Conditioning in a fully nonmonotonic context

P. Baroni, G. Guida, S. Mussi

Presented at Workshop on Conditioning and inference in intelligent systems, Erice, 8-15 Aprile 1997



In this paper we explore how the concept of full nonmonotonicity, introduced by the authors in a recent paper, can be related to the classical and widely accepted view of conditioning and belief revision which has been axiomatically set up by Gardenfors.After recalling the concept of full nonmonotonicity and its importance for a correct modeling of many practical reasoning situations, we show that the classical view of conditioning is unable to encompass all the situations that can be covered by full nonmonotonicity. Then we argue that a correct modeling of fully nonmonotonic reasoning requires a substantial extension of the conceptual model of conditioning. In particular we remark the importance of an explicit representation for the concepts of uncertainty about the applicability of a piece of knowledge and of reasoning attitude. Finally, we introduce a preliminary reasoning formalism featuring the full nomonotonicity property.

PDF file

Return to publication list