نوع دوستی مهندسی: یک تحقیق نظری و تجربی ناشناس ماندن و هدیه دادن

Three double blind treatments of a US$ 10 dictator game are used to examine the role of anonymity and perceptions of anonymity in pro-social behavior. One treatment has dictators viewing pictures of recipients, the second has recipients receiving pictures of dictators and the third is a pictureless control. In each treatment, more than 50 percent of dictators give US$ 0. For those dictators who give positive amounts, the modal gift is US$ 5 in each photograph treatment versus US$ 2 in the control. Twenty-five percent of subjects in each photograph treatment give at least US$ 5 versus 4 percent of subjects in the control.

…no one is willingly just, but only when compelled to be so. Men do not take it to be a good for them in private, since wherever each supposes he can do injustice, he does it (Plato, The Republic 360c–d). Plato examines the relationship between anonymity and moral action with his “Ring of Gyges”, a theoretical construction that cloaks its wearer with impenetrable invisibility. He considers a tale in which the first person to use such a ring murders the king and becomes a cruel tyrant. While Plato believes that unjust actions are bad for the person who performs them, he never refutes the tale-teller’s claim that when people are freed from the constraints of observation and punishment, they usually act unjustly. This experiment investigates the effect of anonymity in a dictator game. The work has three motivations. First, laboratory subjects in most negotiating and altruism experiments do not know the identity of their counterpart. This contrasts with almost all non-laboratory negotiations where the parties negotiate with full knowledge of their counterparts. This discord is relevant because behavior is significantly altered by anonymity in a number of different settings (for a review of these findings, see Roth (1995); pp. 294–304). In this study, photographs are used to selectively reduce anonymity within an otherwise highly private setting. Photographs are a precise step from total anonymity towards the rich and uncontrolled communication of non-laboratory settings. Photographs reduce anonymity but do not allow any verbal messages or cues through facial movement or body language. The second motivation is the disagreement on the effect of shielding subjects’ decisions from the experimenter. Hoffman et al. (1996) (HMS hereafter) find that self-interest is increased when experimenters cannot know the individual decisions of subjects. In contrast, Bolton and Zwick (1995) determine that punishment, not anonymity, is the key to altruistic behavior concluding that, “the punishment hypothesis explains much more of the deviation from perfect equilibrium than does the anonymity hypothesis” (p. 95). The debate on other-regarding preferences provides the third and final motivation for this work. Guth et al. (1982) began the modern discussion by showing that ultimatum game behavior deviates significantly from that predicted for rational, materially self-regarding players. The ultimatum game has such a simple design that Bolton (1991), Rabin (1993), Fehr and Schmidt (in press) and Bolton and Ockenfels (2000) have focused on other-regarding preference structures as explanations for subjects’ behavior and not on deviations from rationality. Subjects with other-regarding preferences attain the selfish goal of maximal happiness by giving money away in some situations (e.g. positive contributions in dictator games) and by foregoing money to hurt others in different situations (e.g. rejections in ultimatum games). These other-regarding preference structures do not predict any behavioral change due to either subject–subject anonymity or experimenter–subject anonymity. The impact of anonymity on behavior thus remains outside the power of these models. More generally, anonymity can be subsumed under the concept of framing effects of Kahneman and Tversky (1984) as one of a long list of factors beyond game structure and information that alter behavior. For those who take framing effects seriously, the challenge is to build models that are both parsimonious and predictive. To be predictive, preference structures should include all the factors that significantly influence behavior, but this works against parsimony. Biology has confronted similar challenges in understanding and modeling the behavior of non-human animals. Specifically, biologists seek to build predictive models that are tractable yet consistent with behavior that varies due to myriad factors. Rather than attempt to directly incorporate all the surface inputs to behavior, biologists have found it productive to hypothesize a limited number of underlying goals that are then implemented by specific mechanisms. One non-human example may serve to illustrate the technique. Holldobler (1977) investigates ants that both tolerate and feed a wide variety of insect parasites, including sometimes preferentially feeding the larvae of beetles over their own siblings. The behavior of these ants would fit a standard definition of altruism as the ant hosts voluntarily help beetles at high cost to themselves. One approach to understanding ant altruism would be to build other-regarding preference structures where ants derive joy from beetle prosperity. Biologists have instead assumed narrow self-interest and sought to understand the exact mechanics of altruism. In the case of ants and beetles, the puzzle of altruism has been solved. Holldobler showed that although beetle larvae are much larger than ant larvae and to the human eye look nothing like their hosts, they mimic the relevant signals of ant larvae. Specifically, ant feeding behavior is conditioned upon a precise set of chemical cues and anything that produces those cues will be fed, regardless of other characteristics. The ant goal of unfettered selfishness is accomplished by mechanisms that use signals to implement behavioral strategies. Accordingly, these mechanisms can be co-opted by other individuals (even from another species) to accomplish different goals. A similar approach is used to understand the behavior of animals much more complicated than ants (see Smith (1978), for a discussion). For example, primates search for food (forage) using contingent strategies that vary based upon a large number of factors including terrain, weather and the location of predators. Optimal foraging models posit a small number of underlying goals including maximizing net energy intake and avoiding predation. All factors that influence behavior are rationalized by their impact on these goals. Thus, the number of surface manifestations is potentially infinite, yet simple and predictive models are possible (and widely used). As compared with other animals, humans have a greater capacity for analytic thought and for overriding instinctual responses. The extent to which the standard toolkit for other species will be useful for human behavior remains to be seen, but the approach has several merits. First, human perceptual systems function identically to those of non-humans in the translation of external information into a limited set of internal signals. Second, the approach retains optimization at its core, but also predicts persistent framing effects. Finally, it has the potential to forecast and explain behavior that systematically deviates from material maximization and equilibrium. As Plato implies, human behavior is affected by considerations of rewards and punishments, which in turn are affected by conditions of privacy. If human behavioral algorithms have continuity with those of non-humans, then the perception of anonymity is performed by mechanisms that use a finite set of signals to estimate the level of privacy. This experiment manipulates those posited anonymity-estimating mechanisms.