A note on uncertainty and cooperation in a finitely repeated prisoner's dilemma

Abstract

An intuitive expectation is that in a finitely repeated prisoner's dilemma, the players will achieve mutual cooperation in at least some periods. Existing explanations for equilibrium cooperation (with agents perfectly informed of one another's characteristics) require that the number of repetitions be unknown, which is in many cases an uncomfortably strong uncertainty assertion. This paper demonstrates that if agents have private information concerning the number of repetitions (as opposed to being completely uninformed), equilibrium mutual cooperation can occur in a finitely repeated game. This appears to be a weaker and more palatable assumption then that of complete uncertainty, and hence provides a natural and useful alternative foundation for mutual cooperation.