1. THREE TYPES OF DEBATE OVER ALTRUISM

One can distinguish three types of debate that make use of the notion of altruism, each of them employing different senses of the term.Footnote ¹ The first type of debate centres on the notion of ‘biological altruism’; it asks how behaviours that increase other organisms’ Darwinian fitness – measured in terms of expected number of offspring – at a cost to the actor's own fitness come to be selected in evolution. Since the publication of Darwin's Origin of species, this question has been one of the main challenges to the theory of evolution. Thanks to the efforts of William Hamilton (Reference Hamilton1964) and others, this difficulty has been resolved.Footnote ²

The second type of debate makes use of the notion of ‘behavioural altruism’; it asks why ordinary people often fail to behave in the way predicted by the neoclassical model of human agency often used in economics. From the behavioural point of view, neoclassical models conceive of human beings as rational maximizers of expected utility. This view is often more or less explicitly combined with a hedonistic assumption of human psychology and with the idea that utility must be understood in monetary terms: according to the ‘false consciousness’ assumption, ‘we all like to think of ourselves as nice, caring, altruistic beings, but then when put in the appropriate circumstances – when money is at stake –, we just cannot help but act as the cynical agents postulated by economic models’ (Guala Reference Guala2005: 241).

This way of conceiving human beings as rationally self-interested and always aiming to maximize profits has been powerfully challenged by work in the field of experimental economics. In a large number of studies, it was shown that ordinary people violate this expected utility paradigm; they are ready to contribute to others’ welfare and to the common good at their own expense, even where a monetary incentive is at stake (Fehr and Fischbacher Reference Fehr and Fischbacher2004a). In the context of this economic debate over human altruism, fitness measures do not enter the picture; behavioural altruism is defined in terms of individual costs and benefits at the end of one or a series of social actions, whereas biological altruism is measured in terms of expected number of offspring at the end of a completed life cycle. More formally, economists label a behaviour ‘altruistic’ if it benefits other individuals or the common good at some cost for the agent and if the cost cannot be compensated in the future. A paradigmatic study case of behavioural altruism is called ‘altruistic punishment’. It consists of a disposition to punish unfair agents even if costly and when the punishment provides neither present nor future material rewards (Fehr and Gächter Reference Fehr and Gächter2002; Fehr and Fischbacher Reference Fehr and Fischbacher2003, Reference Fehr and Fischbacher2005b).

Experimental economists’ research has immensely improved our understanding of human social behaviour. Much of the research is also confirmed by a growing body of data stemming from psychology and economics (Camerer et al. Reference Camerer, Loewenstein and Rabin2003; Kahneman Reference Kahneman2003; Gigerenzer Reference Gigerenzer2008). This has had a major impact on economics and more generally on the future of social sciences.

The biological and behavioural notions of altruism should not be confused with the psychological or philosophical notion of it. The former two focus strictly on a cost–benefit analysis of behavioural outcomes. In contrast, common ‘everyday’ use of the term does not refer to outcomes but to the subjective motivation of the agents. This leads us to the psychological and philosophical debate over altruism (Batson Reference Batson1991; Sober and Wilson Reference Sober and Wilson1998).Footnote ³ The traditional debate over the possibility of ‘psychological altruism’ centres on the nature of primary motives. The notion of motive is a broad category that includes different things, such as desires, intentions, or judgements. There are two sorts of motives: the first are primary motives,Footnote ⁴ usually conceived as the starting points of causal chains that lead to action – they are also the driving force present until actions have come about; the second are instrumental motives whose function is to help achieve the aims of the primary motives. Here is an example:

If a primary motive is directed towards the needs and well-being of other individuals, it earns the label ‘altruistic’. If a primary motive is aimed at some personal benefit for oneself – as in the example of Raymond – it is considered ‘self-interested’.

‘Psychological altruism’ is the view according to which at least some actions are motivated by altruistic primary motives. On the contrary, ‘psychological egoism’ denies the possibility of primary altruistic motives. According to this latter view, human actions are always motivated by the expectation of some personal benefit, usually conceived of in terms of pleasure and avoidance of pain – the hedonistic version – or such things as power, resources, or reputation.

Psychologists (such as Cialdini et al. Reference Cialdini, Schaller, Houlihan, Arps, Fultz and Beaman1987; Batson Reference Batson1991) and philosophers (such as Nagel Reference Nagel1970; Butler Reference Butler and Raphael1991; Hutcheson Reference Hutcheson2004) are the traditional actors of this debate. There is a surge of theoretical arguments and experimental data designed to support one or the other thesis. Excellent studies (Batson Reference Batson1991; Stocks et al. Reference Stocks, Lishner and Decker2009) provide strong evidences in favour of psychological altruism by way of systematically disqualifying some of the most popular ‘egoistic’ interpretations. These important achievements however are not sufficient to settle the dispute. Not all ‘egoistic’ explanatory schemes have yet been rejected (Sober and Wilson Reference Sober and Wilson1998; Stich Reference Stich2007; Stich et al. Reference Stich, Doris, Roedder and Doris2010) and the findings need to be replicated by independent research groups. Since there are not yet decisive arguments for or against the possibility that any primary unconscious motive is necessarily self-directed, the debate still remains open.

