Doxastic preference logic

Abstract

To get a realistic model for describing the epistemic states of cognitive agents and their dynamics, we also have to consider more sophisticated forms of beliefs, e.g. meta, control and default knowledge. As was documented by the seven flaws, approaches linked too closely to the Gärdenforsian tradition cannot deal in an appropriate way with these and other epistemic phenomenons. In particular, iterating belief transformations causes problems. A way to meet this challenge might be doxastic preference logic, which is based on a qualitative, higher order preference concept and a derived flexible, highly but not fully introspective notion of belief. Following some basic intuitions about belief revision, we get a corresponding natural preference-based updating procedure, called doxastic revision. The theory presented in this paper allows, at least to a certain extent, to bring in protection, control and reflection, and to avoid forcing of doxastic preference commitments and ungrounded beliefs. Concerning the remaining gaps, extensions of simple doxastic preference logic seem to be rather well suited for integrating nonmonotonic reasoning and belief revision. For the rest, there are some promising ideas around, but they still have to be worked out.

Anyway, whatever might be the right logical framework for belief dynamics, it will not be found within the limits of the current paradigms.