Abstract
The existence of cooperation among non-kin in many species constitutes an apparent paradox for evolutionary biologists. The most commonly accepted explanation is that cooperation can be enforced by mechanisms that reward cooperators or punish cheaters. Most of the theoretical works in evolutionary game theory, however, aim only at explaining how some cooperation can exist at an evolutionary equilibrium, thanks to these enforcement mechanisms. Here, we aim at showing, instead, that evolutionary game theory can also explain the fine-grained properties of the cooperation that takes place in the living world, especially in the case of the human species. First, we address the question of the origin of enforced cooperation: How can enforced cooperation evolve from an initially non-cooperative state? Using tools from the field of machine learning, we show that enforced cooperation can evolve as a by-product of adaptation to interactions with shared interests. We also show that this process has only two possible evolutionary outcomes. Either all cooperative opportunities are enforced, which corresponds to the human cooperative syndrome, or only a very few number are, which corresponds to non-human cooperation. We also propose a variation of this model to explain why many mutualisms are exaggerated forms of cooperation with shared interests. In a second approach, we focus on one specific enforcement mechanism called partner choice. Using agent-based simulations, we show that, when individuals can freely choose their cooperative partners, the only level of effort invested into cooperation that is evolutionarily stable is the one that maximizes the social efficiency of cooperation. We then build analytical models of partner choice imported from economic matching theory. We show that the only evolutionarily stable distribution of the benefits of cooperation is both independent of bargaining power and proportional to each participant's relative contribution. Thus, partner choice explains two fine-grained properties of human cooperation, namely our preferences for the most socially efficient forms of cooperation and our concerns for fair distributions. Finally, we show that costly signalling models of cooperation can explain several properties of moral reputation, and we conclude by discussing directions for future research.