Some economists concerned with the study of social preferences have become interested in this debate, although they do not frame their questioning in the precise form preferred by philosophers and psychologists. Fehr and Fischbacher for example, claim that ‘the self-interest hypothesis assumes that all people are exclusively motivated by their economic self-interest. (. . .) However, the evidence presented in this paperFootnote ⁵ also shows that fundamental issues in economics and the social sciences in general cannot be understood solely on the basis of the self-interest model’ (Fehr and Fischbacher Reference Fehr, Fischbacher and Gintis2005a: 183).

In fact, the debate over psychological altruism is particularly relevant for experimental economists; a detailed analysis of human motivation can help to describe how people actually make decisions and why and when they tend to violate the narrow model of rational behaviour usually associated with neoclassical economics. Indeed, as Mayr and colleagues note, typical neo classical ‘economic assumptions of stable preferences and rational choices’ imply ‘the assumption of selfish preferences’. One indirect way to challenge such a model is to show that its predictions are at odds with the behaviour observed in laboratory experiments. Another more straightforward and decisive argumentative path is to investigate directly the ‘hidden motives behind altruistic behaviour’ (Mayr et al. Reference Mayr, Harbaugh, Tankersley and Glimcher2009: 303–304). As we shall see, the use by some economists of fMRI technology provides new hope for a better understanding of our unconscious motives, which is a crucial aspect of the altruism debate.

It must be noted at the outset that, despite appearances, this literature is extremely difficult to read because most authors have not carried out the necessary conceptual work and there is no consensus on the various notions of altruism used (West et al. Reference West, Griffin and Gardner2007). Consequently, this literature is filled with conceptual confusions and unwarranted switches from behavioural to psychological altruism.

At times, it is difficult to decide whether the authors are talking about behavioural altruism or altruistic motivation. Here is one example: ‘People are often neither self-regarding nor altruistic. Strong reciprocators are conditional cooperators (who behave altruistically as long as others are doing so as well) and altruistic punishers (who apply sanctions to those who behave unfairly according to the prevalent norms of cooperation)’ (Gintis et al. Reference Gintis, Bowles, Boyd and Fehr2005: 8).

Such an imprecise use of terminology facilitates switches from one domain to the other without further argument. For example, after having explicitly said that ‘throughout the paper we rely on a behavioural – in contrast to a psychological – definition of altruism as being costly acts that confer economic benefits on other individuals’, Fehr and Fischbacher add on the very same page that ‘a combination of altruistic and selfish concerns motivates them [the strong reciprocators]. Their altruistic motives induce them to cooperate and punish in one-shot interactions and their selfish motives induce them to increase rewards and punishment in repeated interactions or when reputation-building is possible’ (Fehr and Fischbacher Reference Fehr and Fischbacher2003: 785). Similarly, in a context in which the authors are clearly talking about behavioural altruism, one finds the statement that ‘altruistic cooperators are willing to cooperate, that is, to abide by the implicit agreement, although cheating would be economically beneficial for them’ (Fehr and Rockenbach Reference Fehr and Rockenbach2003: 137). In neither case are sufficient arguments provided by the authors in support of their claims about human motivation to act altruistically.

Other authors (Mayr et al. Reference Mayr, Harbaugh, Tankersley and Glimcher2009) appear to distinguish between motives and behavioural altruism, but since they do not explicitly define their concepts and use the term ‘altruism’ in both domains of discourse, the reader can easily be confused.

This paper uses simple conceptual distinctions as a tool for sorting out and assessing the specific contribution of experimental and neuroeconomics to the debate over psychological altruism. The paper is organized as follows: section 2 surveys and discusses some studies in experimental economics that seem to be relevant to the controversy; as it will turn out, with respect to the debate over psychological altruism, the results of these studies are less satisfying than those obtained in social psychology; section 3 surveys new developments in neuroeconomics, a young research field which makes use of a technology that might help to overcome the limitations of previous studies; we pay particular attention to a study on altruistic punishment (de Quervain et al. Reference de Quervain, Fischbacher, Treyer, Schellhammer, Schnyder, Buck and Fehr2004) that presents promising results; section 4 presents and discusses various possible interpretations for altruistic punishment; section 5 contains critical conclusions concerning the relevance of the studies discussed as well as constructive suggestions for possible future research.

2. SOME INTERESTING DATA FROM EXPERIMENTAL ECONOMICS

Many experimental economists are concerned with pro-social behaviour, which is defined as behaviour that benefits group interaction and increases cooperation in human societies.Footnote ⁶ In particular, these economists investigate the necessary conditions and the outcomes of behavioural altruism – to avoid confusion in what follows we will use the term ‘b-altruism’ when referring to this notion. More recently, some researchers have also become interested in the psychological aspect of this behaviour. The question we address in this paper is the possible contribution of their studies to the philosophical debate over psychological altruism. Hence, we will not discuss the most well-known and important aspects of experimental economic research, namely its results relating to a better understanding of the modalities and consequences of b-altruism, but their relevance for a better understanding of the causes – the psychological motives – of this behaviour. The hope is that interesting and concurring results can be obtained in a context other than that of social psychology, where the altruism debate has its roots.

