From partner selection to collaboration in information sharing multi-agent systems

Abstract

This research advances distributed information sharing by equipping nodes (e.g., software agents) in a distributed network with (1) partner selection algorithms in cooperative environments, and (2) strategies for providing and requesting information in competitive environments. In cooperative environments, information providers are willing to provide requested information, but information consumers must consider uncertainty in the quality of provided information when selecting appropriate information providers. In competitive environments, if a self-interested agent can be an information consumer and provider at the same time, agents need to determine the best ways to request and provide information so that the information acquisition utility can be maximized. This research defines a set of metrics for evaluating information acquisition utility, and presents a game-theoretic approach for determining the best information sharing strategies based on stochastic games. The results show how agents build collaborative relationships with appropriate agents and how the information acquisition utility is affected by those relationships.