPs: Table 13.2 An alternative hypothesis a. -550 ms:

proximal or roughly proximal urge* to ﬂex emerges

b. -90 to -50 ms:

acquisition of corresponding proximal intention9

c. 0 ms:

muscle begins to move10

Possibly, the intention is consciously acquired. My point here is simply that this ﬁgure is consistent with Libet’s data on type II RPs and on time W. I mentioned that Libet offered a second kind of evidence for “veto control”. Subjects instructed to ﬂex “spontaneously” (in non-veto experiments) “reported that during some of the trials a recallable conscious urge to act appeared but was ‘aborted’ or somehow suppressed before any actual movement occurred; in such cases the subject simply waited for another urge to appear, which, when consummated, constituted the actual event whose RP was recorded” (Libet 1985, 538). RPs were not recorded for suppressed urges. But if these urges ﬁt the pattern

9 Recall that Libet suggests that the activation event occurs between 10 and 90 ms before the onset of muscle motion (1985, 537) and later revises the lower limit to 50 ms (2004, 137–8). 10 In an alternative picture, the acquisition of a proximal intention to ﬂex sends a signal that may be regarded as a command to ﬂex one’s wrist (or ﬁnger), and that signal helps produce ﬁner-grained signals that directly activate the motoneurons to the relevant muscles. This picture moves the time of the acquisition of a proximal intention further from 0 ms, but it does not move it anywhere near -550 ms. See section 4.

Free Will: Action Theory Meets Neuroscience

265

of the unsuppressed ones in cases of “spontaneous” ﬂexing, they appeared on the scene about 550 ms before the relevant muscles would have moved if the subjects had not “suppressed” the urges, and subjects did not become conscious of them for about another 350 to 400 ms. Notice that it is urges that these subjects are said to report and abort or suppress. This coheres with my “urge*” hypothesis about groups V, PS, N and PP. In group V (the veto group), as I have explained, there is excellent reason to believe that no proximal intention to ﬂex is present, and the RPs for this group resembled the type I RPs for these other three groups until “about 150–250 ms before the preset time”. If it is assumed that these RPs represent the same thing for these four groups until the RPs for group V diverge from the others, these RPs do not represent a proximal intention to ﬂex before the point of divergence, but they might represent a growing urge to ﬂex or other items in the urge* collection. And if at least until about the time of divergence there is no proximal intention to ﬂex in any of these groups, we would need a special reason to believe that the type II RPs of the spontaneous ﬂexers indicate that proximal intentions to ﬂex emerge in them around -550 ms. Section 4 shows that there is independent evidence that their proximal intentions emerge much later. Does the brain decide to initiate actions “at a time before there is any reportable subjective awareness that such a decision has taken place” (Libet, Gleason et al. 1983, 640)? Libet and his colleagues certainly have not shown that it does, for their data do not show that any such decision has been made before time W or before the time at which their subjects ﬁrst are aware of a decision or intention to ﬂex. Nothing justiﬁes the claim that what a subject becomes aware of at time W is a decision to ﬂex that has already been made or an intention to ﬂex that has already been acquired, as opposed, for example, to an urge to ﬂex that has already arisen. Indeed, the data about vetoing, as I have explained, can reasonably be used to argue that the urge* hypothesis about what the RPs indicate is less implausible than the decision or intention hypothesis. Now, there certainly seems to be a connection between what happens at -550 ms and subsequent muscle motion in cases of “spontaneous” ﬂexing. But it obviously is not a temporally direct connection. Between the former and latter times, subjects apparently form or acquire proximal intentions to ﬂex, in those cases in which they do intentionally ﬂex. And, for all Libet’s data show, those intentions may be consciously formed or acquired. 3. Free will When Libet’s work is applied to the theoretically subtle and complicated issue of free will, things can quickly get out of hand. The abstract of Haggard and Libet 2001 opens as follows: “The problem of free will lies at the heart of modern scientiﬁc studies of consciousness. An inﬂuential series of experiments by Libet has suggested that conscious intentions arise as a result of brain activity. This contrasts with traditional concepts of free will, in which the mind controls the body” (Haggard and Libet 2001, 47). Now, only a certain kind of mind-body dualist would hold that conscious intentions do not “arise as a result of brain activity”. And such dualist views are rarely advocated in contemporary philosophical publications on free will. Moreover,

266

Intentionality, Deliberation and Autonomy

contemporary philosophers who argue for the existence of free will typically shun substance dualism. If Libet’s work is of general interest to philosophers working on free will, the source of the interest must lie elsewhere than the theoretical location speciﬁed in this passage. Libet writes: “it is only the ﬁnal ‘act now’ process that produces the voluntary act. That ‘act now’ process begins in the brain about 550 msec before the act, and it begins unconsciously” (2001, 61).11 “There is”, he says, “an unconscious gap of about 400 msec between the onset of the cerebral process and when the person becomes consciously aware of his/her decision or wish or intention to act.” (Incidentally, a page later, he identiﬁes what the agent becomes aware of as “the intention/wish/urge to act” [ibid., 62].) Libet adds: “If the ‘act now’ process is initiated unconsciously, then conscious free will is not doing it.” I have already explained that Libet has not shown that a decision to ﬂex is made or an intention to ﬂex acquired at -550 ms. But even if the intention emerges much later, that is compatible with an “act now” process having begun at -550 ms. One might say that “the ‘act now’ process” in Libet’s spontaneous subjects begins with the formation or acquisition of a proximal intention to ﬂex, much closer to the onset of muscle motion than -550 ms, or that it begins earlier, with the beginning of a process that issues in the intention.12 We can be ﬂexible about that (just as we can be ﬂexible about whether the process of my baking my frozen pizza began when I turned my oven on to pre-heat it, when I opened the oven door ﬁve minutes later to put the pizza in, when I placed the pizza on the centre rack, or at some other time). Suppose we say that “the ‘act now’ process” begins with the unconscious emergence of an urge to ﬂex – or with a pretty reliable relatively proximal causal contributor to urges to ﬂex – at about -550 ms and that the urge plays a signiﬁcant role in producing a proximal intention to ﬂex many milliseconds later. We can then agree with Libet that, given

11 When does the action begin in all this – that is, the person’s ﬂexing his wrist or ﬁngers? This is a conceptual question, of course: how one answers it depends on one’s answer to the question “What is an action?” Libet identiﬁes “the actual time of the voluntary motor act” with the time “indicated by EMG recorded from the appropriate muscle” (1985, 532). I favour an alternative position, but there is no need to disagree with Libet about this for the purposes of this article. Following Brand 1984, Frederick Adams and I have defended the thesis that overt intentional actions (that is, intentional actions that essentially involve peripheral bodily motion) begin in the brain, just after the acquisition of a proximal intention; the action is proximally initiated by the acquisition of the intention (Adams and Mele 1992). (One virtue of this view is that it helps handle certain problems about deviant causal chains: see Mele 2003, ch. 2). The relevant intention may be understood, in Libet’s words, as an intention “to act now” (1989, 183; 1999, 54; 2004, 148), a proximal intention. (Of course, for Libet, as for me, “now” need not mean “this millisecond.”) If I form the intention now to start running now, the action that is my running may begin just after the intention is formed, even though the relevant muscular motions do not begin until milliseconds later. 12 A central point of disagreement between Haggard and Libet is usefully understood as a disagreement about when the “‘act now’ process” begins (see Haggard and Libet 2001; also see Haggard and Eimer 1999). Haggard apparently views the onset of lateralized response potentials (LRP), which happens “later than RP onset,” as the beginning of the process (Haggard and Libet 2001, 53; also see Trevena and Miller 2002).

Free Will: Action Theory Meets Neuroscience

267

that the “process is initiated unconsciously, … conscious free will is not doing it” – that is, is not initiating “the ‘act now’ process”. But who would have thought that conscious free will has the job of producing urges (or causal contributors to urges)? In the philosophical literature, free will’s primary locus of operation is typically identiﬁed as deciding (or choosing); and for all Libet has shown, if his subjects decide (or choose) to ﬂex “now”, they do so consciously. Libet asks (2001, 62), “How would the ‘conscious self’ initiate a voluntary act if, factually, the process to ‘act now’ is initiated unconsciously?” In this paragraph I offer an answer. One signiﬁcant piece of background is that an “‘act now’ process” that is initiated unconsciously may be aborted by the agent; that apparently is what happens in instances of spontaneous vetoing, if “‘act now’ processes” start when Libet says they do.13 Now, processes have parts, and the various parts of a process may have more and less proximal initiators. A process that is initiated by an unconscious urge* may have a subsequent part that is directly initiated by the conscious formation or acquisition of an intention.14 “The ‘conscious self’” – which need not be understood as something mysterious – might more proximally initiate a voluntary act that is less proximally initiated by an unconscious urge*. (Readers who, like me, prefer to use “self” only as an afﬁx may prefer to say that the acquisition or formation of a relevant proximal intention, which intention is consciously acquired or formed, might more proximally initiate an intentional action that is less proximally initiated by an unconscious urge*.) Recall that Libet himself says that “conscious volitional control may operate … to select and control [‘the volitional process’], either by permitting or triggering the ﬁnal motor outcome of the unconsciously initiated process or by vetoing the progression to actual motor activation” (1985, 529). “Triggering” is a kind of initiating. In “triggering the ﬁnal motor outcome”, the acquisition of a proximal intention would be initiating an action in a more direct way than does the urge* that initiated a process that issued in the intention. According to one view of things, when proximal action-desires help to initiate overt actions they do so by helping to produce pertinent proximal intentions the formation or acquisition of which directly initiates actions (Mele 1992, 71–7, 143–4, 168–70, 176–7, 190–91).15 What Libet says about triggering here coheres with this. Nothing warrants Libet’s claim that type II RPs are correlated with decisions or intentions rather than with urges strong enough to issue pretty regularly in related intentions and actions. Moreover, that, in certain settings, (roughly) proximal urges to do things arise unconsciously – urges on which the agent may or may not act about half a second after they arise – is no cause for worry about free will. Even if 13 Notice that in addition to “vetoing” urges for actions that are not yet in progress, agents can abort attempts, including attempts at relatively temporally “short” actions. When batting, baseball players often successfully halt the motion of their arms while a swing is in progress. Presumably, they acquire or form an intention to stop swinging while they are in the process of executing an intention to swing. 14 Readers who believe that some item or other in the collection designated by “urge*” cannot be unconscious should exclude that item from consideration when they read “unconscious urge*.” 15 Those who view the connection as direct take the view that actions begin in the brain. See n. 11.

268

Intentionality, Deliberation and Autonomy

one grants Libet much more than his critics tend to grant him, as I have done, it can be shown that his data fall well short of providing good grounds for accepting his main theses.16 4. Further testing I have argued that the urge* hypothesis about what the type II RPs indicate in Libet’s studies is less implausible than the decision or intention hypothesis. Is there an independent way to test these hypotheses – that is, to gather evidence about whether it is (roughly) proximal urges* that emerge around -550 ms in Libet’s studies or instead decisions or intentions? One line of thought runs as follows: (1) all overt intentional actions are caused by decisions (or intentions); (2) the type II RPs, which emerge around -550 ms, are correlated with causes of the ﬂexing actions (because they regularly precede the onset of muscle motion); so (3) these RPs indicate that decisions are made (or intentions acquired) at -550 ms. I have shown that this line of thought is unpersuasive. A lot can happen in a causal process that runs for 550 ms, including a subject’s moving from having an unconscious roughly proximal urge* to ﬂex to consciously deciding to ﬂex “now” or to consciously acquiring a proximal intention to ﬂex. One can reply that, even so, (3) might be true. And, of course, I can run through my argumentation about the veto and related matters again to remind the imaginary respondent why (3) is improbable. But what about a test? If makings of proximal decisions to ﬂex or acquisitions of proximal intentions to ﬂex (or the physical events that realize these things) cause muscle motion, how long does it take them to do that? Does it take about 550 ms? Might reaction time experiments show that 550 ms is too long a time for this? Some caution is in order here. In typical reaction time experiments, subjects have decided in advance to perform an assigned task – to “A”, for short – whenever they detect the relevant signal. When they detect the signal, there is no need for a proximal decision to A.17 (If all decisions are responses to uncertainty about what to do and subjects are not uncertain about what to do when they detect the signal, there is no place here for proximal decisions to A.)18 However, it is plausible that after they detect the signal, they acquire an intention to A now, a proximal intention. That is, it is plausible that the combination of their conditional intention to A when they detect the signal (or the neural realizer of that intention) and their detection of the signal (or the neural realizer of that detection) produces a proximal intention to A. The acquisition of this intention (or the neural realization of that event) would then initiate the A-ing.19 And 16 For example, unlike many critics, I did not challenge Libet’s method for timing the relevant conscious experiences. 17 It should not be assumed that detecting the signal is a conscious event (see Prinz 2003). 18 In a reaction time study in which subjects are instructed to A or B when they detect the signal and not to decide in advance which to do, they may decide between A and B after detecting the signal. 19 Hereafter, the parenthetical clauses should be supplied by the reader. Intentions, in my view, are realized in physical states and events, and their causes are or are realized in physical

Free Will: Action Theory Meets Neuroscience

269

in at least one reaction time experiment (described shortly) that is very similar to Libet’s main experiment, the time between the “go” signal and the onset of muscle motion is much shorter than 550 ms. This is evidence that proximal intentions to ﬂex – as opposed to (roughly) proximal urges* to ﬂex – emerge much closer to the time of the onset of muscle motion than 550 ms. There is no reason, in principle, that it should take people any longer to start ﬂexing their wrists when executing a proximal intention to ﬂex in Libet’s studies than it takes them to do this when executing such an intention in a reaction time study. More precisely, there is no reason, in principle, that the interval between proximal intention acquisition and the beginning of muscle motion should be signiﬁcantly different in the two scenarios.20 The line of reasoning that I have just sketched depends on the assumption that, in reaction time studies, proximal intentions to A are at work. An alternative possibility is that the combination of subjects’ conditional intentions to A when they detect the signal and their detection of the signal initiates the A-ing without there being any proximal intention to A. Of course, there is a parallel possibility in the case of Libet’s subjects. Perhaps the combination of their conditional intentions to ﬂex when they next feel like it – conscious intentions, presumably – together with relevant feelings (namely, conscious proximal urges to ﬂex) initiates a ﬂexing without there being any proximal intentions to ﬂex. (They may treat their initial consciousness of the urge as a “go” signal, as suggested in Keller and Heckhausen 1990, 352.) If that possibility is an actuality, then Libet’s thesis is false, of course: there is no intention to ﬂex “now” in his subjects and, therefore, no such intention is produced by the brain before the mind is aware of it. The reaction time study I mentioned is reported in Haggard and Magno 1999: Subjects sat at a computer watching a clock hand … whose rotation period was 2.56 s. … After an unpredictable delay, varying from 2.56 to 8 s, a high-frequency tone … was played over a loudspeaker. This served as a warning stimulus for the subsequent reaction. 900 ms after the warning stimulus onset, a second tone … was played. [It] served as the go signal. Subjects were instructed to respond as rapidly as possible to the go signal with a right-key press on a computer mouse button. Subjects were instructed not to anticipate the go stimulus and were reprimanded if they responded on catch trials. (Haggard and Magno 1999, 103)

“Reaction times were calculated by examining the EMG signal for the onset of the ﬁrst sustained burst of muscle activity occurring after the go signal” (ibid., 104). “Reaction time” here, then, starts before any intention to press “now” is acquired: obviously, it takes some time to detect the signal, and if detection of the signal helps to produce a proximal intention, that takes some time too. The mean of the subjects’ median reaction times in the control trials was 231 ms (ibid., 104). If a proximal intention to press was acquired, that happened nearer to the time of muscle motion states and events. I leave it open here that although intentions enter into causal explanations of actions, the causal work is done, not by them (qua intentions), but by their physical realizers. I forego discussion of the metaphysics of mental causation, but see Mele 1992, ch. 2. 20 Notice that the interval at issue is distinct from intervals between the time of the occurrence of events that cause proximal intentions and the time of intention acquisition.

