| Most proposals for nonmonotonic reasoning have been oriented towards extensions and/or modifications of classical first-order logic. They usually attempt to construct a model structure in which every proposition in a knowledge base is satisfied. However, this model structure often leads us to unintended conclusions. We argue that the construction of an intuitively attractive and theoretically well-defined nonmonotonic system is possible considering a knowledge base as a set of contexts, were a context is some collection of propositions in the knowledge base. A system built in this way shows higher precision of nonmonotonic reasoning than other approaches.;The intuitive principle behind the above argument is that nonmonotonic reasoning emerges as a result of multicontext interaction. For the reasoning process within a context, we adopt a logic, since it is usually monotonic. In order to capture the interaction between contexts, we develop a meta-model theory which attempts to define some precedence relationship between contexts. Whenever some contexts produce a conflict, the precedence relationship is used to specify which context provides more reliable support. In order to define this relationship between contexts, we develop a seven-valued relevance logic. This logic provides useful ways for defining the precedence relationship between contexts. By providing intermediate truth value (e.g., true by default), meta-model theory is able to compare degrees of certainty between contexts. By providing a new notion of entailment for default inference, meta-model theory is able to compare degrees of specificity between contexts.;In comparison with other approaches, the proposed meta-model theory provides a natural way for encoding default statements. In particular, the qualification problem does not occur and many unwanted conflicts are eliminated. |
