Social Projection Can Solve Social Dilemmas

Abstract

Evidence for cooperation in social dilemmas is empirically robust, socially desirable, and theoretically controversial. We review theoretical positions offering normative or descriptive accounts for cooperation and note the scarcity of critical tests among them. We then introduce a modified prisoner's dilemma to perform a critical test of the social projection hypothesis. According to this hypothesis, people cooperate inasmuch as they believe others respond to the situation as they themselves do. The data from three illustrative studies uniquely support the projection hypothesis. We make the analytical case for the social projection hypothesis in the context of the theory of evidential decision making. We review and rebut critical arguments that have been leveled against this theory. We note that a meta-theoretical benefit of evidential decision making is that the rationality of cooperators in social dilemmas is restored without appeals to murky notions of “collective rationality.”

View correction statement:

Corrigendum

Acknowledgments

Part of this work was completed when the first author was visiting at the University of Marburg, Germany, as an Alexander-von-Humboldt Research Prize winner. We thank Johannes Ullrich and the Social Psychology team at the University of Frankfurt for stimulating discussions and Anthony Evans, Steven Guglielmo, Andrew Monroe, and Jan Rummel for comments on a draft of this manuscript. Christina Wehrli helped with data collection, and David Rand provided valuable references.

Notes

In this article, we assume that game theory prescribes the use of the dominating strategy. Some game theorists, however, avoid prescriptive statements. Reinhard Selten, for example, asserted that “game theory is for proving theorems, not for playing games” (as cited in Goeree & Holt, 1999, p. 10564).

Colman (2003) and Colman et al. (2008a) rejected the idea that team reasoning is a special case of other-regarding social preferences. They pointed out that in a High–Low coordination game (e.g., when both players earn 10 points if they both choose “Left,” both earn 5 points if both choose “Right,” and earn nothing if their choices mismatch) subjective payoff transformations with w do not alter the structure of the game. They only increase payoffs. To solve a coordination game, it is necessary to estimate p_c. Neither social preference theories nor team reasoning provide a mechanism for this estimation (Krueger, 2008).

Each value greater than 1 is set to be equal to 1.

We want to be clear that we are appealing to increases in efficiency only as an auxiliary argument to break the analytical stalemate. As noted in our discussion of morality-based decisions and team reasoning, we do not believe that the efficiency argument can stand on its own.

When done considering all of our arguments, the gentle reader will understand why we believe that evidential thinking is magical in these examples but not in the context of social dilemmas.

One must distinguish between two types of correlation. The correlation over players after choice is zero because the players act independently. However, the expected correlation within a player between own cooperation versus own defection with other's cooperation versus other's defection is positive. Likewise, the diagnostic ratio

Averaging one's own discrepant estimates increases predictive accuracy when both estimates can be modeled as the sum of a single true score plus random error (Herzog & Hertwig, 2009; Vul & Pashler, 2008). In prechoice projection, however, the discrepancy between the estimates is the systematic result of differential conditioning, not random error.

Newcomb's problem can be seen a time-reversed trust game (Evans & Krueger, 2009). The player opens one box if she trusts that the demon filled it. In a regular trust game, the trustee may choose to reciprocate trust, in part, because she honors the trustor's decision to accept vulnerability. In Newcomb, this path to reciprocity is blocked. Trusting a demon should thus be harder than trusting a person. The common cause model, of course, does not care.

To players beliefs matter. Vohs and Schooler (2008) found that belief in determinism (vs. free will) increased the likelihood of cheating on a test, presumably because these participants felt exempted from the moral responsibility they associated with free choice. The social-projection hypothesis suggests the opposite in noncooperative experimental games. Here, belief in determinism should help players to feel exempted from the demands of classic rationality.

In fact, the demon's bias increases the two-boxer's accuracy from .99 to .999.

We thank Thorsten Meiser for this suggestion.