270

Intentionality, Deliberation and Autonomy

than 231 ms and, therefore, much nearer than the 550 ms that Libet claims is the time proximal intentions to ﬂex are unconsciously acquired in his studies. Notice also how close we are getting to Libet’s time W, his subjects’ reported time of their initial awareness of something he variously describes as an “intention”, “urge”, “wanting”, “decision”, “will” or “wish” to move (-200 to -150 ms). If proximal intentions to ﬂex are acquired in Libet’s studies, Haggard and Magno’s results make it look like a better bet that they are acquired around time W than that they are acquired around -550 ms.21 How seriously we should take his subjects’ reports of the time of their initial awareness of the urge, intention, or whatever, is a controversial question, and I will say nothing about it here.22 5. Conclusion In a 2002 article, after writing that “many of the world’s leading neuroscientists have not only accepted our ﬁndings and interpretations, but have even enthusiastically praised these achievements and their experimental ingenuity” and naming twenty such people, Libet adds: “It is interesting that most of the negative criticism of our ﬁndings and their implications have come from philosophers and others with no signiﬁcant experience in experimental neuroscience of the brain” (2002, 292). Later in the article, he writes of one of his critics, “As a philosopher Gomes exhibits characteristics often found in philosophers. He seems to think one can offer reinterpretations by making unsupported assumptions, offering speculative data that do not exist and constructing hypotheses that are not even testable” (ibid., 297).23 When I ﬁrst read the latter passage, I experienced an urge to point out that one does not need any “experience in experimental neuroscience of the brain” to realize that there is a difference between deciding and intending, on the one hand, and wanting – including having an urge – on the other. Also, one who understands Libet’s data and the studies that generate them can see that nothing warrants his claim that the RPs at issue are correlated with decisions or intentions rather than with urges strong enough to issue pretty regularly in related intentions and actions or relatively proximal causes of such urges. Incidentally, as is obvious, I eventually made the transition from having an urge to comment on the quoted remarks to intending to do so. Patrick Haggard, in his contribution to a recent discussion with Libet, asserts that “conceptual analysis could help” (Haggard and Libet 2001, 62). This article may be read as a test of his assertion. In my opinion, the result is positive. Attention not only 21 In a study by Day et al. of eight subjects instructed to ﬂex a wrist when they hear a tone, mean reaction time was 125 ms (1989, 653). In their study of ﬁve subjects instructed to ﬂex both wrists when they hear a tone, mean reaction time was 93 ms (ibid., 658). The mean reaction times of both groups of subjects – deﬁned as “the interval from auditory tone to onset of the ﬁrst antagonist EMG burst” (ibid., 651) – were much shorter than those of Haggard and Magno’s subjects. Day’s subjects, unlike Haggard and Magno’s (and Libet’s), were not watching a clock. 22 For an instructive review of the literature on this, see van de Grind 2002. 23 Incidentally, Gilberto Gomes has informed me that he works in a psychology department.

Free Will: Action Theory Meets Neuroscience

271

to the data but also to the concepts in terms of which the data are analysed makes it clear that Libet’s striking claims about decisions, intentions and free will are not justiﬁed by his results. Libet asserts that his “discovery that the brain unconsciously initiates the volitional process well before the person becomes aware of an intention or wish to act voluntarily … clearly has a profound impact on how we view the nature of free will” (2004, 201). Not so. That, in certain settings, (roughly) proximal urges to do things arise unconsciously or issue partly from causes of which the agent is not conscious – urges on which the agent may or may not subsequently act – is a cause neither for worry nor for enthusiasm about free will.24 References Adams, F. and A. Mele (1992), ‘The Intention/Volition Debate’, Canadian Journal of Philosophy, 22, 323–38. Brand, M. (1984), Intending and Acting, Cambridge, MA: MIT Press. Coles, M. and M. Rugg (1995), ‘Event-Related Brain Potentials: An Introduction’, in M. Rugg and M. Coles (eds), Electrophysiology of Mind, Oxford: Oxford University Press, pp. 1–26. Day, B.; J. Rothwell; P. Thompson; A. Maertens de Noordhout; K. Nakashima; K. Shannon and C. Marsden (1989), ‘Delay in the Execution of Voluntary Movement by Electrical or Magnetic Brain Stimulation in Intact Man’, Brain, 112, 649–63. Gomes, G. (1999), ‘Volition and the Readiness Potential’, Journal of Consciousness Studies, 6, 59–76. Haggard, P. and M. Eimer (1999), ‘On the Relation between Brain Potentials and the Awareness of Voluntary Movements’, Experimental Brain Research, 126, 128–33. Haggard, P. and B. Libet (2001), ‘Conscious Intention and Brain Activity’, Journal of Consciousness Studies, 8, 47–63. Haggard, P. and E. Magno (1999), ‘Localising Awareness of Action with Transcranial Magnetic Stimulation’, Experimental Brain Research, 127, 102–7. Haggard, P.; C. Newman and E. Magno (1999), ‘On the Perceived Time of Voluntary Actions’, British Journal of Psychology, 90, 291–303. Keller, I. and H. Heckhausen (1990), ‘Readiness Potentials Preceding Spontaneous Motor Acts: Voluntary vs. Involuntary Control’, Electroencephalography and Clinical Neurophysiology, 76, 351–61. Libet, B. (1985), ‘Unconscious Cerebral Initiative and the Role of Conscious Will in Voluntary Action’, Behavioral and Brain Sciences, 8, 529–66. —— (1989), ‘The Timing of a Subjective Experience’, Behavioral and Brain Sciences, 12, 183–4. —— (1992), ‘The Neural Time-Factor in Perception, Volition and Free Will’, Revue de Métaphysique et de Morale, 2, 255–72.

24 Parts of this chapter derive from Mele 2006, ch. 2. A draft was presented at the University of Siena in March 2005. I am grateful to the audience and this volume’s editor, Christoph Lumer, for helpful feedback.

272

Intentionality, Deliberation and Autonomy

—— (1999), ‘Do We Have Free Will?’, Journal of Consciousness Studies, 6, 47–57. —— (2001), ‘Consciousness, Free Action and the Brain’, Journal of Consciousness Studies, 8, 59–65. —— (2002), ‘The Timing of Mental Events: Libet’s Experimental Findings and Their Implications’, Consciousness and Cognition, 11, 291–9. —— (2004), Mind Time, Cambridge, MA: Harvard University Press. —— C. Gleason; E. Wright and D. Pearl (1983), ‘Time of Unconscious Intention to Act in Relation to Onset of Cerebral Activity (Readiness-Potential)’, Brain, 106, 623–42. —— E. Wright and A. Curtis (1983), ‘Preparation- or Intention-to-act, in Relation to Pre-event Potentials Recorded at the Vertex’, Electroencephalography and Clinical Neurophysiology, 56, 367–72. —— —— and C. Gleason (1982), ‘Readiness Potentials Preceding Unrestricted “Spontaneous” vs. Pre-planned Voluntary Acts’, Electroencephalography and Clinical Neurophysiology, 54, 322–35. Marcel, A. (2003), ‘The Sense of Agency: Awareness and Ownership of Action’, in J. Roessler and N. Eilan (eds), Agency and Self-Awareness, Oxford: Clarendon, pp. 48–93. Mele, A. (1992), Springs of Action: Understanding Intentional Behavior, New York: Oxford University Press. —— (1997) ‘Strength of Motivation and Being in Control: Learning from Libet’, American Philosophical Quarterly, 34, 319–33. —— (2003), Motivation and Agency, New York: Oxford University Press. —— (2006), Free Will and Luck, New York: Oxford University Press. Prinz, W. (2003), ‘How Do We Know About Our Own Actions?’, in S. Maasen, W. Prinz and G. Roth (eds), Voluntary Action, Oxford: Oxford University Press. Spence, Sean and C. Frith (1999), ‘Towards a Functional Anatomy of Volition’, Journal of Consciousness Studies, 6, 11–29. Trevena, J. and J. Miller (2002), ‘Cortical Movement Preparation before and after a Conscious Decision to Move’, Consciousness and Cognition, 11, 162–90. Van de Grind, W. (2002), ‘Physical, Neural, and Mental Timing’, Consciousness and Cognition, 11, 241–64.

Chapter 14

Belief and Moral Responsibility Carlos J. Moya

Compatibilists and incompatibilists agree that control over our choices and actions is a requirement for moral responsibility. Compatibilists conceive of this control as (roughly) rational choice. Incompatibilists insist that it should have an additional feature, which may be called “ultimate control”. Ultimate control, however, seems to raise an impossible demand and so foster scepticism about moral responsibility. I argue, however, that ultimate control can be shown to be attainable provided that two widespread assumptions about it are rejected. The ﬁrst is a will-centred account of control. The second is a profound individualistic view of human beings. Against these assumptions, I defend a cognitive and social, rather than a volitive and individualistic, approach to (ultimate) control and moral responsibility. Evaluative beliefs play a pivotal role in this approach. We can have over these beliefs a control not based on the will, but deep enough to support responsibility attributions. I conclude that this cognitive approach to moral responsibility is in a better position than traditional, will-centred approaches to overcome scepticism about moral responsibility. Compatibilists and incompatibilists agree that a condition of control over one’s choices and actions has to be satisﬁed if one is to be morally responsible for these choices and actions. An agent’s control over her choices and actions should include at least two aspects: some degree of rationality and some degree of autonomy or self-determination. The rationale for including the ﬁrst aspect is that purely random, arbitrary choices or actions do not seem to be sufﬁciently in the agent’s hands (under her control) for her to be justiﬁably judged as blame- or praiseworthy on account of them. This is why they should meet at least some minimal rationality requirements. Satisfaction of the second aspect, in turn, is designed to ensure that the agent is actually the source or author of the choice or action for which she is judged as blame- or praiseworthy. Actions or choices caused by external forces are not justiﬁably attributed to an agent. This much can be considered as common ground. Any acceptable theory of moral responsibility should include some requirement of rational control and some requirement of autonomy or self-rule. Discrepancies, however, start shortly after these minimal points of agreement. Typically, incompatibilists tend to consider compatibilist accounts of the self-determination aspect as not deep enough to ground moral responsibility, while compatibilists tend to view incompatibilist theories, with

274

Intentionality, Deliberation and Autonomy

their insistence on indeterminism as necessary for moral responsibility, as unable to offer a satisfactory account of the rationality aspect. For early compatibilists, in rough terms, an agent exercises control over her actions on the basis of her desires: an agent controls her actions provided that she does what she wants to do. This view of control does include the two aspects we have pointed to: on the one hand, a choice or action caused by desire has a minimal rational explanation; on the other hand, if an agent’s self – as Hume suggested – is partly constituted by her own desires, determination by her own desires is selfdetermination. However, control, so conceived, is arguably too superﬁcial as a basis of moral responsibility attributions.1 It has been replaced by more sophisticated compatibilist accounts, which incorporate deeper levels of control. The desire on which an agent acts has to be backed by a reﬂective, second-order volition (Frankfurt), or by the agent’s values (Watson), and the agent’s self is conceived as centrally constituted by her second-order volitions or by her values, rather than by her ordinary, ﬁrst-order desires. On other perspectives, the agent has to be able to form objectively correct values and to act on them (Wolf), or to act on a practical reasoning “mechanism” that is her own and appropriately responsive to reasons (Fischer and Ravizza). Incompatibilists may readily acknowledge a progress in compatibilist accounts of control, from the early, simple proposals of Hobbes or Hume to the sophisticated views of Frankfurt, Watson, Wolf or Fischer. But, on their opinion, owing to the project of giving an account of moral responsibility compatible with determinism, even sophisticated views are bound to offer a weak, superﬁcial picture of control, which cannot provide an appropriate ground for moral responsibility. Incompatibilists may agree that rational and volitional control over one’s actions, in some of the ways proposed by compatibilists, may be necessary for moral responsibility, but they will insist that an additional feature is required in order to have something close to a sufﬁcient condition, namely ultimacy. In order for an agent to be morally responsible for her actions, she has to be their true, ultimate origin by having ultimate control over them. An agent enjoys rational and volitional control over her actions by choosing to perform them in the light of such factors as her ﬁrst- and second-order desires, values and even traits of character constitutive of her self. This means that these factors explain, as either necessary or sufﬁcient conditions, why she acted as she did. So, an agent’s control has to extend to these explanatory factors in order for her to effectively control the choices and actions that such factors help to explain and to be truly praise- or blameworthy for such choices and actions. According to the incompatibilist’s intuition, an agent’s rational and volitional control over her actions is too slender a basis for moral responsibility; she also has to control the self that these actions arise from. An agent is truly morally responsible for the way she acts only if she is truly responsible for the way she is. Only then can the agent be said to be the true, ultimate source or origin of her actions and objectively deserve praise or blame for them.

1 Though the point would need more argument, it may sufﬁce to note that control, so understood, would be possessed by higher animals and very young children.

