Quantifying Agent Strategies Under Reputation

Abstract

Our research proposes a simple buyer/seller game that captures the incentives dictating the interaction between peers in resource trading peer-to-peer networks. We prove that for simple reputation-based buyer strategies, a seller's decision whether to cheat or not is dependent only on the length of its transaction history, not on the particular actions committed. Given a finite number of transactions, a peer can compute a utility optimal sequence of cooperations and defections. With the limited information provided by many reputation systems, a peer has incentive to defect on a large fraction of its transactions. If temporal information is used, equilibrium is reached when peers predominantly cooperate.