Part of the literature in experimental economics concentrates on helping and rewarding behaviours. More precisely, the type of behaviour investigated is helping – or rewarding – other players at some personal cost, even in circumstances where no reputation or future gain can be expected – which is a typical instance of b-altruistic behaviour. Repeated studies have shown that people are ready to share their income with others in a b-altruistic manner (Dawes et al. Reference Dawes, Fowler, Johnson, McElreath and Smirnov2007; Croson and Konow Reference Croson and Konow2009). One experimental situation often used to test this caretaking behaviour is the dictator game. It is a two-person game in which the first player receives an amount of money and is asked to divide it between himself and the second player. The other player cannot reject the split proposed.Footnote ⁷ Contrarily to the prediction of classical economics, experiments show that more than half of the dictators are willing to leave on average roughly 30% of their endowment to the other player (Kahneman et al. Reference Kahneman, Knetsch and Thaler1986; Forsythe et al. Reference Forsythe, Horowitz, Savin and Sefton1994; Croson and Konow Reference Croson and Konow2009). Similar results are obtained even in settings where the decision maker's absolute privacy is guaranteed; neither the second players nor the experimenters will be able to trace the identity of the first players (Hoffman et al. Reference Hoffman, McCabe and Smith1996; Charness and Gneezy Reference Charness and Gneezy2008). This result is remarkable because the dictators could keep the whole sum for themselves without expecting any unfavourable consequences, such as punishment or a bad reputation. Moreover, as the game is conducted under anonymity conditions, generous dictators receive neither material benefit nor increase their reputation in return.

One way to interpret these results is to attribute non self-directed fairness motivesFootnote ⁸ to the dictators (Fehr and Schmidt Reference Fehr and Schmidt1999; Bolton and Ockenfels Reference Bolton and Ockenfels2000). However, such an interpretation seems too simplistic; it could be shown that experimental settings that more closely resemble everyday life positively affect the dictator's generosity (Hoffman et al. Reference Hoffman, McCabe and Smith1996; Charness and Gneezy Reference Charness and Gneezy2008). Therefore, there is reason to suspect the interference of other internal self-directed motives, such as guilt aversion, the unpleasant feeling resulting from acting too selfishly (Charness and Dufwenberg Reference Charness and Dufwenberg2006), or aversion to disappointing the second player (Dufwenberg and Gneezy Reference Dufwenberg and Gneezy2000; Dana et al. Reference Dana, Cain and Dawes2006; Koch and Normann Reference Koch and Normann2008), or an expectation of feeling a warm glow, the pleasant feeling associated with the thought of oneself as a caretaker (Andreoni Reference Andreoni1990; Eckel et al. Reference Eckel, Grossman and Johnston2005). The latter hypothesis seems even stronger in the light of Haselhuhn and Mellers’ study (Reference Haselhuhn and Mellers2005); these researchers report a correlation between fair or cooperative behaviour and a self-reported degree of pleasure derived from acting fairly.Footnote ⁹ It might even be argued that one never knows how realistic the anonymous conditions really are from the subjects’ point of view. Despite experimenters’ efforts, the experimental settings might fail to avoid subtle reputation cues (Haley and Fessler Reference Haley and Fessler2005). After all, God's eyes bypass the strongest anonymity conditions. Or, as Hoffman and colleagues suggest, there is the possibility that ‘people have unconscious, preprogrammed rules of social exchange behaviour that suit them well in the repeated game of life's interaction with other people. These patterns are imported into the laboratory’ (Hoffman et al. Reference Hoffman, McCabe and Smith1996: 659).

In light of this review of the literature, one has to admit that current results are not yet fine-grained enough to support either position in the particular debate over altruism. In order to show that altruistic motivation underlies some type of helping behaviour – such as the one observed in the dictator game – the economists should design experiments that allow them to systematically control for the most compelling ‘egoistic’ explanatory hypotheses available; they must aim to rule out any possibility that ‘self-directed’ interpretations are sufficient to fully explain the generous behaviour they elicit. One way to do this would be to apply Batson's methodology, which consists in controlling, one after the other, for all plausible egoistic hypotheses. But this requires painstaking work. Economists would have to control at least for: the selfish desire to get rid of the negative emotion elicited by the perception of social inequity; the fear of feeling guilty for not helping or for disappointing the other player; the expectation of feeling the warm glow of being a caretaker. It is not an easy matter to obtain persuasive evidence on these issues, particularly because egoistic explanations are fully compatible with the presence of motives related to social norms. Subjects might, for example, care about fair distribution and still be acting out of self-interest. This is typically the case of those who expect to feel a warm glow or are afraid of feeling guilty. Thus, identifying the influence of social – or even moral norms – is not enough to refute the egoism hypothesis.

