Abstract

) David W. Etherington James M. Crawford AI Principles Research Department 600 Mountain Ave., P.O. Box 636 Murray Hill, NJ 07974--0636 fether, jcg@research.att.com Abstract Early work on default reasoning was motivated by the need to formalize the notion of "jumping to conclusions". Unfortunately, most existing theories of default reasoning require explicitly considering every possible exceptional case before applying a default rule. They are thus inherently undecidable in the first-order case, and remain intractable in all but the most restrictive cases. One possible approach to tractable default reasoning is to restrict consistency checks (which engender much of the intractability of default reasoning) to a restricted context. Unfortunately, consistency checking is undecidable in the first-order case, so a small context in no way guarantees tractability. Another idea is to use linear or polynomial time, but incomplete, consistency checks. Unfortunately, the known tractable check...

Citations