Testing Behaviors in the Play of an Expected Prisoner’s Dilemma

Abstract

We test the impact of a new form of game structure – the expected game – focusing on whether it improves the outcomes of a prisoner’s dilemma. An expected game is defined by the probability-weighted payoffs of two or more different reference or ‘bookend’ games. We analyze the impact of this structure and several relevant variants through a set of human subject-based experiments on non-na¨ıve, consenting participants. As conjectured, we find that real play under this structure does decrease defection in a one-shot prisoner’s dilemma game. Additionally, we find that such effects can be enhanced through framing and priming treatments. We discuss the micro-economic policy implications of our new results.