In this setting, each of the criminals can of course reduce his or her own sentence by confessing to the police, but if both criminals confess, both of them end up going to jail for a longer time than they would have done, had they both remained silent.

In social science experiments, participants are often asked to interact in computerized versions of this prisoner's dilemma game, and the resulting behavioral data has become a staple in the analysis of cooperative behavior and the factors that may have influenced the evolution of cooperation in human (and non-human) societies.

One of the more surprising findings in these studies has been the conspicuous absence of sex related differences in cooperative behavior, as in most of the conducted experiments to date - and certainly for all results taken together - men and women seem to exhibit virtually identical rates of cooperation. This is surprising, not only because it runs counter to stereotypical clichés (intentional) about sex and gender roles, but also because it appears to conflict with competing theories in evolutionary psychology which support the notion that women and men should exhibit different behavior when it comes to cooperation; if not quantitatively, then at least qualitatively. (For example Kitayama argued in a 1991 paper that

"men rely on an independent self-construal,[while women rely on] an interdependent one [and] this view might suggest that social bonds and interactions are less important for men.",

while Psychologists Baumeister and Sommer argue in a 1997 paper that

"social bonds are [...]equally important for men and women, but in a different way. Men seek social connections in larger groups, with the aim of achieving a favorable position in the social hierarchy, while women seek those connections in smaller, or even dyadic, relationships").

This state of competing theory and apparent discrepancy with empirical evidence seems to have been enticing enough to make a group of economists - Gary Charness from the University of California in Santa Barbara, and Aldo Rustichini from the University of Minnesota- take a closer look at the experimental results and at the paradigm that produces them. In particular, they argued that experiments might fail to support either of the theoretical claims, because laboratory experiments are usually played without an audience. This is important, because evolutionary psychology argues for sex differences in cooperation on the grounds that men and women hope to signal different things to their respective audience of peers. Hence the absence of an audience in the laboratory settings might be crucial to the (missing) results.
In order to correct for this - or actually in order to test the hypothesis that playing prisoner's dilemma games in front of an audience might produce differential results for men and women - the researchers designed an experiment with 234 participants.
All of the participants were first randomly assigned to groups of 20 people, and then they played prisoner's dilemma games against each other and against members of the other groups. In each of the experimental games, the two interacting players were watched by one of the respective groups. That is, one player was watched by his or her own group

(in-group), while the other was watched by the opposing players group (out-group). [Note that random assignment into in-group/out-group produces surprisingly strong behavioral effects in many psychological experiments, with people consistently favoring their randomly assigned in-group peers over out-group members in a varied set of interactions].
By further separating between exclusively male, exclusively female, and mixed groups in this experiment, the researchers were then able to produce a rich data set on the differential behavior of men and women in this prisoner's dilemma with an audience. The results are summarized as follows:
Overall men and women are again found to exhibit similar levels of cooperativeness. However, the tendency to cooperate in this new design is crucially dependent on who is watching. Men cooperate less when being watched by their own group, than when being watched by the opponent's group. Women display the opposite pattern; they cooperate when their peers are watching, but not when only members of an "out-group" are watching. The rate of cooperation when peers are watching is higher for women than it is for men, although both exhibit comparable rates of cooperation when being watched by an out-group audience. The lack of an overall gender difference in cooperativeness therefore appears to be a result of the differing responses to who is watching, and the fact that these responses run in opposite directions for men and women.

In summary, the results support the hypothesis of qualitatively different responses to social dilemmas between men and women. From an evolutionary perspective it may be argued that

"while both males and females wish to gain the approval of their in-group members, the actions that are socially desirable differ across gender. Males wish to signal that they are formidable, while females wish to signal that they are cooperative."

Hence women cooperate at higher rates when watched by their in-group peers, while men cooperate less (thus avoiding the possibility that other's see how they have been taken advantage of - e.g. if the opponent defects, while they themselves cooperate - , or showing that they can take advantage of others - e.g. if they defect against a cooperator). This is shown in more detail in the two graphs below:

The results and the experimental set-up are a great example of scientific research really probing into the relation between theory and empirical results. The original paper can be found at UCSB economics department website.

Note: All quotes are taken from the Charness and Rustichini paper. They are not direct quotes of Kitayama or Baumeister.

Cross-Posted with Ingenious Monkey 20-two-5.

<You can subscribe to the evolved primate blog by clicking here, or follow my research tweets on twitter>