It must also be said that other-regarding helping behaviour is especially difficult to trigger in the context of economic games. Croson and Konow (Reference Croson and Konow2009) have found that people are significantly more responsive to lack of fairness than to kindness; they punish more than they reward. Cherry and colleagues (Reference Cherry, Frykblom and Shogren2002) reveal the ‘other people's money problem’: subjects who are given money at the beginning of the game do not attach much importance to it, whereas if they are first required to earn their endowments, there is a dramatic decrease in their inclination to spend it generously in the dictator game (see also Oxoby and Spraggon Reference Oxoby and Spraggon2008). Frohlich and colleagues (Reference Frohlich, Oppenheimer and Moore2001) point to the difficulty in anonymous games of eradicating the subject's doubt that a real person will actually receive the money. There is also the difficulty of preventing the subject's tendency to consider an economic game as a setting in which the main goal is to win money (Frohlich et al. Reference Frohlich, Oppenheimer and Moore2001; Bowles Reference Bowles2008). Finally, anonymous games are often built in a way that makes them resemble real situations very little. This raises serious questions about the external validity of these studies.Footnote ¹⁰

Another, possibly more promising, part of the experimental economics literature on motivation concentrates on the proximate mechanisms underlying b-altruistic punishing behaviours – as opposed to helping or rewarding behaviour. For example, in an interesting 2002 paper, Ernst Fehr and Simon Gächter conducted an experiment with two tasks. The subjects were first asked to play a ‘public goods game’. In such a game, all participants are free to contribute to a group project and once the group project is realized, every member of the group can benefit from it, even those who did not contribute. In this particular experiment, the public goods game was complemented with a punishing condition: at the end of the game, subjects could punish each of the other players after they were informed about the investments they had made. The results showed that subjects usually adopted b-altruistic punishing behaviour; they sanctioned free-riders – those who benefit from the public good without contributing – even under conditions where it was costly for them and they could not expect any present or future material benefits. In the second part of the experiment, participants were asked to fill in a questionnaire designed to sound out the psychological motivation for b-altruistic punishment. Participants were presented with a hypothetical situation and asked to indicate the intensity of their anger and annoyance towards free-riders. The experimenters hypothesized that strong negative emotions such as anger or disgust were the proximate mechanisms responsible for b-altruistic punishment. This could be confirmed by the participants’ responses: 84% of them indicated that they would feel a high intensity of anger if they were cheated by a free-rider. Fehr and Gächter concluded that negative emotions were the proximate mechanisms behind human b-altruistic punishment. These results are a first step in our understanding of human motivation. Here again, however, the data do not allow us to reach definitive conclusions about the primary motives underlying b-altruism. Negative emotions, such as anger, can be associated with different sorts of motives, such as the desire to enforce justice or the desire for personal revenge.

More promising results are to be found in games using a ‘third party’ condition: a third player, not directly involved in the game, observes the other players and is offered the possibility of using his own money to punish the free-riders. Third-party conditions are particularly interesting because subjects are not personally involved in the situation so their motivation cannot stem from the most obvious self-directed goals. It has already been shown that third-party punishment is a stable behavioural pattern (Fehr and Fischbacher Reference Fehr and Fischbacher2004b)Footnote ¹¹ and it seems that this behaviour is tightly linked to reputation and group membership. A recent study by Kurzban and colleagues (Reference Kurzban, DeScioli and O'Brien2007) reveals that the propensity and degree of b-altruistic punishment decreases dramatically once there is no reputation to gain from such an action. In another interesting field study with indigenous groups in Papua New Guinea, Bernhard and colleagues (Reference Bernhard, Fischbacher and Fehr2006) found a significant influence of group membership on punishing behaviour: the b-altruistic punishers protected in-group victims far more often than out-group victims.

These studies show the existence of some patterns in third party punishment. However, they do not provide the necessary controls for deciding whether some of the punishers – especially those who cannot gain a good reputation and those who protect out-group victims – are motivated by purely other-regarding considerations. To test this hypothesis, one would have to be able to disqualify the most compelling ‘egoistic’ interpretations for this behaviour: for example, the idea that these punishers might seek the warm glow of being a caretaker of justice,Footnote ¹² or the idea that they do not really discriminate between their own welfare and those of others – in this case, they cannot help identifying themselves with the betrayed players and, consequently, feeling angry and wanting to punish in revenge, as if they themselves had been the target of the free-riding behaviour.

Again, more carefully designed experiments are needed in order to discriminate between possible hypotheses. Experimental economists interested in the debate over psychological altruism could either take this line of approach or they might explore an alternative investigatory path recently made available thanks to the development of new technologies: some economists have started to use fMRI methods to see what happens in our brain while we make decisions to act b-altruistically. The following sections deal with this approach.

