Punishment, rational expectations, and relative payoffs in a networked prisoners dilemma

Experimental economics has consistently revealed human behavior at odds with theoretical expectations of rational agents. This is especially true in laboratory games with costly punishment where humans routinely pay to punish others for selfish behavior even though the punisher receives no benefit in return. This phenomenon occurs even when interactions are anonymous and the punisher will never interact with the punishee again. However, costly punishment may not be inconsistent with Darwinian notions of relative fitness. This paper presents exploratory work aimed at a reconciliation between economic and biological expectations of behavior. Agent-based modelling is used to simulate networked populations whose members play the prisoners dilemma while having the ability to altruistically punish one another. Results show that behavior evolving in structured populations does not conform to economic expectations of evolution driven by absolute payoff maximization. Instead results better match behavior expected from a biological perspective in which evolution is driven by relative payoff maximization. Results further suggest that subtle effects of network structure must be considered in theories addressing individual economic behavior.