ارزش یک لبخند: تئوری بازی با چهره انسانی

Many economists and biologists view cooperation as anomalous: animals (including humans) that pursue their own self-interest have superior survival odds to their altruistic or cooperative neighbors. However, in many situations there are substantial gains to the group that can achieve cooperation among its members, and to individuals who are members of those groups. For an individual, the key to successful cooperation is the ability to identify cooperative partners. The ability to signal and detect the intention to cooperate would be a very valuable skill for humans to posses. Smiling is frequently observed in social interactions between humans, and may be used as a signal of the intention to cooperate. However, given that humans have the ability to smile falsely, the ability to detect intentions may go far beyond the ability to recognize a smile. In the present study, we examine the value of a smile in a simple bargaining context. 120 subjects participate in a laboratory experiment consisting of a simple two-person, one-shot “trust” game with monetary payoffs. Each subject is shown a photograph of his partner prior to the game; the photograph is taken from a collection that includes one smiling and one non-smiling image for each of 60 individuals. These photographs are also rated by a separate set of subjects who complete a semantic differential survey on affective and behavioral interpretations of the images. Results lend some support to the prediction that smiles can elicit cooperation among strangers in a one-shot interaction. Other characteristics of faces also appear to elicit cooperation. Factor analysis of the survey data reveals an important factor, termed “cooperation”, which is strongly related to trusting behavior in the game. This factor is correlated with smiling, but is somewhat more strongly predictive of behavior than a smile alone. In addition, males are found to be more cooperative, especially towards female images, whereas females are least cooperative towards female images.

Imagine two persons facing one another for the first time in a social exchange, where there are potential gains to both parties. They know nothing about one another, but each has to make a decision and those decisions will jointly affect their payoffs. How does each anticipate the other's actions? How does one judge whether a partner is trustworthy or predict if trust will be reciprocated? Strategic behavior involves actors choosing strategies contingent on the anticipated actions of others. Actors forecast the choices of others based on a vector of characteristics, which can be either inherent or intentional. Clearly actors embody inherent characteristics, such as gender, ethnicity, or age, that signal a type. Individuals display additional signals through attire, language, and facial expressions. Both inherent and intentional signals provide information and influence strategic choice. Humans perform many actions, purposively taken with full knowledge of their consequences, which appear to be “irrational”; i.e., if they did not perform these actions, they would be better off, at least in the short term. In particular people often put themselves in a position where they must rely on another person to reciprocate a potentially costly trusting move. Examples include ordering software over the internet, leaving an automobile with a valet service, or buying a bottle of fine wine at a new wine shop. In each case we turn over something of value (a credit card number, a car or money) expecting to receive something in return, but at the same time risk exposing ourselves to possible exploitation by “cheaters”. The Internet firm may credit our account and never send the software. The valet may abscond with the car. The wine may be ruined, having been improperly stored. But people routinely trust others and successfully avoid cheaters. How? This paper focuses on a cue that may affect the beliefs held by actors playing a simple game, and affect their willingness to risk an initial trusting move. We argue that the facial expression of a counterpart contains information that is used by an actor in formulating beliefs and subsequent actions. This research presents findings from experiments designed to test the effect of facial expressions on trusting behavior. We examine the effect of a smile on strategy choice in a simple bargaining game, while controlling for sex-pairings. Two questions are addressed: (1) Does smiling elicit trust among strangers? (2) Is there a difference between the sexes in assessing trust? This is the first of a planned series of studies examining the social signaling of cooperation through facial expressions. Our purpose in this initial study is to discover any existing regularity in the relationship between the decision to trust and the facial characteristics of a partner. We isolate the effect of the facial expression of a partner on the decision to trust in a two-person, sequential game; facial expressions are predetermined, and constitute the controlled stimulus in our experiments. We do not examine actual attempts to signal cooperation – whether genuine or deceptive – on the part of partners to a potential exchange, and so do not test for subjects' ability to accurately signal or detect intentions. Plans for future research that will examine these issues are discussed in the final section.