3. SOME INTERESTING DATA FROM NEUROECONOMICS

Recently a new tool has started being used in the social sciences and particularly in economics: functional neuroimaging, most notably, functional magnetic resonance imaging (fMRI). With help of this ‘brain scan’ technique, one can obtain rough measures of brain activity while people perform a task. Experimental economic laboratory games are particularly well-suited for the use of this method. The hope is that this new tool can provide additional information on subjects’ motivation and reactions in social dilemma situations. This tool is particularly interesting for the debate over psychological egoism, since we are at a loss to understand precisely what motivates people to act b-altruistically. In this section, we will survey and discuss some interesting studies in neuroeconomics that seem relevant to our theoretical investigation.

Recall Fehr and Gächter's experiment on the public goods game (Reference Fehr and Gächter2002): the second part of their experiment consisted in asking the subjects what motivated them to b-altruistically punish free-riders; results pointed to the emotion of anger. This result is supported by a complementary study using brain-imaging as a research tool. Alan Sanfey and colleagues (Reference Sanfey, Rilling, Aronson, Nystrom and Cohen2003) investigated the neural substrates of the cognitive and emotional processes involved in decision-making. Subjects were asked to play an ultimatum game while their brains were scanned with the fMRI method. In an ultimatum game,Footnote ¹³ participants have to decide either to accept or reject an offer of money made by another player whose task is to distribute a sum of money between the two of them. If the offer is rejected, no one gets anything. Thus, the offer is called an ‘ultimatum offer’. The results of the study revealed that players confronted with unfair behaviour – unfair distribution of the sum was proposed by the other player – showed increased activity in the ‘anterior insula’, a brain area associated with negative emotional feelings. Moreover, the strength of the negative emotional response was correlated with the rejection rate of the unfair offer. In this context, rejection of a distribution offer can be interpreted as b-altruistic punishment, because a possible gain is sacrificed by the player in order to maintain a fairness norm.Footnote ¹⁴

In another highly interesting study, Dominique de Quervain and colleagues (Reference de Quervain, Fischbacher, Treyer, Schellhammer, Schnyder, Buck and Fehr2004) introduced a neuroimaging tool into the experimental design of a trust game followed by a punishment condition. In such a game, two players receive the same amount of money. The first player is asked to decide how much of his money to pass on to the second player – the trustee. All money passed is increased by a multiplication factor of two to four – depending on the game. The trustee then decides how much of this to return to the first player. She is allowed to keep all the money for herself, in which case she would show free-riding behaviour. As in Sanfey's experiment, the subjects were brain-scanned during the game while they learned that they had been cheated by another player and made the decision whether to punish b-altruistically. Observation of the subjects’ neural circuit activation showed that a brain area linked to anticipation of reward – the ‘caudate nucleus’ – played a prominent role when people decided to punish. Subjects who exhibited stronger activation of the caudate nucleus were ready to incur more personal costs to punish a free-rider in comparison with subjects who exhibited low caudate activation. Moreover, it must be noted that the configuration of the game did not allow the punishers to expect future monetary gains in the course of the game; thus, punishment had to be b-altruistic. Experimen-ters interpreted the reward-directed mechanism underlying b-altruistic punishment in hedonistic terms – people seek satisfaction – and consi-dered it to be a result of evolution.

At first glance, the analysis of the results in terms of hedonistic motives is rather surprising because experimental economists are well known to argue in favour of b-altruism. In fact, there is no contradiction here because the concept of psychological egoism does not deny that actions caused by self-interested motives can have positive effects for others and unfavourable ones for the agent – which is precisely how b-altruism is defined. This allows researchers to argue in favour of psychological egoism while maintaining that the behaviour is b-altruistic. As de Quervain and colleagues put it:

Such a remark makes sense in the light of an argumentative strategy often used by advocates of psychological altruism, which consists in providing paradigmatic examples of actions that are not easily explainable in self-directed terms. De Quervain and colleagues invert this strategy and use an example of an apparently altruistic action, showing that, under closer analysis, it turns out to be self-directed. Such a procedure is often used by advocates of psychological egoism in order to undermine the plausibility of the competing theory.

This paradigmatic way of conceiving of motivation in terms of anticipated pleasure is to be found in many economic studies. De Quervain and colleagues report that people seek satisfaction from re-establishing equity. The same line of thought is applied to generous behaviour: in various studies on the psychological and neural mechanisms underlying reciprocal cooperation, it has been shown that activation of the brain areas linked with reward processing – including the ‘caudate nucleus’ – positively reinforce reciprocity and help to resist the temptation to defect (Rilling et al. Reference Rilling, Gutman, Zeh, Pagnoni, Berns and Kilts2002, Reference Rilling, Sanfey, Aronson, Nystrom and Cohen2004; Decety et al. Reference Decety, Jackson, Sommerville, Chaminade and Meltzoff2004; King-Casas et al. Reference King-Casas, Tomlin, Anen, Camerer, Quartz and Montague2005). Tabibnia and colleagues (Tabibnia and Lieberman Reference Tabibnia and Lieberman2007; Tabibnia et al. Reference Tabibnia, Satpute and Lieberman2008) propose the same theoretical approach and go as far as saying that it is regrettable that Sanfey et al. (Reference Sanfey, Rilling, Aronson, Nystrom and Cohen2003) failed to report activation in the reward regions of the brain.

