Do Reciprocal Strategies Work in Social Dilemmas?

Abstract

According to computer simulations by Parks and Komorita (1997), group-based reciprocal strategies (GBRS; i.e., strategies in which players cooperate only when a sufficient number of others do) are more effective than the strategy of unconditional defection (ALL-D) in 100-person social dilemmas. In this article, theoretical arguments and a replication of the Parks and Komorita simulations are presented, which demonstrate that when strategies are analyzed at the individual level (at which they were defined), GBRS are actually less effective than ALL-D. The divergence of this conclusion from that of Parks and Komorita is attributed to specific selection factors in design and analysis of the original simulations. However, when the strategies are analyzed as strategies for subgroups rather than individuals, the claim that GBRS are more effcetive than ALL-D finds more support, most notably because players using these and other conditionally cooperative strategies tend to defect when single or in small subgroups but to cooperate when in large subgroups.