The evolutionary advantage of conditional cooperation

T he question of how voluntary cooperation can be sustained in human and animal groups has been one of the central theoretical problems of the social and biological sciences for the last several decades. As the problem acquired a more complex mathematical form—having evolved from a simple prisoner’s dilemma game to dynamic evolutionary models—it has also garnered considerable attention from mathematicians (e.g., [1, 2]). The problem’s wide interdisciplinary appeal was due in no small part to the unprecedented impact of Robert Axelrod’s The Evolution of Cooperation [3], one of the most influential books to come out of the social sciences in years. The main message of The Evolution of Cooperation was optimistic: It concluded that cooperation enjoys an evolutionary advantage over defection and so will evolve and stabilize without needing a central authority to monitor and punish defectors as required by the Hobbesian argument. This claim was supported by several findings about the evolutionary superiority of a conditionally cooperative strategy called Tit for Tat (TFT). Some of these findings came from computer simulations where TFT enjoyed spectacular victories; some were deductive results in which TFT was shown to have some ‘‘good’’ properties. It soon became clear, however, that Axelrod’s deductive results left the central issues unanswered. Indeed, while The Evolution of Cooperation is full of interesting insights and observations, it has virtually no deductive answers to the problems it attempted to solve. Hence the problem of cooperation remained open, and nontrivially so. JONATHAN BENDOR AND PIOTR SWISTAK