The main results of this study are that: (1) smiling, at the margin, positively affects trust among strangers, (2) facial features can affect cooperation, regardless of smiling, and (3) both males and females are more trusting toward members of the opposite sex. To our knowledge, ours is the first test of the effect of facial expressions on behavior in a controlled laboratory environment with financial stakes. Smiling increases trust among strangers. Subjects were more likely to trust photographs of smiling persons than non-smiling photographs of the same persons. The result is significant in a one-tailed, matched-pairs proportions test, as well as in regression analysis. We interpret this to mean that subjects were able to detect a difference in facial expression, and that facial expression affected the subjects' beliefs about the trustworthiness of the counterpart represented by the facial image. Thus smiling appears to serve as an informative stimulus to elicit trusting behavior. Our results are somewhat more convincing when we consider that several factors might bias our results toward a null finding. First, the photographs of counterparts were taken under artificial conditions unrelated to the experimental environment. As far as we know, the photographic models were simply instructed to display a particular expression “smile!”, and were not told what their expressions would be used for. Most social signals are tied to a social context, but these photographic models were given no such context. It may well be that if the models knew their expressions would be a signal in a bargaining game, they would have offered facial expressions that could be interpreted more readily. Second, there is considerable variability across smiles. Other than our questionnaire ratings, we made no attempt to directly rate the physical attributes of the smiles, or to categorize them as true or false smiles. This could be achieved by using the Facial Action Coding System (FACS) developed by Ekman and Friesen (1982). If smiles are effective signals, then true smiles should have a much stronger effect than false smiles. Third, the use of still photographs biases the perception of the facial expressions. With still photographs, judgements are derived from permanent physiognomic features rather than transient muscle movements (e.g., facial wrinkles cannot be distinguished from wrinkles caused by muscle action), and the natural flow of behavior may be mutilated into meaningless units (Ekman et al., 1982). Social communication is `a structure not of objects but of events'; therefore dynamic footage of facial expression may be a more appropriate means of eliciting naturalistic viewing patterns to expression-dependent facial features (Nahm, Perret, Amaral, & Albright, 1997). Fourth, we have not considered the attractiveness of faces. Some images might be perceived as more attractive than others, regardless of whether they are smiling. Lau (1982) found that smiling and sex had significant effects on attractiveness. Moreover, physical attractiveness has a powerful effect on perceptions of intelligence and predictions of success (Zebrowitz, 1997, Chapters 6 and 7), and is rewarded in the marketplace across a wide sample of professions (Biddle & Hamermesh, 1998; Hamermesh & Biddle, 1994). We plan an investigation of the effect of attractiveness on trust and other aspects of bargaining. Finally, the questionnaire responses (study 1) make it clear that facial features other than smiling may be important in communicating intentions. Our first factor, which was most strongly correlated with trusting behavior, does more than differentiate between smiling and non-smiling images. It also captures something about the “niceness” of the photographic image. As we have labeled the factor, it appears to be an invitation to cooperate. Another potential shortcoming of our study is its reliance on fictional counterparts. In future work we plan to implement an alternative design in which counterparts can freely send their own signal – a smile or frown. Since each subject controls the facial expression that is sent, we will no longer have the degree of control (two expressions for each face) that the current study provides. The experiment presented here controls one side of the signaling problem. Now that we have established the effect of a smile in a controlled setting we can move on to the more complicated design.20 Relating our results to others, we see that the overall level of trust is somewhat lower than that found in most other studies, though our restricted trust game is sufficiently different from other studies to make that comparison difficult. Our game gives subjects the option to trust their counterpart with 20% of their payoff, with the possibility of doubling the amount if the trust is reciprocated; i.e., subjects give up £0.20 for a potential gain of an additional £0.20. In other studies, subjects choose the amount they wish to entrust their counterpart. In a study where the trusted amount triples in value, Berg et al. (1995) find that 87.5% of subjects trust their counterparts with at least 20% of their endowment; in a face-to-face design, Glaeser, Laibson, Scheinkman, and Soutter (2000) double the trusted amount, and find a somewhat