Belief and Moral Responsibility

275

This deep, ultimate control condition for moral responsibility is what Robert Kane has called “ultimate responsibility” and Galen Strawson “true self-determination”. If this actually is a condition for moral responsibility, the sceptical suspicion easily arises that moral responsibility is not possible. Ultimate control involves two aspects, namely ultimacy of source and rational cum volitional control. And it would seem that this condition is incompatible with either determinism or indeterminism. Determinism may allow for rational cum volitional control, but not for ultimacy of source, for, with the possible exception of a ﬁrst, uncaused cause, there are no ultimate sources or origins in a deterministic world. Indeterminism, in turn, allows for events, such as choices, that, being undetermined, can play the role of fresh, ultimate origins or causes, but now it seems that these ultimate causes cannot be under the agent’s rational cum volitional control. If these events, say choices, are explained by previously existent reasons, they can be rational but hardly ultimate causes; and if they are not so explained, they can be ultimate but not rationally controlled causes. In fact, acceptance of deep, ultimate control as a requirement for moral responsibility has led some thinkers, such as Galen Strawson and Derk Pereboom (2001), to take a sceptical stance towards moral responsibility. Strawson’s wellknown sceptical argument goes roughly as follows. Rational actions are paradigmatic candidates to the status of free and responsible actions, if such there are. Now the way we act when we act rationally, that is, for reasons, depends on our mental constitution, or character. So, unless we are truly responsible for our mental constitution, we will not be truly responsible for our rational actions. But in order to be truly responsible for our mental constitution, we have to have chosen that mental constitution in a rational way, that is, in the light of certain principles of choice or reasons. These reasons or principles explain the choice of our mental constitution. So we cannot be truly responsible for this choice, and so for our chosen mental constitution, unless we are responsible for having such principles to begin with; and this in turn requires that we have chosen them rationally, that is, in the light of a further set of principles of choice or reasons, and so on. According to Strawson, then, true responsibility or true self-determination (ultimate control, in our terms) “is logically impossible because it requires the actual completion of an inﬁnite regress of choices of principles of choice” (Strawson 1986, 29). It seems, then, that no choice can be both an ultimate and a rational source of one’s actions. In the end, we are bound to choose and act on the basis of factors that we cannot have rationally chosen and for which we cannot be truly responsible. Ultimate control (true self-determination, in Strawson’s terms) would seem to involve a self-defeating demand for self-creation. As Randolph Clarke has put the point, according to Strawson “rational free action would be possible only for an agent who was causa sui” (Clarke 1997, 37). If something like ultimate control is in fact both necessary for moral responsibility and impossible to attain, then moral responsibility is not possible. And this is precisely Strawson’s position. Robert Kane agrees with Strawson that ultimate control (“ultimate responsibility” in his terms) is necessary for moral responsibility in the deep sense of true desert: only if an agent is the ultimate source of her actions can it be justiﬁed to consider her as truly, objectively praise- or blameworthy for

276

Intentionality, Deliberation and Autonomy

them. However, unlike Strawson, he thinks that ultimate control can be attained.2 According to Kane, an agent can choose rationally and voluntarily her own character and motives and so be truly responsible, in Strawson’s sense, for having them as well as for the choices and actions that they help to explain. Strawson’s challenge immediately arises: how is it possible for an agent to choose her own character and motives rationally and voluntarily unless she already exists endowed with a previous character and motives? Kane is certainly aware of this difﬁculty. He agrees that his view of ultimate responsibility “appears to lead to a vicious regress … The regress would stop with actions that were not explained by our characters and motives (or by anything else, for that matter), but then in what sense would we be responsible for such actions?” (Kane 1996, 37). In order to meet this crucial objection, Kane resorts to certain choices in a person’s life through which she forms her own character and motives. Kane calls these choices “Self-Forming Willings” (SFWs). If SFWs are to stop the regress that threatens ultimate responsibility, they have to satisfy certain conditions. On the one hand, they must have no sufﬁcient explanation in terms of the agent’s pre-existing character, motives and preferences. They have to be genuinely open and undetermined, relative to the past and the natural laws, for otherwise the agent could not be their ultimate source. And, on the other hand, it is also crucial that the choice, whichever way it may go, remains under the agent’s rational and volitional control. It must be a rational and motivated choice, a result of the agent’s rational will. Irrational or arbitrary choices are not an appropriate foundation of moral responsibility, as compatibilists have always contended against incompatibilists. Can SFWs satisfy these conditions? It is tempting, but wrong, to conceive of SFWs as Buridan’s Ass cases, in which the agent has equal reasons for going one way or another, for in these cases, “instead of one choice … being arbitrary relative to the prior deliberation, both would be arbitrary” (ibid., 109). Rather, what the agent confronts in SFWs is incommensurable sets of reasons for going one way or another, such as moral reasons and reasons of self-interest, or prudential considerations and desires for an immediate pleasure. In cases like these, the agent is “torn between conﬂicting internal points of view that represent different and incommensurable visions of what they want in life or what they want to become” (ibid., 199). The agent’s will is unsettled. She wants to act on one set of reasons and she also wants to act on the other. These choices are self-forming in that, by making them, the agent causally contributes to shaping her own character, motives and will, so that she can be said to be ultimately responsible for those psychological factors and so for further choices and actions that can ﬂow from them. Incommensurability of reasons for each option is a crucial feature of SFW situations, which supposedly allows the ﬁnal choice, whichever it is, to be the result of the agent’s rational will. As Kane writes, “for SFWs, each outcome is rational for different and incommensurable reasons” (ibid., 178). Are Kane’s SFWs able to stop the regress that threatens the possibility of ultimate control? Though they go some way towards doing so, I do not think they 2 For purposes of exposition, I shall present Kane’s work as if it were an attempt to respond to Strawson’s argument. Of course it is much more than that.

Belief and Moral Responsibility

277

go far enough. As we have seen, in an SFW the agent confronts a choice that she has to make on the basis of incommensurable reasons. But if reasons are actually incommensurable, how can the agent choose one of the alternatives rationally? A choice of this kind can be rationally made if the agent faces it with some criterion (a meta-criterion, let’s say) that allows her to rank one set of reasons higher than the other. Such a meta-criterion might be, for example, a Frankfurtian second-order volition, according to which she prefers to be moved to act by, say, moral reasons rather than reasons of self-interest, or maybe a Watsonian valuational system, which ranks the former higher than the latter (or vice versa). However, if the agent confronts the choice with a meta-criterion, her will is not unsettled and the choice cannot be truly self-forming. She can have obtained this criterion through prior choices she faced with incommensurable reasons; but then the rationality problem arises again with regard to these choices. Or she just happens to have the criterion, but then she lacks ultimate control over it and the resulting choice. Moreover, remember that, in SFWs, the ﬁnal choice, whatever it is, has to be the result of the agent’s rational will. But this condition will not be met if the agent confronts the choice with a metacriterion, for then only some of the choices will be rational, namely those which accord with the criterion, but not those which conﬂict with it. Therefore an SFW should be made exclusively on the basis of the incommensurable sets of reasons the agent considers. And then it is hard to see how the choice could possibly be under her rational cum volitional control. The role of a meta-criterion is now played by the agent’s pure decision: “The agents will make one set of reasons or motives prevail over the others then and there by deciding … [B]oth options are wanted and the agents will settle the issue of which is wanted more by deciding” (ibid., 133). Kane’s libertarianism tends to turn into sheer voluntarism or decisionism. The agent’s choice can be ultimate at the cost of her losing rational control over it. In fact, the two aspects of ultimate control, namely ultimacy of source and rational cum volitional control, seem to pull in opposite directions. But then, if ultimate control is so seemingly impossible to attain, why insist on it as a necessary condition of moral responsibility? Why not abide by relative, less radical, non-ultimate forms of control or self-determination? Compatibilists have in fact described a large variety of such forms. Would this insistence on ultimate control not pave the way for scepticism? I can think of two main considerations in favour of this condition. First, the requirement of ultimate control corresponds to the depth of moral responsibility attributions. A serious ascription of moral responsibility is directed to the agent herself, on the assumption that she is the true, ultimate origin, and not a mere instrumental or derived cause, of the action or consequence thereof for which we hold her responsible. These attributions have deep effects on our selfesteem and sense of dignity. So the desire for deep personal control over the grounds on which such attributions are made is clearly reasonable. We do not want our worth and value to depend on factors beyond our reach and control. Second, compatibilist construals of the control condition, which dispense with ultimacy, seem capable of being satisﬁed by agents who prima facie do not seem morally responsible, such as Brave New World citizens or agents in “Covert Non-Constraining (CNC) Control” situations (to use Kane’s terms). These construals would appear to be too weak to ground moral responsibility understood as true, objective desert. So understood,

278

Intentionality, Deliberation and Autonomy

moral responsibility would seem to require some form of deep, ultimate control over our choices and actions. I want to suggest that, appearances to the contrary notwithstanding, deep, ultimate control can be shown to be attainable, provided that some unexamined assumptions about this condition, which can be found in many authors, and especially in Strawson and Kane, are brought to light and questioned. Consider Strawson. According to him, true responsibility for, or ultimate control over, one’s actions requires, at the very end, that one has chosen one’s mental constitution, the way “one is, mentally speaking” (Strawson 1986, 28), as well as the principles on which such a choice is made. For Kane, in turn, ultimate responsibility rests upon Self-Forming Willings, undetermined choices by means of which agents build up their own character and motives. At the root of ultimate control over one’s actions we ﬁnd acts of will or choices. So both Strawson and Kane assume as a matter of course a will-centred view of ultimate control and moral responsibility. For them, an agent cannot have deep, ultimate control over her actions and be truly praise- or blameworthy for them unless she has chosen the springs of those actions. This assumption, in turn, seems to rest upon a more general view, namely that one can be said to control only that which one has a choice about. Only something that is subject to one’s will could be said to be under one’s control and so be an appropriate ground or object for moral responsibility attributions. This assumption, together with a plausible rationality requirement for choices, leads quickly to scepticism about ultimate control. If we accept that ultimate control over one’s choices and actions requires that one has chosen the springs of these choices and actions and that this choice has itself to be based on reasons, then we shall have to accept that, in order for us to have ultimate control over this choice, we should have chosen the reasons on which we made it; but this other choice, in turn, should also be based on reasons, which we should have chosen in the light of further reasons, and so on. We have started the regress of choices of principles of choice that Strawson rightly holds to be impossible to complete. Scepticism looms. Another widespread and unexamined assumption in current conceptions of ultimate control is closely connected with the ﬁrst. Consider Strawson’s claim that in order for true responsibility to be possible, one has to have chosen the very roots of one’s choices, the principles on which one makes them. This claim, I would think, presupposes a deeply individualistic view of human agents as radically self-made, self-contained entities, whose moral responsibility is undermined by the inﬂuence of any factors that are external to them and beyond the scope of their choice. This view transpires also in Kane’s conception of ultimate responsibility as a species of selfcreation. From this individualistic point of view, the social nature of human agents tends to appear as a potential threat to their being ultimate sources or origins of their own actions and so to their moral responsibility for them. This second assumption reinforces the sceptical suspicion about moral responsibility already raised by the ﬁrst one. On the background of these assumptions, my position can be stated as follows. At the foundational root of moral responsibility I shall not place conative phenomena, such as choices, but cognitive ones, such as beliefs. Especially important will be a subset of an agent’s beliefs, namely her evaluative views about what is really worth

Belief and Moral Responsibility

279

pursuing or avoiding in life. Beliefs of this sort are plausibly taken to play a central role in explaining our morally relevant choices and acts. I recommend, then, a cognitive, rather than a conative, approach to moral responsibility. I accept, however, Strawson’s and Kane’s contention that deep, ultimate control is a requirement for moral responsibility. Now, on the assumption we have referred to above, that all control depends on choices or acts of will, acceptance of ultimate control seems to conﬂict with the view that moral responsibility rests on beliefs, unless one embraces some version of doxastic voluntarism, a rather implausible position, in my opinion.3 However, although I share the widely held position that control is necessary for responsibility, I reject the no less widespread view that all control depends on the will. So, though belief is not voluntary, we can rightly be praised or blamed for our beliefs, for we can have over them a form of control that does not rest on acts of will or choices and is deep enough to support true praise- and blameworthiness attributions. As I shall try to show, it may be justiﬁed to grant someone deep control and authorship with regard to her factual or theoretical beliefs, as well as full praiseor blameworthiness for them, even if she has not chosen the reasons and principles on which she has formed them. In fact, in some cases we would plausibly withhold our praise if we discovered that she had chosen those reasons and principles. This suggests that something similar might apply to our evaluative beliefs. If it did, then, provided that these beliefs are among the basic explanatory roots of our actions, no regress of choices would need to start and scepticism about moral responsibility could be resisted. In connection with this, and with regard to the second assumption we have mentioned above, I shall try to show that even if someone’s intellectual achievements are indebted to some external sources, this need not prevent us from justiﬁably granting her full praise- or blameworthiness for them. Again, this suggests that something similar could be the case with regard to an agent’s evaluative beliefs and the actions and choices ﬂowing from them. Let me now defend this cognitive approach to moral responsibility. I do not have a conclusive argument to offer, but I can advance some considerations that show this position to be fairly plausible. Let us note, ﬁrst, that talk about responsibility for one’s beliefs makes perfect sense and is rather common in everyday life. We do not ﬁnd it less natural to praise or blame people on account of their beliefs than of their choices or actions. Think for example of a racist person: under some circumstances, we may hold her as blameworthy for her racist views as for her racist behaviour. Moreover, it would be easy to ﬁnd anywhere remarks like the following: “How could you believe what she told you? Don’t you know how often she lies?” As happens with actions, excuses can be expected in situations of this kind: “Well, you know, this time she really looked sincere.” We sometimes blame people for being careless – or, alternatively, for being too demanding – in forming their beliefs. Attribution of responsibility for beliefs might be interpreted within a view of control as based on choice. On this view, control over our beliefs would be taken to be only indirect and derived from the voluntary control we have over our cognitive activity. It seems true that we sometimes praise 3 Bernard Williams (1973) famously argued that beliefs are not under our direct voluntary control, so that there is not much room for deciding to believe. I ﬁnd Williams’s rejection of doxastic voluntarism very convincing.

280

Intentionality, Deliberation and Autonomy

or blame people for their beliefs on the basis of how they have conducted their inquiries. However, if this were the only way in which we could be said to control our beliefs, the threat of a vicious regress of choices could not be conjured up. But I think we also acknowledge another form of control over our beliefs. This control is neither voluntary nor merely indirect. It does not draw entirely on the voluntary control we may have over our epistemic activity. If we can show that sometimes responsibility for beliefs rests on a form of control over them not based on choices or acts of will, this might be an important step in a defence of moral responsibility against scepticism, at least of a Strawsonian variety. Let us see what a control of this kind is like, whether it actually exists and whether we justiﬁably acknowledge it in some of our ascriptions of praise- and blameworthiness for beliefs. As an initial attempt to characterize the form of control we are after, let us reﬂect on some examples. Think ﬁrst of a secondary school student who is trying to solve a problem of, say, physics. Suppose she performs the task carefully and obtains the right result. She has had control over both her cognitive activity and her belief about the solution to the problem and is praiseworthy on both accounts. It is not by mere luck that she has obtained the right result. But note what having control means in this case. It does not relate to choices or acts of will in any important sense. The student’s control consists rather in her yielding to the internal conﬁguration and structure of the problem, the data, physical laws, mathematical rules and so on. It is, so to speak, a passive form of control, which the student exercises precisely in respecting and being guided by what is there, in the problem itself. She chooses neither the data nor the physical laws or the mathematical rules. In fact, she would lose control over both the task and its result if she chose or decided about these things, and we would rightly blame her for doing so. Moreover, all those factors are external to her self or will: they come “from outside”. It is also true that, without the help of her teachers, she could not have solved the problem. And nonetheless my intuition, which I hope will be widely shared, is that she has control over her belief about the problem’s solution and truly deserves praise for this belief. So in this simple case the two assumptions I emphasized in Strawson’s and Kane’s construals of ultimate control as a requirement for true desert, namely that all control that backs true desert is based on choices, and that the inﬂuence of external factors undermines such a control, are simply absent, but true desert is still there. In order to question these assumptions further and deepen our understanding of this sort of non-voluntary control, consider now more complex cases of belief and belief formation, such as great achievements in science or philosophy. They can harmlessly be considered as belief systems. As a preliminary remark, these cases show that authorship concerning her beliefs may be no less important for a person’s worth and self-esteem than concerning her choices and actions. To mention only a few examples, remember the dispute between Newton and Leibniz about the invention of inﬁnitesimal calculus or, in more recent times, the debate about the true discoverer of the virus causing AIDS. Think also of the strongly negative moral judgement that plagiarism deserves for most of us. Questions about real source or origin, and about corresponding praise- and blameworthiness, have no less signiﬁcance in the cognitive ﬁeld than in the practical one. Let us now come back to our main subject, namely the nature and existence of the form of non-voluntary control over

