A fully model-theoretic semantics for model-preference default systems

Propositional systems of default inference based on the dyadic relation of preference between models have recently been proposed by Selman and Kautz to provide a computationally tractable mechanism for the generation of vivid knowledge bases. In this paper we argue that the formalism proposed, albeit endowed with a semantic flavour, is not a modeltheoretic (or denotational) semantics, as no ontology which is independent of the existence of the knowledge representation language is postulated. Consistently with good modeltheoretic practice we carry on to postulate a language-independent ontology and to use it in the subsequent definition of a fully modeltheoretic semantics for model-preference default systems. This semantics is instrumental in providing guidelines for the development of algorithms that reason on model-preference default systems, and for comparing this with other