Modeling cooperation among self-interested agents: a critique

This paper reviews recent attempts to model cooperation in groups of self-interested unrelated individuals. It concludes that no empirically plausible model has been developed, and the economist's quest for an explanation of human prosociality in terms of self-interested agents has not been successful. We show that the equilibrium criteria employed in standard game-theoretic models have no attractive dynamic properties, and the mechanisms deployed in these models are not generally seen in empirical examples of cooperative groups. We also show that these models are not robust to plausible rates of signaling error and mutation. We present an alternative, involving strongly reciprocal agents, that satisfies all the requirements for a dynamically stable model of cooperation.