Social Projection or the Application of Human Models

Click to increase image sizeClick to decrease image size

Notes

The article reports only the change rates. The cooperation rates before and after the manipulation in each condition should have been reported.

See Kiyonari, Tanida, and Yamagishi (2000) for the reasons these results differed from those identified by Shafir and Tversky (1992). Shafir and Tversky used logic similar to that used in the target article in deriving the prediction that the second players in a one-shot sequential PDG, who have been informed that the first player has already decided to cooperate, would defect more often than the players of a simultaneously played PDG. The second player facing a cooperative first player is in the same position as the participants in the current experiment, who originally cooperated and yet defected when they were given a chance to unilaterally change their choice. As in the current experiment, Shafir and Tversky's prediction was supported by their experiment. For both the current authors and Shafir and Tversky, the first player's choice should not matter to the second player's decision. To counter their prediction, Hayashi and her colleagues (1999) conducted a one-shot sequential PDG experiment and demonstrated that the first player's choice had an enormous effect, as stated in the text. Kiyonari and her colleagues’ explanation for the difference is that in the Shafir and Tversky (1992) experiment, the participants played a large number of sequential games repeatedly with very little incentive per game. In Hayashi et al.'s (1999) experiment, the participants played a sequential PDG only once, with substantial monetary rewards. Another experiment reported in Kiyonar et al. (2000) showed that participants behaved in a more self-regarding and logical manner when no monetary incentives were involved than when monetary incentives were involved. The implication of the finding by Kiyonari and her colleagues is that when participants face an abstract task without tangible rewards, they behave in a logical manner to achieve what they were instructed to do by the experimenter—to maximize their own points. Kiyonari et al.'s findings are consistent with the finding of the current experiment, in which observers/advisers who did not receive monetary incentives themselves gave advice that maximized their client's points.

The ideas presented in this section are based on the goal/expectation theory advanced by Pruitt and Kimmel (1977).

The social projection hypothesis will confront the same kind of challenge: What is the source of individual differences in P_r? The authors provide an account of the greater than chance P_r value based on the ecological correlation. The majority of players cooperate in games with a high level of K, and the majority defect in games with a low level of K. Hence, a correlation is produced between one's own behavior and the behavior of the majority. The logic here is acceptable, but two problems exist. The first problem is the circularity problem, which involves the use of explanandum as a part of explanan. That is, the correlation is explained through the assumption that the high K induces people to cooperate, which is exactly what the authors aspire to explain using the correlation. They argue that cooperation is a gaining strategy when K is high because players with a reasonably high level of P_r would conclude that cooperation is a more beneficial choice than defection. In short, they explain why P_r is greater than chance (the original explanandum) by resorting to the relationship between K and cooperation (the original explanan). At the same time, they explain the K-C relationship (this time, the explanandum) by assuming the presence of the higher than chance P_r (this time, the explanan). The second problem is the question of the source of individual differences in P_r. P_r is a product of the fact that the same individual faces games with different levels of K, sometimes playing a game with high K and sometimes playing a game with low K. Assuming that this is the only source of P_r being higher than chance, the only source of individual differences is the difference in the variability of K in the games each individual faces in his or her life. According to this explanation, those who have a high P_r value and are cooperative face a wide range of games in terms of K, whereas those who have a low P_r and are not cooperative face a limited range of games in their life. This prediction is counterintuitive to me, because people who have been exposed to a variety of games should have learned about the diversity of responses by other players (implying that their P_r should be lower). The prediction that turns out to be true is a matter of empirical study. At the minimum, however, the authors are responsible for providing their own account for the individual differences in P_r in a way that can be subjected to empirical investigation.

I am not presenting the argument concerning the adaptive advantage of being a conditional cooperator. Kiyonari et al. (2000); Yamagishi (2007); and Yamagishi, Terai, Kiyonari, Mifune, and Kanazawa (2007) provided the error management account of the conditionality of cooperation, which is expressed within the context of the reputation management literature.