Commitment and Evolution. Connecting Emotion and Reason in Long-term Relationships

Commitment and Evolution Connecting Emotion and Reason in Long-term Relationships István Back

ii

The research described in this thesis was carried out under the auspices of the Interuniversity Center for Social Science Theory and Methodology (ICS) and the Faculty of Social and Behavioral Sciences (GMW) at the University of Groningen (RuG). Funding was generously provided by the Ubbo Emmius Bursary (2003) and by the Netherlands Organization for Scientific Research (NWO). © 2007 by István Back 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 without written permission of the author. The document was typeset using LATEX 2ε and Stasinos Konstantopoulos’s RuGthesis.cls. Printed by Mesterprint Kft., Budapest, Hungary. ICS Dissertation series (nr. 133)

ISBN 978-90-367-3113-3

R IJKSUNIVERSITEIT G RONINGEN

Commitment and Evolution Connecting Emotion and Reason in Long-term Relationships

Proefschrift

ter verkrijging van het doctoraat in de Gedrags- en Maatschappijwetenschappen aan de Rijksuniversiteit Groningen op gezag van de Rector Magnificus, dr. F. Zwarts, in het openbaar te verdedigen op maandag 17 september 2007 om 14.45 uur

door

István Henrik Back geboren op 19 november 1979 te Boedapest, Hongarije

iv

Promotor:

Prof. dr. T.A.B. Snijders

Copromotores:

Dr. H. de Vos Dr. A. Flache

Beoordelingscommissie:

Prof. dr. M.W. Macy Prof. dr. S.M. Lindenberg Prof. dr. A. Riedl

Acknowledgements I would like to thank my supervisors, Henk de Vos, Tom Snijders, and especially Andreas Flache for their invaluable input throughout the last four years. Their encouragement, inspiring ideas and complementary expertise was instrumental to carrying out this interdisciplinary piece of research. I received further assistance from many others working in the graduate school ICS in Groningen, Utrecht and Nijmegen, most notably Vincent Buskens who always kept a watchful eye on my research, provided me with ideas and helped to carry out a large proportion of my experiments; Károly Takács and Michael Mäs, who were always ready to read and discuss my drafts; Rita Smaniotto who helped me keep my enthusiasm for evolutionary theory; Sigi Lindenberg; Jeroen Weesie; Inneke Maas; Frans Stokman; Richard Zijdeman, Eva Jaspers, Ellen Verbakel, Nienke Moor, Janneke Joly and Stefan Thau; Jessica Pass, Jacob Dijkstra, Christian Steglich, Lea Ellwardt and Jurre van den Berg. I thank Michael Macy and graduate students at the Sociology Department of Cornell University, especially Arnout van de Rijt and Ma Li for making my visit there so rewarding not only professionally but also personally; and David Sloan Wilson at Binghamton University for thought-provoking conversations about human evolution. I owe big thanks to Ji Wenxi (Wendy) who not only provided me with invaluable support during my experimental work in China but continues to be a window into the oriental mind and thinking; Fan Xuejuan at East China Normal University, Xu Bo and Xu Longshun at Fudan University who generously provided the means to carry out my experiments in Shanghai; Zhao Kanglian at Nanjing University. Xu Yu, Gerbren Kuiper and Huixin, Yorgos Vleioras, Justin Park, Simon Dalley, Bori Takács, Tamás Bíró, Ela Polek, Andrea Szentgyörgyi, Gábor Imre, Li Kun and Wang Zhuo for keeping me company in Groningen; and Evelien de Roos, the best land-lady in the Netherlands. Finally, I would like to thank my friends in Hungary, Miki Rosta, Levente Skultéti, Attila Máté, György Hermann, Zoli Gedei, Kristóf Bajnok and Laura Radics, for their friendship which has been one of the key sources of motivation behind this piece of work.

v

Contents 1

Introduction 1.0.1 A brief word on “commitment” . . . . . . . . . . . . . . . 1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1.1 Commitment to a course of action . . . . . . . . . . . . . 1.1.2 Interpersonal commitment . . . . . . . . . . . . . . . . . 1.2 Toward an evolutionary explanation . . . . . . . . . . . . . . . . 1.2.1 Separating ultimate and proximate explanations . . . . . 1.2.2 How evolutionary theory helps to explain seemingly irrational behavior . . . . . . . . . . . . . . . . . . . . . . . 1.2.3 Constructing an ultimate explanation for interpersonal commitment . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2.4 Proximate mechanisms for interpersonal commitment . 1.3 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.4 Outline of chapters . . . . . . . . . . . . . . . . . . . . . . . . . .

17 20 21 22

I

An ultimate explanation

25

2

The Competitive Advantage of Commitment 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . 2.2 Model . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.1 Modeling strategies . . . . . . . . . . . . . 2.2.2 Evolutionary dynamic . . . . . . . . . . . . 2.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . 2.3.1 Simulation setup . . . . . . . . . . . . . . . 2.3.2 The unconditionality of Commitment . . . 2.3.3 Explanation: the importance of strong ties 2.3.4 Sensitivity to initial parameters . . . . . . . 2.4 Discussion and Conclusion . . . . . . . . . . . . .

27 28 32 33 37 38 39 41 46 48 50

vii

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

1 4 5 6 7 12 15 15

viii 3

4

II 5

CONTENTS The Evolutionary Advantage of Commitment 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.1 Modeling strategies . . . . . . . . . . . . . . . 3.2.2 Evolutionary dynamic . . . . . . . . . . . . . . 3.3 Conjectures . . . . . . . . . . . . . . . . . . . . . . . . 3.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4.1 Initial parameters . . . . . . . . . . . . . . . . . 3.4.2 Stability . . . . . . . . . . . . . . . . . . . . . . 3.4.3 The importance of interpersonal commitment 3.5 Discussion and conclusions . . . . . . . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

Fairness and Commitment under Inequality 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.1 Modeling strategies . . . . . . . . . . . . . . . . . . . . . 4.2.2 Evolutionary dynamic . . . . . . . . . . . . . . . . . . . . 4.3 Conjectures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.1 Initial parameters . . . . . . . . . . . . . . . . . . . . . . . 4.4.2 Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.3 The importance of interpersonal commitment . . . . . . 4.4.4 The relative importance of fairness, commitment and capability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.5 Sensitivity to initial parameters . . . . . . . . . . . . . . . 4.5 Discussion and Conclusion . . . . . . . . . . . . . . . . . . . . .

Proximate explanations Commitment Bias 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 Hypotheses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3 Experimental design . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.1 Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.2 Sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4.1 Hypothesis 5.1 and 5.2 – The commitment bias . . . . . . 5.4.2 Hypothesis 5.3 – Effect of affect . . . . . . . . . . . . . . . 5.4.3 Hypothesis 5.4 and 5.5 – Cross-cultural similarities and differences . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4.4 Self-reported reasons for exit . . . . . . . . . . . . . . . . 5.5 Discussion and conclusions . . . . . . . . . . . . . . . . . . . . .

53 54 56 56 58 59 60 61 61 62 67 71 72 73 74 76 77 79 80 80 81 81 84 85

89 91 92 95 98 100 100 100 101 103 103 107 107

CONTENTS

ix

5.5.1 5.5.2

The evolutionary roots of commitment . . . . . . . . . . 109 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . 111

6

Commitment and Networking under Uncertainty 6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2 Theory and Hypotheses . . . . . . . . . . . . . . . . . . . . . . . 6.3 Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3.1 Sample 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3.2 Sample 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3.3 Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3.4 Manipulation of uncertainties . . . . . . . . . . . . . . . . 6.3.5 Manipulation of dilemma . . . . . . . . . . . . . . . . . . 6.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.4.1 Final sample . . . . . . . . . . . . . . . . . . . . . . . . . . 6.4.2 Hypothesis 6.1 – Social uncertainty and commitment . . 6.4.3 Hypothesis 6.2 – Resource uncertainty and commitment 6.4.4 Hypothesis 6.3/6.4 – Interaction between uncertainties . 6.4.5 Hypothesis 6.5/6.6 – Trust and Optimism . . . . . . . . . 6.5 Robustness of results across different dilemmas . . . . . . . . . 6.6 Conclusion and Discussion . . . . . . . . . . . . . . . . . . . . . 6.6.1 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . .

113 114 115 121 122 122 123 124 125 125 126 127 127 128 129 130 131 132

7

Conclusions 7.1 Summary of results . . . . . . . . . . . . . . . . . . . . . . . . . 7.2 General discussion . . . . . . . . . . . . . . . . . . . . . . . . . 7.2.1 In defense of evolutionary theory in the social sciences 7.2.2 Placing our work . . . . . . . . . . . . . . . . . . . . . . 7.2.3 Innovations of the present work . . . . . . . . . . . . . 7.2.4 Possible criticism . . . . . . . . . . . . . . . . . . . . . . 7.2.5 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . 7.2.6 Avenues for future research . . . . . . . . . . . . . . . .

133 134 136 137 138 141 142 143 144

. . . . . . . .

Bibliography

147

Summary

159

Samenvatting - Dutch summary

163

Összefoglalás - Hungarian summary

167

Zhai Yao - Chinese summary

171

A Analytical solution of the simplified dilemma

175

x

CONTENTS

B Parameter values used in ecological simulations

177

C Pseudocode of simulation core

179

D Parameter values used in evolutionary simulations

181

E Pseudocode of evolutionary dynamic

183

F Experiment instructions F.1 Initial instructions . . . . . . . . . . . . . . . . . . . . . . . . . . . F.2 Instruction text from the experiment . . . . . . . . . . . . . . . . F.3 Screenshot from the experiment game . . . . . . . . . . . . . . .

185 185 185 187

G Experiment instructions 189 G.1 Instructions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189 G.2 Screen shots from the experiment . . . . . . . . . . . . . . . . . . 190

Chapter 1

Introduction “We may call the part of the soul whereby it reflects, rational; and the other with which it feels hunger and thirst and is distracted by sexual passion and all the other desires, we will call irrational appetite, associated with pleasure in the replenishment of certain wants... What of that passionate element which makes us feel angry and indignant? Is that a third, or identical in nature with one of those two?” —Plato, The Republic

The tendency to establish lasting personal relationships is a fundamental aspect of human sociality. Throughout life we build friendships, collect acquaintances, forge business alliances, become attached to intimate partners. Many of these relationships follow us through our lives and integrate us into a complex social fabric of interpersonal connections. At the same time, establishing and maintaining long-term relationships involves substantial investment of one’s time, effort and other resources. Moreover, many relationships by definition require exclusivity. For example, we can only have one best friend at a time, in many cultures only one spouse, and in many business settings only one supplier of some product. To a certain extent all relationships, i.e. non-exclusive ones as well, are competitive with each other, given that we have finite attention and resources. This means that we occasionally have to forgo relationships with potentially better alternative partners. And to complicate matters, even when we do our best to invest in a relationship, we have to live with the risk of being dumped for someone else or unknowingly being taken advantage of by our partner. Why do people establish and maintain long-term relationships when these are costly, risky and exclusive? A simple but powerful answer from rational 1

2

Chapter 1. Introduction

choice theory is that it is in their best interest to do so. More precisely, people become committed to each other if and only if the benefits of having a relationship outweigh its maintenance costs and its alternative costs. In particular, having a long-term relationship with a partner provides valuable information about the trustworthiness of the partner compared to other partners (trust explanation, see Kollock, 1994; Yamagishi and Yamagishi, 1994; Yamagishi et al., 1994) and at the same time creates a strategic incentive to cooperate in order to avoid retaliation and stabilize long-term mutual collaboration (reciprocity explanation, see Trivers, 1971; Friedman, 1971; Axelrod, 1984; Fehr and Schmidt, 1999; Fehr and Gächter, 2002; Falk et al., 2001). But at the same time, there seems to be much more to long-term interpersonal relationships than just trust and reciprocity. There are numerous cases, for example, when people keep relationships even after their partner has proved to be untrustworthy (e.g. Roy, 1977; Strube, 1988; Rusbult and Martz, 1995). There are also examples of relationships where a partner has no means of reciprocating in the future (e.g. Monahan and Hooker, 1997). What is it that makes battered wives return to their abusive husband when there are hardly any prospects for change? And why does someone take care of a lifelong partner with Alzheimer’s disease who will never be able to recognize the caretaker? Why do subjects in controlled laboratory experiments give costly gifts to their long-term exchange partners when their identity will never be revealed to each other? A great wealth of empirical evidence suggests that people are engaged in long-term relationships with their full emotional repertoire (cf. Baumeister and Leary, 1995). People create social relationships with great ease even in the absence of materialistic benefits or other ulterior motives, and strongly resist the dissolution of these relationships, well beyond rational considerations of practical advantages. Many of the strongest emotions people experience in their life, both positive and negative, are linked to long-term relationships. The evidence suggests that being accepted, included, or welcomed leads to positive emotions such as happiness, elation, contentment, and calm, whereas being rejected, excluded, or ignored leads to anxiety, depression, grief, jealousy, and loneliness, etc. Indeed, the evidence is sufficiently broad and consistent to suggest that one of the basic functions of emotion is to regulate behavior so as to form and maintain social bonds (Baumeister and Leary, 1995). There is further evidence that people observe and evaluate alternative partners with a biased vision, systematically dependent on how committed their current relationship is (Johnson and Rusbult, 1989). Moreover, we know that even in anonymous exchange settings, positive emotions develop toward frequent exchange partners, and toward the relationship itself, being perceived as an object of value (Lawler and Yoon, 1996). These emotions provide a positive feedback for commitment behavior and lead to a systematic divergence from instrumental rationality.

3 But why is it that our relationship-related emotions are so often out of tune with what is usually regarded as rational? What is the source of emotions that make us consistently more committed than our best interest seems to dictate? Is there, in fact, something fundamentally rational behind seemingly irrational commitments? In order to resolve the paradox between rational and emotional explanations of interpersonal commitment, we put forward an evolutionary explanation. During countless years of prehistoric evolutionary adaptation in the human ancestral environment, people lived together in small groups and fought for daily survival in a world more hostile than today’s (Sterelny, 2003). With many of the formal and informal helping institutions of modern society missing, people had to rely on interpersonal relationships to a much larger extent than today. Sometime during the Pleistocene epoch (roughly 1.8 million years to 12 thousand years before the present) humans moved from rain forests to the savannah, which increased the need for collective hunting and mutual protection from large predators. This in turn created a selection pressure for increased social complexity (c.f. Smaniotto, 2004). At the same time, life-threatening situations produced more opportunities for bonding and deep friendships. Being capable and willing to establish and maintain long-term stable relationships substantially increased one’s survival and reproductive chances. As a consequence, those whose cognitive arsenal was equipped with better tools and stronger preferences for making interpersonal commitments gradually increased their presence in the population over many generations (cf. Nesse, 2001a). In lack of a direct test, evolutionary theories are difficult to empirically falsify and therefore problematic to find convincing support for. Therefore, our strategy in this dissertation is twofold. We first examine a theory of natural selection acting on commitment in closer detail in Part I (An ultimate explanation)1 . The main motivating question for this effort is: Could a trait of interpersonal commitment have been selected for in human evolutionary history, especially in the face of other, more or less cooperative, traits? Building on previous work (especially de Vos et al., 2001) that relies on anthropological knowledge about conditions of the human ancestral environment, we create formal computational models of the ancestral environment. The purpose of these models is to test the internal consistency of an evolutionary theory about deeply rooted (or “hardwired”) emotions that facilitate interpersonal commitments. Then, in Part II (Proximate explanations) we move on to empirically test the existence of an evolved commitment trait. The main question this part addresses is: Are there features of contemporary social behavior that are in line with an ancestral trait for commitment but cannot readily be explained by 1 See

more about the important distinction between ultimate and proximate explanations in evolutionary theory under Section 1.2.1 of this chapter.

4

Chapter 1. Introduction

simpler, existing theories? In order to test the existence of such a proximate mechanism (which we term the commitment bias), we conducted laboratory experiments at six locations in three different countries (the Netherlands, USA and China). In particular, we aimed to find support for mechanisms that are difficult to reconcile with current exchange theoretical and (social) psychological theories but become intelligible in light of the evolutionary explanation. In the remainder of this introductory chapter we are first going to clarify an important issue about the use of the word commitment. We then address the vast literature of commitment from philosophy, economics, game theory, exchange theory, psychology, sociology, and evolutionary psychology, pointing to how the dichotomy of emotional and rational explanations permeates the subject throughout these disciplines. Building on this broad background, we set out to construct an evolutionary theory that aims to bring the two sides of the dichotomy closer to each other and thus present the diverse literature of commitment in a new light.

1.0.1

A brief word on “commitment”

Before we turn to the substantial discussion of commitments, a brief clarification of the term itself is inevitable. The word “commitment” is used excessively in different meanings, within different contexts, which may lead to misunderstandings. It also creates seemingly unrelated research lines in various disciplines across the humanities and behavioral sciences. The first known record of the word entering the English language is from 1386, when Geoffrey Chaucer advised “commit the keeping of your person to your true friends..., who are the best «physicians» and most reliable help and healing” (Wyatt, 1999). Thus, in its original sense, commitment is a promise or threat, pledge, agreement, contract or dedication, made to oneself or to others, to do something or to act in a certain way in the future. “Being committed to protect one’s country from enemies” or “committing oneself to not getting married” are examples. Commitment in these cases is similar in meaning to persistence or consistence (see Section 1.1.1 “Commitment to a course of action” below). By extension of meaning, commitment came to refer to a bond, or loyalty toward a social entity, such as an organization, a group of people, or another person. The basis for this extension is that in such cases one acts in accordance with one’s expressed or understood promise to the entity, and membership therein. A friendship, a marriage vow, an employment contract, or simply refraining from extra-couple romance are examples. Commitment in this sense is related to meanings of belonging, stay behavior, loyalty or faithfulness (see Section 1.1.2 “Interpersonal Commitment” below). Arguably, these different meanings are not independent, and a closer look reveals a number of common characteristics. Firstly, commitment always re-

1.1. Background

5

quires behavioral consistency, in other words acting repeatedly in the same way with regard to the target of commitment. Secondly, commitment entails opportunity costs for the individual due to sacrificing potential rewards from alternative courses of action, that are not explored due to behavioral consistency. Finally, commitment is always temporally embedded – it has a duration in time, or at least it is in some sense about the future. It is by definition continuous in time because one cannot uphold the same commitment in disjoint fractions of time. Given these conceptual similarities behind different forms of commitment, it is surprising that hardly any interdisciplinary research has systematically explored links between commitment in the action and in the interpersonal sense. This is not our major undertaking either but as we will demonstrate through a brief literature review below, there is at least one crucial point on which most theories of commitment converge. This common point is the duopoly of two competitive explanations: one that advocates rational reasons and another that points to deeply rooted emotions. Ignoring either type of explanation leaves a theory potentially vulnerable to criticism by the other side. Our goal is therefore to derive and test hypotheses within an evolutionary framework that is able to accommodate both types of explanations and resolve some of the contradictions arising between them within the context of interpersonal commitment.

1.1

Background

The idea of interpersonal commitment is conceptually embedded into the more general notion of commitment to a course of action. To identify the implications of the more general concept for the more specific, we start our theoretical discussion with this broader idea of commitment. From here we proceed to our core topic of interpersonal commitment, the tendency to maintain long-term relationships. The existing literature of interpersonal commitment can be separated into two, largely disconnected fields. The first field, researched mostly by economists, focuses on exchange and social networks. The second, researched mostly by psychologists, is more directed at close relationships, such as married and romantic couples. We review some of the most important contributions within each field, pointing to the presence of the emotion-rationality dichotomy throughout. Finally, we briefly touch upon a hybrid area, organizational commitment, which is closest to business and management research, although it originally grew out of the psychological field of interpersonal commitment.

6

1.1.1

Chapter 1. Introduction

Commitment to a course of action

Committing ourselves to a course of action means that we voluntarily give up some of our freedom of choice, by agreeing to do (or not to do) something at some point in the future. The willingness to make such commitments has long intrigued scientists and philosophers alike, going back as far as ancient times. The paradoxical benefits of this seemingly self-defeating behavior was already recognized by ancient Greeks. Xenophon, a talented general, when facing a superior enemy, ordered his troops to take up a position with their backs to an impassable ravine (cf. Schelling, 2006) in order to eliminate all their routes of escape. By doing so he signaled both to the enemy and to his own men that there was no alternative for survival, except victory. A major theoretical advance came in 1785 with Immanuel Kant’s “Grundlegung zur Metaphysik der Sitten”, where he proposed to distinguish between two sources of commitment (cf. Levinger, 1999). He argued that commitment, on the one hand, can grow out of desire or affection. In the case of commitment of desire, people act out of inclination, for example, because they like to or enjoy it. Kant considered this type of commitment transient and therefore weak and untrustworthy. The other form of commitment stems from duty or moral obligation, in which case people act in accordance with principles. Kant argued that this type of commitment is more enduring and far better morally. With this theoretical distinction between “having to” and “wanting to”, Kant essentially created the fundamental dichotomy between rational and emotional explanations that still dominates the discourse over commitment. The next major contributor to the theory of commitment was Thomas Schelling with his seminal book “The Strategy of Conflict” (1963). For Schelling, commitment is a strategic tool, deliberate action, the purpose of which is to influence someone else’s choices. Schelling recognized the importance of being able to make commitments in situations where each actor’s outcome mutually depends on other actors’ actions. In such situations each actor needs to take into account what others are likely to do next. The fact that one makes a commitment to act in a certain way radically alters the expectations and decision processes of others (Schelling, 2006). The very possibility to make commitments is a key mechanism for achieving collectively desirable outcomes that are otherwise difficult to agree on (see e.g. Raub, 2004). According to Becker’s side-bet theory (1960), making a commitment links investment in an extraneous interest (a side-bet) with a consistent line of action. In Becker’s example, a man wants to buy a house. The man makes an initial offer of sixteen thousand dollars to the owner. The owner insists on having twenty thousand. Our well-prepared buyer, however, reaches into his pocket to produce certified proof that he had made a bet of five thousand dollars with a third-party that he will not pay more than sixteen thousand for the house. The seller has no choice but to accept the buyer’s standpoint.

1.1. Background

7

In this example, the buyer uses a “credible threat” (a commitment) to modify his own payoff structure, thus also modifying the strategic interdependence between the two. Becker’s theory has received extensive attention, and was widely tested empirically, albeit with mixed success (cf. Cohen and Lowenberg, 1990; Wallace, 1997). Another conceptualization of side-bets is voluntary hostage posting (e.g. Raub, 2004). Hostage posting means surrendering an object of value to a trustor in order to increase trust in the trustee’s willingness to uphold the promise (commitment) made to the trustor. The hostage promotes trust (at least in the economic sense) by binding the trustee through reducing his incentives for abusing trust, by providing compensation for the trustor in case trust is abused, and by serving as a signal for the trustor about unobservable characteristics of the trustee that are related to the trustee’s opportunities and incentives for abusing trust (Raub, 2004; Snijders and Buskens, 2001). An interesting case of commitment is the tendency to escalate investment in a failing course of action, in other words, “throwing good money after bad” (cf. Karlsson et al., 2005; Brockner, 1992; Staw, 1976, 1997). Also known as the sunk cost effect, this motivates people to continue investment in a project despite unsuccessful prior investments of money, effort, or time (Arkes and Blumer, 1985). It is important to recognize that thinking in terms of sunk costs is a departure from rational calculation, in the sense that it distorts actual costs and benefits associated with possible outcomes.

1.1.2

Interpersonal commitment

Interpersonal commitment, or becoming committed to long-term partners, is regarded as a special case of commitment to a course of action by some game theorists, economists and also others (cf. Frank, 1988; Nesse, 2001a). The core idea behind this association is that commitment in long-term relationships is based on an implicit or explicit promise to stay with the partner, and to uphold a general conduct that is aligned with the interests and expectations of the partner. This dissertation focuses on long-term interpersonal relationships, such as marriage, friendship and acquaintanceship, asking the question: why do people become committed to each other when it is seemingly not in their best interest? Just as commitment to uphold a certain course of action entails a seemingly irrational decision to reduce one’s set of available choices in the future, interpersonal commitment involves sacrificing interaction with potentially superior, alternative partners. Yet, as the former type of commitment proves to be not only rational but, in fact, essential for success in society, could the same be said about interpersonal commitment? The tentative answer is yes.

8

Chapter 1. Introduction

Commitment, exchange and uncertainty In social exchange, two or more actors exchange some form of material or social benefit among each other in order to arrive at an advantageous outcome. Exchange theory presumes that people exchange repeatedly with the same actors when success occurs but move to others when failure occurs. The underlying mechanism may be simple reinforcement learning (Homans, 1961; Emerson, 1972; Macy and Flache, 2002) or rational choice (Kollock, 1994; Cook and Whitmeyer, 1992). Exchange often motivates actors to unilaterally modify the balance of exchange to their own advantage without prior knowledge of the partner, in other words to cheat them in some way. This inevitably leads to uncertainty about the outcome of the exchange for both partners. When actors repeatedly exchange resources, they learn more about one another, find each other more predictable, develop mutual trust, and infer that they have similar orientations to the exchange task (Lawler, 2001). Therefore, a standard insight of exchange theory is that frequent exchange with the same partner reduces uncertainty about cheating, and thus decreases the likelihood of exchanging with strangers. More specifically, the uncertainty-reduction hypothesis was tested by Kollock (1994) who showed that commitment is more likely to form in markets where the quality of the products is unobservable at the time of the exchange. Kollock (1994) also simulated different market environments under controlled laboratory conditions. In one condition (high uncertainty), sellers could deceive their potential buyers about the quality of the product they were selling. In the other condition (low uncertainty), it was not possible to deceive buyers. A key finding of Kollock’s experiment was that commitment formation between a particular seller and a particular buyer occurs more frequently in the high-uncertainty condition than in the low-uncertainty condition. In the same vein, Yamagishi and Yamagishi (1994) argue that committed relations give a solution to the problem of uncertainty, for multiple reasons. First, committed partners accumulate information about each other over time. Second, mutually committed people enact “hostage-taking” behaviors (Raub, 2004) – ranging from the formation of mutual emotional attachments to the establishment of relation-specific assets (Helper and Levine, 1992). Hostage-taking behaviors provide deterrence against unilateral defection (Shapiro et al., 1992). Finally, conditionally cooperative strategies such as Tit-for-Tat can be used to control each other’s behavior (Axelrod, 1984). The main underlying argument for the uncertainty hypothesis is that individuals tend to avoid unpredictable or uncertain decision contexts (Tversky and Kahnemann, 1974; Kahneman and Tversky, 1979, 1996), which are created by a lack of first-hand knowledge about a potential partner’s trustworthiness (“social uncertainty”). But is the trust problem the only source of

1.1. Background

9

uncertainty in social exchange? Different exchange partners have different resources and may offer different benefits. The size and range of these potential benefits leads to a conceptually new source of uncertainty. Does this kind of “resource uncertainty” also increase commitment, independently from social uncertainty? We will examine the question of resource-inequality between exchange partners more closely in Chapter 4, and return to the concept of resource uncertainty in Chapter 6. Yamagishi and Yamagishi (1994) list several reasons for the difficulty people have in leaving a committed relationship even when it becomes a liability. One is that the mutual attraction and loyalty that have developed through the relationship keep partners together. Another is that a temporary better offer from outsiders may not be sufficient for someone who has already invested in relation-specific assets to leave the current relationship. Social and psychological assets, such as the warm memory of a pleasant past and mutual understanding, may be considered relation-specific assets that keep people in these relationships. Finally, commitment to a particular partner often reduces the level of trust in “outsiders” (see Kiyonari and Yamagishi, 1996, for experimental support), creating a vicious cycle of distrust of outsiders: those who do not trust “outsiders” tend to stay in committed relationships, and because they avoid “outsiders” they become even less trusting of “outsiders.” Yamagishi et al. (1998) further connect the tendency to form a committed relationship with the individual’s low level of general trust in others. They show in a cross-cultural setting (comparing the USA and Japan) that those who have high trust in others in general are less likely to form committed relationships. In Chapter 5, we follow up with a cross-cultural study (comparing the USA, China and the Netherlands), which shows that simple mere exposure is sufficient to increase commitment, even without an actual solution to the trust problem. Yamagishi et al. (1998) argue that general trust (or trust in people in general) provides a psychological springboard for people who have been “confined” to committed relationships to move out into the larger world of opportunities. However, as we argue in Chapter 6, general trust addresses only one of the concerns about switching to new partners. It mitigates concerns about social uncertainty, but not about resource uncertainty. On the psychological level, a different antidote is required for resource uncertainty, such as general optimism. We argue that general trust and optimism together serve as two mechanisms that help people to explore new relationships with strangers, thus decreasing commitment. Next to the uncertainty reduction mechanism, exchange theorists have recently started to recognize the importance of emotions in exchange commitments. Ed Lawler, the main proponent of the emotion argument postulates that in repetitive exchange, groups and relations become salient social objects that have a cognitive or subjective reality to actors (Lawler et al., 2000; Lawler,

10

Chapter 1. Introduction

2001). As such, these relations or groups may take on objective value and become ends in themselves (cf. Lawler and Yoon, 1996). Lawler and Yoon (1996) contend that success at exchange makes people feel good, while failure makes them feel bad. Their theory of relational cohesion states that individually felt emotions unleash a cognitive process through which the emotion is attributed in part to the relation or group that constitutes the context of the exchange. In this way, groups can become objects of intrinsic value to actors due to the positive emotions generated from exchange. Commitment in close relationships A special case of interpersonal commitment is close relationships, such as marriage and intimate partnership. Close relationships research has been the realm of psychology and social psychology, and so it is little wonder that it has identified the duality of emotional and rational explanations much earlier. Many studies in close relationships psychology refer to this duality as attraction and constraints (Adams and Jones, 1999). According to Goode (1960), the attraction (or “positive pull”) aspect is strong for example in romantic couples having a mutually satisfying and harmonious relationship. Both partners actively work together to ensure the future of the relationship. On the other hand, a constraining mechanism could similarly produce stay behavior. Even a marriage that exhibits no attraction anymore for either partner could nevertheless continue to exist due to external reasons, such as the sake of children’s well-being or to uphold appearances in a society where divorce is unacceptable (an “empty shell” marriage in Goode’s terms). Hinde (1979) creates a similar dichotomy when he distinguishes endogenous from exogenous commitment. Endogenously committed people strive to maximize the outcomes of their relational partner, even at the cost of their own interest. In contrast, exogenous commitment is based on the legal and social environment in which the relationship is embedded. In marriage commitment, Johnson (1973; 1991) introduces a third aspect by distinguishing between personal, structural (constraint) and moral-normative commitments. Personal commitment is the individual preference for staying in the marriage (because one wants to); structural commitment comes from avoiding negative consequences of the dissolution of the relationship (because one has to). Finally, moral-normative commitment arises from a sense of obligation, to do the right thing, to uphold personal behavioral consistency (because one ought to). A key psychological source of moral-normative commitment is the avoidance of cognitive dissonance – divorce may be in conflict with one’s view about marriage, or having made a public declaration through marriage vows. Another source is a sense of obligation to one another, regardless of what others think: one may want to remain true to the promise made in the wedding vow.

1.1. Background

11

From our perspective of an emotional-rational dichotomy in commitment, moral-normative commitment occupies a special position. It could be classified under rational explanations, simply as a factor that modifies instrumental properties of outcomes within a deliberative thought process. On the other hand, it could be part of an emotional explanation, inasmuch as norms are internalized and modify the emotional preferences of the individual. Within interpersonal relationship research, it is perhaps Rusbult who comes closest to establishing a rational choice framework for commitment. Building on Becker’s side-bet theory (1960) and Blau’s work on commitment (1967), Rusbult created an investment theory for interpersonal commitment (1980; 1983). According to the investment theory, the level of commitment to a relational partner is determined by multiple interconnected factors, such as relational satisfaction (the ratio of rewards and costs in the relationship), the quality and availability of alternatives or alternative states (e.g. singleness), and prior investment in the relationship. Having a highly rewarding relationship increases commitment, but so does not having satisfactory alternatives. Yet, in other works, Rusbult gives implicit indication that a rationality framework is insufficient to explain many aspects of interpersonal commitment. Johnson and Rusbult (1989) show, for example, that people unconsciously devalue potential alternatives the more committed they are to their current partner. Doing so, people distort key variables of a rational choice equation. Organizational commitment A large body of research studies commitment to organizations. In the discourse of organizational commitment, commitment refers to the attachment of a member or employee to an organization. It is sometimes used interchangeably with other concepts, such as cooperativeness and stay behavior, or even more broadly, organizational citizenship behavior (see Moorman and Blakely, 1995; Organ, 1988). Organizational commitment research is largely motivated by the insight that members who are more committed, will perform better and regard the interest of the organization as common with their own, are less stressed, and less likely to leave the organization. Meyer and Allen (1991) integrated many of the divergent conceptualizations and measurements of commitment into a coherent theoretical framework. Their model is based on the recognition that there are three main aspects (or “mindsets”) of organizational commitment: 1. Affective Commitment is the employee’s emotional attachment to the organization. It refers to identification with the goals of the organization and a desire to remain a part of the organization. The employees commit

12

Chapter 1. Introduction to the organization because they “want to”. In developing this concept, Meyer and Allen drew largely on Mowday et al.’s (1982) concept of commitment. 2. Continuance Commitment lies behind the commitment of an individual who perceives high costs of losing organizational membership (cf. the side bet theory, Becker, 1960), including economic losses (such as pension accruals) and social costs (friendship ties with co-workers) that would have to be given up. The employees commit to the organization because they “have to”. 3. Normative Commitment is created by feelings of obligation to the organization. For instance, the organization may have invested resources in training an employee who then feels an obligation to put forth an effort on the job and stay with the organization to repay the debt. It may also reflect an internalized norm, developed before the person joins the organization through family or other socialization processes, that one should be loyal to one’s organization. The employees stay with the organization because they “ought to”.

According to Meyer and Herscovitch (2001), an employee has a “commitment profile” at any point in time that reflects high or low levels of all three of these factors, and different profiles have different effects on workplace behavior such as job performance, absenteeism, and the chance to quit. These three factors are thought to jointly determine the overall level of an employee’s commitment to the organization Meyer and Allen (1991). Compare how similar this trichotomy is to Johnson’s model above (1973; 1991) under “Commitment in close relationships”.

1.2

Toward an evolutionary explanation

With the advent of sociobiology, and later the rapid growth of evolutionary psychology, many aspects of human behavior have been convincingly explained from an evolutionary perspective, relying on dynamics of genetic and cultural evolution. The major argument of evolutionary psychology (see Cosmides, 1989; Cosmides and Tooby, 1993) is that human ancestors spent a vast amount of time in a relatively stable environment of the Pleistocene, starting 1.8 million years ago and spanning until about 12,000 years ago. During the time spent in this ancestral environment, human brains and some of the most fundamental sociocultural institutions respectively, underwent a long adaptation process. During evolutionary adaptation (Darwin, 1859), the characteristics of an individual (trait) undergo random changes (mutation) that are inherited by

1.2. Toward an evolutionary explanation

13

their offspring. Through mutation new traits may appear, increase in strength or disappear2 . When the combination of traits (phenotype) of an individual increases reproductive success relative to other individuals, i.e. by increasing the chances of the individual surviving until a reproductive age, the traits of this individual become more prevalent in the population, through the relative increase in the number of offspring possessing the trait (natural selection). Traits that specifically increase mating opportunities, usually through some highly observable physical trait (e.g. the peacock’s colorful tail) may spread even faster (sexual selection). This process led to the stabilization of those cognitive abilities and social preferences which solved problems frequently encountered in our prehistoric ancestral environment. Due to rapid changes in our civilization in the last few millennia, many of these stable adaptations are no longer beneficial but nevertheless continue to influence the behavior of contemporary humans. One example is that, although an estimated 132,687 people sustain gunshot wounds that result in death or emergency treatment in the USA annually (Beaman et al., 2000), and only a handful of people are killed or injured by snakes and spiders, people learn to fear snakes and spiders roughly as easily as a pointed gun, and much more easily than an unpointed gun, rabbit or flowers (Öhman and Mineka, 2001). The explanation from evolutionary psychology is that snakes and spiders were a large threat in the ancestral environment but guns, rabbits and flowers were not. Several attempts have been made to construct a similar evolutionary explanation for commitment (in the general sense of promises and threats) that brings together the emotional and rational sides. As one of the main proponents of this line, Nesse (2001a), puts it: [There are] abundant examples of the importance of commitment in human social life. The evidence is so compelling that one cannot help but wonder why explanations for cooperation have been so narrowly dependent on methodological rationalism and individualism. I suspect the reason is the absence of a framework that can account for actions that seem irrational. In the framework of commitment, such behaviors are not only explicable, they are expected. Certain emotions seem opposed to reason because they are opposed to reason. In the short run they seem mysterious, but in the long run on average they give advantages that shape psychological traits that change the structure of human society. These psychological traits must be incorporated into our model, however difficult that may be (p. 161). 2 Note that according to the theory of cultural evolution (Boyd and Richerson, 1985) such an evolutionary process need not take place on a genetic level. They showed that culture can evolve by a very similar dynamic as genetically based traits evolve by natural selection. Culture also undergoes mutation, individuals have cultural offspring, etc.

14

Chapter 1. Introduction

The answer he proposes is to regard deep rooted emotions related to commitment (to an action) as evolutionary adaptations that serve a good purpose in general and in the long run but due to their hardwiring easily come in conflict with rational deliberation. He argues that the parts of the human brain that evolved latest in our history, the frontal lobes, closely match the abilities needed to use commitment strategies (p. 34). It appears that the frontal lobes are especially well-suited to calculating trade-offs between short-term costs of giving up options and long-term benefits that may or may not be obtained. Such calculations are inherently complex, because they involve considerations about social capital, and would be impossible without specialized mental hardware. According to Nesse, the frontal lobes are also involved in the ability to empathically identify with another person, which is essential to predicting whether the other will fulfill a commitment. The weakness of Nesse’s argument is that it attempts to cram too much under the explanatory umbrella of natural selection. In his book, he integrates works from psychology, game theory, ethology, law, medicine, religion and mythology. Doing so, his argument gets fragmented and lost in the myriad aspects of general commitment. In the end, some of the phenomena and mechanisms considered can only be linked to natural selection through smaller or larger jumps in the argument. In fact, Nesse tries to explain human cooperation and non-kin altruism arguing for a capacity for making threats and promises (commitment in the broad sense) in general, but his argument relies heavily on long-term relationships (commitment in the interpersonal sense). It could possibly strengthen his theory if the evolutionary argumentation were restricted only to the simpler and more specific idea of interpersonal commitment. Another proponent for the crucial role of emotions in commitment is Robert Frank. In his seminal book “Passions within Reason” (1988) he argues that social environments naturally produce situations where commitment could potentially play a pivotal role, yet there is little room for formal commitment devices, such as contracts or other tangible hostages. In these cases, the best solutions are emotional commitments. One of the social emotions Frank argues for, as a relatively hard-to-fake signal of commitment, is sympathy (Frank, 2001). Sympathy enables people to detect other’s emotional state and experience it to some extent. Detecting sympathy in others helps to make promises about future cooperation more credible. Another social emotion that makes commitments credible without tangible assurances is anger. In a world of purely rational self-interested people who have perfect self-control, all acts of defection where the costs of retaliation outweigh benefits would go unpunished. An angry person, however, seldom gets recognized as a rational one, leading to an increase in the credibility of his threat of punishment, and thus decreasing the expected benefits of defection

1.2. Toward an evolutionary explanation

15

in the first place (Frank, 1988).

1.2.1

Separating ultimate and proximate explanations

When attempting to construct an evolutionary explanation for any kind of behavior, it is important to separate parallel explanations on at least two different levels of causality, the proximate and the ultimate level (Mayr, 1961). Proximate explanations identify environmental stimuli that trigger mechanisms within the individual as the causes of physical expression of the behavior. For example, in answer to the question, “why do songbirds sing?”, one might argue that increased daylight in the spring leads to increased testosterone production which activates parts of the brain in male songbirds. This explanation identifies a proximate mechanism (a neurobiological one in this case) in response to a direct stimulus (increased sunshine) to explain behavior (singing). Such a proximate explanation, however, might leave one with a sense of unsatisfied curiosity. In order to answer why such a proximate mechanism came to exist in the first place, one needs to look for an ultimate explanation, on a more general level of evolutionary causation. The reason why male songbirds sing is that singing attracts females and defends territory from other males. Consequently, those males who sing have better chances of reproducing and spreading their habit of singing into the next generation of songbirds, than those males who do not sing. Such an ultimate theory has the advantage of explaining behavior, while at the same time encompassing and justifying the proximate explanation.3

1.2.2

How evolutionary theory helps to explain seemingly irrational behavior

There are at least three systematic4 ways in which evolved behavior may depart from the seemingly rational. The first two result from the fact that ultimate functions are implemented through proximate means, and the third is based on fundamental constraints on information processing. Proximate mechanisms are always imperfect in the sense that they were the first solution, discovered randomly by natural selection, which addressed a specific problem of survival and reproduction in the simplest and most costefficient way in a certain environment. As soon as there is a change in the environment, a proximate mechanism can easily lose its efficiency or even turn 3 Ultimate (also called holistic) explanations are subject to criticism by reductionists who claim that because ultimate explanations are functional, they lack a sufficient causal argument. We provide a counterargument to this criticism in Chapter 7, page 137. 4 By “systematic” we mean that behavior fails to be rational in the same way within the same context for a large number of individuals, i.e. not as isolated occurrences of some random or transient mistake in individual reasoning.

16

Chapter 1. Introduction

against the individual. Consider in the previous example the appearance of a human hunter who learns to imitate the calling of the male bird and thus easily captures female birds. In this case, females who have evolved a preference for males’ songs experience a serious decrease in their survival and reproductive chances. By definition, proximate mechanisms lead to stable behavior across different contexts. But while they create a clear adaptive advantage in one context, they could lead to maladaptive behavior in another. The first possibility for such errors is that a stimulus from the environment is falsely interpreted by the individual as a trigger for a proximate mechanism (a “false positive”, or “type I error” in statistics). The reason why the proximate mechanism could still be left in place by evolution is that the relative cost of the false alarm is smaller than the cost of not recognizing the real stimulus (a “type II error”). According to Error Management Theory (Haselton and Buss, 2000; Haselton and Nettle, 2006), humans acquired a large number of biases that increase the amount of false positives, when false negatives are extremely costly. An interesting example is that people develop a strong aversion to a certain kind of food, if its consumption was closely followed by sickness in the past (Garcia et al., 1966). This mechanism protected ancestral humans against consuming poisonous food sources (cf. Sripada and Stich, 2004). Such behavior could also be regarded as rational if information collection and processing are assumed to be costly. Therefore, the departure from standard rationality in this case is not so much the crude causal approximation between poisonous food and sickness but the fact that the aversion is manifest as a discomforting sensation in the gut, and not as the end-product of a deliberative thought process. An example with regard to interpersonal commitment is the laboratory studies carried out by Lawler and collaborators (Lawler and Yoon, 1993, 1996; Lawler et al., 2000). In these experiments, people became committed to their partners and reaffirmed their commitment with costly gifts when in fact they had never met these partners face to face and were ensured by the experimental setting that they never would. In this case, the bias for interpersonal commitment, a proximate mechanism, misfired in an inappropriate context (terminology from Sripada and Stich, 2004). The second way in which rationality could fail is when a signal is correctly recognized but the response given to it is no longer adaptive due to changes in the environment itself. A core assumption of evolutionary psychology (Cosmides, 1989; Barkow et al., 1992) is that the environment we live in today is radically different from the environment of evolutionary adaptation (ancestral environment). Therefore, some of the evolved stimulus-response mechanisms have become maladaptive. An example from the domain of interpersonal relationships is the very recent phenomenon of Internet addiction taking place among a worryingly large portion of ordinary people. The majority of these people turn to on-line chat

1.2. Toward an evolutionary explanation

17

rooms and role-playing games in search of social support, sexual fulfillment, and an opportunity to safely express forbidden aspects of their personalities. Adverse results include social withdrawal in the real world and loss of control, which are typical of other forms of addiction (Henry et al., 1997). In other words, people follow their evolved need for socialization but given the transformation of our social environment due to rapid technological development, the individual’s fitness is negatively affected. The third way in which human decision-making may depart from rationality is linked to information. In order to make rational decisions by choosing between different actions that lead to different outcomes, one needs information about these outcomes. If evolution ultimately favored rationality, it would also have favored mechanisms that help to obtain and process information accurately. There is mounting evidence that evolution sometimes works in the exact opposite direction. This is most notable in the case of evolved cognitive biases and optical illusions (see e.g. Haselton and Buss, 2000; Haselton and Nettle, 2006; Gigerenzer and Todd, 1999). Among the numerous examples, consider Evolved Navigation Theory. According to this theory, humans were selected to perceive physical characteristics of the environment (e.g. height and altitude) not as precisely as possible, but rather with a factoring in of the dangers they represent for individual fitness. Researchers in an experiment (Jackson and Cormack, 2006) asked one group of people to estimate the height of a very tall lookout point by looking at it from its bottom and another group to do the same from the top. It was found that people on the top consistently overestimated altitude, in proportion to the increased risk of falling. This shows how evolution can build safeguards into our cognitive apparatus that act against standard rational calculation. Consider now the finding of Johnson and Rusbult (1989) from earlier in this chapter, which shows that people systematically underestimate alternative partners, the more committed they are. If evolution ultimately favored the choice for a rational decision, it would have made sure that information about alternative partners is as accurate as possible at the time of making a decision. If, however, evolution aimed at stabilizing interpersonal commitments, it would have biased decisions in exactly this direction.