Let us now focus on the question of whether these studies are meaningful in the context of the altruism debate. Data on the neural basis underlying reciprocal cooperation are only mildly relevant because reciprocal cooperation can hardly be conceived as a case of psychological altruism. These studies do not help to answer the question of whether people can act altruistically when it is costly for them and no future benefit is to be expected. Moreover, it must be noted that most of these studies involve economic games where the goal is to win as much money as possible. In such settings, subjects are inclined to cooperate because they are placed in a setting in which they are expected to behave in ways conducive to reaping the long-term benefits of social interaction.

Studies on the psychological motivation for b-altruism seem more relevant, provided one adopts an interpretation of pro-social behaviour in terms of norm compliance. De Quervain and colleagues, for example, suggest that ‘many people voluntarily incur costs to punish violations of social norms’ (p. 1254) and they feel ‘satisfaction’ and ‘relief’ (p. 1258) when they are able to punish norm violations. They claim that it is precisely this satisfaction that drives them to sustain social norms in their group. The most charitable formal way to understand the causal chain implicitly favoured by the authors is the following:Footnote ¹⁶

It is worth noting that such an interpretation in terms of social normsFootnote ¹⁷ is especially favoured by those who consider the establishment of cooperation and other-directed behaviour through the enforcement of social norms to be a specific feature of human societies, in contrast with animal societies. As de Quervain and colleagues write: ‘The ability to develop social norms that apply to large groups of genetically unrelated individuals and to enforce these norms through altruistic sanctions is one of the distinguishing characteristics of the human species’ (de Quervain et al. Reference de Quervain, Fischbacher, Treyer, Schellhammer, Schnyder, Buck and Fehr2004: 1258). On this occasion, the neuroeconomists are very much in line with the neoclassical tradition in economics and the social sciences; they elaborate the notion of human b-altruism precisely through a characteristic that allows them to posit a sharp distinction between humans and animals: the behavioural characteristic of norm compliance fits perfectly with the ideal of the uniqueness of the human being. If it were emotions that we have in common with animals that explained altruism, it would be harder to argue that human altruism was unique. In these words, ‘Experimental evidence indicates that human altruism is a powerful force and unique in the animal world. [. . .] Human societies represent a large anomaly in the animal world. They are based on a detailed division of labour and cooperation of genetically unrelated individuals in large groups’ (Fehr and Fischbacher Reference Fehr and Fischbacher2003: 785).

4. AN ALTERNATIVE INTERPRETATION

As we have seen, one can garner support for the psychological egoism thesis by showing that a type of behaviour apparently driven by a non-selfish motive – a desire for justice or to promote fairness – is in fact motivated by self-interested considerations – the desire to be relieved of an unpleasant feeling. Confirmation for this interpretation of altruistic behaviour can be found in fMRI studies that show the activation of a hedonistic reward mechanism just before the decision is taken (de Quervain et al. Reference de Quervain, Fischbacher, Treyer, Schellhammer, Schnyder, Buck and Fehr2004). This analysis is indirectly strengthened by further fMRI studies on cooperative behaviour that reveal activation of the same reward-directed area (Tabibnia and Lieberman Reference Tabibnia and Lieberman2007; Tabibnia et al. Reference Tabibnia, Satpute and Lieberman2008).Footnote ¹⁸ According to these studies, the desire to promote fairness is only an instrumental motive. Of course, this presents an interesting argument in favour of psychological egoism. Noble actions that seem to be driven by a desire for justice and fairness prove to be self-directed under close analysis.

Unfortunately, there are good reasons to doubt the validity of this analysis because the various studies discussed provide only a few indicators about the mental considerations – conscious or not – underlying the choice to punish b-altruistically. Thus, it is really hard to provide an overall explanation of what happens in people's minds when they exhibit b-altruistic behaviour. In what follows, we propose to unfold and assess the plausibility of various possible interpretations for the same data.

If we rely on the first two studies mentioned in the third section (Fehr and Gächter Reference Fehr and Gächter2002, Sanfey et al. Reference Sanfey, Rilling, Aronson, Nystrom and Cohen2003), it seems clear that negative emotions such as outrage or anger have an important role to play in the motivation for b-altruistic punishment (see also Dawes et al. Reference Dawes, Fowler, Johnson, McElreath and Smirnov2007). These emotional reactions are obviously elicited by the perception of an inequitable split. Unfortunately, Fehr, Sanfey and colleagues’ results do not allow us to distinguish whether subjects feel angry because (i) a norm has been scorned – which would be an instance of outrage – or because (ii) they are frustrated by a personal loss – which would be a more basic form of anger.Footnote ¹⁹

Under the interpretation (i), the subjects feel outraged because what they observe does not correspond to the social norms of cooperation and equity present in their society; at first glance, this seems to lead them either to the ‘noble’ motive to reinforce these norms by means of punishing the free-riders – a view favoured by de Quervain and colleagues – or to the desire to punish the free-riders simply because, as wrongdoers, they deserve it. However, to make this explanatory scheme compatible with the hedonistic reward mechanism discovered by fMRI studies, a further motive has to be introduced in the causal chain, namely the desire to get rid of the unpleasantness of being outraged. We then end up with a complex and unintuitive causal chain where the feeling of outrage in itself does not play a crucial motivational role. However, the most parsimonious way to explain the observed correlation between the strength of punishment and the emotion of anger – or outrage – consists in attributing a causal role to this emotion. This possibility is overlooked in the abovementioned causal explanation.

