Consolidation of plausiblity relations in multi-agent systems

Abstract

Within the context of multi-agent systems, an agent may often ¯nd itself in a position where it receives information through informants. These informants are independent agents who have their own interests and, therefore, are not necessarily completely reliable. It is natural for an agent to be more inclined to believe one informant over another, especially if the informant has proven itself reliable over a period of time. This preference is stored in a plausibility relation, a partial order indicating the relative credibility of the agent's informants. Through careless expansion or other means, inconsistencies may be introduced into the plausibility relation. A consolidation operator is proposed and characterized through a set of postulates. Alternative constructions are discussed. A non-prioritized revision operator for plausibility relations, based on consolidation, is also presented.