1.2.3

Constructing an ultimate explanation for interpersonal commitment

Although in his 1988 book Frank sets out to summarize empirical support for an evolutionary explanation for commitment in the general sense, many of his examples are more relevant for commitment in the interpersonal sense. Frank refers to marriage as a key example for a commitment dilemma. People search for the perfect mate, but settle for someone after a certain period of exploration despite knowing that there is certainly someone else out there, not

18

Chapter 1. Introduction

yet encountered, who would make a better spouse. And although a marriage contract may create a formal token of commitment, this is hardly the reason why people stop exploring further mates. A far more secure commitment is ensured by emotional bonds of affection (Frank, 2001). These emotional bonds ensure that even if someone kinder, better looking, or richer, who would originally have been preferred over the current partner, comes along now, the threat to the current commitment is diminished. But what does this have to do with evolution? There is growing acceptance among biologists of the idea that marital commitment is a key factor in enhancing the reproductive success of humans (Hrdy, 1999; Martin, 2003; Foley, 1996; Geary, 2000; Pillsworth and Haselton, 2005), indeed more so than in the case of any other primate species. In order to be able to pass through the birth canal of their mother with their large brain unharmed, human infants need to be born at an earlier developmental stage than other primate offspring (Hrdy, 1999). Consequently, they are more helpless and require substantially longer parenting (Martin, 2003). Therefore, finding a committed father who is present and cooperative during this extended period of parenting is instrumental for the reproductive fitness of humans5 (Foley, 1996; Geary, 2000; Pillsworth and Haselton, 2005). Indeed, there is a wealth of empirical findings in psychology and social psychology that gives further support for the existence of a consistently biased emotional-cognitive framework facilitating interpersonal relationships and commitment (cf. Baumeister and Leary, 1995). People in every society on earth belong to small primary groups that involve face-to-face, personal interactions (Mann, 1980). Festinger et al. (1950) found that mere proximity is enough for people to develop social bonds, and is especially suitable to compensate for differences in age or race (Nahemow and Lawton, 1975). Ostrom et al. (1993) showed that people memorize things related to close acquaintances on a person basis, whereas information related to looser contacts is stored and organized based on attribute characteristics (e.g. traits, preferences and duties). There is evidence that forgiving a misconduct of a committed partner directly enhances psychological well-being of the one who forgives (Karremans et al., 2003). It has also been shown that when people evaluate potential alternative partners, they unconsciously devalue potential alternatives the more committed they are to their current partner (Johnson and Rusbult, 1989). Kiyonari and Yamagishi (1996) give experimental support that those who stay committed to steady partners not only increasingly trust their partner, but 5 Indeed there is a possibility that next to natural selection, sexual selection also contributed to the proliferation of a commitment trait. Since human women need to find potentially committed mates to ensure the survival of their offspring, showing interpersonal commitment in social relationships in general could have served as a costly signal of males’ willingness and ability to become committed fathers.

1.2. Toward an evolutionary explanation

19

also increasingly distrust outsiders, leading to a “vicious cycle of distrust in outsiders”. On the one hand, a strategy of commitment appears to be efficient in forging beneficial relationships, yet it also loses out by letting potentially good alternatives slip away, and moreover, it gives way to exploitation within the relationship. To better understand these mechanisms and their interaction under a complex, evolutionary dynamic, we create formal (computational) models of the ancestral environment. Our models for the evolution of deeply rooted emotions underlying interpersonal commitment rely on a series of previous works by Henk de Vos and his collaborators (de Vos and Zeggelink, 1997; de Vos et al., 2001; Zeggelink et al., 2000). These researchers designed an agent-based computational model based on the following minimalistic assumptions about conditions of the ancestral environment: 1. People lived together in relatively small groups. 2. The environment was harsher, its impact less buffered, and resources more scarce than today. 3. In lack of many modern social institutions, help from fellow individuals was more important for survival than today. 4. The environment and subsistence technologies were more stable over an extended period of time than in modern civilizations, which made it possible for evolutionary pressures to hardwire preferences. De Vos and colleagues created a help exchange model, in which members of a relatively small group are dependent on the help of others to survive an event of distress from time to time. They compared two major contestants in their simulations of the evolution of exchange strategies, a strategy based on calculative reciprocal cooperation and a strategy based on commitment. De Vos and collaborators found that when each of the strategies competes against opportunistic players – i.e. actors who are unwilling to help but accept help from others – commitment is more viable than calculative reciprocity. De Vos et al. tentatively concluded from their computational experiments that under conditions of the human ancestral environment, selection pressures might have shaped a tendency towards commitment and largely unconditional cooperation. This tendency may still be present in contemporary humans, even though the pressures that formed it are weakened or no longer in place. However, their studies were strictly limited by the small number of strategy variations they examined. This presents a problem because overly cooperative agents following a commitment strategy could easily fall prey to smart cheaters, a possibility that their model could not account for. Moreover, as Binmore (1998) argued forcefully, the outcome of computer tournaments and simulations of evolutionary dynamics strongly depends on the set of strategies that are initially present in a population.

20

Chapter 1. Introduction

To address whether and to what extent these two potential problems reduce the viability of commitment, we propose in Part I (Chapter 2) a method to considerably and systematically enlarge the set of behaviors examined in the original analysis of the help exchange model. The core idea is to represent behaviors as determined by a set of individual preferences, or traits with respect to possible exchange outcomes. Agents in our model are boundedly and subjectively rational in the sense that they make decisions to cooperate, defect and change partners with the goal of maximizing subjective utility (or satisfaction) given their preferences. However, maximizing subjective utility based on individual preferences in our model does not necessarily lead agents to optimal exchange outcomes. We assume that individual preferences or strategies are subject to evolutionary pressure that selects for successful strategies based on the objective fitness consequences of the behavior resulting from the strategy. This approach is similar to the “indirect evolutionary approach” proposed by Güth and Kliemt (1998).

1.2.4

Proximate mechanisms for interpersonal commitment

Is there support for an ultimate explanation for commitment through a corresponding proximate mechanism? More precisely, do contemporary humans have a stable, hardwired tendency to become committed to their previous interaction partners in an emotional way when it is not in their instrumental self-interest? In Chapter 5 we empirically test the existence of such a potentially hardwired tendency for commitment through a series of cross-cultural laboratory experiments. When arguing for the evolutionary origins of any aspect of sociality, it is usually better to rely on cross-cultural data, in order to rule out cultural explanations. Since the environment of evolutionary adaptation mostly predates the break-up of modern cultures, adaptations associated with the ancestral environment should be present in all cultures. This is not to say, of course, that a cross-culturally stable phenomenon necessitates an evolutionary explanation, or that a lack of cross-cultural evidence rules one out. Culture intricately interplays with how people decide and behave, which itself has implications for biological evolution (Boyd and Richerson, 1985). The idea behind a proximate explanation for commitment is that through repeated positive interactions, people’s view of a committed relationship becomes systematically biased in comparison with a strictly instrumental perspective. When the relationship later takes a negative turn, this positively biased perspective for commitment makes stay behavior and cooperation more likely than otherwise expected. According to the mere exposure effect (originally described by Zajonc, 1968), when being repeatedly subject to a nonrepulsive stimulus, one develops a positive affect toward the stimulus. For example, the more we listen to the same piece of music, the more we appre-

1.3. Methodology

21

ciate it. We argue that such a mere exposure effect exists between long-term interaction partners. There is evidence, for example, that the more we see the same face, the more attractive we find it (Rhodes et al., 2001). What is even more interesting, is the finding that people also trust others more if they have been exposed to them more times, even in the absence of any interaction that could support actual inferences about trustworthiness (Moreland and Beach, 1992).

1.3

Methodology

Studies reported in this volume rely largely on two key methodological approaches, agent-based computational modeling and laboratory experiments. Below is a brief description of both, and an explanation of their usefulness for answering our research questions. Part I is aimed at testing an evolutionary theory that posits the stabilization of a commitment trait under selection pressures of the human ancestral environment. Its goal is to compare the strengths and weaknesses of commitment to other social preferences, such as calculative reciprocity (fairness). Drawing on earlier work in this domain (de Vos et al., 2001) and weighing the complexity of the modeling task, we decided to apply a type of social simulation to our problem, agent-based computational modeling (ABCM). Whereas social scientists usually model social processes as interactions among variables, ABCM studies interactions among adaptive agents who influence one another in response to the influence they receive (Macy and Willer, 2002). In ABCM, all modeling information about the properties of individual agents and their behavioral rules are transformed into a formal language (e.g. a computer program). Subsequently, the dynamics of the model, as well as conclusions on the macro-level can be deduced through step-by-step computation from given starting conditions (Flache and Macy, 2005). The advantages of ABCM are especially apparent when modeling dynamic phenomena in groups that are highly complex, non-linear, path-dependent, and selforganizing. The obvious advantage is that the explanation draws on local interactions among agents and not on predefined global characteristics of the group (Macy and Willer, 2002). A possible alternative methodology in Part I would be game theoretical (e.g. equilibrium) analysis. The benefit of such analysis is that it is able to provide more universal hypothesis tests that benefit from the strength of a mathematical proof. Its drawback is that given the complexity of our model, coupled with the evolutionary dynamic, this methodological approach seems unfeasible. This is also illustrated by the relative complexity of a brief analysis of a strongly simplified version of our model in Appendix A. In Part II, we aim to empirically test the evolutionary argument through

22

Chapter 1. Introduction

identifying peculiarities in decision-making in exchange relationships among contemporary humans. The majority of our hypotheses predict links between specific, well-defined conditions and exact, quantifiable measures of commitment. Therefore, strict control over conditions and measurement is indispensable. Accordingly, we decided to test our hypotheses in laboratory experiments with human subjects, using anonymous, computer-based settings. A possible alternative here would be to rely on representative crossnational surveys, or data collected among contemporary small-scale (e.g. hunter-gatherer) societies. Such secondary data analysis, however, would seriously restrict the scope of hypotheses that we could test.

1.4

Outline of chapters

The rest of this dissertation is organized as follows. Part I lays the theoretical foundation for studying interpersonal commitment using formal computation models. These three chapters investigate whether a commitment trait could have been adaptive under the conditions of the ancestral environment, with each chapter gradually putting commitment under a stricter test. More specifically, Chapter 2 examines whether improving on a known weakness of fair reciprocity (the main contestant of commitment in earlier work of de Vos et al., 2001) eliminates the competitive advantage of commitment. We create an agent-based model that is capable of incorporating previous models, and at the same time offers more flexibility and robustness to study the relative viability of commitment strategies. This chapter also tries to answer whether the exchange network structures, which are formed spontaneously in the simulated populations, help to explain the relative differences in viability. Chapter 3 puts the viability of commitment strategies under a stricter test by extending the previous ecological model through accounting for evolutionary dynamics of selection and mutation. This also means that the strategies examined here are no longer a priori invented and specified by the modeler, but emerge spontaneously through random walks in the strategy space. In comparison with previous work, this can possibly lead to the emergence of more sophisticated opponents that may take advantage of the weaknesses of commitment. A defining characteristic of interpersonal commitment is that one forgoes interaction with potentially better alternatives in favor of a long-term partner. Chapter 4 puts the emphasis on inequalities between potential interaction partners, and asks whether a preference for high-resource (or highly capable) others is more important than a preference for old partners (commitment). It also examines whether becoming committed to average or low-value partners undermines the efficiency of commitment. In order to do so we extend our previous model by introducing an inheritable trait for high-resource others,

1.4. Outline of chapters

23

and non-inheritable inequality in individual capabilities/resources. Part II turns to the experimental investigation of whether an ancestrally evolved commitment trait influences behavior in contemporary humans. Chapters 5 and 6 examine commitment in various exchange situations using laboratory experiments with human subjects. These two chapters take up the two major lines in explaining commitment in exchange: positive emotions and uncertainty reduction. In Chapter 5, we study whether people have a tendency to escalate commitment to previous interaction partners, when it is not in their self-interest. The purpose of this study is to test whether people have a cross-culturally stable emotional preference for previous partners that acts as a decision-making bias even in anonymous, economic setting. At the same time, in Chapter 5 we also find that uncertainty can decrease commitment, which leaves us with a new puzzle. To resolve the puzzle about the effect of uncertainty on commitment, in Chapter 6 we propose to refine the explanatory framework that has been used in the literature of commitment in exchange. More precisely, we identify an important assumption that previous works have left implicit, about the cooperative intentions of people, that qualifies the effect of uncertainty on commitment. In addition, Chapter 6 is also an empirical counterpart of the theoretical Chapter 4, which argued that a preference for previous partners (commitment) should be stronger among contemporary humans than a preference for highresource partners. Accordingly, we argue that partner selection situations can create not only social uncertainty (about trustworthiness) but also resource uncertainty, and that each type of uncertainty has its independent effect on commitment. At the same time, we contend that individual characteristics, such as general trust and optimism, may also influence the level of commitment. Finally, Chapter 7 contains a summary of results, with a general discussion of the findings, and an evaluation of the strengths and weaknesses of the work accomplished. We locate our work within past research and point to possible avenues of future research. Since Chapters 2, 3 and 4 have been accepted for publication in international peer-reviewed journals and a book, and the material that makes up Chapters 5 and 6 is currently under review, these chapters are kept in their original article format. As a result, some overlap may be detected between different chapters, especially in their empirical motivation and model descriptions. On the positive side, this leaves each chapter stand-alone and selfexplanatory.

Part I

An ultimate explanation

25

Chapter 2

The Competitive Advantage of Commitment1

Abstract

A prominent explanation of cooperation in repeated exchange is reciprocity (e.g. Axelrod, 1984). However, empirical studies indicate that exchange partners are often much less intent on keeping the books balanced than Axelrod suggested. In particular, there is evidence for commitment behavior, indicating that people tend to build long-term cooperative relationships characterized by largely unconditional cooperation, and are inclined to hold on to them even when this appears to contradict self-interest. Using an agent-based computational model, we examine whether in a competitive environment commitment can be a more successful strategy than reciprocity. We move beyond previous computational models by proposing a method that allows systematic exploration of an infinite space of possible exchange strategies. We use this method to carry out two sets of simulation experiments designed to assess the viability of commitment against a large set of potential competitors. In the first experiment, we find that although unconditional cooperation makes strategies vulnerable to exploitation, a strategy of commitment benefits more from being more unconditionally cooperative. The second experiment shows that tolerance improves the performance of reciprocity 1 This chapter is based on Back, I. and Flache, A. (2006). The Viability of Cooperation Based on Interpersonal Commitment. Journal of Artificial Societies and Social Simulation 9(1), http://jasss.soc.surrey.ac.uk/9/1/12.html.

27

28

Chapter 2. The Competitive Advantage of Commitment

strategies but does not make them more successful than commitment. To explicate the underlying mechanism, we also study the spontaneous formation of exchange network structures in the simulated populations. It turns out that commitment strategies benefit from efficient networking: they spontaneously create a structure of exchange relations that ensures efficient division of labor. The problem with stricter reciprocity strategies is that they tend to spread interaction requests randomly across the population, to keep relations in balance. During times of great scarcity of exchange partners this structure is inefficient because it generates overlapping personal networks so that often too many people try to interact with the same partner at the same time.

2.1

Introduction

The most prominent explanation of endogenous cooperation in durable relationships is reciprocity under a sufficiently long “shadow of the future” (Axelrod, 1984; Friedman, 1971). In this view, actors engage in costly cooperation because they expect future reciprocation of their investment or because they feel threatened by future sanctions for non-cooperation (Fehr and Schmidt, 1999; Fehr and Gächter, 2002; Falk et al., 2001). Roughly, these analyses show that even in a competitive environment with changing exchange partners, strategies that reciprocate cooperation with cooperation and defection with defection, such as the celebrated “Tit-for-Tat”, are far more successful than strategies that aim to exploit their opponents. Evolutionary game theory has demonstrated that if exchange relations persist long enough, cheaters are outperformed by reciprocators. This is because reciprocators benefit from ongoing mutually cooperative exchanges, while cheaters gain at best a short-term advantage at the outset of the exchange. This, however, cannot offset the longterm losses caused by the early disruption of the exchange relationship. It has been suggested that this reciprocity explanation of cooperation applies to a number of domains ranging from business ties between organizations to interpersonal relationships. However, recent empirical studies of cooperative behavior, in particular in interpersonal relationships, indicate that often reciprocity may be much less strict and actors much less intent on keeping the books balanced than the original reciprocity argument suggests. A short excerpt from Nesse (2001b) offers good examples: Perhaps the strongest evidence that friendships are based on commitment and not reciprocity is the revulsion people feel on dis-

2.1. Introduction

29

covering that an apparent friend is calculating the benefits of acting in one way or another. People intuitively recognize that such calculators are not friends at all, but exchangers of favors at best, and devious exploiters at worst. Abundant evidence confirms this observation. Mills has shown that when friends engage in prompt reciprocation, this does not strengthen but rather weakens the relationship (Mills and Clark, 1982). Similarly, favors between friends do not create obligations for reciprocation because friends are expected to help each other for emotional, not instrumental reasons (Mills and Clark, 1994). Other researchers have found that people comply more with a request from a friend than from a stranger, but doing a favor prior to the request increases cooperation more in a stranger than a friend (Boster et al., 1995). Moreover, there is solid empirical evidence indicating that people have a tendency to build long-term cooperative relationships based on largely unconditional cooperation, and are inclined to hold on to them even in situations where this does not appear to be in line with their narrow self-interest (see e.g. Wieselquist et al., 1999). Experiments with exchange situations (Kollock, 1994; Lawler and Yoon, 1993, 1996) point to ongoing exchanges with the same partner even if more valuable (or less costly) alternatives are available. This commitment also implies forgiveness and gift-giving without any explicit demand for reciprocation (Lawler, 2001; Lawler and Yoon, 1993). One example is that people help friends and acquaintances in trouble, apparently without calculating present costs and future benefits. Another, extreme example is the battered woman who stays with her husband (Rusbult and Martz, 1995; Rusbult et al., 1998). Since the seminal work of Axelrod (1984), a range of studies has used evolutionary game theory to refine the strategy of strict reciprocity and adapt it to empirical criticism. One line of work focused on the advantages of “relaxed accounting” in noisy environments (e.g. Kollock, 1993; Nowak and Sigmund, 1993; Wu and Axelrod, 1995). Broadly, these experiments confirmed the hypothesis that uncertainty favors “tolerant” or “relaxed” conditionally cooperative strategies that do not always retaliate after defection of an opponent. Kollock (1993), for example, found that in noisy environments (with mistakes and miscues), strict reciprocity is prone to needless recrimination that can be avoided by looser accounting systems. However, these studies cannot address the empirical phenomenon of commitment to long-term exchange partners, simply because they apply a repeated game framework in which there is no possibility to exit from an ongoing exchange in order to seek a new partner. A number of authors have explored variations of Tit-for-Tat that combine looser accounting under uncertainty with selective partner choice. Computational analyses of exit effects (Schüssler, 1989; Vanberg and Congleton, 1992; Schüssler and Sandten, 2000) put the role of the shadow of the future for

30

Chapter 2. The Competitive Advantage of Commitment

emergent cooperation into perspective. The route to emergent cooperation that these studies uncover is commitment of cooperators to cooperators, with the consequence of exclusion of defectors from relationships with cooperative partners. This is based on the principle “be cooperative but abandon anyone who defects.” When enough members of a population adopt this strategy, cooperative players stay in stable relationships, leaving defectors with no one but other defectors to interact with. As a consequence, defectors perform poorly and conditional cooperation thrives even under anonymity conditions where unfriendly players can hide in a “sea of anonymous others” (Axelrod, 1984, 100) after they “hit and run”. Considering more complex agent architectures, Schüssler and Sandten (2000) show that strategies that are to some degree exploiters may survive under evolutionary pressure but even then the most successful strategies will have the property of staying with a cooperative partner who turns out to be difficult to exploit. Other computational studies that include partner selection and arrive at similar conclusions are, for example, Yamagishi et al. (1994), or Hegselmann (1996) (cf. Flache and Hegselmann, 1999b). While previous work using evolutionary game theory could demonstrate the viability of relaxed accounting and commitment under certain conditions, it is doubtful whether this suffices to explain how humans may have acquired the deeply rooted emotions and behaviors related to interpersonal commitment that have been empirically observed. This is why de Vos and collaborators (de Vos et al., 2001; Zeggelink et al., 2000; de Vos and Zeggelink, 1997) extended theoretical models with assumptions from evolutionary psychology (Cosmides, 1989; Cosmides and Tooby, 1993). According to evolutionary psychologists, the way our mind functions today is the result of an extremely long evolutionary process during which our ancestors were subject to a relatively stable (social) environment. Individual preferences for various outcomes in typical social dilemmas stabilized in this ancestral environment and still influence the way we decide and behave in similar dilemma situations today. To model a stylized ancestral environment, de Vos and collaborators designed a help exchange game in which members of a relatively small group need the help of others to survive a situation of distress from time to time. More precisely, in their model agents come into distress at random points in time and then ask other members of the group for help. They compared two major contestants in their simulations of the evolution of exchange strategies, a strategy they called “keeping books balanced” (KBB) and a strategy called “commitment”. KBB corresponds to a strategy of strict reciprocity that is willing to help another actor but only as long as the favor is returned by the recipient as soon as possible. Otherwise, KBB will exit the relationship and seek new exchange partners. By contrast, commitment needs only a few successful initial help exchanges with a specific partner to become unconditionally co-

2.1. Introduction

31

operative to its partner further on. Broadly, de Vos and collaborators found that when both strategies need to compete against “cheaters” – i.e. actors who are unwilling to help but accept help from others – commitment is more viable than KBB under a large range of conditions. They conclude that in an environment where unpredictable hazards occur, KBB may be too quick to abandon exchange partners who get into trouble a second time before first reciprocating. As a consequence, a KBB player may often end up with no one willing to help it. A commitment player avoids this problem, because once committed to a cooperative partner it will not leave the partner in times of need and thus will benefit from future help from this partner when it experiences distress. De Vos et al. tentatively conclude from their computational experiments that under conditions of the human ancestral environment, selection pressures may have shaped a tendency towards commitment and largely unconditional cooperation that contemporary humans may still have, even when the pressures that formed it are no longer present. However, it is clearly an important limitation of these studies that only three possible strategies, KBB, commitment and cheating, have been taken into account and confronted with each other in a tournament approach. As Binmore (1998) has argued forcefully, the outcome of computer tournaments and simulations of evolutionary dynamics strongly depends on the set of strategies that are initially present in a population. The small set of strategies used by de Vos and collaborators may hide two potentially severe problems for the viability of the strategy of commitment. The first problem is the unconditionality of the strategy’s willingness to cooperate once it has been committed to a partner. This property obviously makes commitment highly vulnerable to exploitation by strategies that try to take advantage of its willingness to help. The second problem is that commitment may lose out in competition against more tolerant modifications of strict reciprocity. As the work by Kollock (1993) and others suggests, such modifications may avoid the major weakness of strong reciprocity to disrupt potentially cooperative exchanges too readily when problems occur. At the same time, such strategies also are less exploitable than commitment, because they eventually avoid being exploited by a partner who steadfastly fails to reciprocate help. To address whether and to what extent these two potential problems reduce the viability of commitment, we propose in this paper a method to considerably and systematically enlarge the strategy set used in the original analysis of the help exchange dilemma. The core idea is to represent strategies as a set of individual preference parameters, or traits with respect to possible exchange outcomes in a relationship. Agents in our model are boundedly and subjectively rational in the sense that they take decisions to cooperate, defect and change partners with the goal of maximizing utility from their preferences. However, maximizing subjective utility based on individual preference values in our model does not necessarily lead agents to optimal exchange

32

Chapter 2. The Competitive Advantage of Commitment

outcomes. We assume that individual preferences or strategies are subject to evolutionary pressure that selects for successful strategies based on the objective fitness consequences of the behavior resulting from the strategy. This approach is similar to the “indirect evolutionary approach” proposed by Güth and Kliemt (1998). Our approach allows systematic mapping of a range of individual variation in decision-making rules, e.g. variation in the extent of commitment or strictness of reciprocity. With this, we can carry out a stronger test of the viability of commitment than de Vos et al. (2001). We use our model to carry out two sets of simulation experiments designed to assess the viability of commitment in a larger set of potential competitors. For this, we take the original design of de Vos et al. as a starting point but systematically relax the assumption of unconditionality of cooperation in the first set of experiments. In the second set of experiments, we introduce and compare various degrees of relaxed accounting to reciprocity (“fairness”) strategies. In Section 2.2, we motivate and describe the model and our extensions. In Section 2.3, the computational experiments are reported. Section 2.4 contains conclusions and a discussion of our findings.

2.2

Model

Our model is based on a delayed exchange dilemma game, which is very similar to the one originally proposed by de Vos et al. (2001). The game is played by n agents in successive rounds. In the first round all agents are endowed with fi fitness points. In the beginning of each round, Nature selects a number of agents with a given individually independent probability Pd who experience distress and thus become in need of help from other agents in order to preserve their fitness level. These agents who are struck by Nature are the initiators of interactions. They ask others for help which is either provided or not. Providing help costs fh fitness points. Moreover, assuming that help giving is a time-consuming activity, each agent may only provide help once during one round; and only agents who are not distressed themselves may provide help. If a help request is turned down, the distressed agent may ask another agent for help but may not ask more than m agents altogether in the same round. If an agent does not manage to get help before the end of the round, it experiences fd loss in fitness. If the fitness level of an agent falls below a critical threshold fc , the agent “dies”, i.e. it is eliminated from the agent society.

2.2. Model

2.2.1

33

Modeling strategies

Agents in our delayed exchange dilemma face two different types of decision situations from time to time. If they are hit by distress, they have to select an interaction partner whom they believe most likely to be willing and able to help them. On the other hand, when they themselves are asked to provide help they have to decide whether to provide it and in case of multiple requests, whom to provide it to. Thus the mental model, or strategy, of an agent is represented as a combination of two sub-strategies: one for asking help and one for giving help. In previous studies by de Vos and others, behavioral strategies of agents were defined in natural language in terms of a collection of condition-action rules (e.g. for agent ai : if agent aj helped me before when I asked, then help him now) and then translated into a programming language. Even for simpler strategies several such decision rules had to be formulated, and this inherent arbitrariness limited the generalizability of the model. Our most important addition to these models is that we integrate them into a utility-based framework and provide in this way an efficient method to cover a large range of different strategies. In our model, when an agent has to make a decision, it calculates utilities based on some or all of the information available to it without the ability to objectively assess the consequences of the decision on its overall fitness2 . Moreover, we assume that actors are boundedly rational in the sense of being myopic, they evaluate the utility of an action only in terms of consequences in the very near future, i.e. the state of the world that obtains right after they have taken the action. This excludes the strategic anticipation of future behavior of other agents. Since different agents calculate utility differently, there is variation in behavior. Some of the behaviors lead to better fitness consequences than others. In turn, more successful agents have better chances of staying in the game and propagating their way of utility calculus to other agents, while unsuccessful ones disappear. Recent advances in psychological research into interpersonal relationships point to the influence of subjective well-being experienced when making certain relationship-specific decisions (Karremans et al., 2003). Unlike many applications of evolutionary game theory, we define utility calculus such that agents derive an emotional utility from features of a relationship, in addition to materialistic costs and benefits of help exchanges. This emotional utility can be interpreted as feelings and emotions, such as togetherness, belonging, sense of safety, identity, pride, etc., and the lack of it as loneliness, insecurity, shame, etc. We concentrate our modeling efforts on describing and analyzing this additional utility as a function of the history of help exchanges in a relationship. One of our main goals is to determine whether utility calculus based 2 In

certain cases these objective consequences may actually be impossible to foresee for the agents or even for the modeler.

34

Chapter 2. The Competitive Advantage of Commitment

on some form of commitment can lead to beneficial fitness consequences. In our delayed exchange game, agents have a very focused set of information available about their physical and social environment. They are aware of the fact that they got into distress, they follow the rules of the game (e.g. ask for help when in distress), and they remember previous encounters with other agents. This means that they know who and how often helped or refused them and who was helped or refused by them in previous rounds. The implicit assumption we make is that information about interactions between third-party agents is either not (reliably) available to the focal agent or is simply not taken into account in decision-making. We restrict the information available to agents from their earlier interactions to the following situation-specific decision parameters of an agent ai for each interaction partner aj (i 6= j): Definition 1 (Situation-specific decision parameters). EHij ERij AHij ARij

= number of times i helped j (ego helped), = number of times i refused j (ego refused), = number of times j helped i (alter helped), = number of times j refused i (alter refused)

As we mentioned above, agents face two different decisions situations. Accordingly, we define two independently calculated subjective utilities that agents use in these two decisions. The utility of donating that agent ai gains from helping agent aj is defined as a function of the situation-specific parameters: D D D D D + ehD Uij = Um i · EHij + eri · ERij + ahi · AHij + ari · ARij , D D D D where Um expresses materialistic costs of the interaction; ehD i , eri , ahi , ari are agent-specific parameters (or traits) for donation of agent ai that determine the weight of the situation-specific parameters in the total utility. In the actual implementation, every time an agent has to make a decision, there is also a probability Pe that the agent will make a completely random decision. This random error models noise in communication, misperception of the situation or simply miscalculation of the utility by the agent. Taking this random error into account increases the robustness of our results to noise in general3 . For simplicity, we define the utility as a linear combination of situationspecific parameters weighted by agent-specific parameters. The utility of seeking is defined in the same way, the only difference is that agents may put different weights on the situation-specific decision parameters than in the utility of donation: 3 See

more about the problems of involuntary defection in Fishman (2003) and agents getting stuck in mutual defection in noisy environments in Monterosso et al. (2002)

2.2. Model

35

S S Uij = Um + ehSi · EHij + eriS · ERij + ahSi · AHij + ariS · ARij ,

Before agents make a decision, be it help seeking or help giving, they calculate the corresponding one of these two utilities for each possible help donor or help seeker. In case of help giving, they choose a partner with the highest utility, if that utility is above an agent-specific threshold Uit . If the utility of all possible decisions falls below the threshold utility, no help is given to anyone. Otherwise, if there is more than one other agent with highest utility, the agent chooses randomly. As an addition to this rule, if an agent ai is asked to donate by another agent aj with whom ai has had no prior interaction (therefore all situationspecific parameters are 0), ai assumes that AHij = 1. In other words, agents behave as if a successful interaction has already taken place between them. Suspicious (non-nice) strategies can be defined by choosing the utility threshold (U t ) parameter so that without any prior interaction the utility of seeking or donation is lower than the threshold utility. In the case of help seeking, agents also choose a partner with the highest utility but there is no threshold, i.e. agents in distress will always ask someone for help. Using these rules, a strategy S in our model is described by the way utility is calculated. In other words, a strategy can be fully described by the two times four agent-specific parameters and the utility threshold. 4 By specifying ranges for agent-specific parameters we can easily define classes of strategies which correspond to basic personality types. For example, we classify a strategy S as belonging to the group of Commitment-type strategies, if the fact that previously help was received from a certain partner increases the utility an agent derives from donating help to or seeking help from that partner: Definition 2 (Commitment). ahD , ahS > 0 and er, ar = 0 This means that an agent ai of the Commitment-type derives more utility from choosing an agent aj as an interaction partner, the more times aj has helped ai in the past. This is true for choosing from both a group of help seekers and from possible help givers. This also means that the cooperativeness of a Commitment type is unaffected by the fact whether their help was previously refused by an interaction partner. If the utility threshold U t is zero or negative and ahD > U t , the strategy starts by cooperating, if able to.5 Otherwise, it behaves as “Suspicious Commitment”, i.e. it starts with defecting but after some cooperative moves of alter, it becomes cooperative. 4 Since we are interested in the perception of the strength of a relationship between agents rather D and U S are constant in all interactions. than the perception of objective costs, we assume that Um m 5 Note, that for the sake of simplicity in explaining the behavior of strategy classes, we will assume that the probability of a decision making error Pe = 0 throughout this section.

36

Chapter 2. The Competitive Advantage of Commitment

In the remainder of this section, we show how a range of further strategies can be defined with our method. For simplicity, we assume a utility threshold of zero, if not mentioned otherwise. The strategy type of Defection can be modeled with the assumption that it derives zero or negative utility from donation under all conditions: Definition 3 (Defection). ehD , erD , ahD , arD 5 0 and min(ehD , erD , ahD , arD ) < 0 If the utility threshold U t is positive and ahD < U t the strategy always starts by defecting. Otherwise, it only starts defecting after some initial rounds of cooperation. In general, we say that a strategy is a cooperator if at least one of its donation parameters is positive. In all other cases the strategy is a variant of the Defection type. Such a subset of Defection is AllD, which never helps others but when it is in need it randomly chooses others to ask for help: Definition 4 (AllD). Donation: ehD , erD , ahD , arD < 0 Seeking: ehS = erS = ahS = arS = 0 A much discussed strategy (type), especially in the experimental economics literature (see e.g. Fehr and Schmidt, 1999; Fehr et al., 2002) is Fairness.6 This is based on the observation that people may be willing to invest in cooperation initially but will require reciprocation of these investments before they are willing to cooperate further. On the other hand people following the fairness principle are also sensitive to becoming indebted, therefore they will be inclined to reciprocate if they are in debt. In other words, their most important aim is to have balanced relationships. Again, translating this strategy class into our framework is straightforward. Definition 5 (Fairness). Donation: ehD < 0, erD > 0, ahD > 0, arD < 0 Seeking: ehS > 0, erS < 0, ahS < 0, arS > 0 Agents belonging to the Fairness class deduce more negative utility from helping if they helped their partner in the past or if the partner refused them before, and will deduce more positive utility from helping if the partner helped them or if they refused to help the partner earlier. The twist here is that it is actually the ones that are most likely to be selected for giving help to that are different from those that are most likely to be selected for asking. Note moreover, that in case of a “Truly Fair” strategy, we would make the additional assumption about absolute values of traits such that |eh| = |ah| and |er| = |ar|. 6 Social

psychologists also refer to this type of behavior as equity (cf. Smaniotto, 2004).

2.2. Model

37

Suppose, for example, that an agent ai receives two help requests at the same time, one from aj , whom ai has helped twice before but from whom ai received help already three times. The other help request comes from a partner ak who helped ai three times and received help three times. A truly fair-minded person should in this situation help aj and not ak , and this is D exactly what follows from our implementation, because in this case Uij = D D D Uik − eh and eh < 0. Without making the assumption about absolute values, however, we are able to examine a larger class of “Fairness-type” strategies, such as “Tolerant Fairness” which increases credits (ah, er) more than it increases debts (ar, eh). Note that U t must be negative or zero for Objective Fairness, otherwise it requires more cooperation from its partner than it is willing to perform itself. Another way of relaxing the strictness of Objective Fairness is to decrease U t , which allows an asymmetry in favor of alter, in the amount of required reciprocation. For analyzing the individual rationality of cooperation we also define a trigger strategy, Grim Trigger. This strategy is the strictest form of cooperation, in that it permanently retaliates after its partner or itself defected and never cooperates again. Definition 6 (Grim Trigger). Donation: ehD = ahD = 0, erD < 0, arD < 0 Obviously, our approach allows generation of a much larger range of strategies than we discussed above. For our present analysis, it suffices to use these strategy templates but we will explore a larger variety of possible behavioral rules in future work.

2.2.2

Evolutionary dynamic

The heart of our model is an evolutionary dynamic that ensures the selection of objectively successful strategies (preferences). The dynamic we apply in our simulation is based on the replicator dynamics (Taylor and Jonker, 1978). Broadly, the replicator dynamics dictate that after a generation of genotypes (strategies) replicates itself, each different genotype will be represented in the next generation according to its relative success compared to other genotypes in the current generation. This way, unfeasible or self-harming preferences gradually become less widespread in the population, and give way to more “rational” preferences (see also Güth and Kliemt, 1998). To ensure that the size of the group remains constant throughout a simulation run, we apply the replicator dynamics in the following way. Whenever an agent dies, we create a new agent whose probability of belonging to a strategy S is equal to the proportion of collective fitness that is held by the group of agents belonging to S at the time of the new agent’s birth.

38

Chapter 2. The Competitive Advantage of Commitment

The evolutionary dynamic of our model is a strong simplification of the actual genetic reproduction that could have taken place in human evolutionary history. One argument for this simplification is to avoid the unnecessary overparameterization of our model. The central assumption we make is that better exchange outcomes of a strategy type translate into better chances for the propagation of that strategy. To capture this, there is no need to include individual level variables such as average and maximum number of children, age at giving birth etc., which are actually irrelevant for answering our research questions. Thus the great advantage of the replicator dynamics for our purposes is that it keeps the model of reproduction on the macro level. This also means that we only model the evolutionary selection of strategies but not mutation (see more under Discussion and Conclusions). With our explicit model of evolution we improve upon previous work of de Vos et al. (2001) in a number of ways. In their study they did not explicitly model a replication dynamic but instead linked independent tournaments to each other in order to map evolutionary trajectories. More precisely, the authors assumed that in a sequence of evolution the final average distribution of strategies at the end of one generation taken across a series of replications of that generation would also be the initial distribution in all replications in the next generation. This reduces repeatedly the distribution of individual populations to its average trajectory, which may entail a biased picture of the eventual distribution that arises. For example, unlike de Vos et al. (2001), we consider in our analysis also those simulation runs in which the entire population becomes extinct before a generation ends. These runs were originally disregarded by De Vos et al. This may have biased their results towards an overestimation of the survival chances of Commitment because only replications in which Commitment survived could have reached the end of a generation. Moreover, unlike previous work, our model does not suffer from the specification of a “cut-off” parameter, i.e. there is no fixed number of rounds after which we stop our simulations.7 In this way, we can be sure that the evolutionary dynamic reaches an equilibrium state where the population is homogeneous, and avoid biasing our results towards strategies that may be only initially successful.

2.3

Results

De Vos et al. (2001) examined two cooperative strategies, Commitment and Keeping Books Balanced (KBB), both playing against defectors. They showed that Commitment, which is largely unconditionally cooperative to those previous interaction partners who gave help at least once, had better evolution7 To

be precise, a simulation run ends when agents of a strategy have completely pushed out their opponents from population.

2.3. Results

39

ary success than the strictly reciprocal KBB under a large range of conditions. They tentatively interpret this result as evidence for the advantages of being unconditionally cooperative in an environment with scarce and uncertain opportunities to receive help. We argue that another conclusion may also be possible. It is plausible that being so unconditionally cooperative still makes Commitment more exploitable in comparison with conditional cooperators. The relative success of Commitment in comparison with KBB may rather be a result of KBB’s disadvantageous feature to disrupt relationships too readily when some mishap occurs. In order to test this possibility, we first conducted a simulation experiment to assess to what extent it makes a difference in Commitment’s success against defectors, when various degrees of unconditionality are compared. We did this by comparing four different types of Commitment each playing against Defection. We then turned to the possibility that a more tolerant, fairness-based strategy may be more successful against defectors than Commitment or strict fairness (KBB), in an uncertain environment. To test this possibility, we compared Commitment with more tolerant versions of Fairness, and we also compared fairness strategies that vary in their degree of tolerance to each other. Finally, while De Vos et al. measured and compared the individual success of cooperative strategies playing against defectors, they did not consider the possibility of an actual evolutionary invasion against Commitment by a conditional cooperator that is less vulnerable to exploitation by smart cheaters. We also provide results of this type below.

2.3.1

Simulation setup

Our goal with the simulation experiments was to compare different cooperative strategies with each other in terms of viability when there are initially some defectors in society. The most important indicator of a cooperative strategy’s viability was its success in resisting this invasion of defectors under evolutionary pressures of selection and reproduction. More precisely, within each simulation run, we started out with a group of multiple strategies. We allowed this mixed group to play the game for an extended period under an evolutionary dynamic, until only one strategy was present in the population. Within one experiment, we independently repeated such simulation runs from their initial state n times8 , until standard errors of measured variables became sufficiently small in order to be meaningfully interpreted. At first, we kept all environmental and model parameters constant and varied only strategy parameters from experiment to experiment. Even using our compact way of representing strategies (see section 2.2.1), we need to define strategies in a 9-dimensional space (using two times four weight parameters and a threshold parameter). Assuming that parameters and the 8n

= 2000 independent runs were usually sufficient.

40

Chapter 2. The Competitive Advantage of Commitment

threshold can only take 5 possible values (i.e. -2, -1, 0, 1, 2), we are left with a strategy space of 59 = 1953125 individual strategies. Fortunately, vast parts of this strategy-space yield similar behavior and thus can be classified under common concepts such as Defection, Fairness, Commitment, Trigger, etc. (see strategy types described above). For example, multiplying all traits and the utility threshold of a strategy S with a positive value will yield a strategy S 0 that behaves identically to S. More generally, as long as a transformation on the trait parameters does not shift the level of utility below or above the threshold for any given situation-specific parameter and does not modify the ordering of alters, it has no effect on behavior. Therefore, in the analysis that follows we will not vary absolute values of single traits, only traits in proportion to each other.9 To assess the robustness of results derived from the simulation experiments we conducted exhaustive sensitivity tests for all sensibly variable parameters. We report interesting deviations from typical results in section 2.3.4 below. For a list of all parameters see Appendix B. Initial parameters To determine interesting initial parameters for the simulation experiments and to reduce the parameter space that must be explored, we conducted a game theoretical analysis of a simplified version of the dilemma. Our goal was to identify the set of conditions that makes the choice for agents between purposeful defection and (conditional) cooperation as difficult as possible. If cooperation places an excessively high burden on agents, or conversely, if cooperation entails no real sacrifices, the model would hardly yield any interesting insights. To approximate the conditions under which cooperation is rational at all in the delayed exchange dilemma, we calculated expected payoffs in a simplified version of the game using trigger strategies. A trigger strategy behaves so that as soon as its interaction partner or itself defects, it falls back into a period of unconditional defection. The most severe version of trigger strategies is Grim Trigger, which never switches back to cooperation after its partner or itself defected. Even after its own unintended defection (i.e. due to being unable to help), Grim Trigger applies the most severe punishment possible in the game, permanent retaliation. If the sanction imposed by Grim Trigger cannot deter a rational player from unilateral defection, then no cooperative strategy can do so. As a consequence, there exists no Nash equilibrium – that is: a rational outcome – in which both players choose a conditionally cooperative strategy (see Abreu, 1988). The simplifications we make for the sake of the formal analysis 9 In other words, we are not covering exhaustively the entire parameter space, only a very large part of it. It may still be possible that there is a specific trait combination that is superior under some conditions to the ones examined.

2.3. Results

41

are that we reduce the group size to two and that we omit the evolutionary dynamics and the possibility of death due to low fitness. After solving this simplified dilemma situation (see Appendix A), we get a condition for the rationality of cooperation in the form of a relationship between the probability of distress, the cost for helping and the cost for not getting help: fh < fd (1 − Pd ) We used this result to adjust the most important initial parameters of the model. In other words, we always varied the probability of distress, the cost of help and the cost of not getting help in a way that the above inequality remained true. The actual parameters that we used to draw figures below are: Pd = 0.2, fh = 1, fd = 20, fi = 100, fc = 0, N = 25, Pe = 0.05, m = 2. For the entire set of parameter ranges that we tested in the experiments, please refer to Appendix B.

2.3.2

The unconditionality of Commitment

We started our experiments with comparing four different versions of Commitment playing against defectors. The necessary and sufficient condition for a strategy S to be classified under commitment is that its ahD and ahS traits are positive, which means that agents belonging to S will be inclined to choose those alters for cooperation who have helped them in the past. In its simplest form, this is all that Commitment cares about, expressed by e.g. the following traits: [0, 0, 1, 0|1].10 We will refer to this strategy as Weak Commitment from now on. An important question about the behavior of Commitment is whether the fact that ego helped alter before (EH) should also increase ego’s willingness to cooperate. Intuitively, such a preference makes an agent more vulnerable to exploitation and holds no obvious benefits for ego, if it has an effect at all. Therefore, the second version of Commitment we examine has a non-zero value on its eh parameter: [1, 0, 1, 0|1], we denote it Strong Commitment. If we compare now how Weak Commitment and Strong Commitment play against Defection, we see that indeed the eh trait makes a difference. Whereas Weak Commitment managed to eliminate Defectors from the society in 5.1% of all simulation runs, Strong Commitment did so in 67.7%. These results are based on 2000 independent replications of the simulation for both conditions. What is important here are not the actual percentages but the relative success of Strong Commitment compared to Weak Commitment when playing against Defectors. While the survival statistic will shift in favor of defectors when 10 A strategy is described by four trait parameters for donation, four trait parameters for seeking and the utility threshold: [ehD , er D , ahD , arD |ehS , er S , ahS , arS |Ut ]. If parameters for donation and seeking are equal, we provide only four trait parameters instead of eight. Find more information above, in section 2.2.1.

42

Chapter 2. The Competitive Advantage of Commitment

the cost of giving help (fh ) is increased or the cost of not getting help (fd ) is decreased, Strong Commitment remains superior to Weak Commitment. Figure 2.1 shows how the fitness proportion of Weak and Strong Commitment agents playing against Defectors changes over time. In both cases, Defection initially owns 20% of the total societal fitness (and group size), Commitment 80%. Each curve corresponds to an independent simulation run. All runs end with one strategy completely outnumbering the other but each run may be of different length. Shorter runs are complemented with dashed lines for clarity. What is important to observe is the proportion of curves ending in 0% compared to those ending in 100%11 . Additionally, a black curve shows the average fitness proportion held by a strategy at each point in time across multiple runs. Note also how simulation runs tend to last much longer in the case of Strong Commitment. This indicates that it takes less time for Defection to push out Weak Commitment, than it takes Strong Commitment to push out Defection. Although Defection starts from a smaller proportion than its opponent in both cases, it clearly outpowers Weak Commitment in most runs. In the second case, by contrast, Defection hardly ever manages to climb to the fitness level of Strong Commitment. To assess the relative importance of the eh and the ah traits, we examined two “mixed” versions of Commitment: [1, 0, 2, 0|1] and [2, 0, 1, 0|1]. The former version represents a strategy that derives more utility from receiving help than from giving help to a particular partner, whereas the latter version derives more utility from giving than from receiving. Both had high survival statistics. The proportion of replications in which the corresponding Commitment strategy became universal in the simulated group was 64.2% and 72.4%, respectively. The results also hint at a stable positive effect of eh on survival success. Unconditionality and AllC One doubt that might have surfaced in the heedful reader about the characteristics of our model is that the more cooperative a strategy is, the more successful it will become due to the relatively low costs of cooperation. In order to show that this is not true, we provide the results for the strategy AllC playing against Defection. AllC is the upper end of cooperativeness: it always chooses to cooperate. All other strategies are either equally or less cooperative than AllC. If more cooperativeness implied higher survival chances, AllC should be the winner of all. 11 Note that in order the maintain the clarity of the graphs, we reduced the number of simulation

runs actually plotted on the figures.

2.3. Results

43

Figure 2.1: Weak and Strong Commitment playing against Defection This is not what we see in the results: Defectors managed to overthrow AllC in 88.6% of all simulation runs (see also Figure 2.2). Comparing the survival rates of AllC with those of any version of Commitment except Weak Commitment, it is obvious that AllC is less successful. The weakness of AllC is caused partly by its partner selection behavior. Since AllC is blindly cooperating it is also blind in choosing its interaction partners. When deciding who to give help to and who to ask help from, AllC is indifferent between all possible partners.12 The main weakness of AllC compared to Commitment is its lack of an explicit partner selection strategy. Both versions of Commitment are more likely to help those others who have helped them before. Accordingly, a player who tries to exploit Commitment will always be less likely to get help than 12 Therefore, it is important to define AllC as [0, 0, 0, 0|0] and not e.g. as [1, 1, 1, 1|1] or [2, 2, 2, 2|2] since in the latter two cases it would become a variant of commitment (i.e. it will attribute more utility to those he interacted with more times in the past). While the latter two definitions are perfectly identical to each other, the first definition prescribes surprisingly different behavior with regard to partner selection.