Alternatively, we favour the second and more straightforward interpretation (ii), which is more in tune with studies on emotions, does not contain the additional causal step of an instrumental motive, and does not attribute sophisticated thoughts and desires to agents, such as a willingness to re-establish justice. According to this account, subjects are simply frustrated and angry because they expect a more favourable outcome for themselves; consequently, either they aim to compensate their loss – which could be described as a self-directed equity motiveFootnote ²⁰ – or they expect the sweet psychological taste of revenge.Footnote ²¹ Both of these motives are self-directed. The first motive however seems weaker because a study by Falk and colleagues (Reference Falk, Fehr and Fischbacher2005) has shown that in a public goods game, altruistic punishment is used even in circumstances where the cost of punishing equates the cost of being punished – that is, punishment does not obviously help to diminish inequity towards oneself. In such a situation, simple forms of self-directed equity motives are unlikely to be elicited and revenge seems to be a more plausible motive.

There are further reasons to favour the more straightforward interpretation for b-altruistic behaviour – at least in contexts where the subjects are victims of free-riding or inequity. It is well established that people are particularly sensitive to unfair shares toward themselves (Croson and Konow Reference Croson and Konow2009) and report desire for revenge after having been victims of free riding behaviour (Carlsmith et al. Reference Carlsmith, Wilson and Gilbert2008). Moreover, in a recent study, Dawes and colleagues (Reference Dawes, Fowler, Johnson, McElreath and Smirnov2007) found that subjects reported feeling angry towards other players with a high income, even when they knew that this high income had been randomly distributed. This emotional reaction is also correlated with engagement in b-altruistic punishment. This means that one can be frustrated by inequity and feel revengeful towards the lucky beneficiary,Footnote ²² even if she is clearly not responsible for this situation. An interpretation of the proximate mechanism for b-altruism in terms of the desire to maintain social norms or in terms of morally deserved punishment cannot account for these results.

Even more impressively, a recent study by Yamagishi and colleagues (Reference Yamagishi, Horita, Takagishi, Shinada, Tanida and Cook2009) shows that people ‘punish’ in situations similar to the ultimatum game, even when their ‘punishment’ is only costly for them and not costly – or not a ‘punishment’ at all – for the free-riders. In one version of this game, free-riders are not even informed about the ineffective ‘punishments’. In this case, the motive for revenge should be conceived in a broad sense. Subjects merely seek internal retaliation. Their offended pride makes them angry; since there is no real punishing opportunity, they find some satisfaction in showing to themselves that they are independent enough to refuse submission to inequity.

To come back to the particular case of de Quervain and colleagues’ study, if we use the above line of thought, the most plausible explanation seems to be the following: subjects in a trust game get angry when a player they trust does not fulfil their expectations. They trust him to cooperate and share the benefit he has gained thanks to their own trusting behaviour. As soon as they understand that this is not the case, they start feeling frustrated, get angry, seek revenge and begin to think of possible retaliatory actions. Under this interpretation, the following causal chain would hold:

It is worth noting here that we are not denying the role of social norms in the decision to punish. We take the content of norms to depend on various psychological biases as well as widespread practices in particular communities (Kahneman et al. Reference Kahneman, Knetsch and Thaler1986). The norms set a framework that influences people's expectations and also their reactions and behaviour. In a social environment where norms of fairness are not widely held and applied, people may not expect fair behaviour from other agents. Thus, greedy behaviour may not cause anger and revenge reactions; subjects may even react negatively towards excessively generous offers!Footnote ²³ These behavioural patterns have in fact been observed in various studies (see Henrich Reference Henrich2004; Bahry and Wilson Reference Bahry and Wilson2006). Overall, it seems that people's emotional and behavioural tendencies are highly context dependent.

5. DISAPPOINTMENTS AND HOPES

If we favour the second interpretation of de Quervain and colleagues’ study – people seek revenge – it becomes clear that no decisive conclusion can be drawn from it concerning the possibility of psychological altruism. If subjects in the experiment react out of anger and seek revenge, the experiment is no longer dealing with actions apparently motivated by the desire for justice or to enforce social norms. It is simply concerned with the motivation underlying retaliatory actions and no advocate of psychological altruism would deny that motivation for those actions is egoistic. The altruism thesis does not assert that all actions are altruistically motivated; it maintains that at least some actions are. In sum, at least the experimental results discussed in this paper favour neither psychological altruism nor psychological egoism.

