Journal of Neuroscience, Psychology, and Economics © 2011 American Psychological Association 2011, Vol. 4, No. 3, 161–173 1937-321X/11/$12.00 DOI: 10.1037/a0024688 Empathy, Guilt-Aversion, and Patterns of Reciprocity Vittorio Pelligra University of Cagliari, CRENoS, Cagliari, Italy, and UniCa-BERG, Cagliari, Italy This paper reports the results of an experiment aimed at investigating the link between empathy, anticipated guilt, and pro-social behavior. In particular, we test the hypothesis that empathy modulates the anticipatory effect of guilt in bargaining situations and, more specifically, that it correlates with subjects’ willingness to give and to repay trust in an investment game. We also control for the effect of individual risk attitude. Our main results show that empathy significantly influences players’ pattern of restitution in the investment game and that risk-propensity weakly affects the decision to trust; we also find a significant gender difference in the distribution of empathy. These results seem to indicate that empathy affects pro-social behavior in a more complex way than previously hypothesized by existing models of social preferences. Keywords: trust, reciprocity, guilt-aversion, empathy Recent economic research has indicated that intention-based motives (Rabin, 1993; Dufwen- individual behavior in economic interactions berg & Kirchsteiger, 2004; Battigalli & Duf- can be explained, in many cases, by other- wenberg, 2007; Pelligra, 2010). regarding preferences. This evidence has led to Both classes of models include a psycholog- the development of models of agents with “ex- ical element in the extended utility function that tended” utility functions that incorporate both can be interpreted as a guilt factor (Krajbich, material and psychological elements (see, e.g., Adolphs, Tranel, Denburg, & Camerer, 2009; Fehr & Gachter, 2000; Sobel, 2005, and Fehr & Dufwenberg, 2002). When triggered by in- Schmidt, 2006, for recent surveys). Alternative equality or by opportunism, the guilt factor pro- theoretical approaches differ with respect to duces a cost whose negative effects agents tend how those functions are defined. In particular, to anticipate and avoid by behaving prosocially. two main classes of models can be distin- In this interpretation, thus, individual sensitivity guished: models that focus on distributional to guilt should affect, ceteris paribus, the like- concerns (Fehr & Schmidt, 1999; Bolton & lihood of prosocial behavior. Ockenfels, 2000), and models that focus on Psychologists from different perspectives suggest that the cognitive and affective basis for feeling guilt is the capacity to feel or anticipate the suffering and distress of others, in other This article was published Online First July 18, 2011. Vittorio Pelligra, Department of Economics, University words, to empathize with others (Hoffman, of Cagliari, CRENoS (Centro Ricerche Economiche Nord 1982, 2000; Baumeister, Stillwell, & Heather- Sud), Cagliari, Italy, and UniCa-BERG, Cagliari, Italy. ton, 1994; Singer & Fehr, 2005; Tomasello, Financial support from the Regione Autonoma della Sar- Carpenter, Call, Behne, & Moll, 2005). degna (L.R. 7/2007–Project CRP3-223) is gratefully ac- knowledged. To Silvia Aresu, Monica Cugusi, Davide El- In this paper we report on an experiment trudis, and Gianluigi Mazziti goes my gratitude for their designed to investigate the connection be- invaluable research assistance. I would also like to thank tween individuals’ ability to empathize and Florian Artinger, Martin Dufwenberg, Benedetto Gui, An- their prosocial behavior, supposedly driven drea Isoni, Cristina Murroni, and Fabio Tufano who helped by guilt-aversion, in a simple economic game. to improve the paper in a substantial way, as well as audiences at the 5th Nordic Conference on Behavioral and More precisely, we test the hypothesis that Experimental Economic (Helsinki), Annual CIPESS work- the Empathy Quotient (EQ henceforth), a shop (University of Turin), and CRENoS Seminar (Univer- widely used and well-validated psychometric sity of Cagliari). measure (Baron-Cohen & Wheelwright, Correspondence concerning this article should be ad- dressed to Vittorio Pelligra, Department of Economics, Uni- 2004; Lawrence, Shaw, Baker, Baron-Cohen, versity of Cagliari, V.le S.Ignazio 17, 09123 Cagliari, Italy. & David, 2004), is correlated with the sub- E-mail: [email protected] jective sensitivity parameter that appears in 161 162 PELLIGRA models of social preferences, and conse- person’s affective state is the source of our own quently, with players’ tendency to give and to affective state. Condition (a) is particularly im- repay trust in an investment-game (IG). portant as it helps to differentiate empathy from On the trustor’s side we find that players’ mentalizing, which denotes, instead, our ability EQ does not affect trust (the amount sent in to represent others’ mental states without emo- the IG), which is instead weakly correlated tional involvement. with individual risk-propensity. On the trust- Following the perception–action model of ee’s side, the EQ appears to be strongly as- motor behavior and imitation, Preston and de sociated with different patterns of restitution Waal (2002) develop a theory of empathy that (conditional vs. balanced reciprocity). We explains how we can understand what someone also find, in line with the literature on the else feels when he or she experiences simple subject, a significant difference in the empa- emotions such as anger, fear, sadness, joy, or thy distribution across gender. pain, or even more complex ones such as dis- The paper is organized as follows: the next appointment, shame, or guilt. They suggest that section briefly discusses the psychological and the mere observation or imagination of another neuroscientific literature on empathy and antic- person’s emotional state automatically triggers ipated guilt and its associations to the economic a representation of that state in the observer. models of social preferences. The following This theory is supported by recent neuroscien- section describes the experimental design, the tific evidence that shows how the observation or hypotheses and the procedures. Subsequent sec- imagination of another person in a given emo- tions present and discuss the results, and con- tional state activates a representation of a sim- clude the paper. ilar state in the observer through an unconscious and effortful process (Singer, Kiebel, Winston, Dolan, & Frith, 2004). In a fMRI study, Singer Relevant Literature and colleagues find that, as hypothesized, the empathic response is automatic and does not Empathy require any form of engagement of judgment about others’ feelings. This study also finds a Game theory is fundamentally based on the considerable level of heterogeneity across indi- assumption that people are capable of predict- viduals in their ability to empathize.1 ing the actions of others. This ability, usually These individual differences measured by referred to as Theory of Mind, has two distinct questionnaires have been found to be highly components: cognitive (mentalizing) and affec- correlated with differences in the activation of tive (empathy) (see Singer & Fehr, 2005; the bilateral anterior insula and the rostral ante- Singer, 2009). In our study, we focus mainly on rior cingulate cortex, neural circuitry that is the affective component. If we assume that peo- normally activated in the processing of the af- ple’s actions are, at least partially, emotionally fective component of pain. The same affective driven, the ability to anticipate and share emo- pain circuits that are activated when we feel tions and feelings with others—that is, to em- pain are also active when we observe someone pathize—represents a crucial factor of this more else experiencing pain. That suggests that if general process. another person suffers pain, “our brains also Empathy, or emotional perspective-taking, is make us suffer from this pain” (Singer & Fehr, generally defined as our ability to understand 2005, p.342). It is also important to note that other people’s feelings (Preston & de Waal, 2002; Gallese, 2003). A more specific definition is proposed by de Vignemont and Singer 1 It was already known that higher scores in empathy questionnaires are strongly correlated with differences in (2006). In their view, empathy can be defined prosocial behavior such as volunteering and charitable giv- by a set of four conditions: we empathize with ing (Davis et al., 1999). In economic experiments, Ben-Ner, others when we have (a) an affective state, (b) Kong, and Putterman (2004) and Ben-Ner and Halldorsson which is isomorphic to another person’s affec- (2010) also find that a measure of “agreeableness” affects giving choices in the dictator game and trustworthiness in tive state, (c) which was induced by observation the trust game. Gunnthorsdottir, McCabe, and Smith (2002) or imagination of another person’s affective find that a Machiavellian scale (which is related with indi- state, and (d) when we know that the other viduals’ ability in perspective-taking) predicts reciprocity. EMPATHY, GUILT-AVERSION, AND PATTERNS OF RECIPROCITY 163 empathy is not limited to known or significant to anticipate feelings of guilt, people are more others but extends also to unknown or imagined likely to comply with a certain prescription, persons (Morrison, Lloyd, di Pellegrino, & precisely to avoid the guilt that would result Roberts, 2004; Jackson, Meltzoff, & Decety, from noncompliance. 2005). This evidence lends support to