Belief and Moral Responsibility

281

our beliefs that we are after, by reﬂecting on a particular example of a complex and great intellectual achievement, to witness, Descartes’ Meditationes de prima philosophia. I hope we shall agree that Descartes must be considered as the true author and source of this work and that he truly and justiﬁably deserves our praise and gratitude for it. We ﬁnd him praiseworthy not only for his effort and activity, but also for the result, the important ideas and beliefs contained in this outstanding philosophical opus. However, it is interesting to note how many of the factors that made this work possible have not their origin in Descartes himself. It is hard to see, for example, how the Meditations, which are usually considered as a new starting point, as the very beginning of modern philosophy, could have been written without the inﬂuence of medieval philosophy (cf. Gilson 1951). Other important “external” factors include Plato and the Platonic tradition, ancient and modern scepticism and contemporary physiology, to mention only a few. And, nonetheless, this does not incline us to question Descartes’ full authorship and praiseworthiness for his work. This seems to show that our judgements about authorship and responsibility for cognitive accomplishments do not ﬁt Kane’s or Strawson’s conceptions of ultimate control, at least regarding their individualistic assumption. Though Descartes did not give origin to many aspects and elements of the Meditations, we readily consider him as the true, ultimate author and source of his work, and rightly so. This suggests that something similar might be the case in the practical ﬁeld. But Descartes’ example can also be used to dispute the other assumption indicated above, according to which control is constitutively related to acts of will and choices. Though surely the will, in the form of choices, is involved in the process of creation of the Meditations, Descartes’ control over this work is not mainly based on voluntary acts or choices, but, to a large extent, in his respect for the internal requirements and structure of the subject matter itself, in his passively yielding to the relations of justiﬁcation between propositions, to the internal connections between concepts, to the force or necessity of certain steps in the reasoning process, as well as to the empirical data he employs. As happened with our example of the student, if Descartes had made some or all of these aspects depend on his will, he would have had less control over his work and would have been less praiseworthy for it. So reﬂection on matters of authorship and responsibility in the realm of cognitive accomplishments does not validate the assumption that all control depends on the will. Our judgements in this ﬁeld do not correspond to Strawson’s and Kane’s views of ultimate control. And, again, this suggests that something similar might be the case in the practical realm. Our hope can be stated as follows. If, as we have tried to show by means of examples, control over, and true desert for, our beliefs need not rest on choices or acts of will and is not necessarily undermined by the inﬂuence of external factors; and if control over our actions rests ultimately on control over our beliefs, then moral responsibility understood as true desert, as true praise- or blameworthiness, could be shown to be possible and not to fall prey to an inﬁnite regress of choices or to sheer arbitrariness. My suggestion is, in fact, that control over, and true desert for, our actions ultimately rests on control over, and true desert for, our beliefs. A certain class of beliefs is especially relevant in this respect, namely evaluative beliefs. Evaluative beliefs are beliefs with an evaluative content. This evaluative content should include

282

Intentionality, Deliberation and Autonomy

an agent’s conception of a human life that is worth living and have potential effects as a criterion for choice and a guide for action. To insist, a crucial advantage of grounding moral responsibility on evaluative beliefs, rather than choices, is that the problem of an inﬁnite regress of choices, as well as the correlative problem of a groundless, arbitrary choice as a basis for moral responsibility, do not need to arise. Before moving on to substantiate the proposal, however, let me point out that it is not without precedents. A step in this direction is Gary Watson’s emphasis on values instead of – even second-order – desires in his view of moral responsibility (cf. Watson 1982), as well as Susan Wolf’s “Reason View” (the term is hers), according to which the ability to form correct values is necessary for moral responsibility (cf. Wolf 1990, 75), among others. Let me now proceed to a defence of my proposal. If evaluative beliefs are to ground the possibility of ultimate control over our choices and actions, they have to satisfy a number of conditions, which would seem to include at least the following: 1) corresponding to the depth of moral responsibility attributions, they should be a deep, core component of a person’s self; 2) under certain circumstances, the agent could be correctly considered as their true author and source; 3) the agent should have rational control over these beliefs; 4) the preceding condition should hold even if the beliefs are not causally determined (the proposal should not fall prey to some version or other of the so-called Mind argument). I would like to argue that evaluative beliefs are able to satisfy these conditions. Unfortunately, owing to space limits, I can only give some brief remarks in favour of this contention. Evaluative beliefs would certainly seem to satisfy the ﬁrst condition. Our evaluative views are a central core of what we are, mentally speaking. No psychological characterization of a person could be minimally complete unless it included a description of what she ﬁnds worth pursuing or avoiding in life. And if we reﬂect on our serious ascriptions of moral responsibility for choices or actions, we shall ﬁnd ourselves ultimately praising or blaming an agent on account of the evaluative views these choices or actions express. In fact, if we come to see an act of hers as a momentary, passing impulse, not expressing her deep evaluative convictions, our judgement is signiﬁcantly softened, or even withheld. This takes us to the second condition, which corresponds to the “ultimate source” or true authorship aspect of ultimate control. In holding an agent responsible for an action on account of her evaluative views, we certainly seem to assume the she has proper control over them, so that she can be truly considered as her author and source. If we take this assumption to be false, we modify or even withhold our moral responsibility ascription. This happens in CNC manipulation cases, but also in more ordinary situations in which we do not see an agent as truly responsible for her evaluative beliefs. Certain victims of severely deprived childhoods or of a fanatical education can be examples of this predicament. But what requirements should be met in order for an agent to have deep control over, and to be the true author of, her evaluative views? In forming our initial evaluative views we are deeply inﬂuenced by our parents, close relatives or friends. This “external”, social origin is not, as such, a reason to deny our authorship with regard to the evaluative beliefs we end up having, as the student and Descartes examples show. But something else is needed.

Belief and Moral Responsibility

283

If we could do nothing but have the views we receive from our social environment, we would not be truly responsible for having them, and moral responsibility for our actions would not be possible. But we do not only receive beliefs, either evaluative or merely factual. We also acquire general standards that guide us in forming, assessing, accepting and rejecting such beliefs. This is an essential contribution that sociality makes to our constitution as agents. The most basic of these standards has to do with the truth-aiming character of belief. It demands of us that we allow our beliefs to be determined and controlled by the way things actually are. Another important standard for accepting and rejecting beliefs has to do with their logical relations. Contradictions, for example, are important reasons to modify our beliefs in order to avoid them. Applying these and other standards an agent can arrive to a system of beliefs which can truly be said to be her own and which she can be truly responsible for. And this holds for her evaluative beliefs as well. She may discover by her own, often painful, experience that an evaluative view she was taught is not actually true. And she may notice contradictions in her received evaluative views and be led to reshape them. However, as our previous examples indicate, it is not needed, for these beliefs to be truly attributable to an agent, that she has chosen the standards and procedures she employs in forming, retaining or abandoning them. In a modest way, as compared with such great intellectual accomplishments as Descartes’ Meditations, but not in a completely different sense, a system of evaluative beliefs can be truly ascribed to an agent as its source and author, and so as something for which she can truly deserve praise or blame. Moreover, though I shall not develop this important point here, I hold that having available alternatives to her actual evaluative beliefs is also a necessary condition on an agent’s true authorship and responsibility for them. An alternative possibilities condition is applicable to an agent’s evaluative beliefs no less than to her actions and choices if she is to have ultimate control over them. The third condition corresponds to the other aspect of ultimate control, namely rational control. Irrational or arbitrary evaluative beliefs would not be ﬁt to ground moral responsibility. Now, we control our actions rationally and voluntarily by choosing them in the light of our evaluative beliefs, but the control we should have over these beliefs in order for us to have ultimate control over our actions does not rest on a further choice of these beliefs. It is rather a matter of sensitivity to their internal consistency and respect for the facts they aim to capture, namely facts about what is valuable and worthwhile in human life. This control does not relate to our ability to choose and act, but rather to our ability to see what is there to be seen. This sort of non-voluntary, theoretical control, as it might be called, is an appropriate basis for praise and blame. It may relate to some cases in which we are blamed (or praised) for something we involuntarily did or omitted, such as forgetting (or remembering) our partner’s birthday or an appointment we had for dinner. Even if believing, like forgetting, is not voluntary, and not even an action, we can justiﬁably blame someone for her beliefs, in the same way as our partner can justiﬁably blame us for forgetting our appointment. This blaming seems to assume some form of non-voluntary control. In blaming us for forgetting our appointment, our partner is blaming us for not seeing rather than not acting: for our blindness to both our appointment and her, which reveals our lack of consideration and respect for both. And she is assuming that we ought, and could, have remembered the appointment.

284

Intentionality, Deliberation and Autonomy

Something closely analogous, I would think, holds when we blame someone for the evaluative views that an act of hers reveals. In commenting on the last two conditions we have crucially insisted, in analogy with the case of our theoretical beliefs, on a sort of control based not on choices or acts of will but rather on an attitude of humility and respect to what is there, to something endowed with a kind of objectivity. Defending this sort of control seems to commit us to some version of objectivism, to the idea that there are facts of the matter that our beliefs should be guided by and respond to. In the case of evaluative beliefs, we are committed to some version of objectivism about the evaluative. An unrestricted, rampant objectivism in this ﬁeld might appear as a rather implausible position. I do not think, however, that the version of objectivism we are committed to by virtue of our proposal should go that far. We do not need to assert that there are such entities as values. It seems enough to accept that there are facts of the matter in virtue of which evaluative beliefs can be true or false; that someone can be genuinely wrong in her evaluative views; and that evaluative truths are discovered, not invented.4 However, it is important to note that someone can have the sort of control over her evaluative beliefs that is required for true authorship and desert even if these beliefs are not actually true. Essential to the possession of this sort of control is rather the attitude of respect to what is objectively there, as well as the aim of having one’s beliefs guided and determined by it. This holds also for theoretical beliefs. To return to a previous example, consider that many theses contained in Descartes’ Meditations might actually be false, but, even if they were, Descartes would still be rightly considered as the true author of this great intellectual legacy and as truly praiseworthy for it. A second issue raised by my proposal has to do with motivation.5 If evaluative beliefs are to make an agent’s ultimate control over her actions possible, they have to be able to motivate her to perform those actions. A complete causal-explanatory isolation between evaluative beliefs and actions would deprive an agent from control over the latter by means of the former. Motivation, like objectivity, is a big issue, which I cannot deal with in depth here, much less provide something resembling a solution to it. I will restrict myself to a couple of remarks. First, there is something quite bizarre in holding an evaluative belief with no motivating consequence at all, even potential. If, for example, someone seriously holds that causing unnecessary pain to an innocent person is morally wrong but does not feel motivated at all to act (and react) according to this (true) belief, there is reason to resort to psychopathology in order to account for this mismatch. She may feel motivated to act against such a belief (owing maybe to a sadistic tendency) and actually do so, but if she really has the belief, only psychopathology could be of help if its contrary motivating potential did not manifest itself at least as shame or remorse. Second, evaluative facts need the 4 Objectivism about evaluative beliefs or judgements is far from being a hopeless position. I would tend to think that it is rather subjectivism, in its different forms, that is really in trouble. A recent and solid defence of objectivism (and cognitivism) in ethics can be found in Wiggins (2005). 5 Christoph Lumer encouraged me to discuss this and the previous issue concerning objectivism.

Belief and Moral Responsibility

285

participation of emotions in order to be properly appreciated. So, someone unable to feel or react emotionally in seeing someone else causing unnecessary pain to an innocent person is likely to be unable to see the wrongness in this situation and to form the corresponding general evaluative belief. Appreciation of evaluative facts and forming evaluative beliefs is a “thick”, emotionally laden cognitive performance. But the motivating potential of emotions is generally acknowledged. Before moving on to the next condition, let me address another objection that might be raised against my proposal. I have insisted on a form of control over our beliefs based on respect and deference towards objective facts and principles. On this view, control would not consist in choosing these factors, but rather in being guided by them. It might be objected, then, that this view leads rather to heteronomy, to the subject’s being “remotely controlled”, ruled by external forces.6 I would like to respond as follows. First, it is important to point out that the case for my proposal rests in a large degree on the intuitions raised by the examples we considered. Now, are we really prepared to hold that, in following correct reasoning principles, mathematical rules, in accepting certain data and laws of physics as given in order to ﬁnd out the result of the problem she was trying to solve, our student acted in a heteronomous way or that she was “remotely controlled” by those objective, “external” factors? My response, which I hope will be widely shared, would be “no”. And a similar question could be raised and a similar answer given for what regards the example of Descartes. Second, I tend to think that the objection takes for granted the will-centred and individualistic conception of control that I have tried to dispute and so begs the question against my proposal. Someone who ﬁnds this objection powerful is probably in the grip of the individualistic and voluntaristic assumptions that I have been at pains to undermine, according to which an autonomous and morally responsible agent has to be a radically self-made entity, so that the inﬂuence of any factors beyond the scope of her choice is potentially threatening to her authorship and true desert for her deeds and accomplishments. Many factors coming “from outside” are not in the way of our nature as morally responsible agents, but are rather constitutive of it. Again, if someone disagrees, I would ask him to reﬂect on the examples we presented and other similar cases and to judge whether many “objective” factors present in the student’s task of solving the problem or in Descartes’ writing of the Meditations detract from their respective praiseworthiness for those achievements or are rather constitutive conditions thereof. I would certainly say the latter, and I think there are powerful considerations in favour of this judgement. However, as I already pointed out, I do not have, at present, any conclusive arguments to offer to someone who rejects the judgement. Let me ﬁnally address the fourth condition I mentioned. A traditional compatibilist objection to libertarianism is that indeterminism erodes an agent’s rational and volitional control over her choices and actions, which turn into arbitrary, hazardous events, thus undermining her moral responsibility for them. This objection, which is usually known as the Mind argument, has been formulated in several ways. Consider Josephine, a judge who, after careful and relevant deliberation, decides at a certain 6 It was also Christoph Lumer who raised this objection. The terms “heteronomy” and “remotely controlled” are his.