44

Chapter 2. The Competitive Advantage of Commitment

Figure 2.2: AllC playing against Defection

somebody else who cooperated with AllC, all other conditions being equal. Due to its random partner selection method, AllC is the upper end of not only cooperativeness but of unconditionality as well.

Commitment in comparison with fair conditional cooperators The classic fair type (known e.g. from Fehr and Schmidt, 1999; Fehr et al., 2002) keeps a close watch on the balance with its opponent. A fair ego calculates the balance with regard to help donation to a particular alter as follows. Whenever ego helps alter or alter refuses ego, ego subtracts a unit from the balance with alter and whenever alter helps ego or ego refuses alter, ego adds one unit to the balance. The balance with regard to help seeking is calculated in exactly the opposite way. That is, the balance is most favorable with respect to the agent who owes ego the most. In terms of our model, this strategy is defined as [−1, 1, 1, −1|1, −1, −1, 1| − 1]. We will refer to this as Objective Fairness. If we examine how this objective version of Fairness plays against Defection, we see that it has a 2.5% chance of surviving, which is lower than the worst we saw for Commitment. The failure of Objective Fairness (Figure 2.3) can be explained with the large number of rejections out of unwillingness to help. These rejections are due to asymmetries in the number of times helping partners become distressed: e.g. if ego becomes distressed too often compared to alter (i.e. before he can reciprocate help from alter), alter will no longer provide help for ego. This result is consistent with previous game theoretical and simulation analyses that pointed to the disadvantages of strict reciprocity in uncertain environments (e.g. Kollock, 1993). Objective Fairness can be straightforwardly modified in such a way that

2.3. Results

45

Figure 2.3: Objective Fairness playing against Defection

it becomes more tolerant against temporary fluctuations in the frequency of needing help. For our simulation, we define a corresponding strategy of Tolerant Fairness as [−1, 2, 2, −1|2, −1, −1, 2| − 1]. The survival statistics of Tolerant Fairness (see also Figure 2.4) are clearly better than those of Objective Fairness. Tolerant Fairness stayed standing against Defection in 13.2% of all simulation runs. This result, however, is still worse than the result of Strong Commitment.

Figure 2.4: Tolerant Fairness playing against Defection

46

2.3.3

Chapter 2. The Competitive Advantage of Commitment

Explanation: the importance of strong ties

To understand the variation in the success of different cooperative strategies, we first tested how viable they are when they play without defectors, in a homogeneous society composed of only one cooperative strategy. Using identical parameter settings and group size we found that there is still significant variation in success, even between unconditionally cooperative strategies. When Commitment-type agents play against each other, being largely unconditionally cooperative, the most frequently observed type of refusal is when an agent is not able to help another agent. Hence, whether strategies cooperate or not cannot explain variation in success of different commitment types. However, whom players select to cooperate with or to request cooperation from turns out to be decisively different between versions of Commitment. But why should partner selection matter when everybody else is playing Commitment and will never refuse to help out of unwillingness? There are two problems faced by an agent even in this homogeneously friendly world of fellow cooperators. One is if help is sought from an agent who is distressed himself and cannot help; the other is if multiple agents ask the same agent to help. A collectively ideal strategy works so that help requests are evenly distributed among agents who are not distressed. It turns out that commitment comes very close to being such an ideal strategy, due to a phenomenon of dyadization. To understand dyadization, let us consider what happens in the first few rounds. According to individually independent random events with probability Pd , a fraction of the whole society N · Pd becomes distressed. Since nobody has had an interaction before, the distressed agents will all choose randomly whom they ask for help. For a distressed agent the probability of losing fitness by the end of the run is composed of two parts: Pf itnessloss = P1 + P2 , where P1 is the probability that another distressed agent also asked and got help from alter, and P2 is the probability that alter is also distressed. Let us assume now that eh = 0 for all agents. If an agent ai who was not distressed and helped in the previous round becomes distressed now, that agent ai will face a population of equally preferred others to ask for help. Therefore, ai will have to choose randomly, facing the same probability P1 that the other unlucky ones faced in the previous round. Now, if eh > 0, all agents who gave help before will have a preference for those they helped, and thus P1 will be reduced in the second round. To put it more generally, in later rounds, if everybody simply chooses the partner they had the most interactions with, it is likely that there will be few collisions between help requests. This leads to the formation of increasingly

47

2.3. Results

strong ties, and this is what we referred to as dyadization. Graphing the social network of Weak and Strong Commitment clearly shows the difference in the extent of dyadization (see Figure 2.5; the darker a tie the more interactions have taken place between the two agents). In the case of Strong Commitment, we see few but stronger ties (“close friendships”), while in the case of Weak Commitment we see many but weaker ties. A more careful look reveals that most Strong Commitment agents have exactly one strongest tie (“best friend”), while this is less true for Weak Commitment and even less for Fairness (Figure 2.6). What we see in a network of Fairness players are homogeneously weak relationships and nodes with a larger degree in terms of the number of ties. All networks are graphed after 200 rounds, which means that the added strength of all ties (network density) in the networks is roughly13 equal. To precisely measure the dyadization of a network, we first compute the “individual dyadization” for each node in the network by dividing the strongest tie of a node by the total strength of all ties of that node. The strength of a tie is obtained by counting the number of interactions that have taken place along that tie. We average the resulting values over all agents to obtain the dyadization measure of the network. A network dyadization of 1, or perfect dyadization, means that every node is connected to exactly one other node. The dyadization after 200 rounds averaged across 2000 runs for a network of Weak Commitment players is 0.29, whereas it is 0.39 for Strong Commitment. The dyadization measure for Tolerant Fairness is 0.23. Invasion of cooperators on cooperators In a final test of the relative viability of Commitment compared to Fairness we let Strong Commitment play against Tolerant Fairness in the presence of Defectors. Commitment and Fairness both started with equal initial proportions (40% each) while Defectors were in a minority (20%). Although Commitment did better (46.3%) than Fairness (2.7%), it was Defection who won (51.0%) this tournament. Apparently, the cooperative group in this case was considerably weakened by the occasional lack of cooperation of Fairness players. Observing the network structure of all agents we found again the characteristic strong friendships between Commitment players. Defectors, at the other extreme, attempted to interact with as many others as possible, in search of partners to be exploited. Moreover, we also found some strong ties between Commitment and Fairness agents. What happens in these relationships is that the Commitment player becomes attached to the Fairness player after some initial rounds of helping. The problem for the Commitment agent arises if the relationship becomes unbalanced – Commitment will keep asking its Fairness “friend” for help even in the face of repeated refusals. This points to a weakness of Commitment. 13 Depending

on the actual number of stochastically arisen distresses.

48

Chapter 2. The Competitive Advantage of Commitment

Figure 2.5: Emergent social network of Weak and Strong Commitment players after 200 rounds

2.3.4

Sensitivity to initial parameters

To assess the robustness of the results reported above, we examined their sensitivity to the choice of initial parameters. The cost of helping, the cost of not getting help and the probability of distress As we pointed out earlier, the ratios of the three parameters, cost of helping (fh ), cost of not getting help (fd ) and probability of distress (Pd ) are essential to determine the individual rationality of cooperation. Increasing fh in proportion to fd results in better survival chances of defectors. According to the analysis of the simplified dilemma (see Section 2.3.1), increasing the probability of distress makes the conditions under which conditional cooperation is individually rational more restrictive. This is not what we see in the simulations. Increasing Pd from 0.05 to 0.2 actually benefits cooperators, especially Commitment. This result replicates what was found by de Vos et al.

2.3. Results

49

Figure 2.6: Emergent social network between Tolerant Fairness players after 200 rounds

(2001). The explanation is that Commitment players find each other sooner and strengthen their relationships faster, the harsher the environment is, i.e. the larger the probability of distress is. Initial distribution and group size In those simulation experiments reported above, where a group of cooperators played against a group of defectors, we started from an initial population of 20 cooperators against 5 defectors. Decreasing the initial population size to 10 or increasing it to 50, keeping the initial ratio of cooperators to defectors constant, did not result in notable deviations from our results. Changing the initial proportions of cooperators and defectors did change percentages of survival to some extent but it did not reverse our qualitative conclusions. More precisely, although we found that in all experiments increasing the initial proportion of defectors led to lower survival chances for cooperative strategies, the survival chances for strong commitment were still higher than those of weak commitment or fairness strategies. Moreover, decreasing the initial proportion of defectors did not change results qualitatively either because we found that the evolutionary dynamic enabled even a smaller but superior invading strategy to take over the entire population. We can conclude that the choice of these parameters does not affect our results qualitatively. Probability of decision-making error We found no unexpected results when varying the error of decision making between 0.0 to 0.5: the larger the error was, the smaller the difference be-

50

Chapter 2. The Competitive Advantage of Commitment

came between the behavior of different strategies as they all approached a completely random strategy. Defectors suffered most from a high level of decision-making error: although they became more marginally cooperative, their choice of when and with whom to interact was completely haphazard. Number of subrounds and initial fitness By increasing the number of subrounds (m), agents in distress have a higher chance of finding a helping partner within one round. The general effect we expected was that survival becomes easier for all agents, and that connections build up faster as agents encounter more alters during the same number of rounds. It is intuitively not obvious, however, who benefits more from a second (third etc.) chance – cooperators or defectors? Rerunning the simulations, keeping all parameters unchanged except varying m between 1,2 and 3, we found that it is defectors who have better chances of pushing out cooperators, the larger m is. Whereas more subrounds give more chances to get help within one distress period, higher initial fitness keeps the agent alive across multiple distress periods. We expected the same general effect for changing the initial fitness parameter (fi ) as for changing the subrounds parameter because an increase in either parameter results in increased survival chances of all agents. Varying the initial fitness parameter fi , we found an interesting nonlinearity. Increasing the initial fitness from 50 to 100 resulted in defectors becoming more successful against both conditional and unconditional cooperators. Further increasing fi , however, resulted in defectors becoming relatively even less successful against Commitment players than at lower initial fitness. The general result of Commitment being more successfully than Fairness when playing against defectors remained constant throughout this test as well.

2.4

Discussion and Conclusion

Our results suggest that strategies following some form of commitment behavior are highly successful under a wide range of conditions. Broadly, commitment is modeled as the extent to which cooperativeness with a particular partner becomes unconditional after some initial cooperative actions of the partner. Counterintuitively, the faster an agent is inclined to solidify its relationships (see Strong Commitment), the less prone it is to exploitation. The reason is that a relationship between two Strong Commitment agents is built up – probabilistically – at least twice as fast as a relationship between a Strong Commitment and a Defector agent.

2.4. Discussion and Conclusion

51

Our approach shows that the success of Commitment remains stable even when a much larger range of strategy variation is allowed than in the previous computational experiments of de Vos et al. (2001). We find that strategies that base their behavior on fairness principles generally perform much worse than commitment strategies. A truly fair strategy suffers from its lack of tolerance when interacting with its own kind in an unpredictably “unfair” environment, where imbalances in the exchange cannot be avoided due to uncertain hazards. A more tolerant strategy that is nonetheless based on preferences for fair outcomes proves to be more viable. An interesting result of our model is that strategies that take their own past behavior into account – not just that of their interaction partners – make more successful decisions in general. To explicate the reasons for the success of commitment, we also studied the spontaneous formation of exchange network structures in the simulated populations. It turned out that commitment strategies derive a large part of their success from efficient networking: they avoid overloading few designated individuals with interaction requests and instead spontaneously create a structure that ensures an efficient coordination of help requests and help provisions. The problem with fair strategies is that they are predisposed to keeping their relationships in balance so that agents tend to spread interaction requests randomly across the population. During times of great need, this structure is inefficient because fair strategies in small groups generate overlapping personal networks so that often too many people try to interact with the same agent at the same time. While we tentatively conclude that our results support the hypothesis that commitment strategies are evolutionarily viable, we are also aware of a range of potential limitations of our analysis, some of which point in the directions for future research. A first objection to our study might be that we excluded the influences of reputation mechanisms on the relative success of strategies. There are a number of game theoretical studies and agent-based simulations that show how reputation mechanisms can sustain cooperation, because they help cooperators to effectively identify and punish cheaters (see e.g. Takahashi, 2000; Raub and Weesie, 1990). In our analysis, reputation effects were explicitly excluded with the assumption that agents rely only on their own experience about others when making decisions. We argue that taking reputation into account may be an unnecessary complication that would not lead to a qualitative change in the outcome of our comparison of Commitment with other versions of conditional cooperation. There is no particular reason to believe that Commitment would benefit less from reputation than other cooperative strategies. On the contrary, as Commitment players build up ties more readily and never abandon them afterwards, it may be particularly useful for Commitment players to use thirdparty information to identify in advance who is a reliable helping partner and

52

Chapter 2. The Competitive Advantage of Commitment

who is not. A further possible limitation of the present analysis is that we have not yet explored more sophisticated cheating strategies. It is possible that more sophisticated cheating may indeed undermine the viability of Commitment. The Defection strategy derived from AllD that we used in our experiments is not capable of taking advantage of what may be the most decisive weakness of Commitment, its inability to strike back once it is exploited by a partner to whom it has become committed. The only way Commitment punishes opportunists seeking help occasionally is by giving higher priority to friends with whom more numerous successful interactions have taken place. Similarly, Commitment will not even try to get help from its occasional interaction partners, if it has long-standing partners. In other words, instead of detecting cheaters by the number of times they defected, Commitment detects cooperators using the number of successful interactions. Nevertheless, there may exist more viable strategies outside the range that our analysis has covered. In future research we plan to extend our analysis in this direction. In particular, the effect of more sophisticated cheating strategies can be tested by allowing mutations to randomly generate strategies from a large set of possibilities. Clearly, this requires the modification of the evolutionary dynamic and significantly larger computational power than we used for the present study. A final line of future work may follow from resolving a simplifying assumption we made, to ignore possible differences between group members in terms of their attractiveness as exchange partners beyond their strategy. Such differences may for example come from variation in physical strength or more or less favorable local living conditions. Studies by Hegselmann (1996) (cf. Flache and Hegselmann, 1999b) suggested that variation in attractiveness may give rise to core periphery network structures in which the strongest population members exchange help with each other, driving weaker actors to the margin of help exchange networks. However, this work relied on conditionally cooperative strategies that resemble the strategy of Fairness. Accordingly, it is unclear whether variation in individual attractiveness may affect the viability of commitment strategies and also whether commitment strategies would give rise to the exclusion of weak members from exchange networks in the same way as has been found for Fairness-like strategies.

Chapter 3

The Evolutionary Advantage of Commitment1

Abstract

Why are people inclined to build friendships and maintain durable, nonreproductive relationships? Previous computational modeling work showed that it can be an efficient survival strategy to choose interaction partners based on relationship length, even if, as a consequence, individuals become unconditionally cooperative in long-term relationships (interpersonal commitment). Such committed individuals can outperform conditional cooperators who play in a fair, reciprocal manner (e.g. Tit for Tat). However, previous studies did not conduct a sufficiently strict test of the viability of commitment because they did not account for exploiters who specifically take advantage of the tolerance of commitment players. We allow this by extending previous studies with the possibility of randomly mutating strategies under evolutionary pressures, and thus give a much larger coverage of an infinite strategy space. Our results point to the lack of stable strategies: we find that emerging populations alternate between temporarily stable equilibria. We also show that the viability of strategies increases with increasing levels of interpersonal commitment, and that the effect of interpersonal commitment on viability is larger than the effect of fairness. 1 This

chapter is based on Back, I. and Flache, A. (Forthcoming). The Adaptive Rationality of Interpersonal Commitment. Rationality and Society.

53

54

3.1

Chapter 3. The Evolutionary Advantage of Commitment

Introduction

Among the species of the earth, humans exhibit the highest level of cooperation between genetically unrelated individuals (Gintis, 2003). Arguably, cooperation is the de facto key to our evolutionary success. At the same time, cooperation is problematic to explain from a rational actor perspective. Selfinterested actors often face a “social dilemma” (Dawes, 1980) where the rational pursuit of individual interests may lead them towards defection, while this in turn entails collectively undesirable outcomes. Game theory has identified repeated interaction as an important solution for the problem. In a world of harsh competition, repeated encounters reduce uncertainty about the trustworthiness of interaction partners (shadow of the past) while at the same time they create a strategic incentive for cooperation (shadow of the future) (Friedman, 1971; Axelrod, 1984; Buskens and Raub, 2002). Thus durable relationships are expected to be a hotbed of cooperation even in the absence of central enforcement because they provide incentives both to trust others and to honor others’ trust. From this perspective, it is hardly surprising that rational incentives to become committed to long-term cooperative exchange partners are particularly strong in uncertain environments (Schüssler, 1989; Kollock, 1994). A reduction in uncertainty is often more valuable than a probabilistic increase in payoff from a potential new partner, especially if switching itself is risky, costly or alternatives are scarce. This can explain why in situations where uncertainty may otherwise preclude the desirable outcome of mutual cooperation, social actors often restrict their own freedom of action by using commitment devices such as posting a hostage (Raub, 2004). However, this rational explanation of interpersonal commitment behavior is hard to reconcile with the empirical evidence that people tend to stay committed to long-term interaction partners even when (1) alternatives are available, (2) switching costs are low, and (3) uncertainty is of less concern. A growing body of empirical findings from both interpersonal relationships research (Karremans et al., 2003; Wieselquist et al., 1999) and exchange experiments (Kollock, 1994; Lawler and Yoon, 1993, 1996) shows that people have a tendency to remain cooperative with interaction partners who are occasionally uncooperative. Moreover, people tend to keep exchanging with the same partner even if more valuable (or less costly) alternatives are available. Such commitments also imply forgiveness and gift-giving without any explicit demand for reciprocation (Lawler, 2001; Lawler and Yoon, 1993). People help friends and acquaintances in trouble, apparently without calculating present costs and future benefits. Another, extreme example is the case of battered women who stay with their abusive husbands (Rusbult and Martz, 1995; Rusbult et al., 1998). In this paper we seek an explanation by following the general lead of the

3.1. Introduction

55

“indirect evolutionary approach” (Güth and Kliemt, 1998), which posits that individuals act rationally in the light of their preferences but also assumes that in the course of biological and cultural evolution individuals with social preferences and emotions – e.g. for fair distributions (cf. Bolton and Ockenfels, 2000), or for altruistic punishment (cf. Fehr and Gächter, 2002) – may have had a selective advantage, because their preferences produce more viable outcomes than those of pure egoists. This approach aims to integrate the endogenous explanation of such non-selfish preferences with the classical rational choice assumption that humans act (boundedly) rationally, given their preferences. The core idea is that preferences are embodied in the genotype and guide individual actions. Subjective preferences may be harmful to the individual or to the population but genotypes are selected on the basis of objective consequences of the actions that preferences produce. As preferences undergo selection and mutation, unfeasible and harmful preferences gradually become less widespread in the population, giving way to more “rational” sets of preferences. However, while the indirect evolutionary approach has proven to be a fruitful avenue in explaining phenomena such as emotional commitment to a certain course of action, altruistic punishment, or cooperation in the production of collective goods (Frank, 1988; Güth and Kliemt, 1998; Güth and Ockenfels, 2002; Gintis, 2003), the phenomenon of interpersonal commitment has received relatively less attention. Recently, some authors have begun to use the indirect evolutionary approach to explain interpersonal commitment (Back and Flache, 2006; de Vos et al., 2001; Zeggelink et al., 2000; de Vos and Zeggelink, 1997). While these analyses suggested that commitment might have been evolutionarily viable, we argue that the tests they used were not strict enough. De Vos and his collaborators argued in a series of papers that in a stylized “ancestral environment” a strategy based on commitment behavior could outperform a strategy based on calculative reciprocity when both strategies are in competition with one defecting strategy. The researchers modeled commitment as unconditional cooperativeness with a particular partner after some initial cooperative actions of the partner. By contrast, calculative reciprocity (based on fairness principles) continuously keeps track of its interaction balance with alters and adjusts its cooperativeness accordingly. Using an ecological simulation model, Back and Flache (2006) extended the de Vos model introducing variation in the extent to which a strategy follows commitment or calculative reciprocity behavior. This study showed that “strong” commitment strategies outperform “weaker” forms of commitment and various versions of calculative reciprocators under a wide range of conditions. However, a remaining major limitation of these analyses is that the spontaneous emergence of more sophisticated strategies was not considered. In particular, it was precluded that sophisticated cheaters emerge who optimally take advantage of the cooperativeness of commitment. Whether and to what extent this may be possible is crucial for

56

Chapter 3. The Evolutionary Advantage of Commitment

the validity of an explanation of interpersonal commitment behavior in terms of its evolutionary advantages in the human ancestral environment. Accordingly, in the present paper we provide a better test of evolutionary explanations for commitment by extending previous analyses with random mutation of strategies. In Section 3.2, we present our computational model and formulate conjectures. Section 3.3 contains results of simulation experiments, followed by a discussion and conclusions in Section 3.4.

3.2

Model

We use an abstract decision situation that we call the Delayed Exchange Dilemma (see de Vos et al., 2001; Back and Flache, 2006), or DED for short. The DED builds on the well-known repeated Prisoner’s Dilemma but contains two major extensions. First, it puts the problem of cooperation into a sequential exchange perspective, which is essentially a generalization of simultaneous exchange. Second, and more important is that it presents agents with a dilemma to choose interaction partners (see also e.g. Hayashi and Yamagishi, 1998). With these extensions the DED becomes ideal for studying commitment-related behavior in uncertain environments. The DED is played by n agents in successive rounds. Initially, all agents are endowed with fi points. In the beginning of each round Nature strikes a number of agents, each with a given individually independent probability Pd , who become in need of help from other agents. Agents who are struck by Nature are the initiators of interactions. Each of them asks another agent for help which is either provided or not. Providing help costs fh points. Moreover, help giving is time-consuming. Each agent can only provide help once during one round and only agents who are not distressed themselves may provide help. If a help request is turned down, the distressed agent may ask another agent for help but not more than m agents altogether within the same round, due to time restrictions. If an agent does not get help before the end of the round, it experiences fd loss in points. If the points of an agent fall below a critical threshold fc , the agent dies.

3.2.1

Modeling strategies

To explicitly study the evolutionary viability of commitment and fair reciprocity, we model preferences as a combination of commitment-related traits, fairness-related traits and a general cooperativeness trait. These traits determine the extent to which agents base their decisions on commitment- or fairness-related aspects of a decision situation. Equipped with these preferences agents decide about cooperation and also about choosing interaction partners.

3.2. Model

57

In particular, agents may face two different types of decision situations repeatedly in the DED. When they are hit by distress, they have to select an interaction partner to ask help from. On the other hand, when they themselves are asked to provide help they need to decide whether to provide it and in case of multiple requests whom to provide it to. In both cases, agents order possible interaction partners according to the attractiveness of interacting with them. Attractiveness is based on the individual preferences agents have with regard to past interaction histories. The attractiveness of agent aj for giving help to, calculated by agent ai , is formalized as: G G Uij = commG i · IN T F REQij + f airi · IN T BALij + coopi , G where commG i is the preference for commitment in giving, f airi is the preference for fairness in giving, and coopi is the preference for general cooperativeness. IN T F REQij is the proportion of cooperative interactions2 ai had with aj compared to the total number of cooperative interactions ai had. A cooperative interaction is defined as an interaction in which either ai helped aj or ai received help from aj . IN T BALij is the standardized interaction balance between agents ai and aj . To obtain this measure, we took into account both the balance of helps and the balance of refusals. The reason is that neither help balance nor refusal balance alone is sufficient to guarantee an overall balance in the exchange relationship. For example, suppose ai helped aj equally often as aj helped ai but ai refused to help aj ten times more often than aj refused to help ai . Despite the equal amount of help given, this exchange relationship clearly cannot be considered perfectly in balance. Technically, we calculated the measure as follows. We subtracted from 1.0 a measure of the overall standardized imbalance. The overall standardized imbalance is obtained by adding the difference of the number of times ai received help from aj and ai gave help to aj , and the difference of the number of times ai refused to give help to aj and aj refused to give help to ai , and dividing this by the total number of interactions they had. When comparing IN T F REQ and IN T BAL, notice that while a committed agent ai will find an interaction partner aj more attractive the more often it helped aj , a fair agent will be negatively influenced by the same fact. Note also that for simplicity, this model treats the impact of helping and refusing to help on the interaction balance as equally large. In the actual implementation, every time an agent has to make a decision, there is also a probability Pe that the agent will not use the above utility calculation but will choose randomly from the set of available decisions, each 2 Interactions take place always between exactly two agents. Possible interactions are giving help (cooperation) and refusing to help (defection). Asking for help is always followed by one of these.

58

Chapter 3. The Evolutionary Advantage of Commitment

being equally likely. This random error models noise in communication, misperception of the situation or simply miscalculation of the utility by the agent. Taking this random error into account increases the robustness of our results to noise. The attractiveness of agent aj for asking help from is defined in a similar way, the difference is that agents may put different weights on the two historyspecific decision parameters, and that there is no cooperativeness parameter: A A Uij = commA i · IN T F REQij + f airi · IN T BALij ,

Before agents make a decision, be it help giving or help asking, they calculate the corresponding one of these two types of attractiveness respectively for each agent who asked for help (U G ), or for each other agent in the population (U A ). In case of help giving, they choose a partner with the highest attractiveness, if that attractiveness is above an agent-specific threshold uti . Notice that IN T F REQij and IN T BALij are always smaller than or equal to 1. We allow commi , f airi and coopi to take values from [−1; 1]. Thus we allow the attractiveness threshold uti to take values from [−3; 3]. If the attractiveness of all possible agents is below the threshold attractiveness, no help is given to anyone. Otherwise, if there is more than one other agent with highest attractiveness3 , the agent selects one of the others with equal probability. In the case of help seeking, agents also choose a partner with the highest attractiveness but there is no threshold, i.e. agents in distress always ask someone for help. Definition 7 (Strategy). A strategy is a combination of four traits for help giving behavior (commG , f airG , coop, ut ) and two traits for help asking behavior (commA , f airA ).

3.2.2

Evolutionary dynamic

The heart of our model is an evolutionary dynamic that captures the random mutation of strategies and selection of objectively successful ones. The implementation of this process is based on the replicator dynamics (Taylor and Jonker, 1978). Broadly, the replicator dynamics dictates that if a generation of genotypes (strategies) undergoes reproduction, the net reproduction rate of a genotype is proportional to its relative success compared to other genotypes in the current generation. Genotypes that perform below average, in particular, have a negative reproduction rate. In our case, genotypes (strategies) represent subjective preferences. To prevent a population from growing without bounds, thus modeling resource scarcity in an implicit way, we keep the size of the population constant, 3 This

is unlikely, as the preference parameters are high precision real values and interaction histories tend to differ with time.

3.3. Conjectures

59

in the following way. At the end of each round we count how many agents have died and replace them with new agents in the next round. Each new agent A has the same strategy as a randomly selected other agent B, present in the population who has reached a minimum age n (measured in the number of interactions it had). The probability of choosing this other agent B is proportionate to the share of points B holds within the group of all agents older than n. Before A is added to the population, with probability Pmut , its strategy may undergo mutation. A mutation occurs in exactly one, randomly chosen trait, with equal probabilities for all traits, thus P = 91 for each trait. The new value of the trait is a uniformly distributed random value from the interval [−3; 3] for the attractiveness threshold, and from [−1; 1] for all other traits.

3.3

Conjectures

To guide the simulation experiments, in the following we formulate a number of conjectures derived from previous work. Definition 8 (Stability). The stability of a strategy s is equal to the number of consecutive rounds that it existed in a population in a given simulation run, counting from the first round it appeared until the round in which it became extinct. A strategy s is infinitely stable if it does not become extinct. Generalizing from analytical results about the evolutionary stability of strategies in repeated games that are simpler than the DED (cf. Bendor and Swistak, 2001), we expect that there is no single strategy that is superior to all others in the dilemma we study. In other words, for every incumbent strategy there exists another (mutant) strategy that can take advantage of the incumbent’s weakness. Conjecture 3.1. There is no infinitely stable strategy in an infinitely played game of DED. Nevertheless, the length of time a strategy exists (stability) carries an important message about its viability. Since mutations constantly arise and threaten to push other strategies out of the population, stability is an indicator for the number of attacks a strategy can withstand. Therefore, the stability of a strategy will be one of the indicators of its viability4 . The other measure is typical longevity within a strategy (variable longevity, the average age at death of agents belonging to a strategy). Note that in our model, there is no 4 We will not use here stability concepts from the evolutionary game theory literature (e.g. evolutionary stability or asymptotic stability) because they do not allow the expression of the relative stability of strategies.

60

Chapter 3. The Evolutionary Advantage of Commitment

upper age limit on reproducibility, in other words, agents keep reproducing until they die, which makes longevity a suitable measure for viability. Back and Flache (2006) found that the most successful strategies in the DED exhibited some level of interpersonal commitment and that committed agents outcompeted fair reciprocators. These results suggest the following two conjectures, which we will test under the new assumption of random mutation: Conjecture 3.2. Individual preferences for interpersonal commitment and fairness have a positive effect on viability. Conjecture 3.3. The positive effect of commitment preferences on viability is stronger than the effect of fairness preferences. According to de Vos et al. (2001), commitment works best under harsh conditions: the more agents are challenged by Nature to survive, the more compelled they are to cooperate with each other in durable relationships. More technically, they found that the larger the probability of distress, the larger the proportion of commitment strategies surviving, relative to the defector strategy. This leads us to test: Conjecture 3.4. Environmental harshness has a positive interaction effect on stability with the level of cooperation and interpersonal commitment of a strategy.

3.4

Results

Binmore (1998) argued forcefully that the outcome of computer tournaments and simulations of evolutionary dynamics strongly depends on the set of strategies that are initially present in a population. To avoid our results becoming biased by a restrictive set of starting conditions, we ran several hundred replications of our simulation runs, each time with a population whose initial strategy is randomly chosen from the strategy space defined by the six traits. We did not find any significant effects of features of initial starting strategies on the outcomes of simulation runs. The reason is that soon after the initial rounds of a simulation run, mutation ensures the emergence of a large variety of different strategies in the population. We allow this population to play the DED game. In the course of the game agents start to lose points, some of them eventually die, while others reproduce. At some point, random mutations occur in the initial strategy, creating a potential invader. The better a mutated strategy performs in the DED compared to agents of the original strategy, the larger is its probability of reproducing and increasing its proportion within the agent population. The simulation run ends with either the extinction of all agents5 or after an arbitrarily chosen 5 Extinction

is possible if all agents die within one round and thus there is no basis for the distribution of strategies in the next generation.

3.4. Results

61

large number of rounds (10 million). We then repeat the simulation run with another, randomly generated initial population. During each simulation run we record all strategies and their key characteristics that have ever appeared through random mutations. These characteristics include on the individual-level the traits of the strategy (commA , f airA , coop, ut , commG , f airG ); the average longevity measured in rounds of game play on the strategy-level; and finally a population-level variable measuring the overall level of cooperation and defection (SOCCOOP)6 .

3.4.1

Initial parameters

To preserve comparability of our results, we started our simulations with the same initial parameters (where applicable) that were used in earlier work. These are Pd = 0.2, fh = 1, fd = 20, fi = 100, fc = 0, N = 25, Pe = 0.05, m = 2. (For the meaning of each parameter, please consult the Model section above.) These parameters impose a set of conditions under which for strictly instrumental agents the choice between purposeful defection and (conditional) cooperation is as difficult as possible. The parameters are determined such that in a simplified two-person version of the game, perfectly rational actors would be indifferent between choosing for a conditionally cooperative and a fully defecting strategy if they meet a conditionally cooperative partner. In this way, we implement a setting in which the problem of cooperation is particularly hard to solve and provide thus a strict test of the viability of cooperative strategies, including commitment and fairness. If cooperation placed an excessively high burden on agents, or conversely, if cooperation entailed no real sacrifices, the model would hardly yield any interesting insights. (For the detailed game theoretical derivation using trigger strategies, see A.) We refer to this parameter setting as the baseline condition. After obtaining results for the baseline condition, we conduct experiments in which we systematically vary the level of environmental harshness (fd ). Furthermore, we run additional experiments with varying parameter combinations to test the sensitivity of results to variation in model parameters.

3.4.2

Stability

In support of conjecture 3.1, our simulation results show that strategies change endlessly in all initial parameter settings – we found no infinitely stable strategy in the DED. We simulated 175 runs altogether, each of which started with a different randomly chosen initial strategy and consisted of maximally 10,000,000 rounds. During these runs more than 4.7 million mutations took 6 SOCCOOP

measures the difference between the per-round average number of cooperation (helps) and defection (refusals) in the entire population.

62

Chapter 3. The Evolutionary Advantage of Commitment

place altogether, generating as many strategies. However, in none of these runs have we recorded any strategy that existed longer than 220,000 rounds. We may of course simply have not encountered the infinitely stable strategies during our random walks in this vast strategy space. However, judging by the vast coverage of the strategy space by our method, this seems implausible. A plausible explanation for the lack of infinitely stable strategies is that for each strategy there exists a better response that takes advantage of the strategy’s weakness. Sooner or later mutations generate this better response and the original strategy is gradually pushed out of existence. If a strategy is too cooperative, opportunistic exploiters take advantage of this and flourish. Later, in a harsh world of mainly exploiters, where everybody is suffering, two cooperators who appear randomly at the same time and find each other will survive and reproduce more easily than others, given that they have a sound method of excluding defectors from cooperative interactions. That cooperativeness eventually declines again may be explained by the gradual loss of the ability to exclude defectors through “evolutionary drift” (e.g. Bendor and Swistak, 2001) or by the emergence of new defecting mutants who have developed the ability to behave such that they are not excluded from exchanges between incumbent cooperators. Figure 3.1 illustrates these dynamics of average age at death and helping behavior, for a typical simulation run. The upper part of the figure shows how the average age at death (measured in interactions) changes over time within one simulation round. Compare this figure with the level of cooperation, generated for the same simulation run, in the lower part of the figure: periods of high refusal rates coincide with short lives.

3.4.3

The importance of interpersonal commitment

Conjectures 3.2 and 3.3 relate the strength of the commitment preference within a strategy directly to viability, the average length of an agent’s life within a strategy. Commitment is measured by the commG and commA traits, distinguished for giving and asking respectively. The higher these traits are, within the [-1;1] interval, the more an agent is inclined to choose and cooperate with long-term interaction partners. If they are positive, the agent has a preference for commitment; if they are negative, the agent has a preference against being committed; and when the values are close to zero, the agent is indifferent to the concept of commitment. What we find is that among the most stable 1,355 strategies (where stability is at least 50,000 rounds), 776 strategies (57.3%) are positive on both commitment traits. Among the same strategies, only 385 (28.4%) are positive on both f airG and f airA , and 717 (52.9%) on coop. This suggests that if a strategy is highly stable, its decision-making process is likely to be guided by prefer-

3.4. Results

63

ences for unconditional cooperation with old interaction partners. These preferences appear to be far more important for success than being fair or simply being cooperative (coop trait). To get a closer insight into the separate contributions of the traits to a strategy’s success and to compare in particular the relative importance of the f airG and f airA traits to the importance of the commG and commA traits, we conducted a linear regression analysis with average longevity within a strategy as the dependent variable (see Table 3.1). Before performing the analysis we filtered out highly unstable strategies (STAB<2000 rounds) and strategies with low longevity (longevity<75 interactions). The reason to filter out highly unstable strategies is that due to the stochastic nature of the simulation, it often happens that strategies that would otherwise be stable cannot grow to a critical mass in the population to stabilize. In this case, they distort the association between strategy features and viability. The reduced sample consisted of 34,143 strategies. We estimate three models (see Table 3.1), gradually extending the set of independent variables included. With the first model we test the effects of the strength of preferences on viability. The second model adds environmental harshness and its interaction effects; while the third model adds interaction effects between similar preferences.

Figure 3.1: Age at death and cooperation in the baseline condition (single run, initial 1 million rounds)

64 Chapter 3. The Evolutionary Advantage of Commitment

97.035** 1.754** 1.393** 5.625** -0.754** -4.340** -6.798**

(0.088) (0.128) (0.134) (0.126) (0.129) (0.135) (0.051)

R 0.673

97.325** 1.272** 0.143 5.752** -0.651** -4.556** -6.920** 0.038 -0.456** -1.106**

R2 -adj. 0.452

(0.233) (0.144) (0.165) (0.134) (0.129) (0.136) (0.076) (0.019) (0.068) (0.076)

Model 2 Unstandardized Coefficients B Std. Error

R 0.685

97.510** 1.136** 0.043 5.770** -0.560** -3.798** -6.963** -0.006 -0.507** -1.284** 0.639** -2.415**

R2 -adj. 0.469

(0.230) (0.144) (0.162) (0.132) (0.128) (0.136) (0.075) (0.018) (0.068) (0.075) (0.074) (0.076)

Model 3 Unstandardized Coefficients B Std. Error

Table 3.1: OLS regression with dependent variable longevity

R R2 -adj. 0.670 0.448 **Significant at the p<0.001 level.

(Constant) commD f airD coop commS f airS ut fd commD × fd f airD × fd commD ×commS f airD × f airS

Model 1 Unstandardized Coefficients B Std. Error

3.4. Results 65

66

Chapter 3. The Evolutionary Advantage of Commitment

In the first model we see that the effect of having preferences for commitment and for fairness in giving are both positive but the effect of the commitment preference is larger, as expected based on Conjectures 2 and 3. We also see that the cooperation preference has a very large effect. Preferences in asking are negative but for fairness the coefficient is much larger in absolute value. This suggests that strategies that restrict the partner search too much either to old partners or to partners with balanced exchange ratios are disadvantaged because their search space is overly reduced. The attractiveness threshold ut has a very large negative effect. The explanation is that ut is very important in deciding whether a strategy is initially cooperating or defecting (niceness). This in turn is crucial for the ability to bind future helping partners or establish mutually cooperative balanced exchange relationships. In Model 2 we include the main and interaction effects of environmental harshness (fd , cost of not getting help). To test the sensitivity of our result to the choice of environmental harshness, we repeated our simulations for a range of parameters. Namely, we were interested in the effect of variation in the proportion of the cost of cooperation and the cost of being cheated. We reran the simulation with fd = 5, 10, (20), 30 and added these repetitions to the original dataset obtained in the baseline condition. Then we tested for significant effects of fd on the dependent variable LON GEV IT Y . What we see is that although its main effect is not significant, there are negative interaction effects with f airG and commG but with f airG the effect is much larger. The negative interaction with commG is clearly inconsistent with conjecture 4 and thus with the results of previous studies (e.g. de Vos et al., 2001) reported for models without mutation. At the same time, harshness is less of a problem for commitment players than for fairness players. While the net effect of commG remains positive (including the main effect), f airG no longer has a significant main effect. One interpretation is that whatever beneficial effect a preference for fairness may have, the effect is strongly mitigated when the environment is harsh. This is consistent with the explanation that Back and Flache (2006) gave for the weaker performance of fairness strategies in their experiments. Fairness players tend to avoid unbalanced exchange accounts by spreading their help requests across a large number of potential partners. However, the harsher the environment, the more likely it is that help requests are directed by multiple help seekers at the same target. Accordingly, in harsh environments fairness players are likely to lose more points than commitment players, who coordinate their requests in a more efficient way. Finally, in Model 3 we control for the interaction effects between being committed in both giving and asking and for being fair in both giving and asking. This helps us understand whether having consistent preferences for both giving and asking has an impact on the dependent variable. While for commitment the effect is relatively small but positive, the effect for fairness

3.5. Discussion and conclusions

67

is much larger and is in a negative direction. In addition, the interaction effect between the commitment traits slightly outweighs the negative effect of commA . This suggests that, in terms of viability, it is disadvantageous to be committed in giving but not in asking (or vice versa) but such a mismatch in fairness preferences is nonetheless beneficial. This intuitively makes sense: if you have friends you usually ask for help, you also want to give them help when they ask, and vice versa.

3.5

Discussion and conclusions

In this paper we examined the arms race between two tactics for cooperation under evolutionary pressures. One of them is conditional cooperation or reciprocity (cf. Axelrod, 1984), the other one is commitment. The simple idea behind conditional cooperation is this: “Be cooperative but retaliate against those who cheated on you before”. Since Axelrod (1984), conditions that may trigger retaliation (defection) have been refined and sophisticated, adapting conditional cooperation to various challenges, such as asymmetric uncertainty and random noise. In contrast with conditional cooperation, commitment is based on a very different idea: “Be generally cooperative but always favor long-term exchange partners”. Thus, the main question for commitment is not to decide whether to cooperate or defect but to select exchange partners. At first this seems to make commitment excessively cooperative, and vulnerable to exploitation. In a large enough interdependent population, however, partner selection substitutes the need for explicit punishment. Previous computational modeling work (Back and Flache, 2006; de Vos et al., 2001; Zeggelink et al., 2000; de Vos and Zeggelink, 1997) has pointed to the evolutionary advantages of commitment under conditions resembling the human ancestral environment. In this paper, we reported a stricter test of the underlying evolutionary explanation. Unlike previous work, our computational study allowed for the random mutation of competing strategies and thus implemented a much tougher evolutionary selection to which both fairness and commitment strategies were exposed. Under this stricter test, our results still point to certain evolutionary advantages of interpersonal commitment but the findings also highlight weaknesses of commitment and put results of earlier research into perspective. We found that under the postulated conditions of the ancestral environment, both traits for commitment and for fairness in giving increase viability, and as expected, helping old interaction partners (commitment) was more important than helping in a fair, reciprocal way. At the same time, we find that both tendencies for commitment and fairness have negative effects when it comes to seeking help. This suggests that previous studies may have overemphasized the evolutionary advantages of

68

Chapter 3. The Evolutionary Advantage of Commitment

commitment. We find that it is beneficial for agents to bind potential partners (commitment in helping) but it is disadvantageous to restrict search for help too much to these partners (commitment in seeking). In a similar vein, our results are also inconsistent with the argument of de Vos et al. (2001) that environmental harshness strengthens the effects of commitment. Instead, we find that the positive effects of commitment on survival weaken when the environment becomes harsher. However, we did find that commitment players are less affected by harshness than are fairness players. Our study has both supported and refined evolutionary accounts of interpersonal commitment. At the same time, this work has its limitations some of which we believe do not affect the central conclusions, while others point to the need for future research. Previous theoretical work may suggest, in particular, that the viability of commitment is seriously hampered by the “dyadic” nature of this strategy. Bendor and Swistak (2001) have shown that dyadic strategies (strategies that only sanction defections that cause harm to the sanctioner) can never be evolutionary stable, while “social strategies” that also sanction non-cooperation between third parties, are stable. In a nutshell, the reason is that social strategies leave no room for benefiting from second-order free-riding because a second order free-rider would be punished by every compliant group member. However, we believe that this is not a serious problem for our theory of (dyadic) commitment. While dyadic strategies are not eternally stable, it has also been shown that reciprocal dyadic strategies (including commitment) can be relatively more stable (but not perfectly stable) compared to non-reciprocal strategies. At the same time, Bendor and Swistak do not deny that social strategies impose a higher burden of cognitive complexity and information-gathering on agents than do dyadic strategies. To the extent that this creates fitness costs, the advantage of social strategies may turn into a disadvantage. Moreover, social strategies may be relatively more vulnerable to environmental uncertainty and noise, because “erroneous” defections may disrupt more relationships than just the dyad in which they occurred. In sum, while – consistently with our results – Bendor and Swistak’s argument implies that the dyadic strategy of commitment is not eternally stable, it is plausible to assume that at least under uncertainty conditions it also has some fitness advantages as compared to social strategies. A more obvious limitation of our work is the lack of a direct empirical test for the existence of a commitment trait in contemporary societies. To be sure, while we presented a theoretical argument for a preference for building committed relationships, this work was motivated by laboratory research that showed that commitment in exchange is positively related to uncertainty (Kollock, 1994). Moreover, it has been demonstrated, that people attach positive feelings to the mere existence of long-term exchange relationships, in addition to material benefits that result from them (Lawler and Yoon, 1993, 1996; Lawler, 2001). In a similar vein, Smaniotto (2004) showed using sce-

3.5. Discussion and conclusions

69

nario experiments that subjects are more willing to provide help and to report emotions of commitment, if a scenario provides “commitment cues” such as another person being in need, or being a friend. More recently, neurobiology is turning its interest to uncovering traits and mechanisms underlying human sociability and affiliation. Kosfeld et al. (2005) managed to artificially increase the level of trust, a key element in committed relationships, by administering oxytocin, a hormone that acts as a neurotransmitter in the brain, to participants of an experiment. Research on mammals suggests that social bonding can be modulated by various hormones including oxytocin, vasopressin, opioids, corticotropin releasing hormone, dopamine and adrenal steroids, including corticosterone or cortisol (cf. Carter, 2005). Even more to the point, Depue and Morrone-Strupinsky (2005) provide support for the existence of a neurobiological system in humans that regulates reward via opiate functioning when people create and dissolve social bonds. But all these findings do not give sufficient insight yet into the underlying mechanism or trait for interpersonal commitment. In particular they do not allow disentangling conclusively rational commitment and our indirect evolutionary explanation that posits “irrational” emotions as a proximate mechanism driving commitment. It is of course impossible to empirically test an ultimate (evolutionary) explanation for commitment. But future work should devise tests, such as laboratory experiments, that allow ruling out rival hypotheses derived from competing proximate explanations for commitment. In order to test whether a positive feeling, i.e. a preference for commitment, exists as a possible evolutionary remnant in contemporary populations, it is of key importance to find support for at least two hypotheses. The first is that the preference for commitment is stable across situations with varying materialistic payoffs, i.e. people behave according to the preference even when this is not in their rational self-interest. And secondly, that the preference is stable across different cultures. To conclude, there is reason to believe that humans may have been selected for some form of commitment behavior in their evolutionary past. One possible explanation for the success of commitment we offer is the following. Both conditional cooperation and commitment have a tendency to cooperate, which is the only recipe for success under conditions of high interdependence and uncertainty. They both have a method for excluding defectors from the bliss of cooperative interactions: conditional cooperation retaliates against defectors, while commitment leaves them for better partners. This means that commitment does not purposefully defect when it is able to help. Or from another perspective, while conditional cooperation operates by punishment, commitment operates by reward.

