Belief base merging and applications to ontologies

Abstract

We are often faced with different experts of an area whose opinions may differ on a particular subject (a common example is medical consultation). These differences have to be "resolved" in order to reach a consensus and act. This consensus may require each participant to modify some of their opinions/beliefs. In the 80s, the AGM model of belief revision offered a conceptual framework for the dynamics of beliefs of a rational agent, in order to maintain his knowledge base consistent while incorporating new knowledge/beliefs and discarding others. This model gave rise to an area of research that already counts 30 years of development. One of the proposals at the beginning of this century, called "merging", was to extend the AGM model for a set of consensus among multiple agents. However both models make important idealizations, such as representing belief states by sets that include all its logical consequences. Although these models work well in theory, they must be strongly adapted to make their implementation possible. One of the possible adaptations is the study of applying the AGM model agents without the ability to believe (explicitly) the consequences of their knowledge. This area is known as "belief base change". We propose to work on developing a model of belief merging for finite agents, i.e., lacking the consequences of their knowledge. This research has a strong theoretical component to define models and a practical component to assess their behavior in real cases. We propose as an application the "merging" of ontologies, allowing a user to execute a query with different ontologies in order to obtain consistent and coherent information, even if the reference sources are jointly inconsistent. (AU)