286

Intentionality, Deliberation and Autonomy

time, T, not to grant clemency to a convict and to sentence him to life imprisonment.7 Let me now apply to this example Alfred Mele’s recent version of the objection. Imagine a close possible world, with the same past and natural laws as the one Josephine inhabits, in which an identical twin of hers, call her Josephine*, exists. The ﬁrst difference between the two worlds occurs only at time T. At this moment, while Josephine decides to sentence the convict to life imprisonment, Josephine* decides instead to grant him clemency. On the assumption that Josephine’s decision is causally undetermined, this is clearly conceivable. But then, in Mele’s words, “if … there is nothing about the agents’ powers, capacities, states of mind, moral character, and the like that explains this difference in outcome, then the difference really is just a matter of luck” (Mele 1999, 99). From this perspective, Josephine’s decision would appear to be a chancy and arbitrary event, not under her rational and volitional control. This objection may be powerful against will-centred views of ultimate control, but the cognitive approach I recommend might succeed against it. On this approach, it is practical judgement, rather than choice, that plays the pivotal role in practical deliberation. A practical judgement about which action is best, or better than the alternatives, should be understood as the application of an agent’s evaluative beliefs, as normative standards, to the situation she faces. Now, even if choices are not causally determined by practical judgements, it is not a mere accident that an agent’s choice usually accords with her practical judgement. Practical judgement is a normative standard for choice and this is why discrepancy between the two usually counts as irrational and abnormal, as a case of incontinence or weakness of the will. Now with these remarks in mind we are not forced to conclude that Josephine’s decision was a matter of luck just because it differed from Josephine*’s. Since the only difference between the two worlds arises only at T, the decision’s time, we can assume that Josephine and Josephine*, after a similar careful and relevant deliberation, formed the same practical judgement, namely that sentencing the convict to life imprisonment was better than the alternatives. If, thereafter, Josephine decided in accord with this judgement while Josephine* decided against it, this only shows that Josephine*’s decision was irrational or weak-willed, but not that Josephine’s decision was so as well. Josephine, unlike Josephine*, had rational control over her decision, which was not arbitrary or merely lucky. The example may show that Josephine, like any of us, is not immune to irrationality or weakness of the will, but this is much less than is needed to show that indeterminism is incompatible with rational and volitional control over our choices and actions. What needs to be shown is that indeterminism turns virtually all of our choices and actions into lucky, arbitrary events. And the example does not yield this result. I conclude, then, that a cognitive approach to moral responsibility, within the lines we have drawn, may be in a better position than a conative, will-centred approach in order to overcome some traditional difﬁculties of libertarianism, especially those related to ultimate control as a requirement for moral responsibility.8 7 The example is vaguely inspired by an example of Van Inwagen’s (1983, 68–9). 8 This paper is inspired in Moya (2006). I thank Michael Bratman, Christoph Lumer, Alfred Mele and other participants in the conference Intentionality, Deliberation and Autonomy

Belief and Moral Responsibility

287

References Clarke, R. (1997), ‘On the Possibility of Rational Free Action’, Philosophical Studies, 88, 37–57. Gilson, E. (1951), Etudes sur le rôle de la pensée médiévale dans la formation du système cartésien, Paris: Vrin. Kane, R. (1996), The Signiﬁcance of Free Will, Oxford and New York: Oxford University Press. Mele, A.R. (1999), ‘Kane, Luck, and the Signiﬁcance of Free Will’, Philosophical Explorations, 2, 96–104. Moya, C.J. (2006), Moral Responsibility: The Ways of Scepticism, Abingdon: Routledge. Pereboom, D. (2001), Living Without Free Will, Cambridge: Cambridge University Press. Strawson, G. (1986), Freedom and Belief, Oxford: Clarendon. Van Inwagen, P. (1983), An Essay on Free Will, Oxford: Clarendon. Watson, G. (1982), ‘Free Agency’, in G. Watson (ed.), Free Will, Oxford: Oxford University Press. Wiggins, D. (2005), ‘Objectivity in Ethics; Two Difﬁculties, Two Responses’, Ratio (new series), 18, 1–26. Williams, B. (1973), ‘Deciding to Believe’, in B. Williams, Problems of the Self, Cambridge: Cambridge University Press. Wolf, S. (1990), Freedom within Reason, Oxford: Oxford University Press.

for comments and criticisms. This paper is part of the research project HUM2006–04907, awarded by the Spanish Ministry of Education and Science.

This page intentionally left blank

Chapter 15

Autonomy and Weakness of Will Thomas Spitzley

The shorter ﬁrst part of this chapter sketches an account of “autonomy”, following Gerald Dworkin’s characterization of the central idea that underlies the concept of autonomy. According to Dworkin, the core of personal autonomy is the agent’s governing herself: she makes her own laws, rules herself and is not under the control of any conquering power or foreign law. Starting from this basic thesis I develop a bundle of necessary conditions for an action to be autonomous and give some supporting reasons for subscribing to them. Afterwards I present ﬁve contemporary accounts of weakness of will (R.M. Hare, G. Watson, D. Davidson, G. Ainslie and U. Wolf) and investigate whether according to any of these accounts a weak-willed action could count as being autonomous. Even though the answer is always negative, the respective reasons differ considerably. However, claiming that a weak-willed action cannot be autonomous is far from claiming that a weak-willed person is not autonomous. The only thing one can possibly say is that a weak-willed person is not autonomous with respect to her weak-willed action. When one looks up the word “akratic” in the Oxford English Dictionary one ﬁnds among others the following quotation from the Journal of Medical Ethics: “Declaring an akratic agent autonomous classiﬁes him together with an ordinary non-akratic agent; declaring him non-autonomous involves putting him in with nonrational beings such as animals, comatose patients and those in advanced stages of Alzheimer’s Disease. Neither classiﬁcation is comfortable” (McKnight 1993, 206– 10). I fully agree with the tenor of this quotation: there is, indeed, a problem as to how autonomy and weakness of will are compatible, though I think the problem is far more complex than the quotation suggests. When one approaches the question of how autonomy is connected with weakness of will it more or less suggests itself ﬁrst to explain a) the concept of autonomy and b) the concept of weakness of will, and ﬁnally to investigate whether and to what extent autonomy is compatible with weakness of will. However, to follow this strategy in an adequate way would mean to go well beyond the scope of this article. Therefore in what follows, I will, in the shorter ﬁrst part of this chapter, sketch a rough-and-ready working account of “autonomy”. After that I will, in the much longer second part, compare this account of autonomy with some contemporary

290

Intentionality, Deliberation and Autonomy

accounts of weakness of will. To be sure, one could commit oneself with good reasons to a single account of weakness of will, yet passing in review the history of the problem of akrasia from the times of Plato to the twenty-ﬁrst century present, or only looking at the relevant discussions during the last forty-odd years, it seems clear that there is some truth in most accounts and there is not one account which is satisfactory in each respect. Therefore the aim of this chapter is rather modest; to come to a better understanding concerning the compatibility of autonomy and weakness of will, without trying or pretending to give the deﬁnite answer of what either is or whether they are compatible. 1. Autonomy … As a starting point I accept Gerald Dworkin’s characterization of the central idea that underlies the concept of autonomy. In his article “The Concept of Autonomy” Dworkin writes: “What I believe is the central idea that underlies the concept of autonomy, is indicated by the etymology of the term: autos (self) and nomos (rule or law). The term was ﬁrst applied to the Greek city state. A city had autonomia when its citizens made their own laws, as opposed to being under the control of some conquering power” (Dworkin 1989, 57). So the core of personal autonomy is the agent’s governing herself: she makes her own laws, rules herself and is not under the control of any conquering power or foreign law.1 To my mind, “being autonomous” is a property which ﬁrst of all can be ascribed to persons. A person’s autonomy concerns her beliefs and pro-attitudes as well as her actions. Whether a belief or a pro-attitude is autonomous, at least partly, depends on its aetiology. A belief is only autonomous, if it is not induced in the person’s mind by any sort of inappropriate external inﬂuence, for example via brainwashing. However, here we are mainly concerned with the relation between autonomy and weakness of will, and it seems to be undisputed that for an action to be classiﬁed as weak-willed, it means that there is some sort of tension between the action and the beliefs and desires the agent actually has, no matter how she acquired them. Though in principle the aetiology of a person’s beliefs and pro-attitudes is certainly relevant to her autonomy, in the present context it is only of minor importance. The autonomy of beliefs and pro-attitudes poses no special problem for the question of whether weakness of will is compatible with autonomy. Therefore I concentrate on autonomous actions. An agent is autonomous only if most of her actions are autonomous. An action A is autonomous, only if:

1 Cf. “A person governs himself when he is not governed ‘from the outside’ by someone else, and when he does govern from the inside – when he is ‘in control of himself’” (Feinberg 1989, 39; cf. ibid., 51f., n. 26).

Autonomy and Weakness of Will

291

(1) A is performed freely and voluntarily; (2) the pro-attitudes which led the agent to perform A do not contradict those relevant pro-attitudes which the agent wants to be action-guiding in the situation at hand or with which she even identiﬁes herself; and (3) the agent’s relevant pro-attitudes have a certain duration. This is, obviously, no deﬁnition, nevertheless it is meant to illustrate the gist of the notion of autonomy I have in mind. So understood, autonomy is in Dworkin’s words “a purely formal notion” as opposed to “a substantive notion (where only certain decisions [or actions] count as retaining autonomy while others count as forfeiting it)” (ibid., 59). The following brieﬂy tries to explain and to give some supporting reasons for these three conditions. Ad (1) the action is performed freely and voluntarily: if an autonomous agent rules herself and is not under the control of any conquering power, then it seems very plausible to require the action to be performed freely and voluntarily. That the action must be performed freely implies that it is not compulsive, but it does not imply that the agent could do otherwise. If an agent voluntarily puts herself in a situation in which she is compelled to do something or to refrain from doing something, that should still count as a free action or omission in the relevant sense.2 Ad (2) the pro-attitudes which led the agent to perform A do not contradict those relevant pro-attitudes which the agent wants to be action-guiding in the situation at hand or with which she even identiﬁes herself. Even when an agent’s actions are free and can be explained by reference to the agent’s beliefs and desires, it could be that the agent acts in a way that on reﬂection she does not want to. She may have conﬂicting desires and her strongest desire, which effectively inﬂuences her action, may be different from the one she wants to be effective. It may even be incompatible with the pro-attitudes which are so important to her that she identiﬁes herself with them. In cases like this the agent cannot really be described as ruling herself. She may have passed the laws, but she does not enforce them. Ad (3) the agent’s relevant pro-attitudes have a certain duration: if one takes the central idea which underlies the concept of autonomy seriously, and accepts that in her actions an autonomous person follows her own laws, the question about the scope of these laws arises. Should we suppose that they are only valid at a speciﬁc moment in time? I take it that this would be absurd; we should rather assume that these laws are valid for a longer, even though unspeciﬁed period of time. Think for example of a person who in the company of dominant people always shares their views, however, not out of fear or because she expects something to come of it. Instead, she is really convinced of what she says and does; she simply cannot resist the inﬂuence of dominant people. Such a person’s action may fulﬁl conditions (1) and (2), nonetheless it surely falls short of being autonomous. For being autonomous, the agent must not change her pro-attitudes from day to day; rather they must have a certain duration, that is, they must remain constant and coherent for some considerable time. 2

I owe this point to Mele 1995, 166.

292

Intentionality, Deliberation and Autonomy

2. … and weakness of will I now measure some contemporary accounts of weakness of will against the threelevel account of autonomy just outlined. This will illuminate if weakness of will may be compatible with autonomy, and if so, in which respect. As in the case of autonomy, we need a working account of weakness of will, yet in this case it is much easier to provide. In the following, weakness of will should be understood as acting against one’s better judgment. One type of accounts of weakness of will does not deserve special investigation, viz. those accounts which try to explain weakness of will as a conﬂict between ﬁrst-order and second-order desires.3 If a weak-willed agent is somebody who acts on a desire which she does not want to be action-guiding, this characterization is plainly incompatible with condition (2).4 Let us start by drawing attention to the central idea that underlies the concept of autonomy, the idea that the agent makes her own laws, rules herself and is not under the control of any conquering power or foreign law. Compare this to the famous characterization of weakness of will in St Paul’s letter to the Romans (7:23) in which he complains: “I see another law in my members, warring against the law of my mind, and bringing me into captivity to the law of sin which is in my members”, and the tension between autonomy and weakness of will leaps to the eye. 2.1 R.M. Hare St Paul is one of two chief witnesses Richard Hare draws on when he discusses backsliding in his book Freedom and Reason. He quotes more fully from St Paul: “I cannot understand my own behaviour. I fail to carry out the things I want to do, and I ﬁnd myself doing the very things I hate. When I act against my own will, that means I have a self that acknowledges that the Law is good, and so the thing behaving in that way is not my self but sin living in me ….” According to Hare, the typical case of moral weakness is a case of “ought, but can’t” (Hare 1963, 68 and 80). The inability which Hare ascribes to the agent who displays moral weakness is not a physical, but a psychological inability. “[I]t is often true”, says Hare, “that a man is physically in a position and strong, knowledgeable and skilful enough, &c., to do what he thinks he ought … [but] … [i]n a deeper sense the man cannot do the act” (ibid., 82), he cannot bring himself to do what he thinks he ought to do. Judging from this, one might think that the morally weak agent may well be free in doing what she actually does, viz. what she thinks she ought not to do. As an interpretation of Hare’s account, however, that would be wrong. This becomes obvious when we take into account a quotation from Medea, Hare’s second chief witness. Hare quotes approvingly from Ovid’s Metamorphoses a passage where Medea reports her futile attempt to resist her passion for Jason: “[A]n unknown compulsion bears me, all reluctant down. Urged this way – that – on Love’s or Reason’s course, I see the

3 Cf., for example, Jeffrey 1974, 377–91, and Bigelow, Dodds and Pargetter 1990, 39–49. 4 One preliminary remark: when presenting the different accounts of weakness of will I shall try to restrict myself to those elements which are essential for the problem at hand.

Autonomy and Weakness of Will

293

better: do the worse.”5 So Hare’s true thesis concerning moral weakness is that the agent is unable to do anything other than what she actually does, that is, her desires compel her to do what she thinks she ought not to. Signiﬁcantly enough, Hare refers to cases of obsessional neurosis as clearest examples that “in a deeper sense” somebody cannot act. This makes clear that on Hare’s account of moral weakness a morally weak agent does not act autonomously, since not even condition (1) of the explication of autonomy is fulﬁlled. As a matter of fairness I should add that Hare does say more about moral weakness; this, however, rather concerns the special problem which moral weakness poses for his moral prescriptivism and therefore does not bear relevance to our problem. 2.2 G. Watson Contrary to Hare, Gary Watson argues that weakness of will must be distinguished from compulsion. In a nutshell, his proposal is this: if a person does not resist a desire which “the possession of the normal degree of self-control would enable [her] to resist” (Watson 1977, 330), then she is weak. If, however, she does not resist a desire which is “such that the normal capacities of resistance are or would be insufﬁcient to enable the agent to resist” (ibid., 330), it is a case of compulsion. So when we blame an agent for being weak-willed in a certain situation, we intimate that this very person could have resisted her actual desires had she only developed the capacity of self-control which is normally possessed by others. So we do not see a weak-willed agent as a victim of her desires as it is true in cases of compulsion. We rather see her as somebody who, by her own fault, has not acquired or maintained the skills necessary for normal self-control. Here the qualiﬁcation “normal” indicates that the standards of self-control which an agent is expected to fulﬁl are relative to the society in which she lives. When an agent fails to meet the standards of normal self-control, we take her to be responsible for her weak-willed action. To resume: in a case of weakness of will the desire the agent acts upon is indeed resistible, that is, it is resistible for anyone who possesses the normal amount of selfcontrol, yet for the agent in question it is not resistible. Applying this result to our question of whether a weak-willed agent is autonomous, Watson’s answer must be negative. Even though he claims that the weak-willed agent acts intentionally (ibid., 335, n. 20), he denies that she acts freely. So, just as it was with Hare, according to Watson’s account of weakness of will the ﬁrst condition of being autonomous is not fulﬁlled. 2.3 D. Davidson Let us now turn to the potentially most inﬂuential modern account of weakness of will, viz. to the one offered by Donald Davidson. Incidentally, talking of one account of his may be controversial, but I take it that in his later writings he only expanded and reﬁned what he originally proposed in his article ‘How is Weakness of the Will

5

Ovid, Metamorphoses, VII, V. 18–21.