Chapter 4

Fairness and Commitment under Inequality1 Abstract Reciprocity has been a prominent explanation of cooperation in repeated exchange. However, empirical studies indicate that exchange partners are often much less intent on keeping the books balanced than theoretical works suggest. Using an agent-based ecological model, earlier work showed that in competitive environments commitment could be a more successful strategy than fair reciprocity. We now move beyond previous computational models in two respects. First, we extend them with the possibility of randomly mutating strategies under evolutionary pressures. In line with previous simpler models from evolutionary game theory, our results show the lack of evolutionary stable strategies but we also find that commitment strategies still outperform fair reciprocators on average. Our second extension introduces inequality in individual capabilities. We find that the existence of individual differences shifts the balance from commitment towards fair reciprocal strategies. Our explanation is that in a population characterized by inequality, strategies benefit from changing interaction partners from time to time because this gives a larger number of agents access to valuable interactions.

1 This chapter is a shortened version of Back, I. and Flache, A. (In press). Fairness, Commitment and Inequality. In: Hernández, C., Troitzsch, K. G. and Edmonds, B. (eds.). Social Simulation: Technologies, Advances and New Discoveries.

71

72

4.1

Chapter 4. Fairness and Commitment under Inequality

Introduction

The most prominent explanation of endogenous cooperation in durable relationships is reciprocity under a sufficiently long “shadow of the future” (Axelrod, 1984; Friedman, 1971). Briefly, these models show that even in a competitive environment with changing exchange partners, strategies that reciprocate cooperation with cooperation and defection with defection, such as the celebrated “Tit-for-Tat”, are far more successful than strategies that aim to exploit their opponents. Although the idea of Tit-for-Tat, and more generally reciprocity have been refined since the seminal work of Axelrod (1984), a range of (social) psychological studies casts doubt on the empirical validity of these strategies (see Chapters 1-3, and especially Baumeister and Leary, 1995; Nesse, 2001b). Recently, de Vos and collaborators (de Vos et al., 2001; Zeggelink et al., 2000; de Vos and Zeggelink, 1997) extended previous theoretical models with assumptions from evolutionary psychology (Cosmides, 1989; Cosmides and Tooby, 1993) and argued that at least one form of commitment is more viable than fair reciprocity. Back and Flache (2006, see Chapter 2) relaxed some of the most restrictive assumptions of the de Vos model and argued that strategies following some form of commitment outcompete fair reciprocating strategies. This study, however, still suffered from two key simplifications. First, it did not allow for the spontaneous emergence of “smart cheaters” who take advantage of the tolerance of commitment players. This problem has subsequently been addressed by Back and Flache (forthcoming, see Chapter 3). Second, individual differences were neglected between group members in terms of their attractiveness as exchange partners beyond their strategy. This is based on a key idea in interpersonal relationships research – that commitment to a partner depends on the quality of alternative partners (e.g. Rusbult and Martz, 1995). In this paper, we will relax both simplifications and thus put the evolutionary explanation of commitment to a better test. First, we extend previous studies with the possibility of randomly mutating strategies under evolutionary pressures. This model extension imposes the additional pressure of attacks from “smart cheaters” upon commitment players and, more generally, gives a much larger coverage of an infinite strategy space. At the same time, we introduce variation in individual capabilities. Earlier studies based on fairness strategies (Hegselmann, 1996, 1998; Flache and Hegselmann, 1999b,a; Flache, 2001) suggested that individual differences might give rise to core-periphery network structures in which the strongest population members exchange help with each other, driving weaker actors to the margin of the network. However, these studies did not include commitment strategies. Intuitively, we may expect that commitment strategies reduce the exclusion of weak members from exchange networks, in contrast

4.2. Model

73

to what has been found for fair reciprocating strategies, because commitment strategies tend to bind themselves to “old helping partners” irrespective of the balance of help exchanges with those partners. Our evolutionary framework allows us to test this intuition and also to assess to what degree the exclusion of weak members, found in previous analyses, depends upon the assumption of fairness. In Section 4.2, we motivate and describe our model. In Section 4.3, we formulate conjectures based on previous work that are testable with this model. In Section 4.4, we report the results of our computational experiments. Finally, Section 4.5 contains conclusions and a discussion of our findings.

4.2

Model

To improve earlier agent-based models, we rely on an abstract decision situation that has previously been used to study commitment and reciprocity, the Delayed Exchange Dilemma (DED). For the detailed description of this dilemma, consult Chapter 3. To implement individual differences between agents, we introduce random variation in agents’ effectiveness to help. To do this we assign a fixed, random capability value from the [0,1] interval to each agent that describes how much help the agent is able to provide to help-seekers. More precisely, the amount of fitness loss for a distressed agent ai when help is provided by agent aj is equal to fd · (1 − αj ), where αj is the helping capability parameter of agent aj . This implementation of individual differences disentangles two substantively different aspects of inequality that were not separated by Hegselmann (1996, 1998). These earlier studies assumed that individual strength both increases an actor’s capacity to give help and reduces its need for getting help. However, in our current analysis we are mainly interested in the evolutionary viability of commitment behavior. Accordingly, we wish to keep the modeling changes relatively small to previous work in this area (e.g. Back and Flache, 2006; de Vos et al., 2001) and thus include only one source of individual variation in helping capacity. We assume moreover that helping capability, unlike social preferences, is not an inheritable property of agents, but a purely phenotypic one. An agent’s helping capability, α, is randomly assigned at the time of birth and is unrelated to its ancestors’ α value. The reason we make this assumption is that we want to disentangle evolutionary pressures that operate directly on agents’ capacity to help, from those pressures that operate on social preferences for discriminating between strong and weak exchange partners. We are here interested in the latter and thus exclude selective pressures on inheritable helping capability.

74

Chapter 4. Fairness and Commitment under Inequality

4.2.1

Modeling strategies

To explicitly study the evolutionary viability of social preferences, we assume that agents have four types of preferences that determine their helping and asking behavior. The first preference is to be generally cooperative, help whenever asked for help. The second preference is to be fair: do not help or ask agent a1 more often than agent a1 has helped or asked you to help. In other words, fairness introduces conditionality into cooperativeness: agent a1 who was not cooperative enough toward a fair agent a2 in the past is not helped by agent a2 when it asks (retaliation) and is avoided by agent a2 when agent a2 seeks help (exit). The third preference is to build a long-term relationship, to interact with the same agent repeatedly (commitment). Note that this preference takes into account the absolute number of interactions between agent a1 and a2 , and not the relative proportions of cooperative and defective interactions, as in the case of fairness. Finally, the last preference is to interact with the most capable others. Each agent has a combination of hypothetical genes that describe how strong their preferences are along these four dimensions. Whenever an agent ai has to make a choice between a number of other agents, these preferences determine which other agent, if any, agent ai will interact with. The genes are inherited and are subject to mutation2 from generation to generation. This leads to huge variation in possible preferences, and thus in behavior over time. Agents face two different types of decision situations repeatedly in the DED. When they are hit by distress, they have to select an interaction partner to ask help from. On the other hand, when they themselves are asked to provide help they need to decide whether to provide it and, in case of multiple requests, whom to provide it to. In both cases, agents order possible interaction partners according to the overall attractiveness of interacting with them. Attractiveness is based on the individual preferences agents have with regard to past interaction histories. The attractiveness of aj for giving help to, calculated by agent ai , is formalized as: G G G Uij = commG i · IN T F REQij + f airi · IN T BALij + capai · αj + coopi , G where commG i is the preference for commitment in giving, f airi is the preference for fairness in giving, capai is the preference for giving to the most capable others (αj is the help-seeker’s capability here), and coopi is the preference for general cooperativeness. IN T F REQij is the proportion of cooperative interactions3 ai had with aj compared to the total number of cooperative in2 Mutation

in this case can be seen as either genetic or cultural. take place always between exactly two agents. Possible interactions are giving help (cooperation) and refusing to help (defection). Asking for help is always followed by one of these. 3 Interactions

4.2. Model

75

teractions ai had. IN T BALij is the standardized interaction balance between agents ai and aj . To obtain this measure, we took into account both the balance of helps and the balance of refusals. The reason is that neither help balance nor refusal balance alone is sufficient to guarantee an overall balance in the exchange relationship. For example, suppose ai helped aj equally often as aj helped ai but ai refused to help aj ten times more often than aj refused to help ai . Despite the equal amount of help given, this exchange relationship clearly cannot be considered perfectly in balance. Technically, we calculated the measure as follows. We subtracted from 1.0 a measure of the overall standardized imbalance. The overall standardized imbalance is obtained by adding the difference of the number of times ai received help from aj and ai gave help to aj , and the difference of the number of times ai refused to give help to aj and aj refused to give help to ai , and dividing this by the total number of interactions they had. This model disentangles two different aspects of fairness that were not separated by previous studies. The interaction balance expresses solely the balance of the frequency with which help was given and help was received, but it does not take into account the effectiveness of that help. Accordingly, a preference for fair exchanges does not discriminate against weak partners (in terms of capacity). However, discrimination against the weak occurs when an agent has a preference for partners with high capability. Other studies of Hegselmann and Flache did not make this distinction and assumed instead that agents rated the attractiveness of their exchange partners solely based on the balance of help amounts that were effectively given and received. In the actual implementation of our model, every time an agent has to make a decision, there is also a probability Pe that the agent will not use the above utility calculation but will choose randomly from the set of available decisions, each being equally likely. This random error models noise in communication, misperception of the situation or simply miscalculation of the utility by the agent. Taking this random error into account increases the robustness of our results to noise. The attractiveness of agent aj for asking help from is defined in a similar way, the difference is that agents may put different weights on the two historyspecific decision parameters, and that there is no general cooperativeness parameter4 : A A A Uij = commA i · IN T F REQij + f airi · IN T BALij + capai · αj ,

Note also that in this case, αj denotes the capability parameter of the donor. Before agents make a decision, be it help giving or help asking, they calculate the corresponding one of these two types of attractiveness respectively for each agent who asked help (U G ), or for each other agent in the population 4A

general cooperativeness parameter would not make sense in the case of asking for help.

76

Chapter 4. Fairness and Commitment under Inequality

(U A ). In case of help giving, they choose a partner with the highest attractiveness, if that attractiveness is above an agent-specific threshold uti . Notice that IN T F REQij and IN T BALij are always smaller than or equal to one in absolute value. We allow commi , f airi , capai and coopi to take values from [−1; 1]. Thus we allow the attractiveness threshold to take values from [−4; 4]. If the attractiveness of all possible agents is below the threshold attractiveness, no help is given to anyone. Otherwise, if there is more than one other agent with highest attractiveness5 , the agent selects one of the others with equal probability. In the case of help seeking, agents also choose a partner with the highest attractiveness but there is no threshold, i.e. agents in distress always ask someone for help. Definition 9 (Strategy). A strategy is a combination of five genes for help giving behavior (commG , f airG , capaG , coop, ut ) and three genes for help asking behavior (commA , f airA , capaA ).

4.2.2

Evolutionary dynamic

The heart of our model is an evolutionary dynamic that captures random mutation of strategies and selection of objectively successful ones. The implementation of this process is based on the replicator dynamics (Taylor and Jonker, 1978). Broadly, the replicator dynamics dictates that if a generation of genotypes (strategies) undergoes reproduction, the net reproduction rate of a genotype is proportional to its relative success compared to other genotypes in the current generation. Genotypes which perform below average, in particular, have a negative reproduction rate. In our case, genotypes (strategies) represent subjective preferences. Note that while the capa gene randomly mutates from generation to generation, the α parameter of an agent is not subject to mutation. To prevent a population from growing without bounds, thus modeling resource scarcity in an implicit way, we keep the size of the population constant, in the following way. At the end of each round we count how many agents have died and replace them with new agents in the next round. Each new agent ai has the same strategy as a randomly selected other agent aj , present in the population who has reached a minimum age n (measured in the number of interactions it had). The probability of choosing this other agent aj is proportionate to the share of points aj holds within the group of all agents older than n. Before ai is added to the population, with probability Pmut , its strategy may undergo mutation. A mutation occurs in exactly one, randomly chosen gene, with equal probabilities for all genes, thus P = 91 for each gene. The new value of the gene is a uniformly distributed random value from the 5 This

is unlikely, as the preference parameters are high-precision real values and interaction histories tend to differ with time.

4.3. Conjectures

77

interval [−4; 4] for the attractiveness threshold, and from [−1; 1] for all other genes.

4.3

Conjectures

To guide the simulation experiments, in the following we formulate a number of conjectures derived from previous work. Definition 10 (Stability). Stability of a strategy s is equal to the number of consecutive rounds s existed in a population in a given simulation run, counting from the first round in which it appeared until the round in which it became extinct. A strategy s is infinitely stable if it does not become extinct. Generalizing from analytical results about the evolutionary stability of strategies in repeated games that are simpler than the DED (cf. Bendor and Swistak, 2001), we expect that there is no single strategy that is superior to all others in the dilemma we study. In other words, for every incumbent strategy there exists another (mutant) strategy that can take advantage of the incumbent’s weakness. Conjecture 4.1. There is no infinitely stable strategy in an infinitely played game of DED. Nevertheless, the length of time a strategy exists (stability) carries an important message about its viability. Since mutations constantly arise and threaten to push other strategies out of the population, stability is an indicator for the number of attacks a strategy could withstand. Therefore, stability of a strategy will be one of the indicators of its viability6 . The other measure is typical longevity within a strategy (variable LONGEVITY, the average age at death of agents belonging to a strategy). Note that in our model, there is no upper age limit on reproducibility, in other words, agents keep reproducing until they die, which makes LONGEVITY a suitable measure for viability. Back and Flache (2006) found that the most successful strategies in the DED exhibited some level of interpersonal commitment and that committed agents outcompeted fair reciprocators. These results suggest the following conjecture, which we will test under the new assumptions of random mutation and individual differences: Conjecture 4.2. Individual preferences for interpersonal commitment and fairness have a positive effect on the viability of a strategy. 6 We will not use here stability concepts from the evolutionary game theory literature (e.g. evolutionary stability or asymptotic stability) because they do not allow the expression of the relative stability of strategies.

78

Chapter 4. Fairness and Commitment under Inequality

Earlier studies based on fairness strategies (Hegselmann, 1996; Flache and Hegselmann, 1999b) suggested that individual differences in capabilities might give rise to core-periphery network structures in which the strongest population members exchange help with each other, driving weaker actors to the margin of the network. If agents are selected for having a social preference for exchange with partners with high capability, we would expect that the core-periphery pattern of networks also characterizes our setting. If everybody maintains such a preference then matches between highly capable partners are more easily formed than asymmetrical dyads, which in turn forces less capable agents to exchange with each other. However, we believe that the evolutionary advantages of discrimination will actually be relatively small in our framework. The key reason is that we focus on differences in capability that are unrelated to the genetic makeup of agents. In such a world, agents who discriminate against the weak would harm their genetically related but weak “siblings” just as much as they would benefit genetically unrelated but strong others. Accordingly, there is little reason to expect in our setting that there are strong and systematic reproductive advantages from a social preference for exchange with strong partners. This leads to a further conjecture: Conjecture 4.3. Preferences for commitment and fairness are more important for the viability of the strategy than a preference for capability. While we expect that both commitment and fairness tend to preclude exclusion of the weak, we also think that there will be important differences between these two strategy types in terms of their viability in a heterogeneous population. If there are non-heritable differences in capabilities, the fact that commitment segregates the population into closed cliques (especially dyads) could produce an inefficiency on the population level. This is because the superior helping potential of the most capable agents will only benefit their fixed interaction partners and not the entire population. In the case of fairness, however, such superior helpers will distribute their helping potential equally among multiple members of the population. As the evolutionary selection pressures in our model tend to produce populations that are relatively homogeneous in genotypes, behavioral patterns that benefit the population as a whole also tend to benefit the incumbent strategy. Hence, we posit a fourth conjecture: Conjecture 4.4. Variation in capability reduces the degree to which preferences for commitment foster viability and increases the degree to which preferences for fairness sustain viability. If there are non-heritable differences in capabilities, the fact that commitment segregates the population into closed cliques (especially dyads) could produce an inefficiency on the population level. This is because the superior

4.4. Results

79

helping potential of the most capable agents will only benefit their fixed interaction partners and not the entire population. In the case of fairness, however, such superior helpers will distribute their helping potential equally among multiple members of the population.

4.4

Results

To assess the effects of variation in individual capacities, we conducted two sets of experiments. In the first set of experiments, we assumed that all agents in the population have an equal helping capability. Correspondingly, we used in this set of experiments strategies without the two traits capaG and capaA . In the second set of experiments, individual variation in capability was introduced as described above. To avoid our results becoming biased by a restrictive set of starting conditions (cf. Binmore, 1998), in both experiments we run more than a hundred replications, each time with a population whose initial strategy is randomly chosen from the strategy space defined by the eight genes. We did not find any significant effects of features of initial starting strategies on the outcomes of simulation runs. The reason is that soon after the initial rounds of a simulation run, mutation ensured the emergence of a large variety of different strategies in the population. We allowed such a population to play the DED game. In the course of the game agents started to lose points, some of them eventually died, while others reproduced. At some point, random mutations occurred in the initial strategy, creating a potential invader. The better a mutated strategy performs in the DED compared to agents of the original strategy, the larger is its probability of reproducing and increasing its proportion within the agent population. The simulation run ended with either the extinction of all agents7 or after an arbitrarily chosen large number of rounds (10 million). We then repeated the simulation run with another, randomly generated initial population. After results from a large number of runs were acquired, we proceeded to statistically test our conjectures. During each simulation run we record all strategies and their key characteristics that have ever appeared through random mutations. These characteristics include on the individual level the genes describing a strategy (commG , f airG , coop, ut , commA , f airA and in the second set of simulations also capaG , capaA ) and on the strategy level average longevity measured in rounds of game play. 7 Extinction

is possible if all agents die within one round and thus there is no basis for the distribution of strategies in the next generation.

80

4.4.1

Chapter 4. Fairness and Commitment under Inequality

Initial parameters

To preserve the comparability of our results we started our simulations with the same initial parameters (where applicable) that were used in earlier work. These are Pd = 0.2, fh = 1, fd = 20, fi = 100, fc = 0, N = 25, Pe = 0.05, m = 2. (For the meaning of each parameter, please consult the Model section above.) These parameters impose a set of conditions under which for strictly instrumental agents the choice between purposeful defection and cooperation is as difficult as possible (for game theoretical derivation see Appendix A). We refer to this parameter setting as the baseline condition. After obtaining results for the baseline condition, we conduct additional simulation runs in which we systematically vary all interesting parameters of the model. We then add results from these runs to the dataset obtained in the baseline condition.

4.4.2

Stability

In support of Conjecture 4.1, our simulation results show for both sets of experiments that strategies change endlessly in all initial parameter setting – we found no infinitely stable strategy in the DED. We simulated 300 runs altogether, each of which started with a different randomly chosen initial strategy and consisted of maximally 10,000,000 rounds. During these runs more than ten million mutations took place altogether, generating almost as many strategies. However, in none of these runs have we recorded any strategy that existed longer than 500,000 rounds, with the majority of strategies lasting only a few hundreds of rounds. We may of course simply have not encountered the infinitely stable strategies during our random walks in this vast strategy space. However, judging by the vast coverage of the strategy space by our method, this seems implausible. A plausible explanation for the lack of infinitely stable strategies is that for each strategy there exists a better response that takes advantage of the strategy’s weakness. Sooner or later mutations generate this better response and the original strategy is gradually pushed out of existence. If a strategy is too cooperative, opportunistic exploiters take advantage of this and flourish. Later, in a harsh world of mainly exploiters, where everybody is suffering, two cooperators who appear randomly at the same time and find each other will survive and reproduce more easily than others, given that they have a sound method of excluding defectors from cooperative interactions. That cooperativeness eventually declines again may be explained by the gradual loss of the ability to exclude defectors through “evolutionary drift” (e.g. Bendor and Swistak, 2001) or by the emergence of new defecting mutants who have developed the ability to behave such that they are not excluded from exchanges between incumbent cooperators.

4.4. Results

4.4.3

81

The importance of interpersonal commitment

To test Conjectures 4.2 and 4.3 we needed to get an insight into the separate contributions of the genes to a strategy’s success. To compare the relative importance of the genes, we conducted a linear regression analysis with average longevity as the dependent variable. Model A below (see Table 4.1) is based on the dataset obtained in the first set of simulation experiments, conducted without variation in helping capability. Models B1 and B2 were created for the second set of experiments, which contained individual differences in capability. Before performing the regression analysis we filtered out highly unstable strategies (stability<500 rounds). The reason to filter out highly unstable strategies is that due to the stochastic nature of the simulation, it often happens that strategies that would otherwise be more stable cannot grow to a critical mass in the population to stabilize. In this case, they distort the association between features of the strategy and its viability. In Model A (N=34,143) we see that the effect of having preferences for commitment and for fairness in giving are both positive but the effect of the commitment preference is larger, as expected based on Conjecture 4.2. We also see that the cooperation preference has a very large effect. Preferences in asking are negative but for fairness the coefficient is much larger in absolute value. This suggests that strategies that restrict their partner search too much either to old partners or to partners with balanced exchange ratios are disadvantaged because their search space is overly reduced. The attractiveness threshold ut has a very large negative effect. The explanation is that ut is very important in deciding whether a strategy is initially cooperating or defecting (niceness). This in turn is crucial for the ability to bind future helping partners or establish mutually cooperative balanced exchange relationships.

4.4.4

The relative importance of fairness, commitment and capability

To assess how variation in individual capability affects the relative importance of fairness and commitment, we conducted a second set of simulation experiments (corresponding to Models B1 and B2 on Table 4.1, N=344,778) in which all conditions were kept equal to the first set, except that we now assumed variation in α and, accordingly, included the traits of capaA and capaG in the genetic makeup of the agents. After comparing populations where agents are equal in their helping capability (Model A) with populations characterized by inequality (Models B1 and B2), we find a remarkable difference in the relative importance of social preferences. For inequal (heterogeneous) populations, the effects of being fair (f airG ) and cooperative (coop) when giving are the largest, followed by being

82

Chapter 4. Fairness and Commitment under Inequality

committed (commG ), whereas the effect of commitment was stronger than the one of fairness in equal (homogeneous) populations. Model B1 shows this difference for a ceteris paribus replication of the regression analyses we conducted for the homogeneous case. Model B2 demonstrates that the difference is robust even when additional effects are included in the regression. The fact that fairness is more important under inequality than commitment both supports our Conjecture 4.4 and contradicts earlier findings (Back and Flache, 2006). A possible explanation is the following. In a group where capabilities are genetically determined and inheritable, it makes sense to help those who are strongest, as this will ensure the survival and continuous improvement of the group’s genetical pool and thus its survival. If capabilities are not inherited, however, a group with widespread preferences (or norms) for helping weak members as well will essentially fare better. Both commitment and fairness are such preferences. Commitment leads to stable friendships, in which capabilities do not matter, while fairness leads to uniformly equal shares of help among all members. The disadvantage of commitment comes from the fact that once a strong member binds itself to another agent, its superior capability to provide help will no longer benefit the rest of the group, whereas in the case of fairness, this superior agent will distribute its help equally among all members of the group. Interestingly, this suggests that the relative disadvantage of commitment vis-à-vis fairness in a heterogeneous population is based on the same mechanism that Back and Flache (2006) identified as the explanation of the success of commitment in a homogeneous population. In an homogeneous population, the tendency of fairness to change partners inevitably produces a higher rate of collision of help requests than commitment would produce. Accordingly, fairness strategies are more likely to run out of help givers and suffer fitness loss. But in an heterogeneous population this turns into an advantage: in a population dominated by fairness strategies it is more likely that needy agents seek help from sources with different levels of capability. This tendency, in turn, improves the efficiency of the distribution of the overall capacities for giving help that are present in the population.

(Constant) commG f airG coop commA f airA ut capaG capaA

Table 4.1: longevity

R .670

97.035** 1.754** 1.393** 5.625** -0.754** -4.340** -6.798**

R .158

53.793** 3.016** 6.217** 4.843** 1.740** -1.077** -6.442**

B

R2 -adj. .158

(0.068) (0.089) (0.090) (0.089) (0.089) (0.090) (0.029)

Std. Error

R .160

53.805** 2.926** 6.179** 4.792** 1.651** -1.056** -6.423** 1.031** 2.426**

B

R2 -adj. .160

(0.068) (0.089) (0.090) (0.089) (0.089) (0.089) (0.029) (0.090) (0.090)

Std. Error

Unstandardized Coefficients

Model B2 (inequality)

OLS regression models with dependent variable

R2 -adj. .448

(0.088) (0.128) (0.134) (0.126) (0.129) (0.135) (0.051)

Std. Error

Unstandardized Coefficients

Unstandardized Coefficients B

Model B1 (inequality)

Model A (equality)

4.4. Results 83

84

Chapter 4. Fairness and Commitment under Inequality

Regression results for the dataset with inequality in capability also lend support to Conjecture 4.3. The importance of giving to the most capable partner (capaG ) for longevity is substantially smaller than the importance of commitment and fairness. This points to a weakness of this preference. Since the helping capability of an agent i (αi ) benefits every other agent equally, it creates a unanimous ordering among agents, with regard to their attractiveness as helping partners. This naturally leads to collisions of interaction requests and further hampers the efficiency of the capa traits. Note that Commitment and Fairness do not suffer from this problem: a committed agent almost always has exactly one best friend, who will most likely be different from any other agent’s best friend, whereas a fair agent’s interaction balances are likely to be randomized. The picture is somewhat different for asking. The effect of being fair on longevity when searching for help (f airA ) is negative, suggesting that it is harmful to distribute help requests in a uniform manner. The effect of commitment (commA ) is only marginally significant, and the difference between fairness and commitment is probably due to the coordination advantage of commitment over fairness (Back and Flache, 2006). Most beneficial here is (capaA ), showing that it is helpful to turn to the most capable helpers. However, the importance of this preference is still far below that of fairness in giving (f airG ).

4.4.5

Sensitivity to initial parameters

To test our results for sensitivity to the choice of initial parameters, we repeated our simulations for a range of parameters. We then added results from these additional simulation runs to the dataset thus obtaining variation in all theoretically interesting variables. The results presented above are based on this enlarged dataset. For the exact parameter ranges, please consult the Appendix D. One interesting parameter to vary was environmental harshness (fd , the cost of not getting help). We reran the simulation with fd = 5, 10, (20), 30 and added these repetitions to the original dataset obtained in the baseline condition. Then we tested for significant effects of fd on the dependent variable longevity. What we saw was that the main effect of fd is highly negative – when the costs of not getting help are high, agents live shorter lives. What is more interesting is that the effects of the fairness and commitment genes are amplified, and the most important factor in asking for help becomes not capability but commitment, when controlling for harshness of the environment. This is consistent with the findings of de Vos et al. (2001) and also suggests that fairness but especially commitment is more robust to changes in the harshness of the environment than the preference for capability.

4.5. Discussion and Conclusion

4.5

85

Discussion and Conclusion

In this paper we examined the evolutionary arms race between three social preferences. These preferences are fairness or reciprocity (cf. Axelrod, 1984); commitment (cf. de Vos et al., 2001); and the preference for capability. The simple idea behind conditional cooperation is this: “Be cooperative but retaliate against those who cheated on you before”. Since Axelrod (1984), conditions that may trigger retaliation (defection) have been refined and sophisticated, adapting conditional cooperation to various challenges, such as asymmetric uncertainty and random noise. In contrast with conditional cooperation, commitment is based on a very different idea: “Be generally cooperative but always favor long-term exchange partners”. Thus, the main question for commitment is not to decide whether to cooperate or defect but to select exchange partners. At first this seems to make commitment excessively cooperative, and vulnerable to exploitation. In a large enough interdependent population, however, partner selection substitutes the need for explicit punishment. The preference for capability is an elitist striving. Its basic message is: “Try to interact with partners who have the highest capability to provide help.” Such a preference makes sense under widespread individual inequality in capabilities. Obviously, this preference only comes into play when there is individual variation in capability. Previous computational modeling studies of the evolution of commitment – including our own work – have neglected individual differences. In this paper, we have relaxed this restriction. The current study moves beyond previous research also in another important way. Previous work (Back and Flache, 2006; de Vos et al., 2001; Zeggelink et al., 2000; de Vos and Zeggelink, 1997) has pointed to the evolutionary advantages of commitment under conditions resembling the human ancestral environment. In this paper, we reported a stricter test of the underlying evolutionary explanation. Unlike previous work, our computational study allowed for the random mutation of competing strategies and thus implemented a much tougher evolutionary selection to which both fairness and commitment strategies were exposed. Under this stricter test, our results for homogeneous populations still point to certain evolutionary advantages of interpersonal commitment but the findings also highlight weaknesses of commitment and put the results of earlier research into perspective. We found that under the postulated conditions of the ancestral environment, both traits for commitment and for fairness in giving increase viability, and as expected, helping old interaction partners (commitment) was more important than helping in a fair, reciprocal way. At the same time, we find that both tendencies for commitment and fairness have negative effects when it comes to seeking help. This suggests that previous studies may have overem-

86

Chapter 4. Fairness and Commitment under Inequality

phasized the evolutionary advantages of commitment. We find that it is beneficial for agents to bind potential partners (commitment in helping) but it is disadvantageous to restrict search for help too much to these partners (commitment in seeking). This overall advantage of commitment in the evolutionary competition with fairness disappeared, when we relaxed the assumption of homogeneity in capability. Contrary to previous work, we find that fairness is more important than commitment, once heterogeneity is allowed for. The explanation we suggest is that fairness allows a better-coordinated spread of the benefit offered by the most capable helpers, which leads to better collective outcomes. Interestingly, a more efficient coordination of help requests is exactly the reason that Back and Flache (2006) identified as a key advantage of commitment in homogeneous populations. But in heterogeneous populations, commitment’s advantage in avoiding colliding help requests comes together with a more important disadvantage, the exclusion of most population members from the benefit of being helped by the strongest players at least once in a while. Our results also offer an interesting refinement of Hegselmann’s work (see Hegselmann, 1996, 1998) (cf. Flache and Hegselmann, 1999b,a; Flache, 2001) on the dynamics of help exchange networks between unequal agents. This line of work assumed that agents follow fairness strategies that generate a balance of the effective amount of help that agents get and receive from their partners. Those studies showed that on the collective level, the consequence is a core-periphery structure in which strong players form the core of exchange networks and weak players are forced into the margin. However, our analysis questions whether in an evolutionary competition under the assumptions of our stylized ancestral environment, the kind of fairness strategies that produce this pattern would have emerged in the first place. We found that preferences for both fairness and commitment had a larger positive effect on the viability of strategies, than preferences for exchange with highly capable partners. The key difference to Hegselmann et al.’s results is that our notion of fairness does not include discrimination against partners with low capability, it only refers to a balance of the number of times help was given and received, regardless of its effectiveness. Accordingly, our model suggests that despite the prevalence of fairness traits in the evolved social preferences of agents, emergent exchange networks would not or at best weakly exhibit a core-periphery that excludes weak agents from exchanges with strong agents. Our study has both supported and refined evolutionary accounts of interpersonal commitment, as well as studies of the effects of heterogeneity on help exchange networks. At the same time, this work has its potential limitations that point to a need for future research. On the theoretical side, one of the most interesting extentsions to explore in future research is a combination of two sources of variation in helping capa-

4.5. Discussion and Conclusion

87

bility – genetic sources and non-heritable sources. We deliberately excluded heritable differences in the present study, which allowed us to highlight the deficiencies of strategies that discriminate against weak members of the population. However, this picture may change when capability is heritable. When agents vary in heritable capability, those who conditionally discriminate between interaction partners based upon their own strength may become more viable. Suppose, for example, that a genotype is more capable than average and, in addition, it has the social preference to exchange with capable others. Then, on average, this genotype may exchange more with its relatives (who are also stronger than average) than with those who are not genetically related. Hence, genetically capable strategies can be expected to have an advantage but only if they combine this with discrimination against the weak. We believe that testing this intuition in further computational experiments may provide a fruitful avenue towards establishing a theoretical relationship between social preferences and genetic traits of genotypes. A more obvious limitation of our work is the lack of a direct empirical test. To be sure, while we presented a theoretical argument for a preference for building committed relationships, this work was motivated by laboratory research that showed that commitment in exchange is positively related to uncertainty (Kollock, 1994). It was also demonstrated before, that people attach positive feelings to the mere existence of long-term exchange relationships, in addition to material benefits that result from them (Lawler and Yoon, 1993, 1996; Lawler, 2001). More recently, Smaniotto (2004) showed using scenario experiments that subjects are more willing to provide help and to report emotions of commitment, if a scenario provides “commitment cues” such as another person being in need, or being a friend. But all these results do not give sufficient insight yet into the underlying mechanisms, in particular they do not allow disentanglement of conclusively rational commitment and our indirect evolutionary explanation that posits “irrational” emotions as a proximate mechanism driving commitment. Future work should devise laboratory experiments that allow to distinguish these competing explanations of commitment. In order to test whether a positive feeling, i.e. a preference for commitment, exists as a possible evolutionary remnant in contemporary populations, it is of key importance to find support for at least two hypotheses. The first is that the preference for commitment is stable across situations with varying materialistic payoffs, i.e. people behave according to the preference even when this is not in their rational self-interest, and the second is that the preference is stable across different cultures.

Part II

Proximate explanations

89

Chapter 5

Commitment Bias: A tendency to stay with our partners beyond instrumental reasons12 Abstract

There is mounting evidence that humans have an evolved cognitiveemotional system, designed to facilitate the formation and maintenance of long-term interpersonal relationships. However, empirical research that tests corresponding proximate mechanisms is still scarce. Here we propose one such mechanism, termed commitment bias, which is hypothesized to increase people’s willingness to hold on to existing partners beyond instrumental 1 The material of this chapter is based on Back, I. and Smaniotto, R. (2006). The Commitment Bias. Presented at the 18th annual conference of the Human Behavior and Evolution Society (HBES), Philadelphia, PA, USA, July 7-11, 2006. 2 We would like to thank Jurre van den Berg (University of Groningen) and Vincent Buskens (Utrecht University) for their help in conducting experiments in the Netherlands; Xu Longshun (Fudan University), Fan Xuejuan (East China Normal University), and Ji Wenxi for their invaluable help in running experiments in China; Michael Macy (Cornell University) and David Sloan Wilson (Binghamton University) for their insightful comments on the results and their help in conducting experiments in the USA. Further thanks go to participants of the Siena/Vidi discussion group at the University of Groningen; and finally Andreas Flache, Tom Snijders and Henk de Vos. This research was made possible by an Ubbo Emmius grant from the University of Groningen, and a Reisbeurs grant from the Netherlands Scientific Organization (NWO).

91

92

Chapter 5. Commitment Bias

reasons even in anonymous economic settings. To exclude the possibility that the commitment bias is simply an artifact of other already known mechanisms that would not necessitate an evolutionary argument, we distinguish the commitment bias from uncertainty reduction and positive affect. Results from a cross-culturally replicated (Netherlands, USA and China) laboratory experiment provide support for our hypotheses. We conclude by assessing existing evidence for the ultimate explanation, and discussing possible evolutionary trajectories leading up to the development of the proximate mechanism.

5.1

Introduction

Building and maintaining long-term social relationships is a natural part of human life, and a behavior observed universally in all societies (Baumeister and Leary, 1995). People create friendships, romantic relationships, and looser acquaintanceships with great ease but are often reluctant to dissolve them, even after they have turned sour. Why do people hold on to long-term relationships when this is apparently not in their best interest? According to rational explanations of long-term relationships, people engage in repeated interactions when the benefits of a relationship outweigh its maintenance and alternative costs. On the one hand, having a long-term relationship with the same partner provides first-hand knowledge about the past behavior and thus the trustworthiness of the partner (shadow of the past). This makes a previous partner relatively more attractive than an unknown individual. On the other hand, the prospect of recurring benefits from future mutual cooperation creates an incentive for both partners to cooperate in the present, knowing that non-cooperation could trigger retaliation by the partner and thus jeopardize future payoffs (shadow of the future, see Axelrod, 1984). Although rational choice theories have been successful in explaining and predicting behavior in materialistic exchange relationships, their explanatory power outside this narrow context remains weak. With regard to the shadow of the past, we know, for example, that people often keep relationships even after their partner has proved to be untrustworthy. An extreme case is that of battered wives and girlfriends (Roy, 1977; Strube, 1988; Rusbult and Martz, 1995). It has been observed that many women who have been battered run away from their violent partners at first, but as soon as their wounds heal

5.1. Introduction

93

they return and face further abuse. A host of explanations have been proposed, ranging from utility derived from masochistic tendencies to a lack of alternatives. As for the shadow of the future, we know that people care for their longterm partners even when these partners are surely unable to reciprocate help in the future. This is most striking in the case of terminally ill patients, who suffer from Alzheimer’s disease and slowly lose their ability even to recognize their caregiver. Monahan and Hooker (1997) compared people who give care for their spouses with either of two different types of conditions, Parkinson’s disease and Alzheimer’s disease. Whereas the first is mainly a physical decline, the latter is characterized by the loss of cognitive functions, such as memory. One relevant difference between the two groups was the age of the caregiver and the age of the patient, both being higher for the Alzheimer’s group, pointing to the importance of years spent together, even when controlling for other characteristics of the partners and the relationship. It seems almost tautological that interpersonal relationships are governed by a complex mix of emotions and reason, and not simply by instrumental considerations (cf. Baumeister and Leary, 1995). Indeed, a great wealth of empirical evidence suggests that people have a tendency to become emotionally attached to each other, irrespective of materialistic considerations and other ulterior motives. People create social relationships with great ease, and strongly resist the dissolution of these relationships, well beyond rational considerations of practical advantage. The evidence seems sufficiently broad and consistent to suggest that human emotions evolved specifically in order to facilitate the formation and maintenance of social bonds (Baumeister and Leary, 1995). We know that evolutionary adaptation sometimes plays tricks on our senses. According to Error Management Theory (EMT), many cognitive biases that lead to seemingly irrational behavior evolved when the costs of false-positive and false-negative errors were asymmetrical over evolutionary history (Haselton and Buss, 2000; Haselton and Nettle, 2006). In such cases the cognitive bias forced people to avoid large dangers at the cost of unnecessarily avoiding non-dangerous situations as well. An example is the way men often mistakenly overestimate of women’s sexual intent: rejection is less costly in evolutionary terms than letting a good chance for mating slip away. Interestingly, something very similar was found with regard to interpersonal commitments before: Johnson and Rusbult (1989) noticed that when people are in committed relationships, they systematically undervalue alternative partners, in relation to the strength of their current commitment. The researchers followed committed couples over a period of 7 months and found that those who stayed with their partner gave gradually decreasing evaluations of alternatives, while those who left their steady partner gave increasingly positive ones. Johnson and Rusbult also found that the tendency to reject

94

Chapter 5. Commitment Bias

and devalue alternatives was most prominent under conditions of high threat to the committed relationship, and that devaluation was linked more strongly to commitment than to satisfaction. These empirical findings seem only loosely related at first but in fact they can all be derived from an ultimate explanation: an evolved trait for interpersonal commitment. Ancestral humans, like other primates lived in small groups of extended kin. However, as the size of such groups started to increase and joint activity provided increasing benefits, cooperation among nonkin became increasingly important (cf. Smaniotto, 2004). Those who were cognitively better equipped to create and maintain a larger number of relationships fared better, both in terms of survival and avoiding danger, as well as in acquiring suitable mates. A series of recent simulation studies explore the possibility of the evolution of a commitment trait (de Vos et al., 2001; Smaniotto, 2004; Back and Flache, 2006, 2007, forthcoming). These computational models are built on minimalistic assumptions about conditions of the ancestral environment, such as relatively small group size, food shortage and increased need for mutual helping. What these models consistently show is that when compared with strongly selfish or other conditionally cooperative types, interpersonal commitment emerges as the most viable strategy under a wide range of environmental conditions. The role of a consistent cognitive-emotional framework put into place by natural selection in order to facilitate and stabilize long-term interpersonal relationships is now slowly being recognized across the disciplines: in psychology (Baumeister and Leary, 1995), close relationships research (Cassidy and Shaver, 1999), neurobiology (Depue and Morrone-Strupinsky, 2005; Pedersen, 2004), primatology (Silk, 2003), sociology (de Vos et al., 2001; Back and Flache, 2006, 2007, forthcoming), and experimental economics (Kosfeld et al., 2005). However, serious efforts to identify and describe proximate mechanisms derived from an ultimate theory of interpersonal commitment are indispensable and still largely missing. The present work aims to be a modest contribution to filling this gap. Our purpose is to test empirically whether people hold on to their previous interaction partners, not just in complex interpersonal relationships permeated by a host of different emotions and normative constraints, but even in an anonymous, economic market setting. A crucial point in our argument is that the commitment bias exists beyond instrumental factors, which would otherwise be expected to define the situation to its full extent. Therefore, our present purpose is to induce different levels of commitment behavior by manipulating a non-monetary factor. At the same time, we also test whether the commitment bias is merely an artifact of one isolated niche or if it exists universally, across different cultures.

5.2. Hypotheses

5.2

95

Hypotheses

An ultimate theory for commitment predicts the existence of a proximate mechanism that makes people systematically more committed than instrumental reasons would dictate. More precisely, we need to test whether people become committed to their partner with whom they have been interacting, in proportion to how long they have been interacting, and not in proportion to the instrumental benefits they have derived from the relationship. To make the test more convincing, we propose to examine this mechanism in a purely economic setting (buyer-seller relationships), and in a way that interactions remain anonymous (computerized, not face-to-face). These restrictions help to exclude alternative explanations for commitment behavior (or the lack of it), such as physical attraction and liking, social control, and social desirability. A purely rational actor who assumes that everyone else follows their selfinterest would show equal preference for two different interaction partners with whom interaction balances are equal. This is because a rational actor has no more incentive to trust someone with whom he had a mutually cooperative interaction of $20, than another with whom he had two interactions of $10. On the other hand, an actor who has a bias for interpersonal commitment does have a preference for a partner with two previous common interactions over a partner with only one. We argue that such a preference for commitment is due to a hitherto overlooked mechanism in commitment research: the mere exposure effect, originally documented by Zajonc (1968). According to mere exposure theory, it is enough to be subject to a certain stimulus (e.g. a piece of music or a new taste of food) to develop positive dispositions toward the stimulus. As long as there is no strong negative experience associated with the stimulus, the more one is subject to it, the more positive the disposition becomes. An example for the effect of mere exposure in an interpersonal setting, is the finding that repeatedly seeing a person’s face, but without further interaction with the person, leads not only to attraction but also increases trust in the person (see e.g. Rhodes et al., 2001). In a highly realistic experiment, Moreland and Beach (1992) arranged for several women to pose as students in a large college course by attending various class sessions. Each woman visited the class with different frequency. They were present in the class but did not interact with the students. At the end of the term, students on the course were shown slides of the women and asked to evaluate them on several measures of familiarity, attraction, and similarity. Results showed that whereas exposure had relatively weak effects on familiarity and similarity, it had a strong effect on attraction. Moreover, women who attended more class sessions earned higher scores not only on the attraction index, but students also believed that they would be significantly more likely to befriend these women, enjoy spending time with them, and work with them on some project.

96

Chapter 5. Commitment Bias

We argue that the exposure mechanism in turn leads to an increase in commitment to a partner, independently from instrumental reasons: Hypothesis 5.1. The longer the initial exposure between two persons, the more committed they become to each other, holding other aspects (monetary benefits and uncertainty) of their relationship constant. The hypothesized mechanism is what we refer to as the commitment bias. One criticism against this mechanism could argue that during repeated interactions uncertainty about the trustworthiness of the interaction partner is reduced. This reduction in uncertainty3 makes the long-term interaction partner always more attractive than an unknown stranger, holding other characteristics equal (Kollock, 1994). Kollock (1994) compared rice and rubber markets in Southeast Asia, and argued that the reason why rice is sold in open markets but rubber is traded in long-term, committed partnerships is that the quality of rice is readily observable to buyers at the time of the purchase, whereas the quality of rubber cannot be assessed so promptly. Kollock replicated rice and rubber markets in the laboratory. Subjects had to trade goods whose quality was initially known only to the seller. In the high-uncertainty condition the quality of the good was not know to the buyer until after the purchase. Here commitment between trading partners was high – people tended to exchange with the same partner. In the other condition, when quality was readily observable, subjects on average chose to trade more freely with different partners. Kollock concluded that people protect themselves against a perceived uncertainty by forming stable, committed dyadic relationships. Similarly, Yamagishi et al. (1998) proposed that uncertainty promotes commitment formation, and explained the tendency to form committed relationship with the individual’s low level of general trust in others. The researchers showed in a cross-cultural setting (comparing the USA and Japan) that those who trust others less, in general, are more likely to form committed relationships. If the commitment bias is a stable proximate mechanism based on an evolutionary explanation for interpersonal commitments, it would exist independently from the reduction in uncertainty. Since it is impossible to fully eliminate uncertainty from an economic setting, we aim to test that: Hypothesis 5.2. The positive effect of initial exposure on commitment exists even when controlling for the effect of uncertainty. 3 Uncertainty (or social uncertainty) is experienced by the individual if (1) interaction partners have an incentive to hurt the individual by acting in a selfish way, and (2) the individual does not have accurate information about the probability of interaction partners acting in a selfish way (see Yamagishi et al., 1998).

5.2. Hypotheses

97

Previous experiments with commitment showed that during repeated exchange positive affect develops even in an anonymous, economic setting (Lawler, 2001; Lawler and Yoon, 1993, 1996). Researchers devised a series of experiments where they let subjects negotiate prices in repeated exchanges, while they were also asked to report their mood and emotions. They showed that experiences during repeated exchange create emotions that are attributed to the social unit (e.g. group or dyad) in which the exchange takes place. Positive emotions increased the level of commitment between exchange partners and also led to forgiving occasional opportunism. The perception of increased commitment leads to even stronger affect, essentially creating a self-reinforcing feedback loop. Subjects in the experiments came to treasure their long-term relationships per se, as a distinct object of value, and made sacrifices (in the form of gifts) for them that were not motivated by the prospect of reciprocation4 . Here we propose to retest this mechanism and extend it by taking into account the affect that develops toward alternative partners. If positive affect leads to interaction, then positive affect toward alternative partners would decrease commitment to the steady partner: Hypothesis 5.3. Positive emotions toward the steady partner increase commitment to the steady partner, while positive emotions toward alternative partners decrease it. If a cognitive-emotional framework that facilitates interpersonal commitments has been selected for in our evolutionary past, and our conjectured commitment bias is one of its proximate mechanisms, then we should be able to find traces of it across different cultures: Hypothesis 5.4. The commitment bias exists cross-culturally. At the same time, we do not mean to argue that the commitment bias is equally strong in all cultures, especially when comparing collectivist and individualist societies. East Asians, for example, live in closely knit social environments where the importance of the collective frequently supersedes that of the individual. Social relationships are perceived as a key factor in achieving one’s goals (see e.g. Nisbett, 2004). In Western societies, however, people are conditioned to be individualistic and self-reliant. People are taught to trust themselves and to be able to solve their problems on their own. This could lead East Asians more then Westerners to lean on the help of their social contacts, and consequently do more to keep these relationships active: Hypothesis 5.5. Commitment is more important in Eastern than in Western cultures. 4 Gifts could be bestowed in the last round of the experiment, when there was no further chance to reciprocate within the framework of the experiment. Since the experiment was completely anonymous, there was also no prospect of future reciprocation outside the experiment.

