Local Search for Computing Normal Circumstances Models

In this paper, a new method is introduced to check several forms of logical consistency of nonmonotonic knowledge-bases (KBs). The knowledge representation language under consideration is full propositional logic, using “Abnormal” propositions to be minimized. Basically, the method is based on the use of local search techniques for SAT. Since these techniques are by nature logically incomplete, it is often believed that they can only show that a formula is consistent. Surprisingly enough, we find that they can allow inconsistency to be proved as well. To that end, some additional heuristic information about the work performed by local search algorithms is shown of prime practical importance. Adapting this heuristic and using some specific minimization policies, we propose some possible strategies to exhibit a “normal-circumstances” model or simply a model of the KB, or to show their non-existence.