📝 SummarXiv

Abstract

This paper studies an infinite-horizon framework in which two large populations of players are randomly matched to play a Prisoner's Dilemma. Each player lives for two consecutive periods: as a young player from one group, and then as an old player in the other group. Each population has a known fraction of honest types - individuals who always cooperate unless paired with a player who has been observed to defect against a cooperating partner in the past. Because such defections (i.e., breakdowns of trust) are publicly observed, any defector risks carrying a stigma into future interactions. We show that when the benefits from defection are sufficiently large, there exists an equilibrium in which an increase in the fraction of honest types can reduce the likelihood of cooperation. Moreover, we demonstrate that introducing imperfect public memory - allowing past misdeeds to be probabilistically "cleared" - does not enhance cooperation.