A note on the rational closure of knowledge bases with both positive and negative knowledge (Q1387806)

Much of the knowledge of an intelligent agent consists of conditional (defeasible) assertions. A usual representation of such assertions is ``If \(\theta\) then normally \(\phi\)''. Let us assume that a knowledge base \(K\) (of the intelligent agent) consists of the conditional assertions. A natural question arises: what other assertions of this form (that the intelligent agent believe) can we, or should we, conclude? A satisfying answer to the question we can consider as a characterization of rational consequence relation. The paper extends some of the results of \textit{D. Lehmann} and \textit{M. Magidor} (some familiarity with their paper ``What does a conditional knowledge base entail?'' [Artif. Intell. 55, 1-60 (1992; Zbl 0762.68057), erratum: ibid. 68, 411 (1994; Zbl 0818.68137)] is assumed). Lehmann and Magidor provided a characterization of the rational consequence relation for positive conditional assertions. The present authors consider also negative conditional assertions of the form ``not \{If \(\theta\) then normally \(\phi\)\}''. It is shown in the paper that the rational closure exists for satisfiable knowledge bases containing both positive and negative conditional assertions and that it has similar properties to those studied by Lehmann and Magidor. An algorithm of Lehmann and Magidor is adapted to the case of both positive and negative conditional assertions. Two sets of rules and axioms for rational consequence are stated which are complete for the positive and negative consequences of a (conditional) knowledge base.