The coevolution of altruism and punishment: role of the selfish punisher

Punishment is an important mechanism promoting the evolution of altruism among non-relatives. We investigate the coevolution of altruism and punitive behavior, considering four possible strategies: the altruist punisher (AP, a cooperator who punishes defectors), the altruist non-punisher (AN, a pure cooperator), the selfish punisher (SP, a defector who punishes defectors), and the selfish non-punisher (SN, a pure defector). The SP uses a paradoxical strategy as it punishes other defectors. We analyse the effects of SP and AN on the coevolution of altruism and punishment. We study both the score-dependent viability model (whereby the game's score affects survivorship only) and the score-dependent fertility model (whereby the score affects fertility only). In the viability model of a completely mixed population, SP first drives out SN, and hence it helps cooperators (AP and AN) to evolve. In contrast, in the fertility model of a completely mixed population, neither SP nor AN helps the evolution of cooperation. In both the viability and fertility models of a lattice-structured population, SP promotes the spread of AP. In contrast, AN discourages the evolution of AP. These results can be understood that punishment is a form of spite behavior, paying a cost to reduce the fitness of the opponents, and that different models give different magnitude of advantage to spite behavior.