Enigmas of gender and cooperation in social dilemmas
Professor Andrew Colman's team found that cooperation did not steadily decline over many rounds of Prisoner's Dilemma games, and that male pairs cooperated far more than female pairs.
Read the original paper
When the classic Prisoner’s Dilemma (PD) game is played repeatedly over many rounds, is cooperation maintained? And do male pairs cooperate almost twice as much as female pairs? Both findings were reported in a very large US experiment published over half a century ago. Now, an experimental study using more rigorous methodological and statistical techniques and much larger financial incentives have corroborated both major findings in a UK population, despite major shifts in the status of women and gender attitudes over the intervening decades. A research group from the University of Leicester, UK – Andrew Colman, Briony Pulford, and Eva Krockow – investigated cooperation in a computer-controlled experiment in which 150 men and women played 300 rounds of two different PD games in fixed pairs. The work is published in the journal Acta Psychologica.
The PD game represents the strategic structure of any interaction in which two people do better by both cooperating than by both defecting (not cooperating) but are each tempted to defect to get the best possible payoff for themselves and leave the other person with the worst possible payoff. According to game theory, rational players who know in advance the number of rounds that they will be playing should defect on every round, because defection pays better than cooperation whether the other person cooperates or defects but experiments invariably find high levels of cooperation.
It has been widely reported in the literature that cooperation declines in repeated PD, but in most experiments, only a small number of repetitions were used. The apparent declines were not confirmed by statistical time-series analysis and may have been simple endgame effects. It was established many years ago that players tend to cooperate less in the last few rounds, when they see the end approaching, and this can be mistaken for cooperation declining steadily over repetitions. The Leicester researchers confirmed the endgame effect in all six groups that they studied (see the attached graphs), although time-series analysis confirmed that there was no significant overall decline in any of the groups.
The US scientists reported an initial decline in cooperation followed by steadily increasing cooperation, but this can be attributed to an unintentional quasi-endgame effect in their study, according to the Leicester researchers. The US participants were asked to calculate their losses and gains at the end of 25 rounds, and this may have created a quasi-endgame effect, reducing cooperation.
The Leicester researchers also replicated the huge gender difference originally found in the US in the 1960s: women paired with women cooperated much less than men paired with men, with mixed-gender pairs showing intermediate levels of cooperation. This gender difference is a puzzle.
It is much larger than most psychological gender differences and runs counter to conventional sex-role stereotypes. It cannot be explained by players adapting their levels of cooperation to the genders of their co-players, because in the UK experiment, players had no way of knowing the genders of their co-players.
Professor Colman commented to Science Featured: “Our most important finding was that the frequently claimed decline in cooperation over repetitions appears to be a misconception, perhaps driven by an expectation that there ought to be a steady convergence toward the game-theoretic equilibrium of joint defection as players come to understand the game better through experience. That is clearly not the case.” On the gender difference in cooperation, Professor Colman said: “Some researchers have attributed this to risk aversion in women, but that can’t be right, because the same difference occurs in repeated Chicken games, where greater risk aversion would make women more cooperative than men. In PD, cooperation risks the worst payoff, but in Chicken, it is a defection that risks the worst payoff.” How, then, can the gender difference be explained? According to Professor Colman: “It may possibly be related to women being more socially oriented and therefore more concerned than men about relative rather than absolute payoffs. In the PD and Chicken, defecting is the only way of guaranteeing not to come off worse than your co-player.”
Image Credit
Giulia Forsythe, Flickr

About the researchers
Journal reference
Colman, Andrew M., Pulford, Briony D., and Krockow, Eva M. “Persistent cooperation and gender differences in repeated Prisoner’s dilemma games: some things never change.” Acta Psychologica, 187 (2018): 1–8. DOI: https://doi.org/10.1016/j.actpsy.2018.04.014





