Revising and updating probabilistic beliefs
A new perspective of probabilistic belief change is developed in this paper. Based on the ideas of AGM and Katsuno and Mendelzon, the operations of revision and update are investigated within a probabilistic framework, where we assume representation of knowledge to be achieved by a nonmonotonic inference operation. We distinguish between revision as a knowledge adding process, and updating as a change-recording process. A number of axioms is set forth to describe each of the change operations adequately, and we derive representation results. Moreover, we elaborate and deepen the close relationship between nonmonotonic inference and belief change for probabilistic logics by introducing universal inference operations as an expressive counterpart to belief change operations. As an example, we present inference based on the techniques of optimum entropy as an adequate and powerful method to realize probabilistic belief change.
Nutzung und Vervielfältigung:
Alle Rechte vorbehalten