98

5.3

Chapter 5. Commitment Bias

Experimental design

Focusing on the effect of initial exposure on commitment behavior (Hypothesis 5.1) we created an experimental design by manipulating the length of initial exposure between the subject and a steady partner at three levels (short, medium and long). To test the effects of uncertainty (Hypothesis 5.2) and positive feelings (Hypothesis 5.3), we assessed perceived uncertainty in a postexperimental questionnaire, and measured affect at various phases of the experiment. To test the cross-cultural Hypotheses (5.4 and 5.5) we repeated the experiment in two separate locations in three different countries – the Netherlands (University of Groningen and Utrecht University), China (Fudan University and Nanjing University), and the USA (Cornell University and Binghamton University). In the experiment, subjects had to play a trading game. There were two kinds of actors in the game: artists who sell paintings and collectors who buy them. Subjects were told that they would be randomly assigned to the role of seller (artist) or buyer (collector), but in reality, all subjects were assigned to the role of seller, and played against computer-simulated buyers (see instructions in Appendix F.2). This was unavoidable to ensure complete experimental control over pricing mechanisms. We took special care to ensure that subjects would not feel deceived. For example, before starting the actual game subjects had to wait for other subjects to arrive to the same point in the experiment; then again between each round there was a short period of synchronization; we have programmed the timing of incoming offers with probabilistic durations, as if they were made by real people. After the experiment, a non-suggestive5 question showed that only 10% of subjects doubted that they were playing against real opponents present in the laboratory. The game consisted of 17 consecutive rounds, and in each round the subject had to sell a new painting to one of four different buyers who made different bids on it (for a screen shot of the actual game, refer to Appendix F.3). To create a commitment dilemma, those buyers who were not sold a painting by the subject disappeared and were replaced by new buyers in the next round. These buyers would not appear again for the rest of game6 . The commitment dilemma manifested itself when the subject had been selling to the same buyer for a number of rounds and a newly appeared buyer made a better offer. The dilemma was between choosing an old partner with a 5 We asked subjects after the experiment to guess the proportion of all subjects who were acting as buyers in the laboratory at the moment. Those who were suspicious about the deception said 0%. In a follow-up question we also asked them to give their opinion about the behavior of other players whom they interacted with, and whether they thought other players behaved unusually in any way. 6 Since there were always less then 17 x 3+1 subjects in the lab, subjects were told that those partners who once disappeared may reappear in a later round but would have a different name which would make it impossible for both parties to recognize each other.

5.3. Experimental design

99

sub-optimal offer or getting a higher offer from a stranger, thereby sacrificing the relationship with the old partner. Before this commitment dilemma occurred, we manipulated the length of the initial exposure of the subject to the steady interaction partner. Manipulating relationships in an experiment is a challenging task, something that many previous experiments with commitment avoided by explicitly assigning partners to each other or excluding the possibility of partner selection. To manipulate initial exposure, we programmed one of the computer-simulated buyers to give best offers successively, before another buyer would appear with a better offer. At this stage of the game most of the subjects were expected to simply choose the best available option, which kept coming from the same partner. The actual length of the manipulation was 1, 2 or 3 rounds. Out of 326 subjects, only 10 people did not choose the best offer during the manipulation of initial exposure. The manipulation worked so that monetary benefits resulting from interaction with the steady interaction partner were equal across different conditions. In other words, no matter for how many rounds the same steady partner kept giving the best offer, the cumulative difference between the best and second best alternative (the opportunity cost of not choosing the best offer) was kept constant. If the subject remained committed in a commitment dilemma (did not choose a newly appeared buyer with the best offer), the steady interaction partner started alternating between providing best and second best offers in successive rounds. The price gap between best and second best offers continuously widened, resulting in an increasing sacrifice of profit for a committed subject. Note that in this manipulation the subject is not merely exposed to partners, since there is an actual exchange taking place. However, note also that the manipulation concerns the length of interaction and not the size of the exchanges, in other words, there is no difference between conditions in how much the subjects earn, but there is a difference in how often they are exposed to their partner. We measured our main dependent variable, commitment, by the number of interactions the subject kept interacting with the fixed interaction partner before (s)he decided to switch to a new partner. In addition to measuring commitment, subjects were also asked to report how positive, friendly or annoyed they felt towards each buyer in critical rounds that contained a commitment dilemma. The measurement of these emotional dispositions took place after subjects received offers from buyers but before they had to make their choice. After the game we also measured the level of perceived uncertainty on a 5-point scale after the game by asking “How certain did you feel that you could sell your painting at a reasonable price?”.7 7 The limitation of measuring uncertainty in this way, and not manipulating it is that we cannot

100

5.3.1

Chapter 5. Commitment Bias

Procedure

The experiment was administered at several computer laboratories at each university with subjects sitting behind a computer, separated by panels or placed in cubicles. Instructions (see Appendix F.1), except for a brief initial welcome by the experimenter, were received on screen, and all responses were given through the keyboard and mouse. We created the original program in English, and native speakers translated it to Dutch and Chinese (Mandarin). Then another native speaker translated it back to English in order to check for translation errors. For the Mandarin translation an additional round of reverse-translation ensured that all inconsistencies in the original translation were eliminated. After the experiment, subjects received a fixed monetary compensation for their participation and an additional monetary reward based on the amount of money gathered during the game. The amount of monetary payoffs was adjusted in each country to equal on average the price of three meals in the local university canteen.

5.3.2

Sample

Our sample consisted of 326 subjects from three countries, the Netherlands (73 subjects), China (206 subjects) and the USA (47 subjects). 61.4% of subjects were female and 94.5% undergraduate students, mostly in humanity majors. 9.8% of all subjects were suspicious about the deception (in the Netherlands 26%, in China only 4%, in the USA 13%). After a practice game but before playing for real stakes, subjects had to complete a comprehension test about the rules of the game and the use of the game interface. If they incorrectly answered a question, they received feedback and explanation. 10 people (9 in China, 1 in the USA) who did not choose the best price during manipulation of the initial exposure were excluded from the analysis.

5.4

Results

An immediately apparent finding from our experiments is that many subjects failed to spontaneously maximize their profit in an egoistic way. Although, on average, subjects chose more often the relatively more profitable alternative instead of being committed, 26.5% of all subjects chose more often to be committed in situations when their old partner did not offer the best price. argue decisively for the direction of causality. In may be, for example, that people showing a certain level of commitment acted in response to perceived uncertainty, or that they perceived a certain level of uncertainty as a consequence of being committed.

101

5.4. Results

Moreover, 28.1% chose to be committed even in the last round of the game, when any future benefits, such as reciprocation, can be ruled out8 .

5.4.1

Hypothesis 5.1 and 5.2 – The commitment bias

To test hypothesis 1, we need to look at the relationship between initial exposure and commitment. Initial exposure is manipulated at three levels (1, 2 or 3 best offers), while commitment is measured by the number of rounds the subject stayed with the initial partner, who initially made the best offer(s). We found that the longer the initial exposure was, the more committed the subject became (r(314)=.143, p=.011). To control for the effect of uncertainty, we created a linear regression model with dependent variable commitment, and initial exposure and perceived uncertainty (measured on a five-point scale) as explanatory variables. Regression Model A1 (see Table 5.1) shows a positive effect of length of initial exposure on commitment behavior, when controlling for the effect of perceived uncertainty. This means that the longer people were initially exposed to each other, the more committed they became, controlling for the effect of uncertainty and instrumental benefits. This finding clearly supports Hypothesis 5.2.

D.V.: Commitment

Unstd. Coefficients B Std. Error

(Constant) Initial exposure Perceived uncertainty

6.983 0.655 -0.981

0.578 0.293 0.285

t

Sig.

12.084 2.233 -3.438

0.000 0.026 0.001

Table 5.1: Regression Model A1 (N =310, R2 =0.053)

Contrary to previous work (e.g. Kollock, 1994), perceived uncertainty decreased commitment (see Model A1 in Table 5.1). As it turned out from subjects’ verbal explanations (see below “Self-reported reasons for exit”), instead of staying committed to one partner, many subjects protected against uncertainty by trying to diversify their exchange partners. Note that this occurred despite the fact that subjects were not able to return to previous interaction partners once they switched. We also found a positive interaction effect between initial exposure and perceived uncertainty (see Model A2 in Table 5.2). It turns out that 8 The number of the current round and the total number of rounds was clearly defined in the instructions and displayed on the subjects screen during the entire game, leaving no doubts about the time horizon of relationships.

102

Chapter 5. Commitment Bias

although uncertainty decreased commitment, the strength of commitment between different levels of uncertainty converged, and was even reversed in the case of long initial exposure and little uncertainty (see Figure 5.1). In other words, although uncertainty in itself decreased commitment, the commitment-inducing effect of a long relationship was strongest under high uncertainty. Interestingly, if the subject did not feel uncertain at all, the effect of initial exposure turned in the opposite direction. D.V.: Commitment

Unstd. Coefficients B Std. Error

(Constant) Initial exposure Perceived uncertainty Exposure × uncertainty

6.916 0.670 -0.933 0.539

0.575 0.292 0.284 0.239

t

Sig.

12.030 2.299 -3.283 2.253

0.000 0.022 0.001 0.025

Table 5.2: Regression Model A2 (N =309, R2 =0.068)

Figure 5.1: Interaction between initial exposure and perceived uncertainty (0=very certain, 4=very uncertain)

103

5.4. Results

5.4.2

Hypothesis 5.3 – Effect of affect

We found that the friendlier9 the subject initially was toward its steady partner, the more committed it became (r(312)=.122, p=.031). The same was true for initial positivity (r(301)=.121, p=.036). We also found that both initial annoyance, and annoyance in the last measurement before exit made subjects less committed (r(288)=-.139, p=.018 and r(241)=-.285, p<.001, respectively). Other measurements of initial emotional dispositions (e.g. negativity toward worst offer) did not have a significant relationship with commitment behavior. In each round where the steady partner did not make the best offer, we measured how friendly the subject felt toward the player who did make the best offer. This increase in friendliness toward the best offer from the initial measurement until the last measurement (just before exit took place) had a strong negative effect on commitment toward the steady partner (see Model B in Table 5.3). At the same time, the effect of increase in friendliness toward the best offer explained away the effect of initial exposure and decreased the significance of perceived uncertainty. These results are in line with Lawler’s findings, and show that affect develops during exchange and this influences the level of commitment. The results also hint that affect becomes a mediator of other aspects of the relationship and has an important effect on the level of commitment. These findings together lend support to Hypothesis 5.3. D.V.: Commitment

Unstd. Coefficients B Std. Error

(Constant) Initial exposure Perceived uncertainty Increase in friendliness (best off.)

7.087 0.411 -0.840 -2.548

0.948 0.433 0.485 0.840

t

Sig.

7.476 0.948 -1.733 -3.033

0.000 0.345 0.086 0.003

Table 5.3: Regression Model B (N =111, R2 =0.116)

5.4.3

Hypothesis 5.4 and 5.5 – Cross-cultural similarities and differences

To test whether the commitment bias exists across cultures, we looked at the association between initial exposure and commitment in the separate subsamples of each country. We found that the longer the initial exposure was, the more committed subjects became in China (r(195)=.189, p<.001) and in the 9 Friendliness,

positivity and annoyance were measured on 5-point scales. “Initial” refers to the first measurement of each disposition.

104

Chapter 5. Commitment Bias

USA (r(44)=.365, p=.013), but we did not find a significant relationship in the Netherlands. At the same time, we found considerable differences in the average level of commitment between China on the one hand and the USA and the Netherlands on the other. Commitment was highest in China (M=6.731, SD=4.8015), and roughly equal in the USA (M=5.217, SD=3.5208) and the Netherlands (M=5.274, SD=3.5208). The proportion of end-game violators – those who remained committed in the very last round – was also substantially higher in China (37%) than in the Netherlands (11%), and the USA (11%). In addition, the Chinese were almost three times as likely (17.8%) to stay with their initial interaction partner throughout the entire game than the Americans (6.5%), with the Dutch closer to the Americans (6.8%). After including a dummy variable for China in the regression analysis (see Model C), we found a large positive effect of being Chinese on commitment, even when controlling for exposure, uncertainty and their interaction effect.

105

5.4. Results

D.V.: Commitment

Unstd. Coefficients B Std. Error

(Constant) Initial exposure Perceived uncertainty Exposure × uncertainty Chinese

5.818 0.776 -0.890 0.480 1.487

0.676 0.290 0.281 0.237 0.497

t

Sig.

8.610 2.677 -3.167 2.023 2.993

0.000 0.008 0.002 0.044 0.003

Table 5.4: Regression Model C (N =308, R2 =0.095)

We also found differences in reported feelings across countries. On average, the Chinese were more positive to the best, previous (committed) and even to the worst offers (see Figure 5.2). The Chinese also perceived all offers to be friendlier on average than the Dutch, with the Americans in the middle. The Dutch and US subjects reported more annoyance at bad offers than the Chinese. Cross-cultural differences stood out most prominently in differences of self-reported preferences. In a post-experimental questionnaire we asked subjects to rate on three separate five-point scales how important it was for them to play in a committed, fair, and profit-maximizing way. While the Chinese claimed that commitment, fairness but also profit were roughly equally important for them during the game, the US subjects said that especially profit, and also fairness were important, but not commitment. The Dutch claimed that only profit was important (see Table 5). Those who behaved in a more committed way did claim to have a preference for commitment (r(310)=.505, p<0.01), less so with regard to fairness (r(313)=.116, p=0.039), and claimed to not have a preference for profit-maximization (r(312)=.291, p<0.01). These results show that although the commitment bias exists crossculturally, commitment is more important for Eastern than for Western subjects. The difference manifests itself not only in behavior but also in reported emotions and is further backed up by differences in reported preferences.

2.68

Total

312

195 71 46

N

1.15

0.89 1.15 1.16

SD

3.23

3.10 3.42 3.52

profit (M )

314

197 71 46

N

0.61

0.61 0.55 0.55

SD

2.63

3.07 1.82 2.02

fair (M )

Table 5.5: Means of self-reported preferences across-countries

3.13 1.89 1.98

China Netherlands USA

committed (M )

Figure 5.2: Reported emotions across countries

315

197 72 46

N

1.00

0.69 0.97 1.06

SD

106 Chapter 5. Commitment Bias

5.5. Discussion and conclusions

5.4.4

107

Self-reported reasons for exit

During the experiment, we also asked subjects in an open-ended question why they abandoned their steady interaction partners.10 The majority (62%) of answers explained the decision with getting a higher profit. A considerable proportion indicated, however, a preference to build a network of buyers. Note that this was in fact impossible due to the rules of the game11 (e.g. “I will not sell to only one buyer, this way I gain an expanded influence through a little loss.”, “make more friends and do more business”). Other subjects seem to have fallen victim to the fundamental attribution error (Jones and Harris 1967), and made bold inferences about buyers’ personality from the price offered (e.g. “he offered the most money, and seems like a nice person to me” or “the offer is higher than the cost, which shows sincerity, the highest offer among all the buyers”). It was not uncommon among Chinese subjects to belittle their own assets and use this as an explanation for switching partners (e.g. “I am too shy to let buyer 8 buy my paintings at the highest price”, or “I don’t think this painting is very valuable.”). We categorized reasons to abandon the initial interaction into eight types. These are shown in Figure 5.3.

5.5

Discussion and conclusions

A growing body of evidence suggests that humans have an evolved cognitiveemotional framework that facilitates creating and keeping interpersonal relationships, beyond rational considerations (Baumeister and Leary, 1995; Pedersen, 2004; Silk, 2003; de Vos et al., 2001; Back and Flache, 2006, 2007, forthcoming). However, empirical research targeted at uncovering related proximate mechanisms and, in particular, separating these mechanisms from competing explanations such as rational choice in exchange and uncertainty reduction is still lacking. Our work aims to contribute by exploring what we call the commitment bias, a tendency to escalate interpersonal commitment even in an anonymous business setting. We showed experimentally that people have a preference for staying committed to previous interaction partners, and that past or future benefits resulting from such relationships are not sufficient to explain this behavior. We did this by showing that the length of a past relationship (initial exposure) increases subsequent levels of commitment even when controlling for uncertainty reduction (through measurement) and materialistic benefits (through 10 After the subject exited from the relationship with the initial steady interaction partner, the program displayed a question: Why did you choose to sell your painting to [the new buyer]? 11 Once an interaction partner was abandoned, it would not return for the rest of the game, at least not under the same name (number).

108

Chapter 5. Commitment Bias

Figure 5.3: Reasons to leave steady partner

the pricing mechanism). Our results also show that although the level of uncertainty has a negative effect on commitment (i.e. people avoid uncertainty by searching through alternatives), the positive effect of long-term relationship is largest under high uncertainty. This is partially in contradiction with previous literature (Kollock, 1994; Yamagishi and Yamagishi, 1994) and suggests that the relationship between uncertainty and commitment is more complex than hypothesized before. A possible explanation is that different types of uncertainties, i.e. uncertainty about the trustworthiness of partners (social uncertainty) and uncertainty about opportunities, have different effects on networking strategies. More specifically, it seems plausible that whereas social uncertainty leads people to become committed to few, trustworthy partners, uncertainty about opportunities leads people to explore and repeatedly interact with many partners. With regard to emotions, we found that the level of commitment was moderated by both positive affect (general positivity and friendliness) and negative affect (annoyance) toward the steady partner and potential partners during the experiment. Some of these effects explained away the effect of initial exposure, suggesting that the commitment bias is very similar to what Pedersen (2004) calls emotional attachment.

5.5. Discussion and conclusions

109

To strengthen the evolutionary argument, we collected cross-cultural data in three countries. Our results indicate that the commitment bias, the tendency to become committed as a result of initial exposure, exists across different countries although differs in intensity. We also found that the level of commitment was strongest in China, and roughly equal in the USA and the Netherlands. Cross-cultural differences in commitment were correlated with differences in reported emotions, which suggests that cross-cultural differences in commitment behavior cannot entirely be explained with an institutional model of culture, such as Yamagishi et al. (1998) advocate. The difference could lie in the fact that sanctioning and monitoring was not relevant in our setting. The differences are in accordance with the expectation that collectivist cultures value interpersonal commitments more (see Nisbett, 2004). This is because they place increased importance on social ties, and have a greater incentive to hold on to those ties even in the face of growing opportunism. In Western societies, people are conditioned to be individualistic and self-reliant. This leads Western subjects, more than Asians, to attribute their success and potential to themselves, and less to their social environment.

5.5.1

The evolutionary roots of commitment

There are at least two processes that could have led to the evolution of a commitment trait or combination of traits in ancestral humans. One is the increased importance of the pair bond, and another is the importance of non-kin cooperation. Humans are unique among primates in their need for paternal contributions to raising offspring. One of the underlying reasons is that humans have a relatively large brain. In order for this large brain to pass through the birth canal of the mother unharmed, human infants need to be born at an earlier developmental stage than other primate offspring (Hrdy, 1999). Consequently, when they are born they are substantially more helpless, require more attention from their mother, and on the whole need much longer parenting than other primates (Martin, 2003). Therefore, it is imperative for the reproductive success of human mothers to find a committed male who is present and contributes to this extended period of parenting (Foley, 1996; Geary, 2000). Indeed, there is some indication that human females are biologically programmed to require the existence of a committed partner already before conception. The human immune system is designed to attack foreign material inside the body that is not sufficiently related to it genetically. This sometimes happens to the fetus growing inside the mother, a condition related to high blood pressure during pregnancy (preeclampsia). It has been found however, that sexual cohabitation with a steady partner prior to conception reduces the chance of preeclampsia from 40% to less than a 5% (c.f. Pillsworth and Hasel-

110

Chapter 5. Commitment Bias

ton, 2005). Next to natural selection, it is speculated that sexual selection also contributed to the proliferation of a commitment trait, in a way that it became more general, not only present in romantic bonds. Provided that it was imperative for human females to find potentially committed mates, how would they look for one? It is difficult for men to credibly signal pair-bond commitment before the bond is formed, since signaling it from another, existing pair bond could lead to the opposite effect. Therefore, the only context for signaling a tendency for commitment is in the rest of social relationships. In other words, being a good friend, for example, could have served as a costly and hard-to-fake signal for becoming a good father. In this case, a tendency to uphold interpersonal commitments would have enjoyed a strong sexual selection pressure. More research is needed, however, to assess the impact of perceiving a man’s commitment in non-sexual relationships, such as toward friends, and his desirability for mating by women. The other factor that could have led to the evolution of interpersonal commitments is the proliferation of non-kin cooperation. There are many mechanisms that may have governed how humans distributed their cooperativeness among unrelated individuals. Byproduct mutualism (Dugatkin, 1997) for example only works in a very restricted set of circumstances, when cooperation does not entail costs, or when it benefits the donor just as much as the recipient. Another mechanism, reciprocal altruism (Trivers, 1971) requires that people closely keep track of balances: how much was given and how much received. But we know, that people are either not very good at doing this, or not really willing to do it (Mills and Clark, 1982; Silk, 2003). We are then left with largely unconditional cooperation within stable, long-term relationships. As agent-based models have shown (de Vos et al., 2001; Back and Flache, 2006, 2007, forthcoming), cooperation in ancestral groups could have been most efficient in precisely such long-term dyadic relationships. According to these computational models, a simple preference for returning to old partners when in trouble, and favoring old partners over others when distributing help, greatly increases survival chances, much more than other types of cooperation, such as balanced reciprocity or simple generosity (unconditional cooperation). Technically, committed relationships could have evolved as an imitation of kin-relationships. Instead of restricting cooperative and altruistic interactions to those related by blood, which are difficult to identify with certainty anyway, one becomes cooperative to anyone who has been in one’s spatial proximity for an extended amount of time. Indeed, there is some evidence that this happens among other primates as well (Chapais, 2001). There is a number of human abilities that further facilitate interpersonal commitments. One is that humans are extraordinarily good at recognizing faces. Already in early infancy, humans start practicing their skills of interper-

5.5. Discussion and conclusions

111

sonal identification. We know that just a few days after birth, infants have a preference for looking at a human face (Fantz, 1963) and prefer human voices to other sounds (Friedlander, 1970). Ellis (1986) estimates that an adult can name in the order of 700 faces of people they know and can recognize many times more. While other primates are rather bad at recognizing human faces, there is evidence that they similarly excel at recognizing faces of their own species (Parr, 2003). Evidence is also mounting that commitment and bonding are biologically determined processes, pointing to evolutionary origins. Kosfeld et al. (2005) managed to artificially increase the level of interpersonal trust by administering oxytocin, a hormone that acts as a neurotransmitter in the brain, to participants of an experiment. They also showed that the effect of oxytocin on trust is not due to a general increase in the readiness to bear risks. On the contrary, oxytocin specifically affected an individual’s willingness to accept social risks arising through interpersonal interactions. Animal research on mammals suggests that social bonding can be modulated by various hormones including oxytocin, vasopressin, opioids, corticotropin releasing hormone, dopamine and adrenal steroids, including corticosterone or cortisol (cf. Carter, 2005). The effects of these hormones on social bonding are especially apparent following periods of stress or anxiety. Of possible relevance to the commitment bias is the capacity of hormones, such as oxytocin and vasopressin, to overcome anxiety or fear. Brief exposure to oxytocin or vasopressin can facilitate social contact and selective sociality. Neural systems associated with reward, including those that rely on dopamine and possibly the endogenous opioids are also found to help in regulating responses to the presence or absence of a preferred partner (Depue and Morrone-Strupinsky, 2005). The capacity to form social bonds emerges as a function of interactions among genetics and developmental experiences. With a genetic background, social and hormonal experiences apparently have the ability to reprogram the nervous system and thus potentially alter the tendency of individuals to form social bonds.

5.5.2

Limitations

One shortcoming of our experimental design is that we measured uncertainty, instead of manipulating it. The drawback is that the direction of causality cannot be ultimately established between uncertainty and the level of commitment. While it is possible that subjects responded to uncertainty with increasing commitment behavior, it is also possible that people who failed to commit themselves experienced an increased level of uncertainty. Our finding that certain subjects had a strong preference for building a network of interaction partners – even when this was not possible within the framework of the game – clearly contradicts the findings of Kollock (1994) and (Yamagishi et al., 1998)

112

Chapter 5. Commitment Bias

about the positive effect of uncertainty on commitment. Therefore, an attractive path for further research is to examine more closely the relationship between commitment and uncertainty. In particular, prior research has explored the link between commitment and social uncertainty (Kollock, 1994; Yamagishi et al., 1998). Anthropological works show that people in several societies construct their social environment to adapt to uncertainties of the natural environment (see Wiessner, 1982, 2002; Silberbauer, 1981; Kent, 1993; Cashdan, 1985). Whereas previous experimental work explored the link between social uncertainty and commitment, to our knowledge, there is no experimental work that systematically relates commitment to resource uncertainty, or the inability (as opposed to unwillingness) of interaction partners to provide desired outcomes.

Chapter 6

Commitment and Networking under Uncertainty12 Abstract

Previous research explained the tendency to build stable long-term exchange relationships (interpersonal commitment) by the avoidance of uncertainty associated with the trustworthiness of strangers (social uncertainty). Comparisons of markets with different levels of social uncertainty, as well as laboratory experiments with controlled exchange consistently revealed this positive association between commitment and uncertainty. However, there is evidence that commitment exists and systematically varies independently from social uncertainty and that people also hold on to untrustworthy partners. In particular, a wealth of social capital literature and anthropological evidence suggests that people protect themselves against future resource contingencies by establishing committed relationships. To advance the exchange theoretic explanation for commitment, we propose differentiating resource uncertainty 1 This chapter is based on Back, I. and Flache, A. (2007). Commitment and Networking under Uncertainty. Presented at the XXVII. International Sunbelt Social Network Conference, Corfu, Greece, May 1-6, 2007. 2 The authors would like to thank Jurre van den Berg and Vincent Buskens for their help in preparing and conducting the experiments; Tom Snijders and Henk de Vos for their insightful comments. This research was made possible by an Ubbo Emmius grant from the University of Groningen to the first author, and a grant from the Innovational Research Incentive (VIDI) of the Netherlands Scientific Organization (NWO) to the second author.

113

114

Chapter 6. Commitment and Networking under Uncertainty

from social uncertainty. While social uncertainty concerns the trustworthiness of potential partners, resource uncertainty concerns resources of potential partners. We also point to possible explanations for individual differences in commitment behavior by relating it to personality traits. Results from a laboratory experiment and an on-line experiment lend support to our hypotheses.

6.1

Introduction

We live in a world teeming with hazards that carry the threat of unpredictable costs at unforeseen probabilities. A crucial remedy against future contingencies is to rely on the support and cooperation of others. But how do people choose their partners? How does someone decide between returning to a previous partner and switching to a new, unfamiliar one? And why do some people strive to have many partners while others stay committed to their existing ones? Current explanations of commitment in exchange (Kollock, 1994; Yamagishi and Yamagishi, 1994; Yamagishi et al., 1998) suffer from two shortcomings. First, they overemphasize the positive effect of the uncertainty of finding untrustworthy partners (social uncertainty) and second, they explain individual differences in commitment exclusively through the construct general trust. The present work addresses these two shortcomings by testing the limits of social uncertainty while concurrently introducing resource uncertainty, and by extending the scope of individual differences to optimism in life. According to exchange theoretic explanations of commitment (Cook and Emerson, 1978; Kollock, 1994; Yamagishi and Yamagishi, 1994; Yamagishi et al., 1998), people stick to their old interaction partners when they perceive a threat of being cheated by strangers. For example, in markets where cheating is easy (e.g. due to low observability or insufficient punishment opportunities), people reduce the number of partners they exchange with, and stick to only a few long-term partners (Kollock, 1994). This has also been consistently replicated by controlled experimental studies, which show that social uncertainty leads to commitment behavior (Yamagishi and Yamagishi, 1994; Yamagishi et al., 1998; Kollock, 1994). Yet, social capital theory posits that people also create durable relationships in order to gain better access to information and influence over the resources of others (see e.g. Lin, 1999; Flap and Völker, 2004). A wealth of anthropological evidence, for example, suggests that people establish committed relationships even if social uncertainty is missing (Wiessner, 2002; Kipnis, 2002) either because non-cooperation is not a concern, or because an external punishment or reputation system is already in place to enforce cooperation. People nevertheless build strongly committed exchange relationships in these situations, as is especially visible in hunter-gatherer societies of the Kalahari

6.2. Theory and Hypotheses

115

region in Africa (Silberbauer, 1981; Cashdan, 1985; Kent, 1993; Wiessner, 2002). Another problem with the social uncertainty explanation is that we know that people tend to hold on to partners who have already proved to be untrustworthy. An extreme example is wives and girlfriends who are beaten by their partners, but as soon as their wounds heal they return to them and face further abuse (Roy, 1977; Strube, 1988; Rusbult and Martz, 1995). Why do people build committed exchange relationships if not to avoid untrustworthy strangers? To resolve this empirical puzzle, we provide and test a twofold explanation. First, we argue that previous research showing a positive relationship between social uncertainty and commitment (Kollock, 1994; Yamagishi and Yamagishi, 1994; Yamagishi et al., 1998) overestimated the importance of social uncertainty for commitment because it focused on a restricted set of conditions. Second, we point to the importance of another type of uncertainty that has been neglected until now in the exchange-commitment literature. Research in close relationships and interpersonal communication separates uncertainty based on its source – the self, the partner and the relationship (Berger and Bradac, 1982; Berger and Calabrese, 1975; Berger and Gudykunst, 1991). Social dilemma research, in turn, focuses on social uncertainty and environmental uncertainty (Messick et al., 1988; van Dijk et al., 1999; de Kwaadsteniet et al., 2006), the latter comprising uncertainty about features of the dilemma. With regard to commitment in exchange we propose to introduce here a more specific source of uncertainty – partner resources. We argue that such resource uncertainty has a positive effect on commitment behavior, similar to but independent from the effect of social uncertainty. To test the effect of different uncertainties on commitment we created a laboratory experiment, independently manipulating social and resource uncertainty. To explain individual differences in commitment tendencies, we linked the effect of these uncertainties to two personality traits: general trust in people (following Yamagishi et al., 1998), and general optimism in life (for measurement see Scheier et al., 1994).

6.2

Theory and Hypotheses

According to exchange theory, people exchange benefits with each other in order to arrive at mutually satisfactory outcomes. Successful exchange is rewarding and reinforcing, unsuccessful exchange is costly and unattractive (Lawler et al., 2000). People repeatedly exchange with each other when previous exchanges have been successful, and change to other partners when exchange fails, either due to reinforcement learning (Homans, 1961; Emerson, 1972) or rational choice (Kollock, 1994; Cook and Whitmeyer, 1992). Frequent exchange reduces uncertainty about the exchange partner because partners learn about each other’s motives, behavioral patterns, find each

116

Chapter 6. Commitment and Networking under Uncertainty

other more predictable, and develop a similar exchange orientation (Lawler et al., 2000). The same is observed in romantic and close relationships as well. According to the Uncertainty Reduction Theory (Berger and Bradac, 1982; Berger and Calabrese, 1975), uncertainty leads to a lack of confidence about predicting future behavior within the interaction (Berger and Bradac, 1982; Berger and Calabrese, 1975; Berger and Gudykunst, 1991). Research in cognitive psychology and experimental economics also consistently shows that people prefer to avoid unpredictable or uncertain decision situations (Tversky and Kahnemann, 1974; Kahneman and Tversky, 1979, 1996). There is a wealth of evidence suggesting that people prefer outcomes they know more about in advance (Rode et al., 1999), even holding their beliefs constant, because they are averse to uncertainty3 (Cook and Whitmeyer, 1992). In line with these findings, Kollock (1994) demonstrates that the more uncertain the social environment is, the more likely people are to become committed to their existing partners. Relying on the original argument by Cook and Emerson (1978), Kollock contends that instead of always seizing the actual best offer, people form stable, long-term dyadic relationships. Kollock points to the example of rice and rubber trade in Southeast Asia. The difference between rice and rubber is that the quality of rice is immediately apparent upon simple inspection. In other words, the risk of being cheated with regard to the quality of rice is low (low uncertainty). In contrast, the quality of raw rubber is hard to judge at the time of purchase, and can only be assessed once it has been processed. Cheating on the quality of rubber is easy, and therefore the buyer of raw rubber faces high uncertainty. This difference in uncertainty concerning the quality of rice and rubber, Kollock argues, explains the observed difference in the dominant form of trade. Rice is usually traded at open markets between strangers, whereas rubber is often traded between a particular producer and a broker who have formed a long-term relationship, often extending over several generations. Kollock (1994) simulated such markets under controlled laboratory conditions. In one condition (high uncertainty), sellers could deceive their potential buyers about the quality of the product they were selling. In the other condition (low uncertainty), it was not possible to deceive buyers. A key finding of Kollock’s experiment was that commitment formation between a particular seller and a particular buyer occurs more frequently in the high uncertainty condition than in the low uncertainty condition. Similarly, Yamagishi et al. (1998) argue that social uncertainty leads to less exchange partners. They define social uncertainty as “existing for an actor when (1) his or her interaction partner has an incentive to act in a way that imposes costs (or harm) on the actor and (2) the actor does not have enough information to predict if the partner will in fact act in such a way.” 3 Also

known as the ambiguity effect.

6.2. Theory and Hypotheses

117

Yamagishi and Yamagishi (1994) argue that committed relations (a reduction of partners) give a solution to the problem of social uncertainty for multiple reasons. First, committed partners accumulate information about each other over time. Second, mutually committed people enact “hostage-taking” behaviors – ranging from the formation of mutual emotional attachments to the establishment of relation-specific assets (Helper and Levine, 1992; Raub, 2004). Hostage-taking behaviors provide deterrence against unilateral defection (Shapiro et al., 1992). Finally, in long-term relationships conditionally cooperative strategies such as Tit-for-Tat can be used to control each other’s behavior under the shadow of the future (cf. Axelrod, 1984). We argue, however, that there are two very important implicit assumptions in both Kollock’s and Yamagishi’s argument for a positive effect of social uncertainty on commitment (Kollock, 1994; Yamagishi and Yamagishi, 1994; Yamagishi et al., 1998). First, they implicitly assumed that the focal actor in a social dilemma – whose commitment is being explained – either has no option of non-cooperation or always strives for long-term mutual cooperation. But what if this focal actor is not interested in mutually cooperative outcomes in the first place, and instead tries to exploit its partner? Is it still advisable to stay with a partner one has just cheated? If the focal actor has a choice of many partners and cooperation is costly and/or risky, an obvious strategy is to go for a non-cooperative payoff and exit (“hit and run”). Arguably, the fact that one has just cheated its partner, who is now very likely to respond with retaliation, makes social uncertainty, about the likelihood of alternative partners to defect, irrelevant. What is more, the very realization that others are more likely to defect in general (high social uncertainty) would make an otherwise strict cooperator also more willing to abandon partners, even when they are cooperative. The reason for the implicit assumption about the focal actor’s cooperativeness in the existing literature is most probably due to both Kollock and Yamagishi relying originally on Akerlof’s example of the market for lemons (1970). In that example the focal actor, a buyer, faces a unilateral trust problem, to trust a seller or not, which is different most notably from a Prisoner’s Dilemma situation. The second implicit assumption is that the focal actor has already found a cooperative partner. If experience has shown the focal actor that the present partner is trustworthy, then social uncertainty does provide additional incentives to stay with this partner. The problem with the argument that social uncertainty increases commitment is that when people are less trustworthy on average (high social uncertainty), it is more difficult to find a partner who is suitable for a long-term relationship. Therefore, although social uncertainty may increase the strength of commitment, it decreases the chances of establishing it. This could lead, on average, to a negative effect of social uncertainty on commitment. Given these qualifications to the original idea of the effect of social uncer-

118

Chapter 6. Commitment and Networking under Uncertainty

tainty on commitment, we propose to test the following hypothesis: Hypothesis 6.1. Social uncertainty increases commitment more in the case of mutually cooperative relationships, than if either or both partners are non-cooperative. Note that this hypothesis includes the possibility that there is no effect of social uncertainty at all, in the case that the relationship is not mutually cooperative. Yamagishi et al. (1998) also acknowledge that commitment, as a solution to problems of uncertainty, has its own problems. While reducing the risk of being exploited by unfamiliar people, it limits the actor’s choices for exploring better alternatives that might exist outside the current relationship. In forming a commitment with a particular partner, one obtains security (i.e., a reduction in uncertainty) in exchange for opportunities. Commitment is an efficient means for reducing uncertainty in a situation in which outside opportunities are limited (i.e., when the general level of opportunity costs is low). On the other hand, commitments become a liability rather than an asset as people face more and better opportunities outside their current, mutually committed relationships (i.e. when the general level of opportunity costs in the environment is high). According to social capital theory (see Lin, 1999; Flap and Völker, 2004), relationships between humans are a form of capital that needs investment and provides returns. The exponentially growing field of social capital theory (c.f. van der Gaag, 2005) provides at least two key reasons, in addition to the trustproblem, for why people create durable relationships – improved access to information and influence over resources of others (Lin, 1999). Finding highvalue (e.g. high-resource) alternative partners, however, may be just as uncertain as finding trustworthy ones. The fact that potential exchange partners could differ from each other not only in motives but also in resources leads to an entirely different kind of uncertainty, that is independent from social uncertainty. Social networking strategies aimed at reducing this different kind of uncertainty have been richly documented, for example, among peoples of the Kalahari region in Africa. The main sources of uncertainty in this region are, the availability of scarce resources, natural hazards, illness and disability, but also finding a suitable mate in time. Sharing and reciprocity is a key mechanism to purposefully establishing and maintaining long-term relationships among the sedentary Kutse community (Kent, 1993) and Nata River Bushmen (Cashdan, 1985). It is a long-known strategy for the hunter-gather people of the G/wi (Silberbauer, 1981), and also the !Kung (Wiessner, 2002). Polly Wiessner’s work gives a detailed look into the hxaro system, a form of delayed reciprocal exchange among the !Kung. Those who entered into a mutual hxaro relationship regularly give non-food gifts and sometimes share meat with each other. Another use of broader hxaro networks is that they afford

6.2. Theory and Hypotheses

119

people greater access to information of many kinds and assistance for locating spouses and arranging marriages (Wiessner, 2002). People in a hxaro relationship often live far removed from each other. Silberbauer (1981) and Kent (1993) describe reciprocal networks as a cohesive group within which there is a high rate of communication, shared preferences of company and a common interest in cooperative tasks. Throughout Chinese history and up to this day, the institution of guanxi, or informal networking, has been a well-known practice both in business and private life (Kipnis, 2002). Through banqueting, gift-giving, visiting, helping out, etc., the Chinese purposefully create and nourish informal social relationships that can be called upon when something needs to be taken care of. These guanxi relationships are another example of people building long-term relationships not to counter the threat of untrustworthy others, but to exert influence on behalf of others (Kipnis, 2002). In other words, guanxi enables people to share each other’s social network resources. In contradiction with the findings of Kollock (1994) and Yamagishi and Yamagishi (1994), Back (see previous chapter) found in a series of cross-cultural experiments that people can become less committed under higher uncertainty. Each subject in these experiments had to repeatedly choose a partner to do a small trade with in an anonymous computerized setting. Those partners whom the subject did not choose to trade with at a given time point were excluded from further interaction with the subject in the experiment, which the subject knew in advance. In other words, if a subject wanted to interact with a given partner in the future, he/she had to trade with that partner in the current round. Uncertainty was measured by asking respondents how sure they felt about being able to sell their paintings at a good price. Those who were more committed were consistently found to be less certain about being able to sell their paintings at a good price. Back also found that people had a stable tendency to try and build a network of exchange partners, even when it was not in their best interest. A substantial number of subjects were reluctant to trade repeatedly with the same partner (to become committed) and used the first possible opportunity to get out of the relationship, even when they had no financial incentive to do so, e.g. when the payoffs from a new partner were equal to payoffs from the committed partner. In reply to open-ended questions, some of the subjects explained their behavior as an attempt to build a network of cooperative partners. They claimed so despite the fact that they were explicitly told that once they left a partner, they could not return to the same partner in the game any more. There is also evidence that people hold on to relationships with partners who have proven to be untrustworthy. This is most prominent in the case of battered spouses and girlfriends (Roy, 1977; Strube, 1988; Rusbult and Martz, 1995) who stay with their partner even though it would be difficult to run into someone at a lower end of social uncertainty. According to Rusbult and Martz

120

Chapter 6. Commitment and Networking under Uncertainty

(1995), a key factor in holding a relationship together is the availability and quality of alternative partners and alternative states (i.e. no partner). Why do these women hold on to relationships where the partner has already betrayed their trust? This suggests that it is not primarily social uncertainty that explains their commitment because there is little reason for these women to be afraid of running into someone even less cooperative if they abandoned their current relationship. Instead, the reason for their commitment might be uncertainty about the resources available from potential alternative partners or in a life without any partner. Holding the partner’s cooperativeness equal, one should be more inclined to seek out more alternative partners with more resources. What we argue here is that the distribution of resources within a population of potential exchange partners, and the relative standing of the current partner compared to potential partners creates an uncertainty about the expected benefits of switching. The fact that people avoid uncertain decision situations should apply similarly to the uncertainty about resources as it applies to uncertainty about trustworthiness: Hypothesis 6.2. Resource uncertainty increases commitment to partners with sufficiently high resources. Note that the above hypothesis implicitly assumes that people perceive high resource uncertainty more in terms of the risk of running into lowresource others, and not as an opportunity to find partners with even higher resources. This is in line with Kahnemann and Tversky’s theory (1979) that people are risk-averse when contemplating gains. Social and resource uncertainty are both individually hypothesized to lead people to perceive an increased risk of switching to alternative partners. This in turn leads to increased commitment behavior. The combined risk of losing out on switching to a new partner is lowest when both uncertainties are low, and highest when both are high. Therefore: Hypothesis 6.3. Commitment is lowest when both uncertainties are low. Hypothesis 6.4. Commitment is highest when both uncertainties are high. Yamagishi and Yamagishi (1994) mention four reasons for the difficulty people have in leaving a committed relationship even when it becomes a liability. First, committed people are by definition those who stay with the current relationship despite outside opportunities. They may eventually leave the committed relationship, but there typically is a substantial time lag. Secondly, the mutual attraction and loyalty that have developed through the relationship keep partners in the relationship. Third, a temporary better offer from outsiders would not be sufficient for the one who has invested in relationspecific assets to leave the current relationship. Social and psychological assets, such as the warm memory of a pleasant past and mutual understanding,

6.3. Study

121

may be considered relation-specific assets that keep people in these relationships. Finally, commitment to a particular partner often reduces the level of trust in “outsiders” (see Kiyonari and Yamagishi, 1996, for experimental support), creating a vicious cycle of distrust of outsiders; those who do not trust “outsiders” tend to stay in committed relationships, and because they avoid “outsiders” they become even less trusting of “outsiders.” Yamagishi et al. (1998) further extend the argument that uncertainty promotes commitment formation, and explain the tendency to form committed relationship with the individual’s low level of general trust in others. They show in a cross-cultural setting (comparing the USA and Japan) that those who, in general, have high trust in others, are less likely to form committed relationships with few partners. Yamagishi et al. (1998) argue for individual personality differences to explain why some people are more committed than others. They contend that general trust (or trust in people in general) provides a psychological springboard for people who have been “confined” to committed relationships to move out into the larger world of opportunities. Uncertainty leads people to avoid exploring relationships with new contacts. Those who trust strangers more, either experience less uncertainty, or are less affected by it, and thus are more likely to try out relationships with strangers. To retest the finding of earlier works also with regard to this psychological mechanism we propose that: Hypothesis 6.5. General trust decreases commitment. General trust mitigates the effect of social uncertainty on commitment by resolving concerns about the trustworthiness of strangers. However, the concern about running into low-resource partners (resource uncertainty) as opposed to untrustworthy ones (social uncertainty) is not solved by general trust. We argue that stepping out of committed relations by seeking out partners with different resources requires a more general sense of optimism: Hypothesis 6.6. Optimism decreases commitment.

6.3

Study

To test our hypotheses, we conducted an experiment using human subjects playing a repeated dilemma with the choice of partner selection. The design was 2×2×2 between-subjects, factorial. To test main and interaction effects of different uncertainties on commitment, we manipulated social uncertainty (high vs. low) and resource uncertainty (high vs. low). As a robustness test of the effects we also manipulated the dilemma (“fear” vs. “greed”, see later) that people faced in the experiment.

122

Chapter 6. Commitment and Networking under Uncertainty

The experiment was administered in two large batches (Sample 1 and 2) via computer terminals. Sample 2 replicated the design of Sample 1 and extended the scope of the subject pool. Instructions (see Appendix G.1) were always received on a computer screen, and all responses were given through the keyboard or mouse.

6.3.1

Sample 1

The experiments with this sample were conducted at the ELSE laboratory of Utrecht University using 136 subjects from the experimental pool of Utrecht University during December 2006. The experimental pool contains mostly students from different faculties and disciplines, and also some staff members, secretaries, etc. 66% of subjects were female and 95% undergraduate students. 10% of subjects had never participated previously in a laboratory experiment, and 34% had participated at least 3 times in experiments of a different nature. During the experiment, each subject was sitting behind a computer, placed in a cubicle. Subjects received a monetary compensation for their time and an additional reward for the points accumulated during the game. Depending on individual performance, total payments varied between €3 and €14, earned in 30 to 45 minutes.

6.3.2

Sample 2

In addition, another 169 subjects were recruited in the University of Groningen, by advertising the experiment on colorful posters in January 2007. Subjects in Groningen did not have to come to a laboratory, rather they completed the experiment from computer terminals at the university or at home. 44% of these subjects were female and 90% undergraduate students. 50% of subjects had never participated previously in a laboratory experiment, and 17% had participated at least 3 times in experiments of a different nature. Scoring within the experiment was identical to that of Sample 1 but in this case subjects were offered a €200 Grand Prize if they scored highest in the experiment. Everyone was allowed to join the experiment only once. To ensure that the same people did not participate repeatedly, we excluded all subjects who spent less than 20 seconds reading the 3-page instructions of the experiment. The purpose of having Sample 2 was twofold. Firstly, it increased the sample size and widened its representativeness by recruiting from a much larger and varied population. Secondly, with Sample 2 we also measured the effect of personality differences in the level of general trust and optimism on commitment. Therefore, in the case of Sample 2, two short questionnaires were included before the subjects received instructions about the game. The first

6.3. Study

123

questionnaire was an 8-item measurement of general trust, taken from Yamagishi et al. (1998). The second questionnaire was the Revised Life Orientation Test, measuring optimism, taken from Scheier et al. (1994). The experimental design for Sample 1 and 2 was otherwise identical.

