Collaboration of untrusting peers with changing interests

Electronic commerce engines like eBay depend heavily on reputation systems to improve customer confidence that electronic transactions will be successful, and to limit the economic damage done by disreputable peers defrauding others. In a reputation system, participants post information about every transaction, and routinely check the posted information before taking any action to avoid other participants with a bad history. In this paper, we introduce a framework for optimizing reputation systems for objects. We study reputation systems in an asynchronous setting, and in the context of restricted access to the objects. Specifically, we study the cases where access may be restricted in time (objects arrive and depart from system) and in space (each peer has access to only a subset of the objects).