Controlled explanation systems

Abstract

In this paper we extend the definition of explanations to controlled explanations. Traditionally, given a set of facts and a theory, a set of explanations are generated which can be used to infer the facts. When the number of such explanations are more than one, then some criteria of minimality or preference is adapted to select some of the explanations. Most of these selection criteria are syntactic based and are domain-independent. In this paper, we define a system, where the selection can be made using some domain-dependent criteria. We motivate and define controlled explanations, show some of their properties and provide a procedure which generates minimal controlled explanations.