6.3.3

Design

Subjects in both samples played two-person repeated dilemma games (see e.g. Dawes, 1980). After each round of the game there was a random draw that determined whether the game between the two players would go on (continuation probability w=0.5). If the game was determined to be over, subjects were offered the choice to start another game with their current partner or to switch to a new, unknown player. With the chosen partner they played again an indefinite number of rounds before they were offered the chance to switch partners in the same manner again. When choosing partners, subjects were always offered the opportunity to choose from a set of partners that contained all previous partners and a new, unknown player. Subjects played a total of 10 games, each consisting of relatively few rounds due to the low continuation probability. In order to gain the largest possible control over game histories and uncertainties during the experiment, subjects in the laboratory did not play against each other but against computer-simulated opponents. We explained to them that the behavior of each computer-simulated opponent is based on the personality profile of a real person who had played a similar game in a previous experiment. This implementation was identical to that of van Assen and Snijders (2004) and (van Assen, 2001, pp. 99—100). More precisely, the behavior of the computer opponents was determined by a stochastic algorithm, which took into account previous cooperative and defective moves of both the opponent and oneself with decreasing weight in the past (memory parameter). The algorithm and its parameters were tied to a statistical model that had been fitted to data obtained in a previous experiment with real subjects (van Assen and Snijders, 2004). The responses of the computer, given the simple nature of the dilemma, were highly realistic and difficult to distinguish from the behavior of a real person. Most notably, the computer’s behavior was similar to Tit-for-Tat, in that it was both retaliatory and forgiving. However, each computer player was different in their tolerance to the subject’s behavior, e.g. if the subject resumed cooperation after having defected, this did not necessarily induce immediate forgiving from the computer. Another difference between the computer players and Tit-for-Tat was that the computer players were often uncooperative on their first move. Since the behavior of computer players was stochastic, the only experimentally manipulated difference between computer players was the initial probability of cooperation (see below).

124

Chapter 6. Commitment and Networking under Uncertainty

We took special care not to deceive subjects during the experiment for any period of time. Moreover, we also made sure that subjects had complete information about all aspects of the game. Relying on the experience of van Assen (2001), we also ensured that they took the idea of playing against a range of different personality profiles seriously.

6.3.4

Manipulation of uncertainties

Hypotheses 6.1/6.2 and 6.3/6.4 are based on the idea that uncertainty can have different facets, and each of these may have separate effects on commitment. We manipulated social uncertainty (about trustworthiness in the general population) by setting the probability of cooperativeness in the first round4 of the computer-simulated opponents to 68% (low social uncertainty) or 32% (high social uncertainty). In addition, we told subjects before the experiment that there are either very few or many selfish opponents in the game, who only care about their own payoff (low and high social uncertainty, respectively). A manipulation check after the game asked, “If you could keep changing partners, how likely do you think it would be to meet very selfish partners who only care about themselves?” (5-point scale). The association between the manipulation and the check was r(231)=0.420, p<0.001. Resource uncertainty was created by assigning a “resource” score to each partner, which multiplied payoffs from the interaction with that partner. The resource score was different for each new partner, randomly determined right before the first interaction. The actual manipulation was the size of the range from which the resource score took its value: 0.90 to 1.10 (low resource uncertainty) and 0.50 to 1.50 (high resource uncertainty). The subjects were told about the range before the experiment. A manipulation check after the game asked, “If you could keep changing partners, how likely do you think it would be to meet a partner with a very low [resource]?” (5-point scale). The correlation between the manipulation and the check was r(231)=0.127, p=0.053. An additional question, “If you would keep changing partners, how likely do you think it would be to lose some payoff due to meeting a new player with a low [resource]?”, asked only from some subjects, had an even stronger association with the manipulation (r(143)=.224, p<0.01). To eliminate the effect of differences in memory between subjects, visual aids were displayed on the screen during the entire experiment. Subjects saw the moves of previous partners and their own moves from previous rounds in a small table. They also saw the resource value for their current and previous partners and a small graphical bar showing the position of the resource value within the range of possible values (see screen shots of the game in the Appendix G.2). 4 Also

known as the property of niceness in game theory.

6.4. Results

6.3.5

125

Manipulation of dilemma

Different dilemmas imply different incentives. To test for the robustness of our argument under different incentives, we tested our hypotheses using two different dilemmas (between-subjects manipulation). We used two dilemmas, the Fear and the Greed Game, both derived from the Prisoner’s Dilemma. The advantage of these two dilemmas is that they separate the effects of fear and greed that are otherwise confounded in the Prisoner’s Dilemma (Simpson, 2003). At the same time, both games impose the social uncertainty that is important for our analysis of commitment. In both games, there is an incentive to defect but a cooperative player prefers to have a cooperative partner because (s)he is harmed by the parter’s defection. In the Greed Game (see Table 6.1, top, T=“temptation”, S=“sucker”, P=“punishment”, R=“reward” payoffs), T>R, which means that Player A is tempted to cheat on Player B (the greed incentive: T—R). On the other hand, if Player B defects, A’s decision no longer has an influence on A’s payoffs, S=P, which means that Player A has no “strategic fear” that Player B will defect (the fear incentive: P—S=0). Therefore, A is expected to concentrate on the difference between T and R, the greed incentive. Conversely, in the Fear Game (see Table 6.1, bottom), T=R, thus the greed incentive is zero. The choice between cooperation and defection for Player A focuses on the payoff difference in case Player B defects, the fear incentive for A to defect. Subjects had the role of Player A and the computer that of Player B. To avoid the pitfall of completely removing fear from the Fear Game, we followed Kuwabara’s advice (2005), from his follow-up work on Simpson’s paper (2003). Kuwabara suggests that if the Fear Game is left in its original, symmetric form (same payoff structure for both players) then neither player has an incentive to act greedily any more. Thus neither player needs to fear defection. Therefore, we based our Fear Game on Kuwabara’s Fear of Greed Game, where Player B’s payoffs remain the same as in the Greed Game. To increase the stakes in each round and to avoid end-game effects, subjects only received a payoff from the last round of each game, i.e. if the random draw determined that the game with the current partner is over. Since subjects did not know whether they were playing the last round at the time of making their choice, stakes were equally high in each round.

6.4

Results

Subjects in the experiment played 10 games, each consisting of several rounds. After each game, subjects were allowed to choose either a new, unknown partner or return to a known, previous partner. We measured the main dependent variable, commitment, as the maximum number of games a subject stayed with

126

Chapter 6. Commitment and Networking under Uncertainty B cooperates

defects 100

50 cooperates 10 (S)

50 (R)

A

10

10 defects 10 (P)

100 (T)

50

100

cooperates 100 (R)

A

10 (S) 10

10

defects 100 (T)

50 (P)

Table 6.1: Greed Game (top) and Fear Game (bottom) with actual payoffs in eurocents any single partner5 . The distribution of the dependent variable was close to normal, ranging from 1 to 10, with a mean of 4.82 (S.D.=2.10). Since the dependent variable obtains its value at different times for different subjects during the experiment, we took special care in the analysis to avoid making causally incorrect inferences. In particular, we made sure to limit the set of explanatory variables to those that obtained their value before the dependent variable.

6.4.1

Final sample

Since monetary incentives were very similar for Sample 1 and Sample 2, and the distribution of the dependent variable did not show significant differences, the two samples were pooled, and the aggregate sample was used to test the hypotheses. In both samples, after subjects read the instructions but before they started playing the game, they had to complete a comprehension test about the rules of the game and the use of the game interface. Due to the relative difficulty of the game, 62 people (20%) made three or more mistakes on eight of the comprehension questions. Their input was excluded from the analysis. The excluded group roughly corresponds to the 18% of subjects who 5 Since

people were allowed to return to previous partners, games that a subject spent with the same partner are not necessarily successive.

6.4. Results

127

said the instructions were difficult to understand and 15% who said the experiment was difficult or very difficult. The final sample consisted of 244 subjects.

6.4.2

Hypothesis 6.1 – Social uncertainty and commitment

Hypothesis 6.1 predicts that social uncertainty increases commitment more for subjects who are cooperative and meet fellow cooperators than for other subjects. When looking at all subjects, we find that commitment seems to decrease in the high uncertainty group compared to the low uncertainty group (4.6 vs. 5.1 games, t(242)=- 1.944, p<0.53). This clearly contradicts earlier research (Kollock, 1994; Yamagishi and Yamagishi, 1994; Yamagishi et al., 1998), which argued for a general positive effect of social uncertainty on commitment. Looking at the cooperativeness of subjects, we found that in fact many of them followed a strategy of non-cooperation from early on in the experiment. Even on the very first move in the experiment, before the subject experienced any behavior from the first partner, 64% of the subjects made a noncooperative move. More importantly, we found that subjects were less likely to cooperate on their very first move in the high uncertainty condition (0.29, std.err=0.047) than in the low uncertainty condition (0.42, std.err=0.052), most likely as an adaptation to the uncooperative environment. To test Hypothesis 6.1, we need to examine the interaction effect between mutual cooperativeness of the relationship and social uncertainty on commitment. We categorized a relationship mutually cooperative if both the subject and the computer cooperated in at least 75% of their moves within the relationship. Note that the results are robust for the choice of this cut-off point within the 55%-99% range. The interaction effect between social uncertainty and mutual cooperativeness was highly significant, F(1, 243)=6.81, p=.01. The inversion in the effect of social uncertainty on commitment between mutually cooperative and noncooperative relationships is depicted in Figure 6.3. For low mutual cooperation, the effect of social uncertainty turns negative, whereas for high cooperation it is stronger and positive, in line with Hypothesis 6.1.

6.4.3

Hypothesis 6.2 – Resource uncertainty and commitment

Extending the argument from variation in trustworthiness to variation in resources, Hypothesis 6.2 predicts that resource uncertainty should have a similarly positive effect on commitment as social uncertainty. We found that those who played in the high resource uncertainty condition were more committed than those in low resource uncertainty (5.1 vs. 4.5, t(242)=2.114, p<0.05), in other words, subjects were more likely to stay with their most steady partner when variation in the resources of partners was higher.

128

Chapter 6. Commitment and Networking under Uncertainty

Figure 6.1: Interaction between mutual cooperativeness and social uncertainty on commitment

Just as it might be difficult for a person low on trustworthiness to maintain committed relationships, someone low on resources could be less attractive to others for long-term interaction. If we separate the sample according to the resource parameter of the most steady partner, we find that the difference in commitment between high and low resource uncertainty is larger in the subsample where the partner’s resource parameter was higher than 1.00 (5.3 vs. 4.5, t(188)=2.503, p<0.015), and the difference is non-significant in the other subsample (4.3 vs 4.5, t(52)=-0.538, p<0.59), see Figure 2. This confirms our expectation that resource uncertainty only matters for those who manage to find a partner with high enough resources.

6.4.4

Hypothesis 6.3/6.4 – Interaction between uncertainties

The idea that uncertainties increase commitment is based on the insight that people aim to decrease risks by staying with known partners. Accordingly, we hypothesized that both social and resource uncertainty have a positive effect on commitment and their effect is weakest when they are both low, and strongest when both are high. This is not what our analysis has uncovered: we found a strong negative interaction effect between the two types of uncertainties (F(1, 243)=5.96, p=.015). Accordingly, Hypothesis 6.4 has to be refuted. While commitment is lowest when both uncertainties are low (in accordance with Hypothesis 6.3), the effect of resource uncertainty turns negative when

6.4. Results

129

Figure 6.2: Commitment under low and high resource uncertainty

social uncertainty is high. A possible explanation is tied to the fact that both the subjects and the opponent cooperated less in the high social uncertainty condition. Many subjects in the high social uncertainty condition presumably did not aim to build long-term relationships, but instead tried to exploit their opponent and leave. In this case, it makes good sense to try and find opponents with increasingly large resources: the wider the range of resources, the more switching is expected. When we restricted our analysis to mutually cooperative relationships, the interaction effect between the two uncertainties disappeared.

6.4.5

Hypothesis 6.5/6.6 – Trust and Optimism

Stepping out of a committed relationship in favor of a new partner always involves taking some risk that the new partner will not live up to one’s expectations. General trust in people helps to overcome this reluctance and encourages interaction with unknown partners. To measure such general trust we used the 8-item scale created by Yamagishi et al. (1998). Whereas trust makes one optimistic about the trustworthiness of strangers, it does nothing to mitigate other concerns about switching to new partners. Switching to a new partner involves risks of losing the cooperativeness of the previous partner, the problem of ending up with a low-resource alternative partner, etc. In order to overcome such fears, one needs to have a more general sense of optimism. We measured this positive tendency using the 10-item

130

Chapter 6. Commitment and Networking under Uncertainty

Figure 6.3: Interaction between social and resource uncertainty

Revised Life Orientation Test, taken from Scheier et al. (1994). We found that commitment is negatively related to both trust (r(242)=– 0.219, p<0.001) and optimism (r(242)=–0.186, p<0.005), as expected by Hypotheses 6.5 and 6.6. The coefficients suggest that although both trust and optimism decrease commitment, neither one is a primary factor in determining commitment. In other words, someone does not switch partners only because he or she strongly trusts others, but trusting others certainly helps in making the decision. We also found a relatively strong positive association between trust and optimism (r(242)=0.479, p<0.001), suggesting that those who are optimistic in general, are also optimistic about the trustworthiness of strangers.

6.5

Robustness of results across different dilemmas

In order to test whether our results depend on actual payoffs of the dilemma, we used two different dilemmas derived from the Prisoner’s Dilemma, the Fear Game and the Greed Game (Simpson, 2003; Kuwabara, 2005). The advantage of each of these games is that they are simpler than the original Prisoner’s Dilemma because they separate otherwise confounded incentives. Since we did not want to confuse subjects by changing values in the payoff matrix during the experiment, the dilemma manipulation was also betweensubjects, just like the uncertainty manipulations. The results presented above

6.6. Conclusion and Discussion

131

are based on the entire pooled sample, i.e. for both games taken together. Restricting the analysis to the group of subjects for only one game typically resulted in effects becoming insignificant, due to the reduction in the number of observations. More precisely, there was no difference in the level of commitment between subjects playing in the Fear or Greed Fame (t(244)=0.227, p=0.821). Furthermore, multiple regression analyses showed no significant effect of the type of dilemma when controlling for the main effects of social and resource uncertainty manipulations; when controlling for the interaction effects between each uncertainty and type of dilemma; or when controlling for three-way interactions between manipulations. To sum it up, we did not find any inversion in the direction of the effects between different dilemmas. The only differences were reduced significance, apparently resulting from the decreased sample size. This suggests that results are stable across these two different dilemma situations.

6.6

Conclusion and Discussion

In this work we showed that in order to assess the effect of uncertainty on commitment, uncertainty needs to be broken down into at least two components, social uncertainty (related to trust), and uncertainty about resources. Each of these uncertainties may exist in a given exchange situation and increase commitment to previous partners. The underlying mechanism in the case of both uncertainties is that people are generally assumed to be risk-averse, in the sense that they prefer outcomes about which they have more information. Having interacted with a partner in the past provides such extra information in comparison with strangers. Note that especially in the case of resource uncertainty, increased variation among potential partners could induce exploration and decrease commitment, because larger variation gives an opportunity to find partners with higher resources. However, following Kahneman and Tversky’s (1979) argument that people are risk-averse when contemplating gains, we expected that on average a larger range of the resource distribution among potential partners would make subjects more reluctant to abandon sufficiently resourceful exchange partners. The experiments confirmed this expectation. Our results suggest that previous theories of social uncertainty and commitment have an important and unnoticed gap. These theories implicitly assume that focal actors are cooperative and have the opportunity to find cooperative partners. When one intends to cooperate and knows that not everyone else does so, it makes sense to establish a stable committed relationship with a partner who is a fellow cooperator. If one does not intend to cooperate or cannot find a fellow cooperator, social uncertainty could, in fact, have the op-

132

Chapter 6. Commitment and Networking under Uncertainty

posite effect on commitment. Obviously, a similar logic applies to resource uncertainty. We expected and found a significant positive effect of resource uncertainty on commitment when the subject managed to find a partner with sufficiently high resources. In other cases the effect was insignificant. We also replicated Yamagishi’s result that general trust in people, as a personality trait, helps to explain why some people show less commitment behavior than others. The introduction of uncertainty in a more general sense to the theory of commitment in exchange resulted in a more general trait of optimism, which we similarly found to decrease commitment. While trust mitigates the problem of social uncertainty, optimism probably counters uncertainty in the more general sense.

6.6.1

Limitations

Our work shows that resource uncertainty has a similar effect on commitment as social uncertainty. Yet, the comparability of these two uncertainties is hampered by the way we manipulated the two. Social uncertainty concerns the trustworthiness of partners, more precisely the probability that an unknown partner will cooperate. This probability ranges from 0 (never cooperate) to 1 (always cooperate). Resource uncertainty, on the other hand, was implemented by assigning a resource parameter to partners, which multiplied base payoffs. This parameter ranged from 1.00-x to 1.00+x, and thus 1.00 was a natural cut-off point, separating inflated and deflated payoffs. Consequently, the divide between above-average and below-average partners was clear with regard to resources but less so with regard to trustworthiness. Another, more substantial difference between the trustworthiness and the resource dimension in our experiment is that while the subjects did not differ in their resources from each other, they clearly differed from each other in their cooperativeness. Just as we did not wish to impose cooperativeness on the subject, we did not wish to impose resources either. This way, however, our argumentation had to omit an important factor, namely, that the effect of resource uncertainty could depend not only on the resource distribution among potential partners but also on the current amount of resources of the subject. Future work should clarify whether the effects of resource uncertainty that we found are robust if subjects themselves differ in resources.

Chapter 7

Conclusions The present work addresses a fundamental issue about human sociality: Why are human behavior and emotions towards long-term partners and relationships so often seemingly out of tune with rationality? Is there something fundamentally rational behind seemingly irrational commitment? These overarching questions are motivated by a vast and growing body of empirical evidence about the way people make decisions in long-term relationships. The evidence points to a mismatch between predictions of simple self-interested rationality and actual behavior that is influenced by a complex interaction between emotions and rational reasoning. To find an answer for these questions, we point to a common explanatory framework, evolutionary theory, which is capable of integrating theories about emotionality and rationality that would otherwise individually lead to different predictions about behavior in interpersonal relationships. In line with an important distinction in the evolutionary approach, we provided explanations on two levels of causality. First, building on existing research about conditions of the human ancestral environment, we advanced a computational model (Chapters 2-4) to test how a preference for interpersonal commitment could have evolved and out-competed various alternative preferences of opportunism and calculative rationality (ultimate explanation). Second, we turned to empirical examination of the ultimate explanation. We referred to existing evidence and then added to it our own cross-cultural findings. Together these suggest the existence of a direct emotional-cognitive mechanism (proximate explanation) producing commitment behavior, which is most easily explained as a remnant from a long gone era of evolutionary adaptation (Chapter 5). After this, we explicitly focused on exchange theoretic (non-evolutionary) explanations of commitment (Chapter 6) and resolved an empirical puzzle about uncertainty that arose from the contradiction between our cross-cultural findings and existing literature. 133

134

7.1

Chapter 7. Conclusions

Summary of results

To show that commitment could have been an evolved strategy, we simulated human interactions under assumed conditions of the human ancestral environment (Chapters 2-4) and found that strategies that possess a tendency for commitment outperform other strategies, such as fair reciprocation. Our results were stable across simulations, where individual strategies were matched against each other in an ecological competition, as well as in genetic simulations, where genotypes of mutating strategies contested their strengths under evolutionary selection pressures. In the ancestral environment people lived together in relatively small groups. Constant threats from a harsh natural environment led to a much higher frequency of life-threatening situations than today (Sterelny, 2003). In such an unpredictably “unfair” environment, imbalances between exchange partners cannot be avoided due to the uneven occurrence of hazards. Consequently, fair strategies suffer from their lack of tolerance when they interact with their own kind. A more lenient strategy of commitment avoids this pitfall, without being overly tolerant. Another reason behind the success of commitment lies in the structure of the exchange networks that it spontaneously causes to emerge. In a network of committed people, usually each person can easily decide whom to help and everyone is accounted for. Committed agents thus avoid overloading a few designated individuals with interaction requests and instead spontaneously create a structure that ensures an efficient coordination of help requests and help provision. Fair strategies on the other hand are inclined to keep their relationships strictly in balance, which results in spreading interaction requests evenly across the population. During times of need, this structure is inefficient because fair agents in small groups generate overlapping personal networks so that often too many ask help from the same agent at the same time. We also showed that the disadvantage of fair reciprocity increases as the environment becomes harsher (Chapter 3). But could it really have been viable in a harsh and unpredictable ancestral environment to stay committed to people with lower helping capability, instead of investing in relationships with more attractive others? In Chapter 4 we show that even when there are large differences between individuals in helping capability, it is still better to have a preference for helping old friends (commitment) than a preference for helping the most attractive others. Nevertheless, Chapter 4 also emphasizes the importance of fairness, which is probably another strong and cross-culturally stable preference (Fehr and Schmidt, 1999; Heinrich et al., 2001; Fehr et al., 2002). We found that under large inequalities in the population, a preference for fairness is more important for viability than commitment. In lack of a time machine, evolutionary (ultimate) theories are troublesome

7.1. Summary of results

135

to test directly. The next best solution is to derive proximate mechanisms from the ultimate theory, and test the existence of these proximate mechanisms. In addition, such a complex theory based on human evolution will only be attractive if it is able to explain more empirical anomalies than simpler theories that make fewer assumptions. We followed this path with regard to the ultimate theory of interpersonal commitment. If interpersonal commitment indeed evolved and stabilized through selection during countless years of ontogeny, we should be able to detect the resulting proximate mechanisms in contemporary societies. Is there a cross-culturally observable tendency to remain committed to previous partners? Does this tendency remain relatively stable across different situations where rationality would prescribe different behaviors, i.e. is it hardwired? In Chapter 5, we uncovered support for the notion that people possess such a commitment bias: they hold on to their partners simply as a result of exposure, not only as a result of instrumental benefits. We created an experiment where a rational actor would be indifferent between two interaction partners: one that cooperated in a large interaction, and another that cooperated in multiple small interactions over time. The total cost incurred by either partner is equal, therefore there is no reason to rationally suspect better intentions behind one than the other. Still, we found that people were more likely to choose the partner that cooperated over a period of time, suggesting that extended exposure in itself creates a force toward stabilizing a relationship, independently from the size of the benefit. Note that this exposure explanation for commitment might appear to be reducible to a simpler cognitive mechanism that is already proven to be universal among humans and many animals, reinforcement learning (Thorndike, 1911; Homans, 1961; Macy and Flache, 2002). Is it not so that people simply connect a stimulus of being exposed to another person with positive outcomes, and thus reinforce the need for further exposure? Not necessarily. The idea behind our exposure explanation is not that people are more likely to become committed to attractive/valuable partners but that repeated encounters, per se, increase the positive perception of the partner and this is what leads to increased commitment. Another finding of our experiment, reported in Chapter 5, was that uncertainty decreased commitment. This is largely in contradiction with another well-established mechanism that predicts more commitment (Kollock, 1994; Yamagishi and Yamagishi, 1994; Yamagishi et al., 1998). What is the reason for this opposite effect? This was the main question motivating our study reported in Chapter 6. The puzzle also leads to another question: is there an alternative reason based on uncertainty reduction for why people become committed, if not to avoid untrustworthy strangers? We found empirical support that social uncertainty has a much less universal effect on commitment than previously suggested (see Kollock, 1994; Yamagishi and Yamagishi, 1994; Ya-

136

Chapter 7. Conclusions

magishi et al., 1998), because it only affects people who themselves wish to be cooperative, and have an opportunity to meet fellow cooperators. Moreover, we also found that there is at least one other important source of uncertainty that has been neglected in the exchange-commitment literature: resources. Similarly to social uncertainty, resource uncertainty increases commitment, especially when one has an opportunity to meet a high-resource partner. Why are some people more committed than others? In order to explain some of the individual differences in commitment behavior, we linked the effect of resource and social uncertainty to psychological mechanisms. Building especially on Yamagishi’s works, we confirmed that general trust in people has a negative effect on the tendency to become committed to steady partners. Furthermore, we also showed that while general trust decreases commitment, optimism in a more general sense has a similar negative effect. Those who are generally optimistic are more likely to dissolve existing relationships and venture interaction with strangers. In sum, Chapters 2 to 6 answered each research question raised in Chapter 1. When taken together, these answers provide an answer to our overarching questions. Theoretical results from the agent-based computational models, as well as empirical results from the cross-cultural laboratory experiments, give support to the conjecture that humans possess an innate trait for commitment (or attachment) to relationship partners. This trait most likely evolved in the human ancestral environment where it served as an even stronger factor of success (and survival) than in contemporary societies. Although the balance of evidence tips toward an evolutionary explanation for commitment that integrates emotional mechanisms next to rational motives, it must be pointed out that (1) there are other, possibly evolved, preferences, which influence behavior in long-term relationships, and (2) an evolutionary framework is not always necessary to understand or predict commitment behavior. In Chapters 2 to 4 we found that other preferences, such as calculative reciprocity (fairness) have certain advantages compared to commitment, and under certain conditions outcompete it. Then, in Chapter 6 we specifically focused on advancing the rational (exchange theoretic) explanation for commitment, related to uncertainty. We could do so, because in the case of uncertainty reduction, fewer assumptions are sufficient to explain differences in commitment behavior.

7.2

General discussion

Does evolutionary theory help to explain aspects of contemporary human behavior, and commitment behavior in particular? If so, how does our work fit into the broader field of commitment research? What have we added to existing knowledge, and what are the novel aspects of our work? Finally, where do

7.2. General discussion

137

the limits of our theorizing lie? These are the questions covered in this section.

7.2.1

In defense of evolutionary theory in the social sciences

This dissertation advances an evolutionary explanation for why people make seemingly irrational decisions, based partially on emotions, about relationships. There is disagreement within the social sciences, especially in sociology, whether human behavior can at all be explained from an evolutionary perspective. While sociobiology has had great success in explaining seemingly irrational behaviors, such as altruism, among animals (Hamilton, 1964, 1972; Trivers, 1971, 1974; Maynard Smith, 1974), attempts to extend the same arguments to humans attracted fierce scientific debate (see Holcomb, 1993). Introducing a domain-specific approach in addition to the domaingenerality of sociobiology, a new field, evolutionary psychology, has created new momentum in this debate. Evolutionary psychology has proved successful in using evolutionary theory to derive and empirically corroborate a range of hypotheses about human preferences and behavior, such as cheater detection, mating preferences, language acquisition, incest avoidance, etc. One of the strengths of evolutionary psychology is its elegance in explaining many types of human behaviors that were previously thought to be simply irrational or erroneous. Examples range from explanations for cognitive biases (e.g. Error Management Theory, see Haselton and Buss, 2000; Haselton and Nettle, 2006) to sensory illusions (e.g. Evolved Navigation Theory, see Jackson and Cormack, 2006). These seemingly irrational tendencies are explained within a clear functional framework that is increasingly well supported by empirical evidence. Functional (or holistic) theories have been heavily criticized, leading many to argue that they are not real theories at all, failing to meet standards of the logical positivist philosophy of science. The root of the criticism is that a function, intention or goal, on which these theories are based, becomes manifest only at a later point in time. And as future things cannot be considered antecedent conditions, functional explanations cannot be considered causal ones (Looijen, 1998). It is important to realize that evolutionary theory, which modern biology is based upon, is not the usual kind of functional theory. In fact, arguments based on evolutionary selection only appear to be functional but in fact they are perfectly acceptable, causally adequate theories. To understand why, consider the following simple idea: Putting aside the question of how life originally appeared, we can formulate two basic assumptions: 1. Living organisms can only be created by other living organisms, through

138

Chapter 7. Conclusions reproduction1 .

2. When organisms (parents) reproduce, the new organism (offspring) will resemble the original organism to a high degree (inheritance). Now, an argument, for example, that an offspring has an eye in order to see and thus better be able to survive, appears to be a functional explanation, since the offspring had to have an eye before the function of seeing came into existence. But taken together with the previous two assumptions, the argument can easily be turned into a perfectly valid, causal explanation: an eye served the parent well, so much that it managed to stay alive and reproduce, creating an offspring, which then inherited the eye. This explanation is not only causally correct but benefits from the additional credibility it gains from the functionalist aspect. This evolutionary functionalism becomes a substitute for the optimality (e.g. maximal utility) that rational choice theories usually require to explain behavior. Our explanation of interpersonal commitments draws on the increased functional value that committed relationships had in our evolutionary past, in order to explain their existence in situations when immediate rational optimality is missing.

7.2.2

Placing our work

How do our model and findings fit into a theoretical framework of interpersonal relationships in the ancestral environment? Table 7.1 shows a rough classification of important interpersonal relationship types that are likely to have been present in the ancestral environment, along with the ultimate challenge they addressed (i.e. adaptive benefit) and some of the underlying emotions they are associated with (proximate mechanisms). Notice that the type of commitment that our simulations most closely resemble is friendship. A specific adaptation for commitment is especially important in the case of this non-kin, non-reproductive type of relationship. However, note also that once a trait for interpersonal commitment is in place it has a stabilizing effect on all types of relationships listed. This notion is indeed supported by research arguing that interpersonal experiences in infancy with close kin act as a foundation for the capacity for stable relationships with non-kin in adults (Lundeen, 1999).

1 Note that cloning constitutes a special exception from this argument, and shows that technology clearly has the potential to disrupt the natural dynamic of evolutionary selection. Nevertheless, this in itself does not invalidate the subsequent argument about our prehistoric evolutionary past.

attach-

sense of fairness

bonding, ment

sexual desire passionate love, sexual desire, attachment familial love

creation of offspring creation of offspring with increased fitnessa increasing fitness for a subset of own genesb increase own fitness through unconditional exchange, especially high-value exchange such as saving one’s life increase own fitness through reciprocal exchange, especially repeated, accountable exchange

proximate mechanisms

adaptive benefit

uncertainty reduction

and

commitment bias

a There is mounting evidence that long-term reproductive relationships increase the survival chances of offspring not only when they are young but even before they are born. Research on contemporary hunter-gatherer societies shows that children between 1 and 5 years of age are 2.6 times more likely to die if their fathers are dead than if their fathers are alive, and 2.9 times more likely to die if their parents are divorced than if they are together (Hurtado and Hill, 1992). Moreover, it has been found that sexual cohabitation prior to conception reduces the chance from 40% to less than a 5% of preeclampsia, a condition that could even lead to the death of the fetus (cf. Pillsworth and Haselton, 2005). b This happens through the a mechanism known as kin selection (see Hamilton, 1964, 1972).

Table 7.1: Typical relationship types in the ancestral environment, their adaptive benefit and proximate mechanisms

acquaintance

friend

kin

relationship type sexual partner spouse

7.2. General discussion 139

140

Chapter 7. Conclusions

There are two perspectives in the existing literature about the seemingly irrational tendency to hold on to long-term partners. One is to regard it as a devotion or promise to stay in a relationship, that is kept beyond rational incentives, being internally enforced by a host of emotions such as love, attraction or a desire to keep a clean conscience. This line is apparent in the works of Frank (1988) and Nesse (2001a). The other approach is to regard this seemingly irrational tendency as a basic emotional attachment that is a fundamental characteristic of humans, extended not only towards other people but also toward objects and abstract ideas. This line is advocated in particular by Depue and Morrone-Strupinsky (2005) and Pedersen (2004). Note that while we tried to give credit to both approaches throughout this work, the use of the word “commitment” unfairly biases attention toward the first approach. The argument for the evolutionary origins of interpersonal commitment is increasingly credible and convincing, being able to rely on findings from other disciplines. Outside the social sciences there are at least two important disciplines that provide a synergy for the evolutionary argument: primatology and the cognitive neurosciences. Although several animal species, especially birds and some mammals, exhibit long-term relationships among kin or monogamous sex partners (Hrdy, 1999; Carter, 1998), it is among primates that emotionally attached relationships are especially prominent. Friendships and other alliances between individuals who are not relatives or sexual partners have been identified and found to increase social complexity (De Waal, 1996). Studies suggest that early human language evolved in order to facilitate long-term cooperative interpersonal relationships. Researchers found a clear positive association between group size and the call-repertoire of primates, as well as between grooming and call-repertoire, suggesting that the original purpose of language was to efficiently groom multiple individuals at the same time (McComb and Semple, 2005). Proliferation of facial expressions, gestures, vocalizations and other types of social communication created a need for increased cognitive capabilities, and possibly led to the increase of intelligence in human ancestors (Pedersen, 2004). Fortunately, the cognitive neurosciences are also increasingly interested and capable of uncovering mechanisms in the brain related to sociality in general, and bonding and attachment in particular. Studies found that when people fall in love, serotonin levels plummet and the brain’s reward centers are flooded with dopamine. This gives an emotional high similar to an addictive drug, creating powerful links in our minds between pleasure and the object of our affection, making people addicted to the loved one (Aron et al., 2005). Other hormones, such as oxytocin and vasopressin, express their effect later in the relationship, and are crucial in forming long-term partnerships. Researchers found that couples that have been together for several years show

7.2. General discussion

141

increased brain activity associated with these chemicals, when they looked at pictures of their long-term partner (Aron et al., 2005). When administering oxytocin to experimental subjects, researchers were able to artificially increase trust in social interactions (Kosfeld et al., 2005). Oxytocin is also known to be produced when couples have sex and touch, kiss and massage each other. There is also evidence that similar biological mechanisms are triggered in the absence of a romantic partner as in infants who are separated from their mother, leading to various levels of separation anxiety (Carter, 1998). A consensus is starting to emerge that our neural systems exhibit built-in functions that are designed to respond to the presence or absence of social bonds (Depue and Morrone-Strupinsky, 2005).

7.2.3

Innovations of the present work

This work is an attempt to reconcile existing models of rationality with people’s seemingly irrational, emotionally based tendency to keep existing relationships going. As such, it is among the first attempts to combine deep-seated emotional preferences and rationality into an evolutionary argument that explains interpersonal commitments. While there have already been forceful attempts to argue for the evolutionary origins of a more general form of commitment, the tendency to uphold promises and threats (Frank, 1988; Nesse, 2001a), interpersonal commitment has received surprisingly little attention. This is even more surprising, when one looks closer at Frank’s and Nesse’s work. In one of their key examples they both argue that humans have evolved a capacity that facilitates keeping promises through emotions, which explains why people remain committed to their spouses, despite more “rational” alternatives. But is it not more plausible that humans have evolved the capacity to remain in relationships directly, not as a bi-product of honesty, avoidance of guilt, etc.? Our work is also pioneering in testing competitive advantages of commitment and calculative reciprocity under different conditions of social inequality. Our simulation studies (Chapters 2-4) outline a possible evolutionary pathway for the emergence of commitment behavior under minimalistic assumptions about the human ancestral environment, strengthening the ultimate answer for our original overarching research question. To our knowledge, the mere exposure effect has not been linked to interpersonal commitment before, despite their very obvious connection. Our core argument is that (1) repeated exposure to the same person increases positive evaluations and trust toward the person, holding actual positive experiences and uncertainty about trustworthiness constant; (2) this necessarily leads to becoming more committed to the person; and (3) such behavior was especially adaptive in the ancestral environment where people were much more reliant on the help of other individuals. To strengthen the leap from (2) to (3)

142

Chapter 7. Conclusions

we also rule out cultural explanations for the existence of such a tendency by replicating our experiments in three countries, the Netherlands, China and the USA. Our main endeavor to reconcile emotionality with rationality concentrates on describing and examining the emotional aspects of decision-making. In addition, we also advance existing explanations for the instrumental rationality of interpersonal commitment by refining its link with uncertainty. Previous literature concentrated exclusively on the positive effect of social uncertainty on commitment. We argue that the effect of social uncertainty is less general than previous literature suggested and point to another form of uncertainty, about resources, that similarly leads to commitment behavior.

7.2.4

Possible criticism

A key source of criticism of our work is that we advocate an evolutionary explanation for social behavior. This stems from two more specific problems, one methodological and one substantial. First, evolutionary explanations, including the original ones by Darwin (1859), are difficult to empirically test because we are short of direct evidence about actual evolutionary trajectories. The strength of the evolutionary framework itself lies in the countless sub-theories that are all based on the simple but powerful dynamics of reproduction, mutation and selection and manage to give a coherent explanation for what we see in the biological world around us today. Another, more specific problem is that by explaining human behavior through evolved preferences we implicitly refer to underlying biological mechanisms. This is bound to draw fierce criticism, especially from those who argue for the primacy of culture and society as determinants of human behavior. But the idea that behavior is biologically determined to a certain extent, is receiving increasing support from research in the cognitive neurosciences. Whenever relevant, we pointed out such links throughout preceding chapters and in Section 7.2.2 above. One related epistemological problem is that we argue for the existence and relative strength of a universal human characteristic. Research in the social and behavioral sciences usually tries to pinpoint and explain differences between individuals. In our case, we argue about a difference between the entire population and a fictive reference group. More precisely, we need to argue that everyone is universally more committed than would be rational, but there is no real “rational population” to compare to. Therefore, what we tried to test (in Chapter 5) is whether people behave differently under different conditions that otherwise do not differ when viewed from a purely rational perspective. We found support for the idea that this difference in behavior is cross-culturally stable. An explanation based on natural selection does not posit that every member of the population

7.2. General discussion

143

necessarily possesses the evolved trait, nor that the trait is equally manifest in each individual. Rather, it argues about the population mean or frequency of the trait, which itself is subject to selection, leaving ample room for individual differences. Although showing such individual differences was not our main focus, we uncovered support in Chapter 6 that commitment is systematically related to individual personality characteristics, such as trust and optimism. Another major point for criticism is that interpersonal commitment, as a collective concept for many types of relationships that people may have, is too general. It could be argued that an explanation for friendships should have little or nothing in common with an explanation for marriage. This is, however, precisely what we propose. Although different types of relationships have different purposes, there are many common features that make them inherently similar. First, relationships by definition comprise repeated interaction. Second, there is always a trust issue between partners. Third, many relationships require exclusivity, which leads to a dilemma of choice between alternative partners. The fact that our proposed mechanisms, emotional commitment, mere exposure, and uncertainty reduction are assumed to exist across most interpersonal relationships makes them very general. Therefore, it is important to note that we do not argue for the exclusivity of these mechanisms in producing each of these relationship types but regard them as a few of many factors that create and stabilize interpersonal commitments. Indeed, there is indication in the literature that our concept of emotional commitment is not broad enough. Recent work argues that humans have evolved a general emotional attachment drive, which in itself helps to develop a bond not only to children, sexual partners and groups, but also to cultural ideas and abstract concepts as well, resulting in the evolution of love and increased human intelligence (Pedersen, 2004). Moreover, recent laboratory experiments show that children develop strong emotional preference for objects they have become attached to, independent of the objects’ physical characteristics, and are unwilling to substitute them for perfectly identical duplicates (Hood and Bloom, 2007).

7.2.5

Limitations

A general limitation of the theoretical part of our work is that although our evolutionary simulations were capable of examining and comparing millions of different strategies that randomly emerged and competed, we still assumed a mental model only along the dimensions of commitment, fairness, cooperation and attractiveness. We had to do so in order to reduce the complexity of our model, and to retain the interpretability of the results. However, this also reduced the complexity of strategies explored, and could have inflated the success of commitment strategies. Moreover, the simulations do not tell us much about how feasible it was to initially invent the idea of commitment.

144

Chapter 7. Conclusions

A second limitation lies in the difficulty of separating mere exposure from uncertainty reduction. There are two components that create uncertainty reduction in committed relationships, one is instrumental, the other is temporal. Displaying one’s trustworthiness through sacrificing one’s own immediate interest for the sake of the relationship is the instrumental component. Reiterating the display of trustworthiness over an extended period of time is the temporal component. While we argued for the effect of mere exposure on commitment by separating it from the instrumental component of uncertainty reduction (Chapter 5), we did not separate it from the temporal component. This raises the question whether mere exposure is an independent factor or simply an aspect of uncertainty reduction.

7.2.6

Avenues for future research

To address the first of the above limitations, the simulation model can be further extended by making the model of the individual agent less specific and more flexible. A more realistic model based on our current understanding of how the human brain functions is perhaps the most attractive direction. One of the currently applied paradigms for this approximation is neural networks (see, for example Haykin, 1994). Its drawback is an exponential increase in required computational power compared to our existing models. The second limitation is more difficult to address. In order to separate the effect of mere exposure and the temporal component of uncertainty reduction, a more precise, controlled measurement or manipulation of uncertainty is required. Alternatively, the examination of uncertainty reduction from an evolutionary perspective seems desirable. Our current theory classified the effect of uncertainty on commitment under the rational motives. It is possible, however, that the avoidance of uncertainty itself is an evolved preference. Further examination of its theoretical viability and empirical characteristics could shed light on the credibility of this conjecture. Our volume is among the initial attempts (see also Baumeister and Leary, 1995; de Vos et al., 2001; Pedersen, 2004) to provide an evolutionary account for interpersonal commitment. It outlines an ultimate theory and at least one proximate mechanism for how natural selection might have shaped a capacity and willingness to become committed to long-term partners. To make the evolutionary claim stronger and more credible, much further research is needed. One promising direction is to study how people cooperate in different types of relationships, such as kinship, friendship and acquaintanceship. For example, it is likely that each type of relationship is designed to solve cooperation problems of a different size. Such work would help to further our understanding of the relative importance and place of interpersonal commitment among other arguably cross-culturally stable tendencies of fair reciprocity, and attachment toward kin. At the same time, it could provide further evidence

7.2. General discussion

145

for the evolutionary account of social bonding. An important issue that we have not explored in this volume is the counter-mechanism of mere exposure: satiation. Although mere exposure to a stimulus creates a positive disposition, it can also lead to boredom or even aversion from the stimulus. The same could probably be argued for interpersonal relationships in some contexts. Thus it is likely that there is a mechanism that acts in the opposite direction as our commitment bias. Deriving such a mechanism from the evolutionary theory of interpersonal commitment requires further theoretical work and empirical examination. Despite these limitations and given the potential criticism against our approach that we discussed above, we believe that our work has made a contribution to the comprehension of seemingly irrational decisions in durable relationships. We found support for the notion that people instinctively stick to their existing interpersonal relationships, more so than would seem rational given the circumstances. We argued that this tendency could be the result of a long-term evolutionary process. Furthermore, we advanced previous research on the relationship between commitment and one of its key rational sources, uncertainty. Our efforts also testify to the importance of interdisciplinary research. Without combining previous research and insights from psychology, sociology, economics and evolutionary theory, most alternative explanations of commitment remain limited in their power and scope. Together, they promise to further our understanding of the wonderful and mysterious complexity of human nature.

Bibliography Abreu, D. (1988). On the theory of infinitely repeated games with discounting. Econometrica, 56(2):383. Adams, J. M. and Jones, W. H., editors (1999). Handbook of interpersonal commitment and relationship stability. New York: Academic/Plenum. Akerlof, G. (1970). The market for ‘lemons’: Qualitative uncertainty and the market mechanism. Quarterly Journal of Economics, 84:488–500. Arkes, H. R. and Blumer, C. (1985). The psychology of sunk cost. Organizational Behavior and Human Decision Processes, 35:124–140. Aron, A., Fisher, H., Mashek, D. J., Strong, G., Li, H., and Brown, L. L. (2005). Reward, motivation, and emotion systems associated with early-stage intense romantic love. Journal of Neurophysiology, 94:327–337. doi:10.1152/jn.00838.2004. Axelrod, R. (1984). The Evolution of Cooperation. Basic Books, New York. Back, I. and Flache, A. (2006). The viability of cooperation based on interpersonal commitment. Journal of Artificial Societies and Social Simulation, 9(1):12. Back, I. and Flache, A. (2007). Commitment, fairness and inequality. In Hernández, C., Troitzsch, K. G., and Edmonds, B., editors, Social Simulation: Technologies, Advances and New Discoveries. Idea Group Inc. Back, I. and Flache, A. (Forthcoming). The adaptive rationality of interpersonal commitment. Rationality and Society. Barkow, J., Cosmides, L., and Tooby, J., editors (1992). The Adapted Mind: Evolutionary Psychology and the generation of culture. Oxford University Press, Oxford. Baumeister, R. F. and Leary, M. R. (1995). The need to belong: desire for interpersonal attachments as a fundamental human motivation. Psychological Bulletin, 117(3):497– 529. Beaman, V., Annest, J., Mercy, J., Kresnow, M., and Pollock, D. (2000). Lethality of firearm-related injuries in the united states population. Annals of Emergency Medicine, 35(3):258–66.

147

148

Bibliography

Becker, H. S. (1960). Notes on the concept of commitment. American Sociological Review, 66:32–40. Bendor, J. and Swistak, P. (2001). The evolution of norms. American Journal of Sociology, 106:1493–545. Berger, C. R. and Bradac, J. J. (1982). Language and social knowledge: Uncertainty in interpersonal relationships. London: Edward Arnold. Berger, C. R. and Calabrese, R. J. (1975). Some explorations in initial interaction and beyond: Toward a developmental theory of interpersonal communication. Human Communication Research, 1:99–112. Berger, C. R. and Gudykunst, W. B. (1991). Uncertainty and communication. pages 21–66. Norwood, NJ: Ablex. Binmore, K. (1998). Review of “R. Axelrod, the complexity of cooperation: Agent-based models of competition and collaboration; Princeton: Up 1997”. Journal of Artificial Societies and Social Simulation, 1(1). http://www.soc.surrey.ac.uk./JASSS/1/1/review1.html. Blau, P. M. (1967). Exchange and power in social life. New York, Wiley. Bolton, G. and Ockenfels, A. (2000). ERC – a theory of equity, reciprocity and competition. American Economic Review, 90(1):166–193. Boster, F., Rodriguez, J., Cruz, M., and Marshall, L. (1995). The relative effectiveness of a direct request message and a pregiving message on friends and strangers. Communication Research, 22(4):475–484. Boyd, R. and Richerson, P. J. (1985). Culture and the Evolutionary Process. Chicago, IL: University of Chicago Press. Brockner, J. (1992). The escalation of commitment to a failing course of action: Toward theoretical progress. Academy of Management Review, 12:39–61. Buskens, V. and Raub, W. (2002). Embedded trust: Control and learning. Advances in Group Processes, 19:167–202. Carter, C. (2005). Biological perspectives on social attachment and bonding. In Carter, C., Ahnert, L., Grossmann, K., Hrdy, S., Porges, S., and Sachser, N., editors, Attachment and Bonding. A New Synthesis, pages 85–100. MIT Press. Carter, C. S. (1998). Neuroendocrine perspectives on social attachment and love. Psychoneuroendocrinology, 23:779–818. Cashdan, E. (1985). Coping with risk: Reciprocity among the Basarwa of Northern Botswana. Man, 20(3):454–474. Cassidy, J. and Shaver, P. R. (1999). Handbook of attachment: theory, research, and clinical applications. The Guilford Press, New York.