294

Intentionality, Deliberation and Autonomy

Possible?’. Davidson characterizes an action that reveals weakness of will in the following way: In doing x an agent acts incontinently if and only if: (a) the agent does x intentionally; (b) the agent believes there is an alternative action y open to her; and (c) the agent judges that, all things considered, it would be better to do y than to do x. (Davidson 2001a, 22)

According to (a) a weak-willed action is performed intentionally and, as Davidson would add, freely.6 So if a person acts incontinently in a Davidsonian sense, condition (1) of acting autonomously is fulﬁlled too. To ﬁnd out whether condition (2) is or at least can also be fulﬁlled, we have to understand Davidson’s account more deeply. Davidson’s proposed solution of the problem of weakness of will makes use of a distinction between two types of evaluative judgements, viz. conditional or prima facie evaluative judgements on the one hand and unconditional evaluative judgements or evaluative judgements sans phrase on the other hand (ibid., 40). The easiest way to characterize a conditional evaluative judgement is by means of an example: if I want to buy a car, if there are two cars on offer, a Ford and a VW, and if the Ford has my favourite colour, then, as far as the colour is concerned, I prefer buying the Ford. If, however, the Ford is more expensive than the VW, that may result in my preferring to buy the VW. In this example just two things have been considered, viz. colour and prize. In the case of weakness of will, however, Davidson maintains that the agent forms an all things considered judgement, that is a judgement of the following type: judging on the basis of all the relevant evidence which is available, action y is better than action x. Bear in mind, this still is a conditional judgement, because the agent’s preferring action y to action x is conditional on the evidence which she takes into account. An unconditional judgement on the other hand would be, for example, “action y is better than action x” – sans phrase, that is without reference to the evidential basis for this judgement. The unconditional judgement, Davidson claims, corresponds to an intention, and therefore only a judgement of this type can directly lead to action. One would expect an agent who holds the conditional all things considered judgment that action y is better than action x to form the corresponding unconditional judgment “action y is better than action x”. This would be justiﬁed and in normal cases that is exactly what happens. In cases of weakness of will, however, the agent who holds that, all things considered, action y is better than action x proceeds to the unconditional judgement “action x is better than action y”. To be sure, something went wrong with this agent, yet the two judgements, the conditional and the unconditional one, do not contradict each other logically. But what did go wrong? According to Davidson there is a principle which every “rational man will accept in applying practical reasoning: perform the action judged best on the basis of all 6 For the thesis that the incontinent action is free cf.: “Some distance back, I tried analysing, ‘A is free to do x (or can do x)’ in terms of the conditional, ‘He would do x intentionally if he had attitudes that rationalized his doing x’” (Davidson 2001b, 79), and “I hold that there is a basic sense in which we are free to — (can —) provided all that is needed for the action (in addition to other present circumstances) is the right attitudes and beliefs” (ibid., 73, n. 9).

Autonomy and Weakness of Will

295

available relevant reasons” (ibid., 41). This principle, Davidson suggests, may be called the principle of continence, and it is this principle that the weak-willed agent does not obey. The conceptual framework just presented allows Davidson – at least so he thinks – to give a consistent description of weak-willed action: the weak-willed agent, indeed, judges that, all things considered, it would be better to do y than to do x and she does x intentionally, since she acts on her unconditional judgement that it is better to do x than it is to do y. So the weak-willed agent has indeed a reason for doing x, yet she has no reason for acting against her better reasons, which are in favour of doing y. After this exposition of Davidson’s account there is still the open question as to whether this account is compatible with autonomy as sketched above. We have seen that condition (1) is fulﬁlled, but what about condition (2)? Is it conceivable that in a case of weakness of will the pro-attitudes which lead the agent to perform x do not contradict those pro-attitudes the agent wants to be action-guiding? I must admit that I am not completely sure, but for all I can see the answer must be afﬁrmative. However, Davidson claims that every creature which is “in a position to decide anything” accepts the principle of continence, because only then it is possible “that we can intelligibly attribute propositional attitudes to it, or that we can raise the question whether it is in some respect irrational” (Davidson 1985a, 352).7 Is it not necessary for the agent as a rational being to want the principle of continence to be effective in her acting? To this question the answer is deﬁnitely “no”, since, as Davidson argues, “it makes sense to imagine that a person has the principle without being aware of it or able to articulate it” (Davidson 1985b, 141). If, however, the agent need not even be aware of her having or accepting the principle of continence, she certainly need not want it to be action-guiding. So the conﬂict that is sketched in condition (2) does not necessarily arise. Must we therefore conclude that on Davidson’s account acting from weakness of will is compatible with acting autonomously? Again the answer is “no”, but that is due to condition (3). To this we will return soon. 2.4 G. Ainslie Traditionally, tackling the problem of weakness of will means trying to describe and explain how such irrational behaviour can occur. Rather than sticking to that traditional procedure, George Ainslie, the next author I will consider, urges us to investigate how people behave rationally. Whereas Davidson asked “How is Weakness of the Will Possible?” Ainslie emphasizes we should rather ask “How is Strength of Will Possible?”. In his view, which is based on empirical research, there is nothing mysterious in acting in a way weak-willed agents do, yet there is much to explain as to how one can act rationally. Empirical research conﬁrms that people indeed try to maximize their prospective rewards, but, contrary to the assumption of utility theorists, “they discount their prospects using a different formula from the one that’s obviously rational” (Ainslie 2001, 28). If Ainslie is right, people are by nature 7

Cf. Davidson 1985a, 351, and Davidson 1985b, 142.

296

Intentionality, Deliberation and Autonomy

predisposed to discount expected values hyperbolically and not exponentially, but only the latter conforms to our standards of rationality. Disregarding mathematical details, the gist of the matter is that according to Ainslie the evaluation of future events is a function of their distance in time. One of Ainslie’s examples may be helpful for clariﬁcation: If I ask a roomful of people to imagine that they’ve won a contest and can choose between a certiﬁed check for $100 that they can cash immediately and a postdated certiﬁed check for $200 that they can’t cash for three years, more than half of the people usually say they would rather have the $100 now. If I then ask what about $100 in six years versus $200 in nine years, virtually everyone picks the $200. But this is the same choice seen at six years’ greater distance. (Ibid., 33)

The consequence of this ﬁnding for the problem of weakness of will can be understood best when we think of situations where at t1 an agent knows that she must choose at some future time t2 between a minor, but at t2 present good x and a major, but at t2 still future good y. If at t1 she values x less than y, this should not have changed at t2 – at least ceteris paribus and according to a standard assumption of utility theory. However, due to the hyperbolic character of her discount curves, that is, due to the fact that hyperbolic discount curves are more bowed than exponential discount curves, there may be an interval of time during which the agent actually values x more than y. This is perfectly compatible with the fact that at time t3 the agent again values x less than y. As can be seen, our preferences concerning a choice between two goods need not stay constant, yet this result is contrary to a basic assumption of utility theory. What happened, then, between t1 and t2 and between t2 and t3 respectively? Did the agent change her mind? To this one could answer “yes” or “no”, but I am inclined to say “no”. The agent did not come to new insights, she did not weigh the alternatives anew, she simply did not do anything! Instead, the observable change in her preferences is built right in to her evaluation of the alternatives at t1. According to Ainslie’s model, weakness of will is, as he puts it, “just maximizing expected reward, discounted in highly bowed curves” (ibid., 39), and the fact that the preferences of a weak-willed agent oscillate is nothing to be surprised about. Weakness of will, therefore, must be described neither as a case of psychological inability or compulsion, nor as being overwhelmed by or yielding to a desire. In displaying what we call “weakness of will” the agent acts exactly according to the desires she has at the time of action. Nothing at all contradicts the assumption that her action is free, as is required in condition (1), and equally nothing speaks against condition (2) being fulﬁlled. But would her weak-willed action count as being autonomous? What follows now holds true for Ainslie’s account as well as for Davidson’s. Even though in this case the agent’s pro-attitude which led her to perform the weak-willed action does not contradict those pro-attitudes which the agent wanted to be actionguiding (condition 2), this pro-attitude does not stay unchanged over a sufﬁciently long period of time. Instead, even according to Ainslie’s model the weak-willed agent typically regrets what she has done and after performing the weak-willed action she very often explicitly does not want the pro-attitude on which she acted to be actionguiding any more. What is more, when one observes in cases of weakness of will

Autonomy and Weakness of Will

297

which of her pro-attitudes the agent wants to be effective, she typically oscillates between the pro-attitudes in favour of the weak-willed action on the one hand, and the pro-attitude which supports the rival action on the other. So condition (3) for autonomous actions is not fulﬁlled and again we have a negative result.8 Let us pause for a moment and consider where we have come so far. According to two of the four accounts of weakness of will we have examined, that is, according to Hare’s and Watson’s account, a weak-willed action fails to be autonomous because it cannot count as free or non-compulsive. As I mentioned in passing, when one accepts higher-order preferences accounts of weakness of will, a weak-willed action may well count as free, but by deﬁnition the second condition for autonomous actions is violated. And as far as the last two accounts, viz. Davidson’s and Ainslie’s, are concerned, to be sure, it was possible to think of conditions (1) and (2) as being fulﬁlled, yet condition (3) provided an insurmountable obstacle. What is needed for weakness of will to be consistent with autonomy is an account of weakness of will in which it is allowed to ascribe to the agent relevant pro-attitudes that are sufﬁciently persistent. This is almost inconceivable, though, one should think, when one recalls that weakness of will is typically accompanied with regretting and being ashamed. However, there is one account which seems to allow exactly what we are looking for. 2.5 U. Wolf Ursula Wolf objects to Hare’s thesis that typical cases of weakness of will are cases of “ought but can’t”, that is, cases in which the agent is psychologically unable to do what she thinks she ought to do, and she equally turns against Davidson’s thesis that the weak-willed agent acts freely against her better judgement. If the action somebody freely performs differs from what she claims to be the better course of action, then, Wolf maintains, it follows analytically that she does not really take the alternative course of action to be better. Note, therefore, that to Wolf’s mind there is no such thing as acting against one’s better judgement.9 Cases of weakness of will are rather cases which at ﬁrst glance look as if the agent acted against her better judgement. Wolf argues that in cases of weakness of will we are confronted with a cognitive mistake in the self-image of the weak-willed agent, and that this mistake is what she calls an “interested mistake”, that is, a mistake the weak-willed agent has an interest to make (Wolf 1985, 29). On the one hand, a weak-willed agent wishes to be, in her own eyes as well as in the eyes of other people, a person who follows certain principles in her actions. On the other hand, she does not want to abstain from satisfying certain of her ﬁrst-order desires, though some of them conﬂict with her principles. When she claims to think one course of action best, although she follows a different course of action, she gets tangled up in a web of contradictions, because that is like wanting to have the cake and eat it. Since Wolf works on the assumption 8 If Ainslie is right, and people by nature discount hyperbolically, that would ﬁt quite well with the plausible assumption that people have to learn to act autonomously. 9 Of course, Ainslie would not analyse weakness of will in this way either.

298

Intentionality, Deliberation and Autonomy

that deeds speak louder than words, this contradiction can only be resolved by admitting that one is not the person one would like to see oneself as (ibid., 29). To avoid or block this unpleasant self-assessment – at least temporarily – the agent may use her ability to deceive herself. Wolf points out two important strategies of self-deception (cf. ibid., 30ff.). The ﬁrst strategy for preserving an acceptable self-image makes use of self-deception in exploiting the difference in time between, say, resolving to follow a certain moral principle and not acting accordingly. At least during the time which passes between her forming the resolution to act on that moral principle and the time when she could perform the act she can cherish the illusion that she really wants to follow that moral principle. When it comes to acting accordingly, however, this illusion is likely to collapse, yet then a second strategy of self-deception is ready at hand. The second strategy may be called “rationalization”. For our purposes there are two different techniques to be distinguished: a) The agent who employs the ﬁrst kind of rationalization makes use of the fact that every principle admits of well-founded exceptions. One very popular and widespread variety of deceiving oneself in this way is special pleading. Though the agent claims to accept a moral principle, when the time of action arrives, she lifts, as Hare once called it, a corner of the net (Hare 1963, 53), so that she escapes from the principle’s scope of regulation. Rationalizing one’s actions consists either in redescribing them in a way that they ﬁt perfectly well to one’s beliefs and proattitudes, or in adjusting one’s beliefs appropriately. And it is clear from the start that the result of the rationalisation will be that acting on the ﬁrst-order desire which is incompatible with that moral principle was the best thing to do. In point of fact, however, what in those cases the agent claims to be her reasons for acting are only a pretext. Wolf maintains that though this technique may be quite successful at the time of acting, it becomes transparent to the agent at some later time when she regrets her action. I think this may very well often be the case, but “it ain’t necessarily so”. Some people succeed in rationalizing their actions over a pretty long time. b) The second kind of rationalization that Wolf points our attention to allows an agent to follow her ﬁrst-order desires openly and without reservation, since she declares just this way of behaviour to be part of how she views herself. As Wolf puts it: When we let ourselves be determined by our immediate wishes, without taking further consequences into account, we often talk in a way that we declare that in the end it is not very important what we do and who we are; that basically, it does not mean much or it does not really matter what one does. In this way we need not say that our immediate wishes as such are most important to us, but can instead hold up our behaviour as expression of our distanced relation to ourselves and the world. (Wolf 1985, 31; my translation)

Wolf hastens to add that it is quite possible, though, to have such a self-image without intending to rationalize one’s behaviour. However, when this kind of dissociation from one’s self is exercised in situations in which it comes in handy to the agent, it gives cause for suspicion.

