On heterogeneous model-preference default theories

Default systems based on the notion of "modelpreference" have recently been proposed by Selman and Kautz to give a semantic account of the phenomena involved in default reasoning and to provide a formal justification for the limited cognitive load that default reasoning seems to require of human beings. In this paper we argue that the way these formal systems have been denned makes them inadequate for the task of reasoning in the presence of both certain information and defeasible information. We propose a.modification to the original framework and argue that it formalizes correctly the interaction between these two fundamentally different kinds of information. We then show that the proposed modification has also a positive effect on the complexity of model-preference default reasoning.