This paper investigates the influence of access to (and use of) different sources of information about potential task- partners behavior on the performance of self-interested agents that are operating in an open environment. The paper focuses on scenarios in which agents must choose partners to accomplish tasks and on task settings in which there is a small group of potential partners, including both people and software agents. The investigations consider three different sources of information about potential partners: (1) individual knowledge from past direct interactions with another agent; (2) gossip, information gained by querying another agent about the behavior of a potential partner; and, (3) reputation system information, information from a centralized repository. The paper considers strategies for contributing information to reputation systems or in response to gossip-queries and the influence of truthfulness as well. To establish a baseline for performance, equilibrium analyses and a set of experiments in which agents were given complete, accurate knowledge of other agents' behavior were performed.
|Original language||American English|
|Title of host publication||Intelligent Agent Technology, 2007. IAT'07. IEEE/WIC/ACM International Conference on.|
|State||Published - 2007|