Abstract
The Repeated Prisoner’s Dilemma (RPD) is a metaphor for the difficulties of achieving cooperation in social life. We present results from an agent-based model (ABM) of the RPD in which agents interact in a “market,” equivalent to the non-organizational interactions within the standard RPD, or have the option of interacting within networks, which allow agents to acquire information from other agents with whom they have been paired or to select agents with whom to interact, or hierarchies, in which cooperation is enforced by the threat of third party punishment. The ABM sheds new light on when and how organizations affect cooperation. In relatively nice worlds, agents join networks and thereby insulate themselves against the initial and often devastating defections of nasty players. In relatively nasty worlds, agents enter hierarchies and cooperate sufficiently often to also preserve their share of the population. In moderate worlds, on the other hand, contingent strategies are vulnerable and typically decline as a share of the population. The information advantages are not large enough to justify joining networks and the population of nasty agents is not threatening enough to drive TFTs into the hierarchy. As a consequence, contingent players are defenseless and easily exploited. These middling worlds are, surprisingly, the most dangerous for TFTs. Organizations also improve the welfare of all agents – both nasty and nice – but only in relatively nice worlds. We find the benefits of civil society are contingent on the characteristics of the population in which it emerges.