Autonomy and Weakness of Will

299

Let us now return to the question of whether weakness of will is compatible with autonomy. As stated earlier, Wolf denies that there is such a thing as acting against one’s better judgement; but of course there is a phenomenon which we are used to calling “weakness of will”, and Wolf analyses this phenomenon by referring to the agent’s self-image and the strategy of self-deception. When the agent makes use of a time difference (as in the ﬁrst strategy for preserving an acceptable self-image), she deceives herself insofar as she thinks she adheres to a moral principle; she intends to do x at t1 and that would mean to follow the moral principle, but at t1 she does not do x, instead she does y. Wolf emphasizes that the agent performs her weak-willed action freely, so condition (1) is met. Since Wolf holds that deeds speak louder than words, to her mind, at least, the agent’s doing y provides sufﬁcient evidence for ascribing to the agent the belief that in the actual situation y is for her the best thing to do. Therefore, it seems, condition (2) could be taken to be fulﬁlled. And when we take Wolf’s position seriously, we may also assume that the agent’s relevant pro-attitude is relatively persistent: judging from Wolf’s characterization of the agent we may safely assume that the agent will regularly slide back and fail to put that moral principle into action, so that condition (3) may be satisﬁed too. However, as has become clear, part of the agent’s self-image consists in her wish to follow the moral principle. This pro-attitude is one with which she identiﬁes herself, yet this very pro-attitude conﬂicts with the pro-attitude which, in fact, is effective when she displays weakness of will. So it turns out that, contrary to ﬁrst appearance, condition (2) is not fulﬁlled. When the agent rationalizes her behaviour by way of special pleading (as in case (a) of the second strategy for preserving an acceptable self-image), she deceives herself in the same way as just discussed: she mistakenly thinks she adheres to a moral principle. In this case, however, she does not intend to do x at t1, because due to her deceiving herself she does not believe that the moral principle is properly applicable at t1; therefore she does not do x, but instead she does y. Even if in this case conditions (1) and (2) could be regarded as being fulﬁlled, remember that Wolf claimed the self-deception would become transparent to the agent when she regrets her action. If this is true, I think the relevant pro-attitude which was action-guiding in the case of weakness of will could not really count as being sufﬁciently persistent and therefore condition (3) would not be fulﬁlled. If, however, my hypothesis is correct, according to which some people succeed in rationalizing their actions in this way over a pretty long time, this objection would break down. However, as in the case just discussed, part of the agent’s self-image consists in her wish to follow the moral principle, a wish with which she identiﬁes herself. Since this pro-attitude conﬂicts with her action-guiding pro-attitude, condition (2) is not fulﬁlled. Finally, when the agent rationalizes her behaviour in the sense of pretending indifference, she deceives herself with respect to the wishes that are important to her, or rather she deceives herself with respect to the wishes she wants to be actionguiding. Here I take sides with Joel Feinberg: To the degree to which a person is autonomous he is not merely the mouthpiece of other persons or forces. Rather his tastes, opinions, ideals, goals, values, and preferences are all

Intentionality, Deliberation and Autonomy

300

authentically his. … A person is authentic to the extent that … he can and does subject his opinions and tastes to rational scrutiny. He is authentic to the extent that he can and does alter his convictions for reasons of his own, and does this without guilt or anxiety. (Feinberg 1989, 32)

Since in the example under discussion the agent succumbs to self-deception and so deﬁnitely does not subject her opinions, tastes and so on to rational scrutiny, it may well be that she does not realize a conﬂict between the pro-attitudes which lead her to act in the way she does and the pro-attitudes which she truly wants to be actionguiding.10 Therefore, I think, it is clear that in this case too, condition (3) is not met, and the agent’s doing y cannot count as an autonomous action. 3. Summary We started with a rough-and-ready working account of autonomy or rather of autonomous actions. The conditions of autonomy mentioned were only necessary conditions and I admit that I hesitate to claim they are jointly sufﬁcient. Afterwards we presented ﬁve contemporary accounts of weakness of will and investigated whether according to any of these accounts a weak-willed action could count as being autonomous. The result was that on no account of weakness of will a weakwilled action could count as being autonomous, though the respective reasons differed considerably. To be honest, this result is not extremely surprising. As long as weakness of will is characterized as acting against one’s better judgement and experienced as being at least unwelcome, one should have thought that it would take a considerable number of arguments and persuasiveness to show that a weak-willed action might be autonomous. Anyway, claiming that a weak-willed action cannot be autonomous is far from claiming that a weak-willed person is not autonomous. Autonomy is a sort of general property. A person can be more or less autonomous, but if somebody acts non-autonomously in one situation that does not mean she is non-autonomous tout court. It goes without saying that one weak-willed action does not have the effect that the person is not autonomous. The only thing one can safely say is that she is not autonomous with respect to that particular action.11

10 Cf. “[T]he self-deceived person believes what he believes to be true and wants what he believes to be true because of his placing a distorted construction on the total evidence. … Self-deception, like neurosis, is something which any person who aspires to autonomy would prefer to avoid” (Young 1980, 38). 11 I thank Christoph Lumer, Michael Saxl and especially Ralf Stoecker for helpful comments and discussions. I am grateful to Michael Saxl for adapting my paper to the editorial guidelines of this volume.

Autonomy and Weakness of Will

301

References Ainslie, G. (2001), Breakdown of Will, Cambridge: Cambridge University Press. Bigelow, J.; S. Dodds and R. Pargetter (1990), ‘Temptation and the Will’, American Philosophical Quarterly, 27, 39–49. Davidson, Donald (1985a), ‘Incoherence and Irrationality’, Dialectica, 39, 345–54. —— (1985b), ‘Deception and Division’, in E. LePore and B.P. McLaughlin (eds), Actions and Events: Perspectives on the Philosophy of Donald Davidson, Oxford: Basil Blackwell, pp. 138–48. —— (2001a), ‘How is Weakness of the Will Possible?’, in D. Davidson, Essays on Actions and Events, 2nd edn, Oxford: Clarendon, pp. 21–42. —— (2001b), ‘Freedom to Act’, in D. Davidson, Essays on Actions and Events, 2nd edn, Oxford: Clarendon, pp. 63–81. Dworkin, G. (1989), ‘The Concept of Autonomy’, in J. Christman (ed.), The Inner Citadel: Essays on Individual Autonomy, Oxford: Oxford University Press, pp. 54–62. Feinberg, J. (1989), ‘Autonomy’, in J. Christman (ed.), The Inner Citadel: Essays on Individual Autonomy, Oxford: Oxford University Press, pp. 27–53. Hare, R.M. (1963), Freedom and Reason, Oxford: Oxford University Press. Jeffrey, R. (1974), ‘Preference among Preferences’, Journal of Philosophy, 71, 377–91. McKnight, C.J. (1993), ‘Autonomy and the Akratic Patient’, Journal of Medical Ethics, 19, 206–10. Mele, A. (1995), Autonomous Agents: From Self-Control to Autonomy, New York: Oxford University Press. Watson, Gary (1977), ‘Scepticism about Weakness of Will’, The Philosophical Review, 86, 316–39. Wolf, Ursula (1985), ‘Zum Problem der Willensschwäche’, Zeitschrift für philosophische Forschung, 39, 21–33. Young, R. (1980), ‘Autonomy and the “Inner Self”’, American Philosophical Quarterly, 17, 35–43.

This page intentionally left blank

Index of Names Adams, F. 5, 15, 17, 24, 27, 32, 76, 88, 266, 271 Agazzi, E. 48, 64 Ainslie, G. 208, 295–297, 301 Allen, C. 54, 64 American Law Institute 114 Anscombe, E. 47, 64, 91–94, 98, 114, 144, 155, 231, 240 Aquinas, T. 120, 132 Ariew, A. 54, 64 Aristotle 1, 107, 120, 154, 159, 167, 227, 231–233, 235, 238 Armstrong, D. M. 74–76, 88 Audi, R. 10, 20, 32, 49, 64, 78, 88, 127–128, 132, 135–138, 141–146, 148, 150, 153–155, 162, 183 Ayer, A. J. 247, 251, 255 Bach, K. 49, 64 Batson, C. D. 168, 170, 183 Bayertz, K. 215, 225 Beck, L. W. 81, 88 Bedau, J. 54, 64 Bekoff, C. M. 54, 64 Bennett, D. 71 –72, 92, 97, 114 Bennett, J. 13, 92 Berlyne, D. E. 230, 240 Bettman, J. R. 185 Bigelow, J. 292, 301 Bishop, J. 47, 64, 72, 76 –79, 88 Bittner, R. 45 Bovary, E. 83 Brand, M. . 21, 32, 57, 58, 55, 64, 72–73, 75, 77, 88, 99, 114, 266, 271 Brandt, R. B. 177, 183 Bratman, M. 10, 93, 96, 99–101, 114, 159, 160, 183, 187–190, 193–194, 199, 202, 203, 286 Brink, D. O. 6, 180, 183 Buchanan, A. 216, 225 Buller, D. 54, 64

Camerer, C. F. 164, 184 Castaneda, H. N. 145, 155 Chan, D. K. 103, 114 Chang, R. 194, 203 Chappell, T. 103, 114 Chisholm, R. M. 49, 53, 60, 71, 73–74, 88. Christman, J. 225 Churchland, P. S. 48, 62, 64 Churchlands, P. M. 48, 62, 64 Cialdini, R. B. 168, 170, 183 Clark, S. 230, 240 Clarke, R. 53, 65, 275, 287 Cohen, J. 194–195, 203–204 Coles, M. 259, 271 Connell, F. J. 93, 114 Cooper, W. E. 75, 88 Cosmides, L. 22, 32 Craig, E. 155 Crick, F. 65 Crozier, R. 164, 183 Cullity, G. 144, 155 Cummins, R. 54, 64 Currim, I. S. 164, 183 Dan-Cohen, M. 203 Dancy, J. 147, 155, 157, 158, 183 Davidson, D. 27, 44, 45, 47, 55, 65, 69–73, 75, 79–81, 83, 85, 88, 107, 141, 144, 155, 159, 174, 183, 184, 228, 240, 289, 293–297, 301 Davis, N. 94, 97, 114, Davis, W. H. 172, 184 Day, B. 270, 271 Deci, E. L. 168, 184 Dennett, D. 48, 62, 65 Derpmann, S. 225 Descartes, R. 281–285 Dodds, S. 292, 301 Donagan, A. 154, 184 Dretske, F. I. 61, 65

304

Intentionality, Deliberation and Autonomy

Ducasse, C. J. 60, 65 Duff, R. A. 105, 115 Dworkin, G. 289–291, 301 Edwards, P. 38, 45 Eimer, M. 230, 240, 266, 271 Enç, B. 65 Engel, A. K. 62, 65 Engelhard, C. 225 Epicurus 1 Farah, M. J. 215, 225 Feinberg, J. 290, 299, 300, 301 Finnis, J. 93, 105, 106, 115 Fischer, J. M. 203, 210, 212, 217, 220–224, 226 Flaubert, G. 83 Fletcher, G. 104–107, 115, Foley, R. 145, 156 Føllesdal, D. 75, 89 Frankfurt, H. 67, 78, 89, 188–194, 196, 198– 203, 204, 209, 218, 226, 236, 240 Freeman, A. 49, 65 Frey, R. G. 92, 115 Frith, C. 258, 262, 272 Fultz, J. 168, 170, 183 Gaut, B. 144, 155 Gauthier, D. 182, 184 Geach, P. 36, 45 Gibbard, A. 187–188, 204 Gibbons, J. 65 Gigerenzer, G. 164, 184 Gilson, E. 281, 297 Ginet, C. 12, 47, 65, 75, 87, 89, 183, 243, 247, 251–255 Glass, L. 62, 65 Gleason, C. 257, 260, 265, 272 Goethe, J. W. von 236 Goldman, A. 47, 65, 67, 74–76, 89, 159, 184 Gomes, G. 263, 270, 271 Gray, C. M. 62, 67 Grice, H. 20, 32, 74–78, 88 Grifﬁn, J. 177, 184 Habermas, J. 159, 184 Hacking, J. 239, 240 Haggard, P. 230, 240, 257, 263, 265, 270–272 Hailsham, Lord 106

Hameroff, R. 239, 240 Harackiewicz, J. M. 168, 185 Hare, R. 293, 297, 298, 301 Harless, D. W. 164, 184 Harman, G. 96, 99, 115, 145, 156, 159, 184, 193, 200, 204 Harris, J. 94, 115 Harsanyi, J. C. 177, 184 Hart, H. 35, 36, 38, 45, 92, 107, 115 Heckhausen, H. 168, 184, 269, 271 Heckman, D. 90 Hegel, G. W. F. 1 Hempel, C. G. 52, 65 Hinchman, R. 200, 204 Hintikka, M. B. 90 Hobart, R. E. 243, 251, 255 Hobbes, T. 228, 231, 234, 237, 240 Hogan, B. 105, 116 Holmstroem-Hintikka, G. 47, 65 Holton, R. 193, 200, 204 Horgan, T. 123, 132 Hull, C. L. 230, 240 Hume, D. 1, 3, 181, 228, 234, 244, 247, 255, 274 Ingarden, R. 83, 89 Isen, A. M. 168, 170, 183 Jackson, F. 61, 62, 65 Jeffrey, R. 292, 301 Johnson, E. J. 185 Kahneman, D. 164, 176, 184, 185 Kalogeras, J. 230, 240 Kane, R. 133, 212, 212–220, 222, 225, 226, 239, 240, 248, 255, 275–281, 287 Kant, I. 1, 4, 6, 109, 127, 154, 180, 181, 198, 229, 237, 240 Kapitan, T. 222, 223, 226 Kaplan, D. 62, 65 Keil, G. 5, 48, 49, 65, 69, 77, 85, 89 Keller, I. 269, 271 Kenny, A. 75, 89, 106, 115, 130, 132 Knobe, J. 17–32 Koch, C. 65 Köhler, W. 123, 132 Korsgaard, C. M. 7–8, 13, 128, 133, 193, 204 Kuhl, K. 106, 115

Index of Names Lackner, K. 106, 115 Lauder, G. 54, 64 Law Commission 115 Leibniz, G. W. 11, 13, 234, 240 LePore, E. 65 Lewis, D. 57, 65, 69, 89 Libet, B. 49, 65, 230, 240, 257–272 Locke, J. 1, 188–189, 191–192, 196–197 Lumer, C. 1, 10, 45, 49, 65, 73, 75, 87–89, 114, 133, 155, 157–159, 162, 167– 168, 172–173, 178, 180–185, 203, 271, 284–286, 300 MacCorquodale, K. 230, 240 Mackie, J. L. 98, 115, 122, 123, 133 Maertens de Noordhout, A. 270, 271 Magno, E. 269, 270, 271 Malle, B. 21, 23, 32 Mandelbaum, M. 123, 125, 133 Mangan, J. T. 97, 111, 115 Manucia, G. K. 170, 185 Marcel, A. 258, 272 Marconi, D. 48, 65 Marsden, C. 270, 271 Maslow, A. 159, 168, 185 McCann, H. 10, 17, 21, 24–26, 29–33, 119, 127, 133, 155, 183 McCarthy, D. 112, 115 McClelland, D. 62, 66, 168, 185 McConnell, S. 255 McDowell, J. 128, 133 McKnight, C. J. 289, 301 McLaughlin, B. P. 65 Mead, G. H. 173, 185 Meehl, E. 230, 240 Melden, A. I. 51, 66 Mele, A. R. 12, 17, 20, 29–33, 49, 66, 72– 76, 89, 99, 114, 115, 159, 162, 183, 185, 222, 223, 225, 226, 248, 255, 266, 271, 272, 285, 287, 291, 301 Miller, J. 266, 272 Millgram, E. 203 Millikan, R. G. 54, 61, 66 Mitchell, D. 75, 78, 89 Moore, G. E. 60, 65, 66 Morillo, C. 168, 185 Morton, A. 81, 89 Moya, C. 12, 47, 55, 66, 273, 286, 287