Bibliography

149

Chapais, B. (2001). Primate nepotism: what is the explanatory value of kin selection? International Journal of Primatology, 22:203–220. Cohen, A. and Lowenberg, G. (1990). A reexamination of the side-bet theory as applied to organizational commitment: a meta-analysis. Human relations, 43:1015–1050. Cook, K. and Whitmeyer, J. (1992). Two approaches to social structure: Exchange theory and network analysis. Annual Review of Sociology, 18:109–27. Cook, K. S. and Emerson, R. M. (1978). Power, equity and commitment in exchange networks. American Sociological Review, 43:721–39. Cosmides, L. (1989). The logic of social exchange: Has natural selection shaped how humans reason? Studies with the Watson selection task. Cognition, 31:187–276. Cosmides, L. and Tooby, J. (1993). Cognitive adaptations for social exchange. In Barkow, J. H., Cosmides, L., and Tooby, J., editors, The Adapted Mind: Evolutionary Psychology and the Generation of Culture, pages 163–228. Oxford. Darwin, C. (1859). On the origin of species by means of natural selection, or the preservation of favoured races in the struggle for life. Atheneum, London, 1967 edition. Dawes, R. M. (1980). Social dilemmas. Annual Review of Psychology, 31:169–193. de Kwaadsteniet, E. W., van Dijk, E., Wit, A., and de Cremer, D. (2006). Social dilemmas as strong versus weak situations: Social value orientations and tacit coordination under resource size uncertainty. Journal of Experimental Social Psychology, 42(4):509– 516. de Vos, H., Smaniotto, R., and Elsas, D. A. (2001). Reciprocal altruism under conditions of partner selection. Rationality and Society, 13(2):139–183. de Vos, H. and Zeggelink, E. P. H. (1997). Reciprocal altruism in human social evolution: the viability of reciprocal altruism with a preference for ’old-helping-partners’. Evolution and Human Behavior, 18:261–278. De Waal, F. (1996). Good natured: the origins of right and wrong in humans and other animals. Harvard University Press, Cambridge. Depue, R. A. and Morrone-Strupinsky, J. V. (2005). A neurobehavioral model of affiliative bonding: Implications for conceptualizing a human trait of affiliation. Behav Brain Sci, 28(3):313–350. Dugatkin, L. A. (1997). The evolution of cooperation. Bioscience, 47(6):355–63. Ellis, H. (1986). Processes underlying face recognition. In Bruyer, R., editor, The Neuropsychology of face perception and facial expression. Mahwah, NJ: Lawrence Erlbaum. Emerson, R. M. (1972). Exchange theory, part 1: A psychological basis for social exchange. In Berger, J., M. Zelditch, J., and Anderson, B., editors, Sociological Theories in Progress, volume 2, pages 38–57. Boston: Houghton-Mifflin.

150

Bibliography

Falk, A., Fehr, E., and Fischbacher, U. (2001). Driving forces of informal sanctions. In NIAS Conference Social Networks, Norms and Solidarity. Fantz, R. (1963). Pattern vision in newborn infants. Science, 140:296–297. Fehr, E., Fischbacher, U., and Gächter, S. (2002). Strong reciprocity, human cooperation, and the enforcement of social norms. Human Nature, 13(1):1–25. Fehr, E. and Gächter, S. (2002). Altruistic punishment in humans. Nature, 415:137–140. Fehr, E. and Schmidt, K. (1999). A theory of fairness, competition and cooperation. Quarterly Journal of Economics, CXIV:817–851. Festinger, L., Schachter, S., and Back, K. (1950). Social pressures in informal groups: A study of a housing community. Palo Alto, CA: Stanford University Press. Fishman, M. A. (2003). Indirect reciprocity among imperfect individuals. Journal of Theoretical Biology, 225(3):285–292. Flache, A. (2001). Individual risk preferences and collective outcomes in the evolution of exchange networks. Rationality and Society, 13(3):304–348. Flache, A. and Hegselmann, R. (1999a). Altruism vs. self-interest in social support. computer simulations of social support networks in cellular worlds. In Advances in Group Processes, volume 16, pages 61–97. Flache, A. and Hegselmann, R. (1999b). Rationality vs. learning in the evolution of solidarity networks: A theoretical comparison. Computational and Mathematical Organization Theory, 5(2):97–127. Flache, A. and Macy, M. (2005). Social life from the bottom up: Agent modeling and the new sociology. In Social Dynamics: Interaction, Reflexivity and Emergence. Proceedings of the Agent 2004 Conference. Argonne National Laboratory, The University of Chicago. Flap, H. and Völker, B., editors (2004). Creation and Returns of Social Capital. A New Research Program. London and New York: Routledge. Foley, R. A. (1996). An evolutionary and chronological framework for human social behaviour. In Runciman, W. G., Smith, J. M., and Dunbar, R. I. M., editors, Evolution of Social Behaviour Patterns in Primates and Man. Oxford University Press. Frank, R. (2001). Cooperation through emotional commitment. In Nesse, R. M., editor, Evolution and the Capacity for Commitment. Russel Sage Foundation, New York. Frank, R. H. (1988). Passions Within Reason. W. W. Norton & Company. Friedlander, B. (1970). Receptive language development in infancy. Merrill-Palmer Quarterly, 16:7–51. Friedman, J. (1971). A non-cooperative equilibrium for supergames. Review of Economic Studies, 38:1–12.

Bibliography

151

Garcia, J., Ervin, F. R., and Koelling, R. A. (1966). Learning with prolonged delay of reinforcement. Psychonomic Science, 5:121–122. Geary, D. C. (2000). Evolution and proximate expression of human paternal investment. Psychological Bulletin, 1:55–77. Gigerenzer, G. and Todd, P. (1999). Simple Heuristics That Make Us Smart. New York: Oxford University Press. Gintis, H. (2003). Solving the puzzle of prosociality. Rationality and Society, 15(2):155– 187. Goode, W. (1960). A theory of role strain. American Sociological Review, 25:483–496. Güth, W. and Kliemt, H. (1998). The indirect evolutionary approach. Rationality and Society, 10(3):377–399. Güth, W. and Ockenfels, A. (2002). The coevolution of morality and legal institutions an indirect evolutionary approach. Discussion Papers on Strategic Interaction 200206, Max Planck Institute of Economics, Strategic Interaction Group. Available at http://ideas.repec.org/p/esi/discus/2002-06.html. Hamilton, W. (1964). The genetical evolution of social behaviour, i-ii. Journal of Theoretical Biology, 7:1–16. Hamilton, W. (1972). Altruism and related phenomena, mainly in the social insects. Annual Review of Ecology and Systematics, 3:193–232. Haselton, M. G. and Buss, D. M. (2000). Error management theory: A new perspective on biases in cross-sex mind reading. Journal of Personality and Social Psychology, 78:81– 91. Haselton, M. G. and Nettle, D. (2006). The paranoid optimist: An integrative evolutionary model of cognitive biases. Personality and Social Psychology Review, 10(1):47–66. Hayashi, N. and Yamagishi, T. (1998). Selective play: Choosing partners in an uncertain world. Personality and Social Psychology Review, 2:276–289. Haykin, S. (1994). Neural Networks: A Comprehensive Foundation. Macmillan College Publishing. Hegselmann, R. (1996). Solidarität unter ungleichen. In Hegselmann, R. and Peitgen, H.-O., editors, Modelle sozialer Dynamiken — Ordnung, Chaos und Komplexität, pages 105–128. Hölder–Pichler–Tempsky, Wien. Hegselmann, R. (1998). Experimental ethics – a computer simulation of classes, cliques and solidarity. In Fehige, C. and Wessels, U., editors, Preferences, pages 298–320. De Gruyter, Berlin. Heinrich, J., Boyd, R., Bowles, S., Camerer, C., Fehr, E., Gintis, H., and McElreath, R. (2001). In search of homo economicus: behavioral experiments in 15 small-scale societies. American Economic Review Papers and Proceedings, 91:73–78.

152

Bibliography

Helper, S. and Levine, D. (1992). Long-term supplier relations and product-market structure. Journal of Law, Economics and Organization, 8:561–81. Henry, S., Lehrman, S., and Motte, P. (1997). Losing yourself on-line. Health, 11(8). Holcomb, H. (1993). Sociobiology, Sex, and Science. Albany: State University of New York Press. Homans, G. L. (1961). Social Behavior: Its Elementary Forms. New York: Harcourt Brace Jovanovich. Hood, B. M. and Bloom, P. (2007). Children prefer certain individuals over perfect duplicates. Cognition. doi:10.1016/j.cognition.2007.01.012. Hrdy, S. B. (1999). Mother nature: a history of mothers, infants, and natural selection. Pantheon Books, New York. Hurtado, A. M. and Hill, K. R. (1992). Paternal effect on offspring survivorship among ach’e and hiwi hunter-gatherers: Implications for modeling pair-bond stability. In Father-child relations: Cultural and biosocial contexts, pages 31–55. Hawthorne, NY: Aldine de Gruyter. Jackson, R. E. and Cormack, L. K. (2006). predicted by evolved navigation theory. http://journalofvision.org/6/6/962/.

Previously unknown illusion Journal of Vision, 6(6):962a.

Johnson, D. J. and Rusbult, C. E. (1989). Resisting temptation: devaluation of alternative partners as a means of maintaining commitment in close relationships. Journal of personality and social psychology, 57(6):967–980. Johnson, M. P. (1973). Commitment: A conceptual structure and empirical application. The Sociological Quarterly, 14:395–406. Johnson, M. P. (1991). Commitment to personal relationships. In Jones, W. H. and Perlman, D., editors, Advances in personal relationships: A research annual, volume 3, pages 117–143. London: Jessica Kingsley. Kahneman, D. and Tversky, A. (1979). Prospect theory - analysis of decision under risk. Econometrica, 47(2):263–291. Kahneman, D. and Tversky, A. (1996). On the reality of cognitive illusions. Psychological review, 103(3):582–591. Karlsson, N., Gärling, T., and Bonini, N. (2005). Escalation of commitment with transparent future outcomes. Experimental Psychology, 52(1):67–73. Karremans, J., Lange, P. V., Ouwerkerk, J., and Kluwer, E. (2003). When forgiving enhances psychological well-being: the role of interpersonal commitment. Journal of Personality and Social Psychology, 84(5):1011–1026. Kent, S. (1993). Sharing in an egalitarian Kalahari community. Man, 3:479–514.

Bibliography

153

Kipnis, A. (2002). Practices of guanxi production and practices of ganqing avoidance. In Gold, T., Guthrie, D., and Wank, D., editors, Social Connections in China: Institutions, Culture and the Changing Nature of Guanxi. Cambridge: Cambridge University Press. Kiyonari, T. and Yamagishi, T. (1996). Distrusting outsiders as a consequence of commitment formation. Japanese Journal of Experimental Social Psychology, 36:56–67. Kollock, P. (1993). An eye for an eye leaves everyone blind - cooperation and accounting systems. American Sociological Review, 58(6):768–786. Kollock, P. (1994). The emergence of exchange structures: An experimental study of uncertainty, commitment, and trust. American Journal of Sociology, 100(2):313–45. Kosfeld, M., Heinrichs, M., Zak, P. J., Fischbacher, U., and Fehr, E. (2005). Oxytocin increases trust in humans. Nature, 435(7042):673–676. Kuwabara, K. (2005). Nothing to fear but fear itself? Fear of fear, fear of greed and gender effects in two-person asymmetric social dilemmas. Social Forces, 84(2):1257– 1272. Lawler, E. and Yoon, J. (1993). Power and the emergence of commitment behavior in negotiated exchange. American Sociological Review, 58(4):465–481. Lawler, E. and Yoon, J. (1996). Commitment in exchange relations: Test of a theory of relational cohesion. American Sociological Review, 61(1):89–108. Lawler, E. J. (2001). An affect theory of social exchange. American Journal of Sociology, 107(2):321–52. Lawler, E. J., Thye, S., and Yoon, J. (2000). Emotion and group cohesion in productive exchange. American Journal of Sociology, 106:616–57. Levinger, G. (1999). Duty toward whom? Reconsidering attractions and barriers as determinants of commitment in a relationship. In Adams, J. M. and Jones, W. H., editors, Handbook of interpersonal commitment and relationship stability, pages 37–52. New York: Kluwer Academic/Plenum. Lin, N. (1999). Building a network theory of social capital. Connections, 22(1):28–51. Looijen, R. C. (1998). Holism and reductionism in biology and ecology : the mutual dependence of higher and lower level research programmes. PhD thesis, University of Groningen, the Netherlands. Lundeen, M. G. (1999). Interpersonal experience in infancy as a foundation for capacity in adults for stable relationships. In Adams, J. M. and Jones, W. H., editors, Handbook of interpersonal commitment and relationship stability, pages 91–108. New York: Kluwer Academic/Plenum. Macy, M. and Flache, A. (2002). Learning dynamics in social dilemmas. Proceedings of the National Academy of Sciences U.S.A, 99(10):7229–36.

154

Bibliography

Macy, M. W. and Willer, R. (2002). From factors to actors: Computational sociology and agent-based modeling. Annual Review of Sociology, 28:143–66. Mann, L. (1980). Cross-cultural studies of small groups. In Triandis, H. and Brislin, R., editors, Handbook of cross-cultural psychology: Social psychology, volume 5, pages 155–209. Boston: Allyn and Bacon. Martin, R. D. (2003). Human reproduction: A comparative background for medical hypotheses. Journal of Reproductive Immunology, 59:111–135. Maynard Smith, J. (1974). The theory of games and the evolution of animal conflicts. Journal of Theoretical Biology, 47:209–221. Mayr, E. (1961). Cause and effect in biology: Kinds of causes, predictability, and teleology are viewed by a practicing biologist. Science, 134:1501–1506. McComb, K. and Semple, S. (2005). Coevolution of vocal communication and sociality in primates. Biology Letters, 1:381–385. Messick, D. M., Allison, S. T., and Samuelson, C. D. (1988). Framing and communication effects on group members’ responses to environmental and social uncertainty. In Maital, S., editor, Applied Behavioral Economics, volume 2, pages 677–700. New York: New York University Press. Meyer, J. P. and Allen, N. J. (1991). Component conceptualization of organizational commitment. Human Resource Management Review, 1:61–89. Meyer, J. P. and Herscovitch, L. (2001). Commitment in the workplace: Toward a general model. Human Resource Management Review, 11:299–326. Mills, J. R. and Clark, M. S. (1982). Communal and exchange relationships. In Wheeler, L., editor, Annual Review of Personality and Social Psychology. Beverly Hills, Calif.: Sage. Mills, J. R. and Clark, M. S. (1994). Communal and exchange relationships: Controversies and research. In Erber, R. and Gilmour, R., editors, Theoretical Frameworks for Personal Relationships. Hillsdale, N.J.: Lawrence Erlbaum. Monahan, D. J. and Hooker, K. (1997). Caregiving and social support in two illness groups. Social work, 42(3):278. Monterosso, J., Ainslie, G., Mullen, P. A. C. P. T., and Gault, B. (2002). The fragility of cooperation: A false feedback study of a sequential iterated prisoner’s dilemma. Journal of Economic Psychology, 23(4):437–448. Moorman, R. H. and Blakely, G. L. (1995). Collectivism as an individual difference predictor of organizational citizenship behavior. Journal of Organizational Behavior, pages 127–142. Moreland, R. L. and Beach, R. (1992). Exposure effects in the classroom: The development of affinity among students. Journal of Experimental Social Psychology, 28:255–276.

Bibliography

155

Mowday, R., Porter, L., and Steers, R. (1982). Employee-Organizational Linkages. New York: Academic Press. Nahemow, L. and Lawton, M. P. (1975). Similarity and propinquity in friendship formation. Journal of Personality and Social Psychology, 32:205–213. Nesse, R. M., editor (2001a). Evolution and the Capacity for Commitment. New York : Russell Sage Foundation. Nesse, R. M. (2001b). Natural selection and the capacity for subjective commitment. In Nesse, R. M., editor, Evolution and the Capacity for Commitment. New York : Russell Sage Foundation. Nisbett, R. E. (2004). The Geography of Thought: How Asians and Westerners Think Differently...and Why. New York: Free Press. Nowak, M. A. and Sigmund, K. (1993). A strategy of win-stay, lose-shift that outperforms tit-for-tat in the prisoner’s dilemma game. Nature, 364:56–58. Öhman, A. and Mineka, S. (2001). Fears, phobias, and preparedness: Toward an evolved module of fear and fear learning. Psychological review, 108(3):483–522. Organ, D. W. (1988). Organizational Citizenship Behavior - The Good Soldier Syndrome. Lexington, Massachusetts/Toronto: D.C. Heath and Company. Ostrom, T. M., Carpenter, S. L., Sedikides, C., and Li, F. (1993). Differential processing of in-group and out-group information. Journal of Personality and Social Psychology, 64:21–34. Parr, L. A. (2003). The discrimination of faces and their emotional content by chimpanzees (pan troglodytes). Ann NY Acad Sci, 1000:56–78. Pedersen, C. A. (2004). How love evolved from sex and gave birth to intelligence and human nature. Journal of Bioeconomics, 6:39–63. Pillsworth, E. G. and Haselton, M. G. (2005). The evolution of coupling. Psychological Inquiry, 16:98–104. Raub, W. (2004). Hostage posting as a mechanism of trust: Binding, compensation, and signaling. Rationality and Society, 16:319–365. Raub, W. and Weesie, J. (1990). Reputation and efficiency in social interaction: an example of network effects. American Journal of Sociology, 96:626–654. Rhodes, G., Halberstadt, J., and Brajkovich, G. (2001). Generalization of mere exposure effects to averaged composite faces. Social Cognition, 19(1):57–70. Rode, C., Cosmides, L., Hell, W., and Tooby, J. (1999). When and why do people avoid unknown probabilities in decisions under uncertainty? Testing some predictions from optimal foraging theory. Cognition, 72:269–304. Roy, M., editor (1977). Battered women. New York: Van Nostrand.

156

Bibliography

Rusbult, C. and Martz, J. (1995). Remaining in an abusive relationship – an investment model analysis of nonvoluntary dependence. Personality and Social Psychology Bulletin, 21(6):558–571. Rusbult, C., Martz, J., and Agnew, C. (1998). The investment model scale: Measuring commitment level, satisfaction level, quality of alternatives, and investment size. Personal Relationships, 5(4):357–391. Rusbult, C. E. (1980). Commitment and satisfaction in romantic associations: A test of the investment model. Journal of Experimental Social Psychology, 16:172–186. Rusbult, C. E. (1983). A longitudinal test of the investment model: The development (and deterioration) of satisfaction and commitment in heterosexual involvements. Journal of Personality and Social Psychology, 45:101–117. Scheier, M. F., Carver, C. S., and Bridges, M. W. (1994). Distinguishing optimism from neuroticism (and trait anxiety, self-mastery, and self-esteem): A reevaluation of the life orientation test. Journal of Personality and Social Psychology, 67:1063–1078. Schelling, T. C. (1963). The Strategy of Conflict. Oxford University Press, New York. Schelling, T. C. (2006). Strategies of Commitment and Other Essays. Harvard University Press. Schüssler, R. (1989). Exit threats and cooperation under anonymity. The Journal of Conflict Resolution, 33:728–749. Schüssler, R. and Sandten, U. (2000). Exit, anonymity and the chances of egoistical cooperation. Analyse & Kritik, 22(1):114–129. Shapiro, D. L., Sheppard, B. H., and Cheraskin, L. (1992). Business on a handshake. Negotiation Journal, 8:365–77. Silberbauer, G. (1981). Hunter and habitat in the Central Kalahari Desert. Cambridge: Univ. Press. Silk, J. (2003). Cooperation without counting: The puzzle of friendship. In Hammerstein, P., editor, Genetic and Cultural Evolution of Cooperation, pages 37–54. MIT Press, Cambridge, MA. Simpson, B. (2003). Sex, fear, and greed: A social dilemma analysis of gender and cooperation. Social Forces, 82(1):35–52. Smaniotto, R. C. (2004). ‘You scratch my back and I scratch yours’ versus ‘love thy neighbour’: two proximate mechanisms of reciprocal altruism. PhD thesis, ICS/University of Groningen. Snijders, C. and Buskens, V. (2001). How to convince someone that you can be trusted? The role of hostages. Journal of Mathematical Sociology, 25(4):355–383.

Bibliography

157

Sripada, C. and Stich, S. (2004). Evolution, culture and the irrationality of the emotions. In D., E. and Cruse, P., editors, Emotion, Evolution and Rationality. Oxford University Press. Staw, B. M. (1976). Knee-deep in the big muddy: A study of escalating commitment to a chosen course of action. Organizational Behavior and Human Performance, 16:27–44. Staw, B. M. (1997). The escalation of commitment: An update and appraisal. In Shapira, Z., editor, Organizational decision making, pages 191–215. New York: Cambridge University Press. Sterelny, K. (2003). Thought in a Hostile World. New York, Blackwell. Strube, M. J. (1988). The decision to leave an abusive relationship: Empirical evidence and theoretical issues. Psychological Bulletin, 104:236–250. Takahashi, N. (2000). The emergence of generalized exchange. American Journal of Sociology, 105(4):1105–1134. Taylor, P. D. and Jonker, L. B. (1978). Evolutionary stable strategies and game dynamics. Mathematical Biosciences, 40:145–156. Thorndike, E. L. (1911). Animal intelligence: experimental studies. Macmillan, New York. Trivers, R. (1971). The evolution of reciprocal altruism. Quarterly Review of Biology, 46:35–57. Trivers, R. (1974). Parent-offspring conflict. American Zoologist, 14:249–264. Tversky, A. and Kahnemann, D. (1974). Judgment under uncertainty: Heuristics and biases. Science, 185:1124–1131. van Assen, M. A. L. M. (2001). Essays on Actor Models in Exchange Networks and Social Dilemmas. Amsterdam: Thela Thesis. van Assen, M. A. L. M. and Snijders, C. (2004). Effects of risk preferences in social dilemmas: a game-theoretical analysis and evidence from two experiments. In Ramzi, Suleiman, Budescu, Fischer, and Messick, editors, Contemporary Psychological Research on Social Dilemmas, volume Cambridge: Cambridge University Press. van der Gaag, M. (2005). Measurement of Individual Social Capital. PhD thesis, University of Groningen. van Dijk, E., Wilke, H., Wilke, M., and Metman, L. (1999). What information do we use in social dilemmas? environmental uncertainty and the employment of coordination rules. Journal of Experimental Social Psychology, 35:109–135. Vanberg, V. and Congleton, R. (1992). Rationality, morality and exit. American Political Science Review, 86:418–431. Wallace, J. E. (1997). Becker’s side-bet theory of commitment revisited: Is it time for a moratorium or a resurrection? Human Relations, 50(6):727–749.

158

Bibliography

Wieselquist, J., Rusbult, C., Agnew, C., Foster, C., and Agnew, C. (1999). Commitment, pro-relationship behavior, and trust in close relationships. Journal of Personality and Social Psychology, 77(5):942–66. Wiessner, P. (1982). Risk, reciprocity, and social influences on !Kung San economics. In Leacock, E. and Lee, R. B., editors, Politics and history in band societies, pages 61–84. Cambridge: Cambridge University Press. Wiessner, P. (2002). Hunting, healing, and hxaro exchange: A long-term perspective on !Kung (Ju/’hoansi) large-game hunting. Evolution and Human Behavior, 23(6):407– 436. Wu, J. and Axelrod, R. (1995). How to cope with noise in the iterated prisoner’s dilemma. Journal of Conflict Resolution, 39:183–189. Wyatt, P. K. (1999). Concieving couple commitment: Choice, promise, and fulfillment. In Adams, J. M. and Jones, W. H., editors, Handbook of Interpersonal Commitment and Relationship Stability. Kluwer Academic/Plenum Publishers, New York. Yamagishi, T., Cook, K. S., and Watabe, M. (1998). Uncertainty, trust, and commitment formation in the United States and Japan. American Journal of Sociology, 104(1):165– 194. Yamagishi, T., Hayashi, N., and Jin, N. (1994). Prisoner’s dilemma networks: selection strategy versus action strategy. In Schulz, U., Albers, W., and Mueller, U., editors, Social Dilemmas and Cooperation, pages 311–326. Heidelberg: Springer. Yamagishi, T. and Yamagishi, M. (1994). Trust and commitment in the United States and Japan. Motivation and Emotion, 18:9–66. Zajonc, R. B. (1968). Attitudinal effects of mere exposure. Journal of Personality and Social Psychology Monographs, 9(2). Zeggelink, E., de Vos, H., and Elsas, D. (2000). Reciprocal altruism and group formation: The degree of segmentation of reciprocal altruists who prefer old-helping-partners. Journal of Artificial Societies and Social Simulation, 3(3). http://www.soc.surrey.ac.uk/JASSS/3/3/1.html.

Summary Building long-term personal relationships is an ubiquitous tendency of human beings. Throughout life we build friendships, collect acquaintances, forge business alliances, become attached to intimate partners. Many of these relationships follow us through our lives and integrate us into a complex social fabric of interpersonal connections. At the same time, establishing and maintaining lasting relationships involves substantial investments of one’s time, effort and other resources. Moreover, many relationships by definition require exclusivity. For example, it’s only possible to have one single best friend at a time, in many cultures only one spouse, and in many business settings only one supplier of some product. This means that we have to forgo relationships with potentially better alternative partners. And to complicate matters, even when we do our best to invest in a relationship, we have to live with the risk of being dumped for someone else or unknowingly being taken advantage of by our partner. Why do people establish and maintain long-term relationships when these are costly, risky and exclusive? Explanations based on rational choice (Kollock, 1994; Yamagishi and Yamagishi, 1994; Yamagishi et al., 1994; Trivers, 1971; Friedman, 1971; Axelrod, 1984; Fehr and Schmidt, 1999; Fehr and Gächter, 2002; Falk et al., 2001) appear to be insufficient to account for some of the consistent behaviors people express in long-term relationships. There are numerous cases, for example, when people keep relationships with partners who have proved to be untrustworthy (e.g. Roy, 1977; Strube, 1988; Rusbult and Martz, 1995). There are also examples for relationships where a partner has no means of reciprocating in the future (e.g. Monahan and Hooker, 1997). What is it that makes battered wives return to their abusive husbands when there are hardly any prospects for change? And why does someone take care of a life-long partner with Alzheimer’s disease who will never be able to recognize the caretaker? Why do subjects in controlled laboratory experiments give costly gifts to their long-term exchange partners when their identity will never be revealed to each other? Apparently, people create social relationships with great ease even in the absence of materialistic benefits or other ulterior motives, and strongly re159

160

Summary

sist the dissolution of these relationships, well beyond rational considerations of practical advantages (Baumeister and Leary, 1995). Many of the strongest emotions people experience in their life, both positive and negative, are linked to long-term relationships. Furthermore, there is evidence that people observe and evaluate alternative partners with a biased vision, systematically dependent on how committed their current relationship is (Johnson and Rusbult, 1989). We know that even in anonymous exchange settings, positive emotions develop toward frequent exchange partners, and toward the relationship, which gradually becomes an object of value (Lawler and Yoon, 1996) in itself. In order to resolve the paradox between rational and emotional explanations of interpersonal commitment, we put forward an evolutionary explanation. During countless years of prehistoric evolutionary adaptation in the human ancestral environment, people lived together in small groups and fought for daily survival in a world more hostile than today’s (Sterelny, 2003). With many of the formal and informal helping institutions of modern society missing, people had to rely on interpersonal relationships to a much larger extent than today. Being capable and willing to establish and maintain longterm stable relationships substantially increased one’s survival and reproductive chances. As a consequence, those whose cognitive arsenal was equipped with better tools and stronger preferences to make interpersonal commitments gradually increased their presence in the population over many generations (cf. Nesse, 2001a). And these evolved preferences and abilities still influence how people make decisions about their partners today (see e.g. Cosmides, 1989; Cosmides and Tooby, 1993).

Results Since evolutionary theories are difficult to empirically test, our strategy is twofold. We first examine a theory of natural selection acting on commitment in closer detail (Part I). Building on previous work (especially de Vos et al., 2001) that relies on anthropological knowledge about conditions of the human ancestral environment, we created formal computational models of the ancestral environment. In these simulations (Chapters 2-4) we found that strategies that possessed a tendency for commitment outperformed other strategies, such as fair reciprocation. Our results were stable across simulations, where individual strategies were matched against each other in an ecological competition, as well as in genetic simulations, where genotypes of mutating strategies contested their strengths under evolutionary selection pressures. We found (in Chapter 4) that even when there were large differences between individuals in helping capability, it was still better to have a preference for helping old friends (com-

Summary

161

mitment) than a preference for helping the most attractive other. At the same time, we also emphasized the importance of fairness, which is probably another strong and cross-culturally stable preference (Fehr and Schmidt, 1999; Heinrich et al., 2001; Fehr et al., 2002). In order to empirically test the existence of an evolved commitment trait (Part II), we conducted cross-cultural laboratory experiments at six locations in three different countries (the Netherlands, USA and China). In particular, we aimed to find support for decision mechanisms that are difficult to reconcile with current exchange theoretical and (social) psychological theories but become intelligible in light of the evolutionary explanation. In Chapter 5, we uncovered support for the notion that people possess such a commitment bias: they hold on to their partners simply as a result of repeated encounters, and not necessarily as a result of instrumental benefits accumulated during encounters. Another finding of our experiments was that uncertainty decreased commitment, which is largely in contradiction with previous literature (Kollock, 1994; Yamagishi and Yamagishi, 1994; Yamagishi et al., 1998). To explain this anomaly, in a follow-up experiment (Chapter 6) we tested the hypothesis that social uncertainty only affects people who themselves wish to be cooperative, and have an opportunity to meet fellow cooperators. Furthermore, we found that there is at least one other important source of uncertainty that has been neglected in the exchange-commitment literature: resources. Similarly to social uncertainty, resource uncertainty increases commitment, especially when one has an opportunity to meet a high-resource partner. But why are some people more committed than others? In order to explain some of the individual differences in commitment behavior, we linked the effect of resource and social uncertainty to psychological mechanisms. Building especially on Yamagishi’s works, we confirmed that general trust in people has a negative effect on the tendency to become committed to steady partners. Furthermore, we also showed that while general trust decreases commitment, optimism in a more general sense has a similar negative effect. Those who are generally optimistic are more likely to dissolve existing relationships and venture interaction with strangers. Our work aimed to make a contribution to the comprehension of seemingly irrational decisions in durable relationships. We found support for the notion that people instinctively stick to their existing interpersonal relationships, more so than would seem rational given the circumstances. We argued that this tendency could be the result of a long-lasting evolutionary process. Furthermore, we advanced previous research on the relationship between commitment and one of its key rational sources, uncertainty. Our efforts testify to the importance of interdisciplinary research. Without combining previous research and insights from psychology, sociology, economics and evolutionary theory, most alternative explanations of commitment

162

Summary

remain limited in their power and scope. Together, they promise to further our understanding of the wonderful and mysterious complexity of human nature.

Samenvatting Het is een ingebakken neiging van mensen om langdurige persoonlijke relaties aan te gaan met anderen. Gedurende ons leven bouwen we vriendschappen op, krijgen steeds meer kennissen, leggen zakelijke contacten en raken gehecht aan onze partners. Veel van deze relaties houden stand gedurende ons hele leven en zorgen ervoor dat we deel uitmaken van een complexe sociale structuur van interpersoonlijke relaties. Tegelijkertijd vergt het aangaan en het onderhouden van deze langetermijnrelaties een substantiële investering wat betreft tijd, moeite en andere hulpbronnen. Bovendien vergen veel van deze relaties, per definitie, een hoge mate van exclusiviteit. Zo kun je bijvoorbeeld maar één beste vriend tegelijkertijd hebben, in veel culturen kun je maar één huwelijkspartner hebben en in zakelijke “settings” kun je maar één toeleverancier van een bepaald product hebben. Dit betekent dat we soms geen nieuwe relaties aan kunnen gaan met anderen die wellicht beter zouden zijn dan onze huidige partners. En om het nog ingewikkelder te maken, zelfs wanneer we onze best doen om een relatie in stand te houden, moeten we leven met het risico dat we gedumpt of misbruikt worden. Waarom gaan mensen langdurige persoonlijke relaties met anderen aan wanneer deze relaties zowel kostbaar en riskant zijn? Verklaringen die gebaseerd zijn op de rationele keuze theorie (Kollock, 1994; Yamagishi and Yamagishi, 1994; Yamagishi et al., 1994; Trivers, 1971; Friedman, 1971; Axelrod, 1984; Fehr and Schmidt, 1999; Fehr and Gächter, 2002; Falk et al., 2001) lijken onvoldoende rekening te houden met gedragspatronen van mensen in dergelijke langetermijnrelaties. Zo zijn er bijvoorbeeld vele gevallen bekend waarin mensen relaties in stand houden met partners waarvan is gebleken dat ze onbetrouwbaar zijn (e.g. Roy, 1977; Strube, 1988; Rusbult and Martz, 1995). Er zijn ook voorbeelden van relaties waarbij de partner niet de hulpbronnen heeft om in de toekomst een keer wat terug te doen voor de ander (e.g. Monahan and Hooker, 1997). Waarom keren mishandelde vrouwen terug naar hun mishandelende echtgenoten als er nauwelijks vooruitzichten zijn dat hun echtgenoot zich in de toekomst beter zal gedragen? En waarom neemt iemand de levenslange zorg op zich voor een partner met Alzheimer, als die partner zijn / haar verzorger nooit zal herkennen? Waarom geven deelne163

164

Samenvatting - Dutch summary

mers in laboratoriumexperimenten waardevolle giften aan hun langetermijn handelspartners als hun identiteit geheim blijft? Blijkbaar gaan mensen met groot gemak sociale relaties aan, zelfs wanneer materiële of andere belangen geen rol spelen, en hebben mensen een sterke weerstand tegen het beëindigen van deze relaties. Dit gaat veel verder dan het rationeel afwegen van de praktische voordelen van het beëindigen of in stand houden van een relatie (Baumeister and Leary, 1995). De sterkste emoties die mensen in hun leven beleven, zowel positief als negatief, hebben te maken met hun langetermijnrelaties. Bovendien zijn er aanwijzingen dat mensen alternatieve partners met een bevooroordeelde blik bekijken en evalueren, afhankelijk van de kwaliteit (hechtheid) van hun huidige relatie (Johnson and Rusbult, 1989). We weten dat mensen, zelfs in anonieme onderhandelingssituaties, positieve emoties ervaren ten opzichte van hun ’vaste’ handelspartners en ook ten opzichte van de relatie op zich (Lawler and Yoon, 1996). Om het tekortschieten van de rationele verklaring en de kennelijk belangrijke rol van emoties in de menselijke toewijding en trouw aan langetermijnrelaties te verhelderen, brengen wij een evolutionaire theorie onder de aandacht. Gedurende vele millenia van prehistorische evolutionaire selectie in de zogenaamde Omgeving van Evolutionaire Aangepastheid, leefden onze verre voorouders samen in kleine groepen en vochten voor hun voortbestaan in een omgeving die in sommige opzichten veel riskanter was dan de huidige (Sterelny, 2003). Omdat de vele formele hulpinstanties die we in onze moderne samenleving kennen, toen nog ontbraken, waren mensen vroeger in veel hogere mate afhankelijk van hun interpersoonlijke relaties dan tegenwoordig. De capaciteit en bereidheid om stabiele langetermijnrelaties met anderen aan te gaan en te onderhouden verhoogde je kans om te overleven en je voort te platen aanzienlijk. Dit had waarschijnlijk als gevolg dat het aantal mensen die beter waren ’uitgerust’ met cognitieve en emotionele capaciteiten om interpersoonlijke relaties aan te gaan en te onderhouden in de loop van vele jaren langzaamaan toenam (c.f. Nesse, 2001a). En het is aannemelijk dat deze cognitieve en emotionele capaciteiten ook tegenwoordig nog altijd beinvloeden hoe mensen beslissingen nemen over de omgang met anderen (see e.g. Cosmides, 1989; Cosmides and Tooby, 1993).

Resultaten Omdat evolutionaire theorieën moeilijk empirisch te toetsen zijn, was onze strategie tweeledig. Eerst (Deel 1) onderzoeken we in meer detail een theorie over het effect van natuurlijke selectie op ’commitment’ (noot van vertaler: commitment is niet goed te vertalen naar het Nederlands, het kan onder andere betekenen: betrokkenheid, loyaliteit en trouw). Voortbouwend op eerder werk (met name de Vos et al., 2001) dat gestoeld is op antropologische kennis

Samenvatting - Dutch summary

165

over de leefomstandigheden van mensen in de Omgeving van Evolutionaire Aangepastheid (ongeveer het Pleistoceen), creëerden we formele computer modellen van deze leefwereld van onze verre voorouders. Uit de simulaties (Hoofdstukken 2-4) bleek dat strategieën die neigden naar ’commitment’ beter presteerden dan andere strategieën, zoals ’iets aardigs terug doen’ of ’iemand helpen die jou ook heeft geholpen’. Onze resultaten waren hetzelfde in de verschillende simulaties, waarin individuele strategieën tegen elkaar werden afgezet in een ecologische competitie, evenals in genetische simulaties, waarin genotypen van gemuteerde strategieën elkaar beconcurreren onder druk van evolutionaire selectie. Zelfs wanneer er grote individuele verschillen waren in de capaciteit om te kunnen helpen, bleek het nog steeds beter te zijn om een voorkeur te hebben voor het helpen van oude vrienden (commitment) dan een voorkeur te hebben voor het helpen van de meest aantrekkelijke ander (Hoofdstuk 4). Tegelijkertijd benadrukten we het belang van eerlijk zijn, wat waarschijnlijk ook een sterke cross-culturele voorkeur is (Fehr and Schmidt, 1999; Heinrich et al., 2001; Fehr et al., 2002). Om het bestaan van een ontwikkelde commitment eigenschap empirisch te toetsen (Deel 2) hebben we cross culturele laboratoriumexperimenten uitgevoerd op zes verschillende locaties en in drie verschillende landen (Nederland, Amerika en China). Het doel was om aanwijzingen te vinden voor besluitvormingsmechanismen die moeilijk in overeenstemming zijn te brengen met huidige onderhandelingstheorieën en (sociaal) psychologische theorieën maar begrijpelijk worden in het licht van evolutionaire verklaringen. In Hoofdstuk 5 vinden we aanwijzingen voor de idee dat mensen een commitment vooroordeel hebben: ze blijven trouw aan hun partners simpelweg omdat ze elkaar vaak zijn tegen gekomen en niet noodzakelijkerwijs omdat die vele ontmoetingen instrumentele voordelen hebben opgeleverd. Een andere bevinding van onze experimenten is dat commitment afnam naarmate de onzekerheid toenam, dit in tegenstelling tot eerdere bevindingen (Kollock, 1994; Yamagishi and Yamagishi, 1994; Yamagishi et al., 1998). Om deze afwijking te verklaren hebben we in een vervolgexperiment (Hoofdstuk 6) de verwachting getoetst dat sociale onzekerheid alleen die mensen beïnvloedt die zelf coöperatief wensen te zijn en de kans hebben om anderen te ontmoeten die ook coöperatief wensen te zijn. Bovendien, ontdekten we dat er tenminste één andere bron van onzekerheid genegeerd is in de literatuur over onderhandeling en commitment, namelijk onzekerheid over de hoeveelheid hulpbronnen waarover anderen beschikken. Vergelijkbaar met sociale onzekerheid, doet onzekerheid over hulpbronnen commitment toenemen, met name wanneer de kans aanwezig is om een onderhandelingspartner met veel hulpbronnen te ontmoeten. Maar waarom zijn sommige mensen meer gecommiteerd dan anderen? Om enkele individuele verschillen in commitment gedrag te kunnen verklaren, relateren we de effecten van sociale onzekerheid en onzekerheid over

166

Samenvatting - Dutch summary

hulpbronnen aan psychologische mechanismen. We bouwen met name voort op het werk van Yamagashi en bevestigen dat het hebben van een algemeen vertrouwen in mensen een ondermijnend effect heeft op de neiging om je te commiteren aan vaste partners. Bovendien laten we zien dat niet alleen algemeen vertrouwen in mensen commitment vermindert, maar optimisme in het algemeen een vergelijkbaar negatief effect heeft. Mensen die in het algemeen optimistisch zijn, zijn eerder geneigd om bestaande relaties af te breken en zich te wagen aan nieuwe contacten met onbekenden. Het doel van ons werk was om een bijdrage te leveren aan de verklaring van de schijnbare irrationele beslissingen die gemaakt worden in langetermijnrelaties. We vonden aanwijzingen voor de idee dat mensen instinctief bij hun bestaande interpersoonlijke relaties blijven, meer dan wat rationeel te verklaren is, rekening houdend met de omstandigheden. We beargumenteren dat deze neiging het resultaat zou kunnen zijn van een langdurig evolutionair proces. Bovendien, zijn we een stap verder gegaan dan eerder onderzoek door de relatie tussen commitment en onzekerheid, als een mogelijke rationele bron daarvan, te bestuderen. Onze inspanningen laten zien wat het belang is van interdisciplinair onderzoek. Wanneer eerdere studies en inzichten op het gebied van de psychologie, sociologie, economie en evolutionaire theorie niet gecombineerd worden, blijven de meeste alternatieve verklaringen van commitment zwak en hebben een klein bereik. Al deze studies en inzichten samen beloven onze kennis en begrip van de wonderlijke mysterieuze en complexe menselijke aard te vergroten.

Összefoglalás A hosszútávú személyes kapcsolatok kialakítása alapvet˝o emberi foglalatosság. Életünk során számos barátságot kötünk, ismer˝osöket szerzünk, üzleti kapcsolatokat létesítünk, intim viszonyokba bonyolódunk. E kapocsolatok végigkísérnek életünk jórészén, beágyazva a társas kapcsolatok bonyolult szövetébe. Ugyanakkor a hosszantartó kapcsolatok létrehozása és fenntartása jelent˝os befektetést kíván akár id˝oben, pénzben vagy egyéb er˝oforrásokban kifejezve. Ráadásul számos kapcsolat már definíciója alapján is kizárólagosságot ír el˝o, hiszen egyszerre csak egyetlen legjobb barátunk lehet, sok kultúrában csak egyetlen házastársunk, bizonyos üzleti szituációkban csak egyetlen beszállítónk, szolgáltatónk. Ez egyben azt is jelenti, hogy néha le kell mondanunk potenciálisan jobb, alternatív kapcsolatokról. S még ha mindent meg is teszünk az adott kapcsolat ápolásáért, együtt kell élnünk annak kockázatával, hogy partnerünk egyszerre lecserél minket valaki másra, vagy egyéb módon használja ki gyanútlan jóindulatunkat. Miért alakítanak ki és tartanak fenn az emberek hosszútávú kapcsolatokat, ha ezek kötségesek, kockázatosak, s˝ot kizárólagosak? A racionális választáson alapuló elméletek (Kollock, 1994; Yamagishi and Yamagishi, 1994; Yamagishi et al., 1994; Trivers, 1971; Friedman, 1971; Axelrod, 1984; Fehr and Schmidt, 1999; Fehr and Gächter, 2002; Falk et al., 2001) látszólag nem adnak kimerít˝o magyarázatot a hosszútávú kapcsolatokban hozott tipikus döntések egy jelent˝os részére. Sok példa van olyan kapcsolatokra, melyek annak ellenére is fennmaradnak, hogy az egyik partner már eljátszotta a másik bizalmát (e.g. Roy, 1977; Strube, 1988; Rusbult and Martz, 1995). Vannak olyan kapcsolatok is, ahol a jöv˝obeni viszonosság ki van zárva (e.g. Monahan and Hooker, 1997). Milyen er˝o késztet egy megvert feleséget arra, hogy visszatérjen er˝oszakos férjéhez? És miért gondoskodik valaki Alzheimer-kórban szenved˝o társáról, ha az még felismerni sem lesz többé képes gondvisel˝ojét? Miért adnak kontrollált laboratóriumi kísérletek résztvev˝oi költséges ajándékokat hosszútávú partnereiknek annak ellenére, hogy azok személyazonossága örökre rejtve marad? Az emberek nyilvánvaló könnyedséggel létesítenek új kapcsolatokat mindennemu˝ anyagi érdek és egyéb hátsó szándék nélkül, és gyakran mindent 167

168

Összefoglalás - Hungarian summary

megtesznek e kapcsolatok megszunése ˝ ellen, messze túl a gyakorlati ésszeru˝ ség határain (Baumeister and Leary, 1995). Az életünk során tapasztalt leger˝osebb érzelmek nagy része – akár negatív, akár pozitív – hosszútávú kapcsolatainkkal van összefüggésben. Ezenkívül az is bizonyított, hogy az emberek más mércével mérnek fel potenciális partnereket annak függvényében, hogy a jelenlegi kapcsolatuk mennyire elkötelezett (Johnson and Rusbult, 1989). Tudjuk, hogy még anoním cserekapcsolatokban is er˝os érzelmek alakulhatnak ki a gyakori partnerek, illetve a kapcsolat irányában, és így az már önmagában is értékké válhat a partnerek számára (Lawler and Yoon, 1996). A racionális választáson, illetve az érzelmeken alapúló elméletek közötti ellentmondás feloldására a jelen munka egy evolúciós magyarázattal szolgál. Az o˝ störténeti törzsfejl˝odés megszámlálhatatlan évei során az emberek kis csoportokban éltek együtt és egy, a mainál ellenségesebb természeti környezetben kuzdöttek ˝ napi fennmaradásukért (Sterelny, 2003). A jelenlegi társadalom számos formális és informális segtítségnyújtó intézményének hiányában az emberek jóval inkább rá voltak utalva személyes kapcsolataikra, mint napjainkban. A hosszútávú kapcsolatok fenntartására leginkább képes és hajlandó egyedek túlélési és szaporodási esélyei így nagyban megnövekedtek. Következésképpen azok, akiknek szellemi képességei leginkább el˝osegítették a személyes kapcsolatok ápolását, generációról generációra fokozatosan növelték jelenlétüket a populációban (cf. Nesse, 2001a). Az ekkor kialakult preferenciák és képességek pedig a mai napig hatással lehetnek hosszútávú kapcsolatainkban alkalmazott döntési stratégiáinkra (lásd pl. Cosmides, 1989; Cosmides and Tooby, 1993).