Let us emphasize here that we are not denying the relevance of these studies for a better understanding of the proximate mechanisms underlying cooperative and b-altruistic behaviour. We have only shown that these studies cannot be easily integrated with the particular debate over psychological altruism. This is bad news for those who want to use this debate as a lever to link economic studies with investigations into human morality. This rather disappointing conclusion leads us to more general considerations on the contribution of experimental and neuroeconomic research for the debate over psychological altruism. We will start with some sceptical remarks before concluding with constructive ideas and hopes for future research.

First, there are good reasons to be sceptical about associating psychological motives with behaviour on the basis of brain scans. The possibility of teasing out selfish motivation from other directed motivation by looking at brain activation is controversial. This technology is immensely helpful in understanding highly modular machineries such as vision or smell. But it is unclear whether we know enough about the physical mechanisms underlying people's thoughts – and even their motivation – to allow for interesting inferences.Footnote ²⁴ A mental state or an emotional reaction is never located in one single brain region and each brain region is involved in more than one process (LeDoux Reference LeDoux2002). Neuropsychologists are very well aware of the difficulty. For example, Golnaz Tabibnia and Matthew Lieberman admit that ‘we cannot confidently infer from the observation of increased signal in a region that activity in that region evoked one mental process rather than another’ (Tabibnia and Lieberman Reference Tabibnia and Lieberman2007). This being said, it seems plausible to think that observation of neural activity in particular regions of the brain sometimes enables one to rule out specific hypothesesFootnote ²⁵ or to strengthen some others. In the latter case, neuroimaging data could converge with interpretations elaborated in other sciences – such as psychology, philosophy or economics.

In sum, one should carefully avoid overstating neurological findings without being over-suspicious. After all, a brief look back into history of science reminds us that new scientific discoveries can stun people at first and that it often takes a long time for the broader scientific community to admit and integrate new knowledge (Camerer Reference Camerer2008).

To conclude, we propose one possible line of research that could be undertaken in order to help solve the debate. Neuroimaging is a particularly interesting tool since, in principle, it could allow us to address some of the limitations encountered by more traditional psychological enquiries. Consider the example of studies on the ‘caudate nucleus’. For the sake of the argument, let us take for granted that activation of the ‘caudate nucleus’ is necessarily linked to self-directed thoughts – more precisely, to anticipation of reward for oneself. Under this assumption, brain-imaging methods could be used to scan the brains of subjects who seem to act without seeking reward. We have seen that ultimatum games are not particularly well designed for that, but there are other games where apparently altruistic behaviour occurs: for example, third party games would provide an excellent setting for this kind of investigation.

In section 2, we have seen that classical experimental economics helped to reveal some behavioural patterns in third-party punishment. However, existing studies still fail to provide data fine grained enough to decide whether any of the punishers are motivated by purely other-regarding considerations. At least two ‘egoistic’ interpretations need to be dismissed: either the punishers seek the warm glow of being a caretaker of justice or they do not really discriminate between their own welfare and those of others and consequently feel revengeful. Scanning the brains of the players who seem to behave as ‘impartial’ justice makers might be a way to control for these explanatory hypotheses. If a brain scan can show that the ‘caudate nucleus’ is not significantly activated in relation to some types of third-party punishment, both of these hypotheses might be ruled out – of course, under the condition that this brain area can truly be associated with reward expectation. B-altruistic punishers would more accurately be described as impartial observers and caretakers of justice and social order, thus, altruists in the psychological sense.

If, on the contrary, brain scans show activation of the ‘caudate nucleus’, the psychological egoism thesis could not be refuted. A further step would then be to determine whether the same kind of self-directed mechanism is activated before other apparently altruistic actions. One type of behaviour that should obviously be investigated is the motivation to help or reward other players at some personal cost, even in circumstances where no reputation or future gain can be expected. The dictator game would provide an excellent setting for this kind of investigation. Here, the most compelling ‘egoistic’ hypotheses to control for would be the desire to get rid of the negative emotion elicited by perception of social inequity, the expectation to feel a warm glow of being a caretaker,Footnote ²⁶ and the fear of feeling guilty for not helping or for disappointing the other player. If a brain scan can show that the ‘caudate nucleus’ is not significantly activated in relation with generous dictators’ choices – that is when dictators transfer up to half of their endowment to the second player – it would be difficult to maintain the two first hypotheses. However, the fear of feeling guilty or of disappointing would still be available alongside the other-regarding interpretation.Footnote ²⁷ Further studies would then be needed to discriminate among these explanations.Footnote ²⁸

At this stage of the debate, it seems to us that the most promising way to decide empirically whether psychological altruism or psychological egoism is true is to run a carefully designed series of experiments where all plausible alternative hypotheses are systematically controlled for. Here economists could take inspiration from the excellent work done in social psychology (Batson Reference Batson1991). Both classical experiments and fMRI studies can be combined depending on the hypothesis tested. Brain imaging studies have the disadvantage – at least nowadays – that fMRI data can easily be over-interpreted, casting doubt on the procedure itself. However, if reliable, such a technique can provide an unprecedented tool for discarding – or providing confirmation for – various hypotheses at once.