Learning how to cooperate when contingencies are ambiguous

This paper studies how economic agents learn to cooperate when the details of what cooperation means are ambiguous. It considers a dynamic game in which one player’s cost for the cooperative action is private information. From the perspective of the other player, this cost is an unknown but stationary function of observable states of the world. Initially, because of information asymmetries, full cooperation can be sustained only at the cost of inefficient punishment. As players gain common expe-rience, however, the uninformed player may learn how to predict her partner’s cost, thereby resolving informational asymmetries. Once learning has occured, players can sustain cooperation more efficiently and reduce the partnership’s sensitivity to adverse economic conditions. Nevertheless, because inducing information revelation has an efficiency cost, it may sometimes be optimal for the uninformed player to remain un-informed even though that limits the amount of cooperation that can be sustained in equilibrium