COOPERATION IN A ONE-SHOT PRISONERS-DILEMMA

In this paper, we model social interactions which are characteristic of large economies. The key properties of this model are: (1) agents are randomly matched over time to engage in a Prisoners' Dilemma; (2) each agent routinely interacts with a proper subset of the other agents; and (3) each agent has highly imperfect information about the past conduct of other agents. For this setting, we show the optimality of a rule of thumb which does not discriminate between encounters with agents that one regularly meets and encounters with agents that one never expects to meet again. This rule of thumb generates cooperative behavior in all encounters. (C) 1995 Academic Press, Inc.