Drawing reasonable inferences from inconsistent premises

Inconsistencies are likely to occur in any non-trivial Knowledge Based System (KBS), yet no satisfactory, principled way of recovering from inconsistencies exists. This paper discusses two possible approaches to repair contradictions and to reason from inconsistent premises. Our algorithm combines a reason maintenance system with an algorithm based on Coherence Theory - a Cognitive Science model of how the human mind forms and uses knowledge - to return an inconsistent KBS to a consistent state and to evaluate counterfactual conditionals.