305

Nagel, E. 54, 66 Nagel, T. 103, 115, 128, 133, 157, 185 Nakashima, K. 270, 271 Nannini, S. 5, 47, 48, 52, 61, 66 Nelson, J. 172, 185 Newman, C. 271 Nichols, S. 32 Nossek, A. 225 Nowell-Smith, P. 243, 251, 255 Nozick, R. 171–172, 185, 189, 200, 203, 204, 219, 226 O’Connor, T. 53, 66 O’Quinn, C. 168, 170, 183 Parﬁt, D. 147, 155–156, 188, 204 Pargetter, R. 292, 301 Pascal, B. 94, 95 Payne, J. 164, 165, 185 Peacocke, C. 66, 78, 90 Pearl, D. 272 Penrose, R. 239, 240 Pereboom, D. 275, 287 Perlman, M. 54, 64 Perry, J. 81, 90 Pitcher, G. 38, 45 Plato 138, 154, 187, 194, 196, 201–204, 232, 235, 236, 281, 290 Powers, M. 93, 115 Prinz, W. 268, 272 Quante, M. 12, 209, 212, 215, 218, 225, 226 Quine, W. V. O. 195, 204 Quinn, W. 93, 94, 108, 115 Ranyard, R. 164, 183 Ravizza, M. 210, 212, 217, 220–224, 226 Rawls, J. 194 –195, 204 Raz, J. 115, 194, 204 Remnant, P. 13 Rescher, N. 54, 66 Rorty, R. 47, 66 Roth, G. 49, 57, 66 Rothwell, G. 270, 271 Roughley, N. 5, 91, 100, 115, 231, 240 Rowe, W. 234, 240 Rubenfeld, J. 197, 204 Rugg, M. 259, 271

306

Intentionality, Deliberation and Autonomy

Rumelhart, D. E. 62, 66 Russell, B. 60, 66 Ryan, M. 168, 184 Ryle, G. 66 Saarinen, R. 230, 240 Sansone, C. 168, 185 Sarin, R. K. 164, 183 Scanlon, T. M. 102, 115 Schaber, P. 110, 180, 185 Schaller. M. 168, 183 Schauer, F. 201, 204 Schefﬂer, S. 4, 13, 203 Schlick, M. 228, 238, 240 Schopenhauer, A. 1, 182, 228, 231, 237, 237, 238, 240 Schweikard, D. 225 Searle, J. 47, 48, 50, 66, 75, 78, 90, 158, 185, 231, 240 Seebass, G. 12, 107, 116, 227, 230–234, 240 Sehon, S. R. 78, 90 Seidman, J. 201, 203, 204 Seinowski, T. J. 62, 65 Setiya, K. 121, 133 Shaffer, J. 73–74, 90 Shannon, K. 270, 271 Shugan, S. M. 164, 165, 185 Singer, W. 62, 67 Skorupski, J. 195, 205 Smart, J. J. C. 247, 251, 255 Smilansky, S. 243, 255 Smith, J. C. 105, 106, 116 Smith, M. 157, 185, 195, 205 Sober, E. 59, 67 Soeder, J. R. 234, 241 Spean, S. 258, 262, 272 Spitzley, T. 12–13, 45, 289 Stampe, D. 156 Steadman, A. 5, 17, 24, 27, 32 Stoecker, R. 5, 35, 40, 44, 45, 72, 82, 89–90, 114, 300 Stoutland, F. 49, 51, 67 Strawson, G. 243, 248, 255, 275–281, 287 Sutherland, K. 49, 65 Svavarsdòttir, S. 128, 133 Svenson, O. 164, 165, 183, 185 Swain, M. 156 Sylwanowicz, M. 233, 241

Tatarkiewicz, W. 172, 185 Taylor, C. 67 Taylor, R. 53, 67, 238, 241 Thalberg, I. 78 Thompson, P. 270, 271 Timmons, M. 123, 132 Tooby, J. 22, 32 Trevena, J. 266, 272 Tugendhat, E. 231, 241 Tuomela, R. 47, 50, 53, 65, 67 Tversky, A. 164, 176, 184, 185 Unger, P. 147, 156 Valla, L. 227, 241 Van de Grind, W. 270, 272 Van Inwagen, P. 243, 249–255, 286, 287 Vanderplas, M. 168, 170, 183 Vargas, M. 203 Vassallo, N. 48, 64 Velleman, J. D. 100, 116, 155, 159, 185, 188, 191–192, 197, 205 Vermazen, B. 88, 90 Waller, B. 248, 255 Walter, S. 90 Waterman, A. 168, 172, 186 Watson, G. 13, 191, 205, 274, 277, 282, 287, 289, 293, 297, 301 Wegner, D. M. 49, 53, 67 Wendel, H.J. 67 Widerker, D. 255 Wiggins, D. 201, 205, 284, 287 Wilker, D. 172, 185 Williams, B. 128, 133, 182, 186, 279, 287 Williams, G. 106, 116 Williamson, T. 147, 156 Wittgenstein, L. 47, 51, 57, 67 Wlash, D. M. 48, 67 Wolf, S. 274, 282, 287 Wolf, U. 289, 297–299, 301 Wright, E. 257, 260, 265, 272 Wright, G. H. von 41, 47, 48, 52, 67, 159, 186, 228, 241 Wright, L. 54, 67 Yaffe, G. 203 Young, R. 172, 186, 300, 301

Subject Index

absolutism, moral 97–98, 109 ‘act now’ process 257, 266 action(s) / acting 257–272 against one’s own better judgement 297, 299–300; see also weakness of will autonomous 289–294, 296–297, 299 incontinent 294; see also weakness of will intentional(ly) see intentionality vs. acting knowingly see intentionality, vs. acting knowingly legal concept of see intentionality, legal concept of negative 43 rational see rational action; see also practical rationality standard conception of 38 action explanation 3, 6–10, 37, 40–45, 148, 150 causal 49, 51–52, 59 intentional 49–51 by reasons 251–252 action plan 73–74, 81–83, 87 action sentences 35, 39 action theory 1–13, 35, 66 advice giving, practical–technological 3, 10–13 as basis for practical philosophy 1, 4–13 causal 47, 49, 51, 58–59, 64, 69–88 history of 47–49 methods 2–4 surveys 20–21, 26, 29–32 parts of 2–4 action trees 43 active 229, 237, 239 activity see active agency 43 agent causation 53–54, 59, 76, 245, 250–251 agential authority see authority, agential

aggregation of intrinsic desires / valuations see desire(s), intrinsic, aggregation of akrasia / akratic see weakness of will alternative possibilities see possibility, alternative analysis, conceptual 2, 4–5 anomalism of the mental (Davidson) 70, 80, 83 ascribability of action 4, 247 , 250 ascriptivism 35–47 asymmetry of intentionally helping and harming 18–19, 22–26, 31 attributing autonomy see autonomy, attribution of attributing responsibility see responsibility, attribution of authority, agential 187–193, 196–197, 202 bootstrapping 200 Frankfurtian model 189–193, 202 Lockean model 189, 191–192 autonomy / autonomous 10–13, 289–292, 297, 300; see also heteronomy attribution of 209–210, 219 for real people 209 backsliding 292 basic actions 94, 106 belief 135–140, 143–146, 149, 150–155, 278 evaluative 273, 278–279, 281–285 relation to desire 136–138, 140, 144, 155 responsibility for 279–284 blame and intentionality see intentionality, condition for blaming Buridan’s problem 161, 213, 276 causal theory of action see theory of action, causal causality

308

Intentionality, Deliberation and Autonomy

intentional 59–60 natural 49, 53–54, 59 causation 244–245, 269 deterministic 244, 250–254 indeterministic 244, 250–254 purposive 54, 59 chance 227, 235, 239, 248–250, 254 character 236–237 cognitive turn (in philosophy of mind and action) 47, 62, 64 commitment 99–101 compatibilism 227, 229, 234, 238 compulsion / compulsive 291–293, 297 consequence argument (against compatibilism) 212–213 content of mental states and of propositional attitudes 49, 59, 135–136, 140–141, 143, 145–147, 151 continence, principle of 295 control 273–274, 280–285 ultimate 274–284 culpability 103, 105–106 deciding / decision 100–102, 230, 236–237, 239, 257–273 explanation 3, 6–10 how to decide 165 decision explanation see deciding / decision, explanation decision theory, rational 164, 176–177 deliberation (practical) 128–130, 163–167; see also practical reason(s); practical syllogism deriving from social practice 41–43 minimal 100, 165 density of the real world 81–83 dependency thesis for ethics and theory of practical rationality 6–9, 11 desirability see desire desire(s) 257–260 as apprehension of objective good 121–125, 131–132 distinct from judgment 120; see also belief, relation to desire intrinsic 163–173 aggregation of 163–167 dynamics of 172–173, 176 hedonic 168–169, 171–172 induced by feelings 169–171, 179

prospect desire 163–167, 172–173, 178–179 total 163–167 and value supervenience 124 variability of 125–127 determinism (and indeterminism) 227–229, 233–239 not in psychology 239 and quantum mechanics 245 spots of indeterminacy see spots of indeterminacy deviant causal chains 44, 49, 54–57, 69–88 basic and nonbasic deviance 72–73 causal immediacy strategy 77–79 de re judgement 84, 88 Gricean deference 74–77 method for deﬁning them 2 normality clause 75, 79, 82, 84 tertiary waywardness 73, 77 direction of ﬁt 144, 146, 176 discounting (of future events) / discount curves 295–296 doctrine of double effect 91–118 doing vs. allowing (to happen) 95 education 210, 214, 220, 222–224 emotions their role in valuation 285 enhancement 209, 215 ethics, basis of 6, 180–183 explanation of action see action explanation externalism in ethics see internalism facticity 147–149, 151–152 fatalism 239 folk concepts 17, 20–21, 30–31 folk judgments 17–18 free will 11, 227–230, 233–237, 252, 257–272 and neurosciences 228, 257–272 freedom / free 227, 230–232, 237, 239, 243–244, 250–252 of action 11, 252 of decision see free will of the will see free will full information account of practical rationality see practical rationality, full information account of

Subject Index goodness 142 harm vignette 18, 25–26 hedonism, psychological 168–169, 171–172, 174, 179 help vignette 18–20, 22–24 heteronomy 285 history of action theory see action theory, history of inability to act 292, 297 psychological 292 indexicality 84 inference, practical see practical syllogism intending see intention intention(s) / intending 23–27, 105, 159–163, 229–231, 258–273 and acceptance 95–96, 99–103, 105–114 as all things considered judgement 294–295 British legal intention (blintention) 106 doxastic theories of 160–161 empirical features of 160–162 optimality belief theory of 159–162, 164–167 as unconditional judgement 294–295 versus downstream theory 92, 100–102 intention-action relation 49, 51–55, 59 intentional action see intentionality intentionality 4, 17–24, 26–27, 29–33, 59, 63 Brentano’s sense 59 in common sense surveys 17–18, 20–21, 26, 29–32 condition for blaming 17–19, 22–31, 94–95 dissociated from intentions? 20 legal concept of 21; see also intention, British legal intention moral considerations in ascribing 20, 32 pragmatics of ascription 17, 20, 22–23, 25–32 semantics of the concept 19–20, 25–26, 30 versus acting knowingly 17–31, 91, 103–104, 108–109; see also intention and acceptance of weak-willed actions 300 intentionally see intentionality internalism and externalism 6, 159

309

in ethics 6–7, 180–183 effective 6–7 foundational 6–7 motivational 18–19, 28, 128–131, 138, 144 in theory of practical rationality 8 effective 8 foundational 8–9 judgement 294–295 all things considered see intention, as all things considered judgement conditional / prima facie 294–295 evaluative 294–295 practical see practical judgement unconditional / sans phrase see intention, as unconditional judgement justiﬁcation, moral 109–113 knowingly see intentionality, vs. acting knowingly law (legal) see intentionality, legal concept of; see also intention, British legal intention law of nature 244, 250, 254 Logical Connection Argument 51–53 luck 247–248, 255 manipulation 209, 212–213, 217–218, 221, 224–225 manipulation argument (for incompatibilism) 212–213, 222–225 means, and permissibility of actions 92, 96–97 Mind Argument (against compatibilism) 285 moral considerations in ascribing intentionality see intentionality, moral considerations in ascribing moral reasons see reasons, moral motivation (by evaluative beliefs) 284 motivational internalism see internalism, motivational 18–19, 28 motives for moral action 170, 173–174 murder, legal deﬁnition 105 naturalism 48, 58 negative actions see action(s) negative

310

Intentionality, Deliberation and Autonomy

neurosciences and voluntary 49, 57, 62; see also free will and neurosciences noncognitivism 122–123 normativity 137 objectivism, evaluative see value objectivism and subjectivism optativity, optative attitudinizing 95, 99–102, 106 optimality belief theory of intention see intention, optimality belief theory of package deal 96, 100–102 permissibility 93 person 227, 230, 236–237, 239 personality 209, 218, 222 philosophy of action see action theory philosophy of mind (and action theory) 47, 49, 62 plan / planning 199–203 Platonic model 202 pluralism of values 194–195, 202 possibility 227, 232–234, 237 alternative 283 practical deliberation see deliberation (practical) practical judgement 286 practical philosophy 10–13 basis of 1–13 practical rationality 176–179; see also practical reaon(s) basis of see practical reason(s), basis of full information account of 177–178 reconstructive approach to 178–179 practical reason 135, 142, 145, 154 content skepticism about 7–8 practical reason(s) 135, 139, 141, 144–146, 149–150 , 152–154, 157–159 axiological character 180 basis of 175–183 explanatory 157–159 normative 157 ontology of 141, 158–159 practical reasoning see deliberation (practical) practical syllogism 131, 165 practical thinking see deliberation (pratical) propositional attitudes 140, 143, 151 prospect desire see desire(s), prospect desire

quantum mechanics see determinism, and quantum mechanics random 247–249 rational action 135, 151–153, 186; see also practical rationality rationality / rational 137–142, 150–152, 154–155, 235 practical see practical rationality rationalization Davidson 71, 79, 85 and weakness of will 298 reaction time 257, 268 readiness potential (RP) 259–272 onset 259–266 type I 261–272 type II 261–272 realism, moral 181 reason(s) 135, 139, 141–149, 151–155 moral 153–154 practical see practical reason(s) reasoning and free will 235–236 reasons explanation of action see action explanation, by reasons recklessness 105–106 reconsideration of plans and intentions 200–202 representation, mental 62 responsibility 11, 37, 229, 238–239 attribution of 209–211, 215–216, 218, 222 moral 103, 105–109, 113, 243–244, 252 taking 209–211, 215–216, 218, 220–222, 224–225 ultimate 214–216 RP see readiness potential self-control 293 self-deception 297–300 simple theory 84 simple view 17–18, 26, 31–32 Socialization 218, 222 spots of indeterminacy (Ingarden) 82–83 stability of practical attitudes 186, 190, 192–203 surveys about meaning of ‘intentional’ see intentionality, in common sense, surveys

Subject Index as method in action theory see action theory, methods, surveys syllogism, practical see practical syllogism theoretical reasons 135, 139, 141, 143–144, 147, 150, 152–155 theory of action see action theory theory of perception, adverbial 49, 58, 60–64 thought experiment 49–51, 55–58, 61–62 truth–groundability 151–154 universal means 196–198 urge 257–272 urge* hypothesis 268–272 utilitarianism 179 utility theory 164, 295–296

311

value objectivism and subjectivism 9, 159, 284 valuing 189, 194–196, 199, 201–202; see also desire(s) vetoing 260–272 volition 244, 247; see also will / willing volitional necessity 190–193, 196, 198–199, 202 wanting 232, 257–259, 270; see also desire; intention wayward causal chains see deviant causal chains weakness of will 289, 294–300; see also action, incontinent; akrasia will / willing 11–12, 124, 138, 143–144, 147, 155 , 228–232, 235–237, 292; see also volition active and passive 119