Eredmények Mivel az evolúciós elméletek empírikus bizonyítása rendkívül körülményes, az általunk itt alkalmazott stratégia két részb˝ol állt. El˝oször közelebbr˝ol megvizsgáltuk az elméletet, mely szerint az o˝ sközösségben a természetes szelekció hatására az emberek közötti elkötelezettségek és személyes kapcsolatok egyre er˝osödnek (I. rész). Ehhez olyan szimulációs modelleket állítottunk fel, melyek az o˝ sközösségbeli életkörülményekr˝ol rendelkezésre álló antropológiai ismereteket használtak fel (lásd de Vos et al., 2001). Szimulációink (2-4. fejezet) tanusága szerint az elkötelezettségekre hajlamosabb stratégiák sikeresebbek a többi, például az igazságos viszonosságon alapuló stratégiáknál. Ez az eredmény szilárdan domborodott ki mind az egyszerubb ˝ versenyhelyzetek modelljéb˝ol, mind a genetikus szimulációkból, melyek során mutáló stratégiák mérték össze erejüket evolúciós nyomások közepette. Úgy találtuk, hogy még ha nagy egyéni különbségek jellemzik is a társadalmat segítségnyújtási potenciálban, akkor is hasznosabb a régi barátok segítése (elkötelezettség, „commitment”), mint az éppen legvonzóbb partneré

Összefoglalás - Hungarian summary

169

(4. fejezet). Ugyanakkor hangsúlyoztuk az igazságosságra törekvés jelent˝oségét is, ami valószínuleg ˝ egy másik, stabil kultúra-közi preferencia (Fehr and Schmidt, 1999; Heinrich et al., 2001; Fehr et al., 2002). A törzsfejl˝odés során kialakult elkötelezettségi preferencia empírikus vizsgálatát (II. rész) kultúra-közi kísérletek segítségével végeztük három ország (Hollandia, USA és Kína) hat különböz˝o helyszínén. Célunk az volt, hogy bizonyítékot találjunk olyan döntéshozatali mechanizmusok jelenlétére, melyeket nem magyaráznak jól a jelenlegi csereelméleti és (szociál)pszichológiai tézisek, ám az evolúciós elmélet fényében egycsapásra érthet˝ové válnak. Az 5. fejezeteben arra utaló bizonyítékról számolunk be, hogy az emberek valóban rendelkeznek egy ilyen elkötelezettségi hajlammal: a partnereikkel való együttmaradást már pusztán az ismételt találkozások kiváltották, függetlenül a találkozások során felhalmozott anyagi el˝onyökt˝ol. Kísérleteink arról is árulkodtak, hogy a bizonytalanság csökkenti az elkötelezettségeket, ami nagyrészt ellentmond a korábbi irodalomnak (Kollock, 1994; Yamagishi and Yamagishi, 1994; Yamagishi et al., 1998). Hogy feloldjuk az ellentmondást, egy rákövetkez˝o kísérletben (6. fejezet) megvizsgáltuk a hipotézist, miszerint a társadalmi bizonytalanság csak azokra van hatással, akik maguk is kooperatívak, és lehet˝oségük van magukhoz hasonló szándékú partnert találniuk. Továbbá azt is megtudtuk, hogy a bizonytalanságnak van legalább egy másik fontos forrása is – melyet a korábbi csere-elkötelezettség irodalom figyelmen kívül hagyott – az er˝oforrások. A társadalmi bizonytalansághoz hasonlóan az er˝oforrások bizonytalansága szintén növeli az elkötelezettséget, különösen akkor, ha lehet˝oség van er˝oforrásokban gazdag partnerek megszerzésére. De miért hajlamosak az elkötelezettségre bizonyos emberek inkább, míg mások kevésbé? Az egyéni különbségek magyarázata érdekében az er˝oforrások és társadalmi bizonytalanság hatását pszichológiai mechanizmusokkal hoztuk összefüggésbe. Különösen Yamagishi munkásságára támaszkodva meger˝osítettük azt az elméletet, hogy az emberekbe vetett általános bizalom negatív hatással van az állandó partnerekkel szembeni elkötelezettségek er˝osségére. Továbbá megmutattuk azt is, hogy nemcsak az általános bizalom, de az optimizmus is csökkenti az elkötelezettségek erejét. Azok akik általában optimisták, hajlamosabbak felbontani meglév˝o kapcsolataikat és újakat létesíteni idegenekkel. Munkánk célja, hogy segítse a hosszútávú kapcsolatokban tapasztalt, látszólag irracionális döntések megértését. Meger˝osítettük az elméletet, miszerint az emberek ösztönösen ragaszkodnak meglév˝o személyes kapcsolataikhoz, jobban mint azt a körülmények ismeretén alapuló racionalitás diktálja. Elméletünk szerint ez a hajlam egy rendkívül hosszú evolúciós folyamat eredménye. Emellett új eredményekkel szolgáltunk az elkötelezettség és bizonytalanság irodalma számára is. Er˝ofeszítéseink tanusítják a tudományközi kutatás fontosságát. A pszichológia, szociológia, közgazdaságtan és evolúciós elméletek egyedi meglátásai-

170

Összefoglalás - Hungarian summary

nak kombinációja nélkül az elkötelezettségek legtöbb alternatív magyarázata csak korlátozott ereju˝ lehet. Ezek együttese azonban hozzájárulhat ahhoz, hogy jobban megértsük az emberi természet csodálatos és kifürkészhetetlen bonyolultságát.

摘要 建立长期的人际关系, 对于人类来说, 是普遍存在的现象。例如，一生中，我 们与他人建立友谊，扩大自己的人际网，达成生意上的联盟，成为亲密伴侣。 在人的一生中会有许多这样的关系，并且，它将我们融入一个复杂的社会关系 中。同时，建立并保持这种持久的关系需要投入大量的时间，精力和其他资 源。此外，这种关系不言而喻地具有排他性。例如，一个人在一个时间内只会 有一个最好的朋友；在许多社会中实行一夫一妻制；在生意场上有时一种产品 只有一个供应商。它意味着我们不得不放弃一些潜在的更加选择。更糟糕的 是，我们甚至不得不承担和忍受被他人抛弃或利用的风险。 这种关系的成本昂贵，具有风险性，但为什么人们仍然想建立和保持这 种长期的人际关系呢？理性选择（Kollock, 1994, Yamagishi and Yamagishi, 1994, Yamagishi etal., 1994, Trivers, 1971, Friedman, 1971, Axelrod, 1984, Fehr and Schmidt, 1999, Fehr and Gchter, 2002, Falk etal., 2001）显得不足以解释人 类在表达长期关系中的一贯行为. 有很多案例可以证明，例如，很多人的同伴不值得信任，但这些人仍旧 与其保持长期的关系（e.g. Roy, 1977, Strube, 1988, Rusbult and Martz, 1995 ）；又例如，双方并没有互惠利益但仍保持关系（e.g. Monahan and Hooker, 1997）；是什么原因使受虐待的妻子在几乎没有可能改变现状的情况下仍 旧回到她粗暴的丈夫身边？又是什么原因使人能够终生照顾患有阿而采目 （Alzheimer）病的伴侣，即使病者根本不可能再认出对方？ 显然，即便在缺少物质利益或其他外界刺激时，人们仍可以轻易地建立 他们的社会关系，并且强有力的抵御实际利益对其的瓦解（Baumeister and Leary, 1995）。在生活中经历的积极或消极的强烈情感往往与长期关系有关。 而且，有证据证明人是带着一种偏好去观察和评价可供选择的同伴的，并且 由当前他们关系的忠诚度而决定（Johnson and Rusbult, 1989 ）。科学研究指 出，即使在匿名的交易情况下，那些和对方交易次数较多的人会对其以及他们 之间的关系产生好感，这种关系也逐渐会变成一种价值产物(Lawler and Yoon, 1996)。 为了解决对于这种人际关系理性和感性解释之间的矛盾，我们提出一个进 化论的解释。在史前进化适应时期，人们在一个比现代有更多敌人的环境中以 小部落的形式群居在一起，为每天的生存而战斗（Sterelny, 2003）。由于那时 没有现代正式或非正式的援助机制的存在，在很大程度上与今天相比，人们不 得不更加依靠人际关系。能够并愿意建立和维持长期稳定的关系极大地提高了 171

172

Zhai Yao - Chinese summary

一个人的生存机会和繁殖潜力。因此，如果人们对于人际关系的认知积累被更 好的工具和更强的倾向“武装”，那么这些人的后代或基因在今后的几代中会 更加普遍（cf. Nesse, 2001a ）。这些从古代进化来的意愿和能力仍然影响今 天的人们怎样决定他们的同伴（见e.g. Cosmides, 1989, Cosmides and Tooby, 1993）。

结论 既然进化论的解释很难去实际测量，我们的策略是双重的。首先在第一部分我 们解释自然选择理论怎样作用于偏好。在先前基于人类学关于史前条件的研究 基础上（尤其是deVos etal., 2001 ），我们建立了一个史前环境计量模型。 在这个模拟实验中（章节2－4），我们发现那些具有交友倾向的策略胜于 其他一些策略，例如公平互惠策略。在整个模拟实验中我们的测试结果是稳定 的，实验中的个体策略无论是在生态竞争中还是在遗传仿真模拟中都是相互匹 配的，而且在进化选择的压力下基因变异策略论证了他们的优势。我们还发现 （第四章节）即使当个体在协助能力上有很大差异时，优先选择帮助老朋友 （长期关系）要比那些优先选择帮助最吸引人的人更有优势。同时我们也强调 公平的重要性，它可能是另一种强大的跨文化的稳定偏好（Fehr and Schmidt, 1999, Heinrich etal., 2001, Fehr etal., 2002）。 为了从测试进化的关系特点的存在，我们分别在三个不同国家（荷兰、美 国和中国）的六个地方进行了一系列跨文化实验。特别是我们有目的地去证实 决定机制的论证，当前的交互心理学理论和社会心理学理论并不能解释决定机 制，但是由进化论来解释是可以被理解的。 在第五章，我们揭示了人们具有偏好暗示（commitment bias）观念的论 证：人们维系着和对方的关系仅仅是因为不断相遇的结果，而不一定是相遇 中积累的利益关系的结果。我们实验的另一个发现是不确定性会减弱人们的 关系，这在很大程度上与以前的文献相抵触（Kollock, 1994, Yamagishi and Yamagishi, 1994, Yamagishi etal., 1998）。为了解释这种异常性，在后续的实 验中（第六章节）我们测试了以下假设：社会不确定性只影响那些本身希望是 协作的和希望是有机会去遇见合作者的人群。并且，我们还发现在交换关系学 文献中至少有一个不确定性因素被忽视：资源。和社会不确定性类似，资源的 不确定性使人们的关系增强，尤其是当人们有机会遇见拥有很多资源的同伴。 但是为什么有些人比其他人更加忠诚呢？为了解释在关系行为中的个体 差异，我们将资源和社会不确定性的效益与心理机制结合起来。尤其是基 于Yamagishi的研究成果，我们证实了人们的普遍信任感在对稳定同伴的忠诚 趋势上有消极影响。而且我们表明了信任减弱人们之间的关系，同时，在很大 程度上乐观主义也有同样的消极作用。一般，乐观的人更可能减少现存的关系 而冒险与陌生人交往。　 我们的目的是对长期关系中表面非理性决定做出一些解释。即便看起来在 并不合理的情况下，人们本能地更加坚持他们现存的人际关系。我们论证了这 种趋势是长期进化过程的结果。而且我们提出了先前关于关系建立和其关键的 理性资源—不确定性之间的关系。

Zhai Yao - Chinese summary

173

我们的努力证实了各学科间研究的重要性。如果不把先前的研究与对于心 理学、社会学、经济学和进化论的理解结合起来，大多数可供选择的关系建立 解释仍然在力度和广度上是有限的。总之，它们对于促进关于人类本性奇妙和 神秘复杂性的理解给予了广大的前景。

Appendix A

First we calculate the expected number of times that a trigger player will help a trigger player, nh , and multiply this by the costs of giving help, fd . This yields the total loss, nh · fh , of playing trigger against trigger (which is not incurred by an ALLD player playing against trigger). Let nT T be the expected number of times before both players revert to eternal defection, i.e. before both trigger players get into distress simultaneously: nT T =

1 −1 Pd2

Let nh be the number of those rounds within these initial nT T rounds that the trigger alter is in distress, while ego is not. Then nh = Ph · nT T where

Pd (1 − Pd ) 1 − Pd2 The conditional probability Ph is calculated as follows. The unconditional probability of the event "alter in distress while ego not" is Pd (1 − Pd ) but what we need to know is the probability that this event occurs under the condition that the event "both are in distress" has not yet occurred. To obtain this, we divide the unconditional probability by the probability that the condition occurs. Next, we calculate the expected number of times nd that an ALLD player does not get help from a trigger opponent, while a trigger player would get help at the same time. Multiplied with the costs of not getting help, fd , this amounts to the expected loss, nd · fd , that an ALLD player incurs which is not incurred by a trigger player. Let n0DT be the length of the period in which a trigger player has not yet lost the help of his opponent, while an ALLD player gets no more help. Let Pn be the probability that in any single round of this part of the game, the ALLD ego is in distress, while alter is not. Then Ph =

nd = n0DT · Pn where Pn =

Pd (1 − Pd ) 1 − Pd2 175

176 Here Pn is obtained in the same way as Ph above, only this time the roles of ego and alter are reverted (ego in need but alter not), which does not affect the result in this case. To obtain n0DT , we need to find the difference between nT T , the expected duration until a trigger player would stop getting help from his trigger opponent, and nDT , the number of times until an ALLD player would stop getting help from this opponent. Furthermore, we need to take into account that, if there is a second phase of the game, the first round of that second phase cannot be a round in which an ALLD ego is refused by alter because it is the round in which ego reveals itself as a defector, i.e. ego cannot be in distress. Hence, to obtain n0DT we need to subtract from nT T − nDT exactly that probability. To obtain that probability, we need to consider that there may never be a second phase of the game in which an ALLD player would not get help, because the first phase may already end with a round in which both players get into distress simultaneously, in which case also a trigger player would get no more help. The probability for that latter event is again obtained as a conditional probability as follows: the unconditional probability that alter gets into distress, but ego not, is Pd (1−Pd ). There are only two possible events that can occur in the first round after the first phase of the game. Either it is the first round of the second phase, with an unconditional probability of Pd (1 − Pd ), or there is no second phase and both players are in distress simultaneously, with probability Pd2 . All other events are excluded by virtue of the precondition that this is the round following the first phase of the game. So the probability we need to subtract from nT T − nDT is Pd (1 − Pd ) divided by the probability that the condition occurs, Pd (1 − Pd ) + Pd2 . All in all this amounts to n0DT = nT T − nDT − where nDT =

Pd (1 − Pd ) Pd (1 − Pd ) + Pd2

1 −1 Pd

Finally, we need to find the condition under which the loss of a trigger player against trigger is equal to that of an ALLD player against trigger. That is, we need to solve nd · fd = nh · fh This condition yields after some rearrangement the following result: fh = fd (1 − Pd )

Appendix B

Parameter values used in ecological simulations We acquired and tested our results using the following parameters:

Model parameters probability of distress (Pd ) = 0.05, 0.2, 0.5 probability of decision making error (Pe ) = 0.0, 0.05, 0.1, 0.5 cost of helping (fh , measured in fitness) = 1 cost of not getting help (fd , measured in fitness) = 5, 10, 20 initial fitness (fi , measured in fitness) = 50, 100, 200 critical fitness (fc , measured in fitness) = 0 group size (N ) = 10, 25, 50 number of sub-rounds in a round (m) = 1, 2, 3

Simulation parameters number of rounds in a run = variable, depending on how long it took the winning strategy to push its opponent(s) out number of runs in an experiment = 2000

177

Appendix C

Pseudocode of simulation core /** * Main cycle of simulation */ begin simulation for each experiment initialize result variables and charts for each run initialize parameters, run-level result variables, charts initialize society for each round initialize round-level result variables for each agent A generate a random event R with probability P_d if R occurs distress A end for for each subround for each agent A if A is distressed call decideWhomToAskForHelp of A end for for each agent A call decideWhomToGiveHelp of A end for end for for each agent A update fitness if fitness < f_c remove A from society end for call replicator_dynamics() update round-level result variables update charts if necessary if society is empty end run if there is only one group left in society end run end for /* end of rounds */ update run-level variables update charts end for /* end of runs */ update experiment-level variables update charts end for /* end of experiments */ close charts and show results

179

180 end simulation /** * Agent deciding whether/whom to give help. */ begin decideWhomToGiveHelp if nobody asked for help return nobody if agent is distressed return nobody if agent already gave help return nobody generate a random event R with probability P_err if R does not occur determine group G of agents for whom U_d is maximal if U_d < U_t return nobody /* threshold utility */ else return random agent from G else return random agent from {helpseekers+nobody} end if end decideWhomToGiveHelp /** * Agent deciding whom to ask for help. */ begin decideWhomToAskForHelp remove itself from possible helpers remove those who refused before in this round if possible helpers is empty return nobody generate a random event R with probability P_err if R does not occur determine group G of agents for whom U_s is maximal return random agent from G else return random agent from possible helpers end if end decideWhomToAskForHelp /** * Evolutionary selection process (based on the replicator dynamics) */ begin replicator_dynamics for each dead agent calculate society_fitness_sum for each strategy S in society calculate strategy_fitness_sum generate a random event R with... probability strategy_fitness_sum / society_fitness_sum if R occurs create and add agent A with... strategy S to new_generation /* in order to condition successive probability calculations on previous ones */ else decrease society_fitness_sum with strategy_fitness_sum end for end for add new_generation to society end replicator_dynamics

Appendix D

Parameter values used in evolutionary simulations probability of distress (Pd ) = 0.1, 0.2, 0.4 probability of decision making error (Pe ) = 0.05 cost of helping (fh , measured in fitness) = 1 cost of not getting help (fd , measured in fitness) = 5, 10, 20, 30 initial fitness (fi , measured in fitness) = 50,100 critical fitness (fc , measured in fitness) = 0 group size (N ) = 25,50 length of simulation runs = 10,000,000 rounds number of help-seeking subrounds in a round (m) = 1,2,3

Parameters of the evolutionary process evolution frequency min. childbearing age (measured in interactions) = 20 mutation probability (Pmut ) = 0.05

181

Appendix E

Pseudocode of evolutionary dynamic The evolutionary process, executed at the end of each run: ... end of round begin evolutionary process for each dead agent A choose with a probability equal to its fitness... share among N-old agents an agent B... who is alive and is at least N-old generate a new agent C with a strategy... identical to that of B mutate the strategy of C with a probability Pmut end for end evolutionary process start of new round ...

183

Appendix F

Experiment instructions F.1

Initial instructions

F.2

Instruction text from the experiment

The experiment takes place in an imaginary market of art. In this market Artists make paintings and sell them to Collectors. In the following, all participants will be classified into two groups. Some of the participants will be assigned the role of Artist and some will be assigned the role of Collectors. It is important that you try to imagine yourself in the 185

186 role you are assigned to as well as possible. Both Artists and Collectors receive an initial amount of money. Artists use their money to buy brushes, paint and canvas so they can create a painting and sell it to one of the Collectors. Collectors use their money to buy paintings from the Artists. The experiment consists of 17 rounds. In the beginning of each round Artists create a painting and interested Collectors make a bid. Then Artists decide whom to sell the painting to. After you press the button below computer will determine whether you have to act as Artist or as Collector. You will act as an Artist in the rest of the experiment. In the beginning of the game you will receive an initial amount of $100. In each round of the game, you create a piece of art at a certain cost price (based on the price of brushes, paint, canvas etc). Next, you need to sell the painting you created. Out of all Collectors in the experiment, always four are selected to bid on your painting. Collectors are linked to several Artists at the same time and bid on multiple paintings but they cannot see each other`s offers. Moreover, Collectors don't have information about your cost price but they know something that Artists don't. They know the objective value of the painting. When all Collectors made their offers, you have to choose one whom you want to sell your painting to. You always have to sell your painting. After you sell the painting, the amount offered by the Collector is added to you balance, deducting the cost price. After the experiment you will receive $2.6 US dollars for each $100 you make in the experiment. In each round you sell the painting to one of the Collectors. Those Collectors whose offer you did not accept disappear and are replaced by new ones in the next round. The Collector whose offer you accepted will stay with you in the next round and will make an offer for your next painting as well. Collectors who once disappear may reappear later in the game as Collectors but under a different name. You will also appear to them under a different name, so that you cannot recognize each other. At different time points during the 17 rounds, the experiment will be briefly paused to ask some questions on screen. After the last round, you will be asked to answer some final questions. At first you will play a practice game. The purpose of this practice game for you is to learn how to play the game. The results of this game are discarded and do not count into your final score. You will be clearly notified before the real game begins.

F.3

Screenshot from the experiment game

187

Appendix G G.1

Instructions

189

G.2

Screen shots from the experiment

190

191

ICS dissertation series The ICS-series presents dissertations of the Interuniversity Center for Social Science Theory and Methodology. Each of these studies aims at integrating explicit theory formation with state-of-the-art empirical research or at the development of advanced methods for empirical research. The ICS was founded in 1986 as a cooperative effort of the universities of Groningen and Utrecht. Since 1992, the ICS expanded to the University of Nijmegen. Most of the projects are financed by the participating universities or by the Netherlands Organization for Scientific Research (NWO). The international composition of the ICS graduate students is mirrored in the increasing international orientation of the projects and thus of the ICS-series itself. 1. C. van Liere, (1990), Lastige Leerlingen. Een empirisch onderzoek naar sociale oorzaken van probleemgedrag op basisscholen. Amsterdam: Thesis Publishers. 2. Marco H.D. van Leeuwen, (1990), Bijstand in Amsterdam, ca. 1800 - 1850. Armenzorg als beheersings- en overlevingsstrategie. ICS dissertation, Utrecht. 3. I. Maas, (1990), Deelname aan podiumkunsten via de podia, de media en actieve beoefening. Substitutie of leereffecten? Amsterdam: Thesis Publishers. 4. M.I. Broese van Groenou, (1991), Gescheiden Netwerken. De relaties met vrienden en verwanten na echtscheiding. Amsterdam: Thesis Publishers. 5. Jan M.M. van den Bos, (1991), Dutch EC Policy Making. A Model-Guided Approach to Coordination and Negotiation. Amsterdam: Thesis Publishers. 6. Karin Sanders, (1991), Vrouwelijke Pioniers. Vrouwen en mannen met een 'mannelijke' hogere beroepsopleiding aan het begin van hun loopbaan. Amsterdam: Thesis Publishers. 7. Sjerp de Vries, (1991), Egoism, Altruism, and Social Justice. Theory and Experiments on Cooperation in Social Dilemmas. Amsterdam: Thesis Publishers. 8. Ronald S. Batenburg, (1991), Automatisering in bedrijf. Amsterdam: Thesis Publishers. 9. Rudi Wielers, (1991), Selectie en allocatie op de arbeidsmarkt. Een uitwerking voor de informele en geïnstitutionaliseerde kinderopvang. Amsterdam: Thesis Publishers. 10. Gert P. Westert, (1991), Verschillen in ziekenhuisgebruik. ICS dissertation, Groningen. 11. Hanneke Hermsen, (1992), Votes and Policy Preferences. Equilibria in Party Systems. Amsterdam: Thesis Publishers. 12. Cora J.M. Maas, (1992), Probleemleerlingen in het basisonderwijs. Amsterdam: Thesis Publishers. 13. Ed A.W. Boxman, (1992), Contacten en carrière. Een empirisch-theoretisch onderzoek naar de relatie tussen sociale netwerken en arbeidsmarktposities. Amsterdam: Thesis Publishers. 14. Conny G.J. Taes, (1992), Kijken naar banen. Een onderzoek naar de inschatting van arbeidsmarktkansen bij schoolverlaters uit het middelbaar beroepsonderwijs. Amsterdam: Thesis Publishers. 15. Peter van Roozendaal, (1992), Cabinets in Multi-Party Democracies. The Effect of Dominant and Central Parties on Cabinet Composition and Durability. Amsterdam: Thesis Publishers. 16. Marcel van Dam, (1992), Regio zonder regie. Verschillen in en effectiviteit van gemeentelijk arbeidsmarktbeleid. Amsterdam: Thesis Publishers. 17. Tanja van der Lippe, (1993), Arbeidsverdeling tussen mannen en vrouwen. Amsterdam: Thesis Publishers. 18. Marc A. Jacobs, (1993), Software: Kopen of Kopiëren? Een sociaal-wetenschappelijk onderzoek onder PC-gebruikers. Amsterdam: Thesis Publishers. 19. Peter van der Meer, (1993), Verdringing op de Nederlandse arbeidsmarkt. Sector- en sekseverschillen. Amsterdam: Thesis Publishers. 20. Gerbert Kraaykamp, (1993), Over lezen gesproken. Een studie naar sociale differentiatie in leesgedrag. Amsterdam: Thesis Publishers. 21. Evelien Zeggelink, (1993), Strangers into Friends. The Evolution of Friendship Networks Using an Individual Oriented Modeling Approach. Amsterdam: Thesis Publishers. 22. Jaco Berveling, (1994), Het stempel op de besluitvorming. Macht, invloed en besluitvorming op twee Amsterdamse beleidsterreinen. Amsterdam: Thesis Publishers. 23. Wim Bernasco, (1994), Coupled Careers. The Effects of Spouse's Resources on Success at Work. Amsterdam: Thesis Publishers.

192 24. Liset van Dijk, (1994), Choices in Child Care. The Distribution of Child Care Among Mothers, Fathers and Non-Parental Care Providers. Amsterdam: Thesis Publishers. 25. Jos de Haan, (1994), Research Groups in Dutch Sociology. Amsterdam: Thesis Publishers. 26. K. Boahene, (1995), Innovation Adoption as a Socio-Economic Process. The Case of the Ghanaian Cocoa Industry. Amsterdam: Thesis Publishers. 27. Paul E.M. Ligthart, (1995), Solidarity in Economic Transactions. An Experimental Study of Framing Effects in Bargaining and Contracting. Amsterdam: Thesis Publishers. 28. Roger Th. A.J. Leenders, (1995), Structure and Influence. Statistical Models for the Dynamics of Actor Attributes, Network Structure, and their Interdependence. Amsterdam: Thesis Publishers. 29. Beate Völker, (1995), Should Auld Acquaintance Be Forgot...? Institutions of Communism, the Transition to Capitalism and Personal Networks: the Case of East Germany. Amsterdam: Thesis Publishers. 30. A. Cancrinus-Matthijsse, (1995), Tussen hulpverlening en ondernemerschap. Beroepsuitoefening en taakopvattingen van openbare apothekers in een aantal West-Europese landen. Amsterdam: Thesis Publishers. 31. Nardi Steverink, (1996), Zo lang mogelijk zelfstandig. Naar een verklaring van verschillen in oriëntatie ten aanzien van opname in een verzorgingstehuis onder fysiek kwetsbare ouderen. Amsterdam: Thesis Publishers. 32. Ellen Lindeman, (1996), Participatie in vrijwilligerswerk. Amsterdam: Thesis Publishers. 33. Chris Snijders, (1996), Trust and Commitments. Amsterdam: Thesis Publishers. 34. Koos Postma, (1996), Changing Prejudice in Hungary. A Study on the Collapse of State Socialism and Its Impact on Prejudice Against Gypsies and Jews. Amsterdam: Thesis Publishers. 35. Jooske T. van Busschbach, (1996), Uit het oog, uit het hart? Stabiliteit en verandering in persoonlijke relaties. Amsterdam: Thesis Publishers. 36. René Torenvlied, (1996), Besluiten in uitvoering. Theorieën over beleidsuitvoering modelmatig getoetst op sociale vernieuwing in drie gemeenten. Amsterdam: Thesis Publishers. 37. Andreas Flache, (1996), The Double Edge of Networks. An Analysis of the Effect of Informal Networks on Cooperation in Social Dilemmas. Amsterdam: Thesis Publishers. 38. Kees van Veen, (1997), Inside an Internal Labor Market: Formal Rules, Flexibility and Career Lines in a Dutch Manufacturing Company. Amsterdam: Thesis Publishers. 39. Lucienne van Eijk, (1997), Activity and Well-being in the Elderly. Amsterdam: Thesis Publishers. 40. Róbert Gál, (1997), Unreliability. Contract Discipline and Contract Governance under Economic Transition. Amsterdam: Thesis Publishers. 41. Anne-Geerte van de Goor, (1997), Effects of Regulation on Disability Duration. ICS dissertation, Utrecht. 42. Boris Blumberg, (1997), Das Management von Technologiekooperationen. Partnersuche und Verhandlungen mit dem Partner aus Empirisch-Theoretischer Perspektive. ICS dissertation, Utrecht. 43. Marijke von Bergh, (1997), Loopbanen van oudere werknemers. Amsterdam: Thesis Publishers. 44. Anna Petra Nieboer, (1997), Life-Events and Well-Being: A Prospective Study on Changes in Well-Being of Elderly People Due to a Serious Illness Event or Death of the Spouse. Amsterdam: Thesis Publishers. 45. Jacques Niehof, (1997), Resources and Social Reproduction: The Effects of Cultural and Material Resources on Educational and Occupational Careers in Industrial Nations at the End of the Twentieth Century. ICS dissertation, Nijmegen. 46. Ariana Need, (1997), The Kindred Vote. Individual and Family Effects of Social Class and Religion on Electoral Change in the Netherlands, 1956-1994. ICS dissertation, Nijmegen. 47. Jim Allen, (1997), Sector Composition and the Effect of Education on Wages: an International Comparison. Amsterdam: Thesis Publishers. 48. Jack B.F. Hutten, (1998), Workload and Provision of Care in General Practice. An Empirical Study of the Relation Between Workload of Dutch General Practitioners and the Content and Quality of their Care. ICS dissertation, Utrecht. 49. Per B. Kropp, (1998), Berufserfolg im Transformationsprozeß. Eine theoretisch-empirische Studie über die Gewinner und Verlierer der Wende in Ostdeutschland. ICS dissertation, Utrecht.

193 50. Maarten H.J. Wolbers, (1998), Diploma-inflatie en verdringing op de arbeidsmarkt. Een studie naar ontwikkelingen in de opbrengsten van diploma's in Nederland. ICS dissertation, Nijmegen. 51. Wilma Smeenk, (1998), Opportunity and Marriage. The Impact of Individual Resources and Marriage Market Structure on First Marriage Timing and Partner Choice in the Netherlands. ICS dissertation, Nijmegen. 52. Marinus Spreen, (1999), Sampling Personal Network Structures: Statistical Inference in EgoGraphs. ICS dissertation, Groningen. 53. Vincent Buskens, (1999), Social Networks and Trust. ICS dissertation, Utrecht. 54. Susanne Rijken, (1999), Educational Expansion and Status Attainment. A Cross-National and Over-Time Comparison. ICS dissertation, Utrecht. 55. Mérove Gijsberts, (1999), The Legitimation of Inequality in State-Socialist and Market Societies, 1987-1996. ICS dissertation, Utrecht. 56. Gerhard G. Van de Bunt, (1999), Friends by Choice. An Actor-Oriented Statistical Network Model for Friendship Networks Through Time. ICS dissertation, Groningen. 57. Robert Thomson, (1999), The Party Mandate: Election Pledges and Government Actions in the Netherlands, 1986-1998. Amsterdam: Thela Thesis. 58. Corine Baarda, (1999), Politieke besluiten en boeren beslissingen. Het draagvlak van het mestbeleid tot 2000. ICS dissertation, Groningen. 59. Rafael Wittek, (1999), Interdependence and Informal Control in Organizations. ICS dissertation, Groningen. 60. Diane Payne, (1999), Policy Making in the European Union: an Analysis of the Impact of the Reform of the Structural Funds in Ireland. ICS dissertation, Groningen. 61. René Veenstra, (1999), Leerlingen - Klassen - Scholen. Prestaties en vorderingen van leerlingen in het voortgezet onderwijs. Amsterdam, Thela Thesis. 62. Marjolein Achterkamp, (1999), Influence Strategies in Collective Decision Making. A Comparison of Two Models. ICS dissertation, Groningen. 63. Peter Mühlau, (2000), The Governance of the Employment Relation. A Relational Signaling Perspective. ICS dissertation, Groningen. 64. Agnes Akkerman, (2000), Verdeelde vakbeweging en stakingen. Concurrentie om leden. ICS dissertation, Groningen. 65. Sandra van Thiel, (2000), Quangocratization: Trends, Causes and Consequences. ICS dissertation, Utrecht. 66. Rudi Turksema, (2000), Supply of Day Care. ICS dissertation, Utrecht. 67. Sylvia E. Korupp (2000), Mothers and the Process of Social Stratification. ICS dissertation, Utrecht. 68. Bernard A. Nijstad (2000), How the Group Affects the Mind: Effects of Communication in Idea Generating Groups. ICS dissertation, Utrecht. 69. Inge F. de Wolf (2000), Opleidingsspecialisatie en arbeidsmarktsucces van sociale wetenschappers. ICS dissertation, Utrecht. 70. Jan Kratzer (2001), Communication and Performance: An Empirical Study in Innovation Teams. ICS-dissertation, Groningen. 71. Madelon Kroneman (2001), Healthcare Systems and Hospital Bed Use. ICS/NIVELdissertation, Utrecht. 72. Herman van de Werfhorst (2001), Field of Study and Social Inequality. Four Types of Educational Resources in the Process of Stratification in the Netherlands. ICS-dissertation, Nijmegen. 73. Tamás Bartus (2001), Social Capital and Earnings Inequalities. The Role of Informal Job Search in Hungary. ICS-dissertation Groningen. 74. Hester Moerbeek (2001), Friends and Foes in the Occupational Career. The Influence of Sweet and Sour Social Capital on the Labour Market. ICS-dissertation, Nijmegen. 75. Marcel van Assen (2001), Essays on Actor Perspectives in Exchange Networks and Social Dilemmas. ICS-dissertation, Groningen. 76. Inge Sieben (2001), Sibling Similarities and Social Stratification. The Impact of Family Background across Countries and Cohorts. ICS-dissertation, Nijmegen. 77. Alinda van Bruggen (2001), Individual Production of Social Well-Being. An Exploratory Study. ICS-dissertation, Groningen.

194 78. Marcel Coenders (2001), Nationalistic Attitudes and Ethnic Exclusionism in a Comparative Perspective: An Empirical Study of Attitudes Toward the Country and Ethnic Immigrants in 22 Countries. ICS-dissertation, Nijmegen. 79. Marcel Lubbers (2001), Exclusionistic Electorates. Extreme Right-Wing Voting in Western Europe, ICS-dissertation, Nijmegen. 80. Uwe Matzat (2001), Social Networks and Cooperation in Electronic Communities. A theoretical-empirical Analysis of Academic Communication and Internet Discussion Groups, ICS-dissertation, Groningen. 81. Jacques P.G. Janssen (2002), Do Opposites Attract Divorce? Dimensions of Mixed Marriage and the Risk of Divorce in the Netherlands, ICS-dissertation, Nijmegen. 82. Miranda Jansen (2002), Waardenoriëntaties en partnerrelaties. Een panelstudie naar wederzijdse invloeden, ICS-dissertation, Utrecht. 83. Anne Rigt Poortman (2002), Socioeconomic Causes and Consequences of Divorce. ICSdissertation, Utrecht. 84. Alexander Gattig (2002), Intertemporal Decision Making, ICS-dissertation, Groningen. 85. Gerrit Rooks (2002), Contract en Conflict: Strategisch Management van Inkooptransacties, ICSdissertation, Utrecht. 86. Károly Takács (2002), Social Networks and Intergroup Conflict. ICS-dissertation, Groningen. 87. Thomas Gautschi (2002), Trust and Exchange, Effects of Temporal Embeddedness and Network Embeddedness on Providing and Dividing a Surplus. ICS-dissertation, Utrecht. 88. Hilde Bras (2002), Zeeuwse meiden. Dienen in de levensloop van vrouwen, ca. 1850 – 1950. Aksant Academic Publishers, Amsterdam. 89. Merijn Rengers (2002), Economic Lives of Artists. Studies into Careers and the Labour Market in the Cultural Sector, ICS-dissertation, Utrecht. 90. Annelies Kassenberg (2002), Wat scholieren bindt. Sociale gemeenschap in scholen, ICSdissertation, Groningen 91. Marc Verboord (2003), Moet de meester dalen of de leerling klimmen? De invloed van literatuuronderwijs en ouders op het lezen van boeken tussen 1975 en 2000. ICS-dissertation, Utrecht. 92. Marcel van Egmond (2003), Rain Falls on All of Us (but Some Manage to Get More Wet than Others): Political Context and Electoral Participation. ICS-dissertation, Nijmegen. 93. Justine Horgan (2003), High Performance Human Resource Management in Ireland and the Netherlands: Adoption and Effectiveness. ICS-dissertation, Groningen. 94. Corine Hoeben (2003), LETS' Be a Community. Community in Local Exchange Trading Systems. ICS-dissertation, Groningen. 95. Christian Steglich (2003), The Framing of Decision Situations. Automatic Goal Selection and Rational Goal Pursuit. ICS-dissertation, Groningen. 96. Johan van Wilsem (2003), Crime and Context. The Impact of Individual, Neighborhood, City and Country Characteristics on Victimization. ICS-dissertation, Nijmegen. 97. Christiaan Monden (2003), Education, Inequality and Health. The Impact of Partners and Life Course. ICS-dissertation, Nijmegen. 98. Evelyn Hello (2003), Educational Attainment and Ethnic Attitudes. How to Explain their Relationship. ICS-dissertation, Nijmegen. 99. Marnix Croes en Peter Tammes (2004), Gif laten wij niet voortbestaan. Een onderzoek naar de overlevingskansen van joden in de Nederlandse gemeenten, 1940-1945. Aksant Academic Publishers, Amsterdam 100. Ineke Nagel (2004), Cultuurdeelname in de levensloop. ICS- dissertation, Utrecht. 101. Marieke van der Wal (2004), Competencies to Participate in Life. Measurement and the Impact of School. ICS-dissertation, Groningen. 102. Vivian Meertens (2004), Depressive Symptoms in the General Population: a Multifactorial Social Approach. ICS -dissertation, Nijmegen. 103. Hanneke Schuurmans (2004), Promoting Well-Being in Frail Elderly People. Theory and Intervention. ICS-dissertation, Groningen. 104. Javier Arregui (2004), Negotiation in Legislative Decision-Making in the European Union. ICSdissertation, Groningen. 105. Tamar Fischer (2004), Parental Divorce, Conflict and Resources. The Effects on Children’s Behaviour Problems, Socioeconomic Attainment, and Transitions in the Demographic Career. ICSdissertation, Nijmegen.

195 106. René Bekkers (2004), Giving and Volunteering in the Netherlands: Sociological and Psychological Perspectives. ICS-dissertation, Utrecht. 107. Renée van der Hulst (2004), Gender Differences in Workplace Authority: An Empirical Study on Social Networks. ICS-dissertation, Groningen. 108. Rita Smaniotto (2004), ‘You Scratch My Back and I Scratch Yours’ Versus ‘Love Thy Neighbour’. Two Proximate Mechanisms of Reciprocal Altruism. ICS-dissertation, Groningen. 109. 109) Maurice Gesthuizen (2004), The Life-Course of the Low-Educated in the Netherlands: Social and Economic Risks. ICS-dissertation, Nijmegen. 110. Carlijne Philips (2005), Vakantiegemeenschappen. Kwalitatief en Kwantitatief Onderzoek naar Gelegenheid en Refreshergemeenschap tijdens de Vakantie. ICS-dissertation, Groningen. 111. Esther de Ruijter (2005), Household Outsourcing. ICS-dissertation, Utrecht. 112. Frank van Tubergen (2005), The Integration of Immigrants in Cross-National Perspective: Origin, Destination, and Community Effects. ICS-dissertation, Utrecht. 113. Ferry Koster (2005), For the Time Being. Accounting for Inconclusive Findings Concerning the Effects of Temporary Employment Relationships on Solidary Behavior of Employees. ICSdissertation, Groningen. 114. Carolien Klein Haarhuis (2005), Promoting Anti-Corruption Reforms. Evaluating the Implementation of a World Bank Anti-Corruption Program in Seven African Countries (1999-2001). ICS-dissertation, Utrecht. 115. Martin van der Gaag (2005), Measurement of Individual Social Capital. ICS-dissertation, Groningen. 116. Johan Hansen (2005), Shaping Careers of Men and Women in Organizational Contexts. ICSdissertation, Utrecht. 117. Davide Barrera (2005), Trust in Embedded Settings. ICS-dissertation, Utrecht. 118. Mattijs Lambooij (2005), Promoting Cooperation. Studies into the Effects of Long-Term and Short-Term Rewards on Cooperation of Employees. ICS-dissertation, Utrecht. 119. Lotte Vermeij (2006), What’s Cooking? Cultural Boundaries among Dutch Teenagers of Different Ethnic Origins in the Context of School. ICS-dissertation, Utrecht. 120. Mathilde Strating (2006), Facing the Challenge of Rheumatoid Arthritis. A 13-year Prospective Study among Patients and Cross-Sectional Study among Their Partners. ICS-dissertation, Groningen. 121. Jannes de Vries (2006), Measurement Error in Family Background Variables: The Bias in the Intergenerational Transmission of Status, Cultural Consumption, Party Preference, and Religiosity. ICS-dissertation, Nijmegen. 122. Stefan Thau (2006), Workplace Deviance: Four Studies on Employee Motives and Self-Regulation. ICS-dissertation, Groningen. 123. Mirjam Plantinga (2006), Employee Motivation and Employee Performance in Child Care. The effects of the Introduction of Market Forces on Employees in the Dutch Child-Care Sector. ICSdissertation, Groningen. 124. Helga de Valk (2006), Pathways into Adulthood. A Comparative Study on Family Life Transitions among Migrant and Dutch Youth. ICS-dissertation, Utrecht. 125. Henrike Elzen (2006), Self-Management for Chronically Ill Older People. ICS-Dissertation, Groningen. 126. Ayse Güveli (200 New Social Classes within the Service Class in the Netherlands and Britain. Adjusting the EGP Class Schema for the Technocrats and the Social and Cultural Specialists. ICSdissertation, Nijmegen. 127. Willem-Jan Verhoeven (2007), Income Attainment in Post-Communist Societies. ICSdissertation, Utrecht. 128. Marieke Voorpostel (2007), Sibling support: The Exchange of Help among Brothers and Sisters in the Netherlands. ICS-dissertation, Utrecht. 129. Jacob Dijkstra (2007), The Effects of Externalities on Partner Choice and Payoffs in Exchange Networks. ICS-dissertation, Groningen. 130. Patricia van Echtelt (2007), Time-Greedy Employment Relationships: Four Studies on the Time Claims of Post-Fordist Work. ICS-dissertation, Groningen. 131. Sonja Vogt (2007), Heterogeneity in Social Dilemmas: The Case of Social Support. ICSdissertation, Utrecht. 132. Michael Schweinberger (2007), Statistical Methods for Studying the Evolution of Networks and Behavior. ICS-dissertation, Groningen. 133. István Back (2007), Commitment and Evolution: Connecting Emotion and Reason in Long-term Relationships. ICS-dissertation, Groningen.

Index agent-based simulation, see simulation ancestral environment, 12, 19, 110 anger, 15 attachment, 93, 140

forgiving, 18 friendship, 3, 12, 29, 47, 95, 103, 105, 110, 139 frontal lobes, 14 functionalism, 137 fundamental addtribution error, 107

battered wives, 92, 115, 120 birds, 15 birth, 18, 109 bonding, 69, 111 brain size, 18 bushmen, 115, 119

game theory, 6, 21 grooming, 140 guanxi, 119 help exchange, 19, 31 holistic theories, see functionalism hormones, 69, 111, 140 hostage posting, 7, 8, 54, 117 hunter-gatherers, 115, 119 hxaro, 119

cognitive bias, 2, 17, 94, 107, 137 cognitive dissonance, 10 collectivism, 97, 109 commitment close relationships, 10 escalation, 7 etimology, 4 in exchange, 8, 29, 54, 73, 114 organizational, 11 strategic, 6

indirect evolutionary approach, 20, 32, 55 individual capability, 73 individualism, 97, 109 inequality, 73, 85 informal networking, see guanxi intelligence, 140 internalization, 11 Internet addiction, 17

divorce, 10, 139 duty, 6 dyadization, 46 East Asians, 97, 105 emotion, 2, 14, 29, 93, 97, 108 empathy, 14 environmental harshness, 60, 62, 84 error management, 16, 93 evolution, 3, 30, 137 evolutionary drift, 80 evolutionary psychology, 30, 137 evolved navigation theory, 17 exchange, 10, 29, 115 experiment, 22 exposure, 21, 95, 96, 135

love, 18, 140 loyalty, 4

face recognition, 111 fairness, 36, 44, 51, 67, 78, 81 fear, 13, 111, 125

Pleistocene, see prehistory preeclampsia, 109, 139 prehistory, 3, 12, 93

marriage, 10, 18 modesty, 107 morality, 6, 10 natural selection, 13 opiates, 69, 111 optimism, 115, 130 oxytocin, 111, 140

196

197 proximate mechanism, 15 rational choice, 2, 15, 17, 92, 137 reciprocity, 2, 28, 72 reinforcement learning, 115, 135 relational cohesion, 10 relaxed accounting, 30 replicator dynamics, 37 risk-aversion, 131 sexual selection, 13 shadow of the future, 28, 72, 92 side-bet theory, 6 signaling, 110 simulation, 21, 32, 56, 73 smart cheaters, 72 social capital, 14, 114 social dilemma, 54, 99, 125 social network, 46 social simulation, see simulation stability, 59, 62, 77, 80 stay behavior, 4 strong ties, 46 sunk cost effect, 7 survey, 22 technological development, 17 tit-for-tat, 28, 72, 123 tournament, 60 trust, 2, 7, 9, 21, 96, 117, 136 general trust, 9, 115, 121 type I and II error, 16 ultimate mechanism, 15, 18, 135 uncertainty, 8, 9, 96, 99, 108, 114, 136 environmental, 115 resource, 118, 124 social, 114, 117

© 2007 by István Back [email protected] 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 without written permission of the author. ICS Dissertation series (nr. 133) Printed by Mesterprint Kft., Budapest, Hungary. ISBN 978-90-367-3113-3