Heuristic reasoning and relative incompleteness

Abstract

In this paper an approach is presented in which heuristic reasoning is interpreted as strategic reasoning. This type of reasoning enables one to derive which hypothesis to investigate, and which observable information to acquire next (to be able to verify the chosen hypothesis). A compositional architecture for reasoning systems that perform such heuristic reasoning is introduced, called SIX (for Strategic Interactive eXpert systems). This compositional architecture enables user interaction at strategic level. It is formally specified in the specification framework DESIRE (DEsign and Specification of Interacting REsoning modules). We show that this reasoning method can adequately be applied to deal with relative incompleteness in domains that have a (layered) empirical basis. A logical analysis of domains is presented making these concepts more precise. A theorem is presented stating that a domain has an empirical basis if and only if a decisive knowledge base is possible. Moreover, a simple criterion is given to test whether or not a domain is empirically founded: implicit definability of hypotheses in terms of observables. © 1993 Elsevier Science Publishing Co., Inc.