受限互动与惯例演化：一项实验分析

摘要

In this paper we study how constraints on social interactions can promote the emergence of efficient conventions via decentralized interactions of individuals who can choose their partners of interaction in a social network. We use the stag-hunt game to represent the tension between payoff-dominant and risk-dominant conventions. Theoretical models suggest that stricter linking constraints favor the coordination on the payoff-dominant equilibrium, while less stringent constraints promote coordination on the risk-dominant one. We test and confirm these predictions in a laboratory experiment finding that under tight linking constraints individuals predominantly coordinated on the payoff-dominant action, while looser constraints increased the frequency of the risk-dominant equilibrium being played. Behavioral analysis shows most participants’ decisions followed the myopic best-response model. However, non-negligible deviations are observed which appear to be influenced by both recent payoffs and past deviations, challenging the assumption of random mistakes applied in many theoretical models. These nuances explain why some groups coordinated effectively on the payoff-dominant action even under loose linking constraints.