higher level of trust among Harvard University undergraduates. Both of these studies have average payoffs about five times the level of our study. While Berg et al., does not examine individual differences in amounts entrusted, the Glaeser et al., study tests for sex differences (among other things) and finds no significant differences in behavior of women and men or across sex pairings. Croson and Buchan (1999) finds that while women and men are equally likely to trust an anonymous partner, women are more likely to reciprocate trust. One similar study (Burnham, McCabe, & Smith, 2000) finds very similar overall levels of trust in a somewhat more complex restricted trust game. They find that when subtle difference in instructions (describing subjects as “partners” rather than “opponents”) are used to cue cooperative intentions, subjects are more likely to trust and reciprocate trust. Since our counterparts are simulated players, the relationship between facial characteristics and reciprocity cannot be examined: that is a topic for further study. None of the previous studies examines facial expressions, nor compares face-to-face with anonymous pairings in a common design. There is considerable evidence that face-to-face interaction leads to different results from anonymous interaction in bargaining games. For example, Frohlich and Oppenheimer (1998) show that face-to-face communication in comparison to e-mail contact improved the joint prisoner's dilemma outcome of subjects. Roth (1995, 295–304) notes that face-to-face interaction is more likely to lead to efficient outcomes and equal distributions. In a time when more and more interactions take place via telephones and electronic mail, it is important to understand the role of nonverbal communication in face-to-face interactions. Video-conferencing and video-telephones might allow sufficient facial communication resulting in similar cooperative outcomes as face-to-face interactions. Further research is required in this direction as well. In addition to the role of facial expressions, our results indicate that both male and female subjects are more trusting of members of the opposite sex. Our analysis provides some evidence that both women and men respond differently to male than to female facial expressions. Consistent with numerous findings from social psychology, these findings might be explained by subjects having a sexual interest in their counterparts. For example, Tidd and Lochard (1978) show that men gave larger tips than women to a smiling waitress, whereas there was no difference in tips when the waitress was not smiling. The simplest explanation might be that men earn more than women, but Tidd and Lochard also suggest that men might be motivated to give larger tips as a way of signaling interest to a women who is issuing a friendly (smiling) expression. We also find that male subjects discriminate more between smiling and non-smiling counterparts whereas female subjects cooperate at similar levels regardless of facial expression. An explanation of this result might be an effect of dominance. Individuals in subordinate roles smile more than individuals in power positions, regardless of their gender, as shown by Deutsch (1990) in simulated interview situations. Since men are more commonly in dominant positions, they might be more receptive to smiles than females. Therefore men might discriminate more between smiling and non-smiling faces. It may also be that women, who are known to smile more than men (Hinsz & Tomhave, 1991), do not regard smiling as an important, or honest, signal, because it has been overused. Our research suggests that actors draw meaning from facial expressions. They use that meaning to condition their own behavior. The change in their own behavior may be due to the effect of facial expressions on beliefs about the counterpart, which are then used to formulate behavioral strategies. Inviting facial expressions may lead subjects to be more likely to forego a safe, best response (Nash equilibrium) strategy in favor of a risky, cooperative (trusting) behavioral strategy. Thus, smiling is used by humans who are interacting in a one-shot game with strangers. This scenario is self-consistent, and would be evolutionarily stable if the signal were more costly to produce in the absence of the appropriate emotion. But as we discussed above, this assertion leaves the reason for the emotional reliability of smiling obscure. One likely explanation relates to the conditions in which human social behavior evolved. Humans are universally social and tended to live in small groups, and hence one-shot interactions may have been a rarity in their evolutionary history. Group living subjects interact repetitively, and reputations count because a fake smiler will be eventually recognized through her record, and hence discriminated against. When reputation for truthful expression of emotion is important, one might expect the evolution of psychological mechanisms that enforce it. If this interpretation is true, the evolved mechanisms will still be present in modern social conditions, when we sometimes interact with waiters once or play a one-shot trust game in the computer with a perfect stranger. Smiling to strangers may be a relic of our tribal past.