Abstract
This thesis proposes formalisms to model the dynamics of knowledge on agents' belief bases in the environment of a multi-agent system where agents interact with peers, exchanging information about their beliefs. Our proposal focuses on multiple source belief revision, where agents can receive information through multiple independent informants with their own interests and that might not be completely reliable. To that end, the use of a partial order on the set of informants is introduced to represent their credibility as perceived by the agent being modeled. This order can be changed through agent's updates on its partial order relation reflecting new perceptions on the credibility of the informants.