Résumé
This paper proposes a new way of handling the inconsistency of a knowledge base when answering to a query about the validity of a formula. The idea is inspired by human behavior in front of inconsistency, namely, try to never encounter it. For this purpose, we encode a kind of compartmentalization of the working memory. More precisely, given a query and a potentially inconsistent knowledge base, called long term memory, our system only loads in working memory the consistent knowledge which is the most related to the query. We position this system with regard to a major reference in the field, Brewka's preferred subtheories, and study its efficiency by providing complexity and experimental results.