Learning belief revision operators

Abstract

The beliefs of an agent change in response to new information. Formal belief change operators have been introduced to model this change. Although the properties of belief change operators are well understood, there has been little work on specifying exactly where these operators come from. In this paper, we propose that belief revision operators can be learned from data. In other words, by looking at the behaviour of an agent, we can use basic machine learning algorithms to determine exactly how they revise their beliefs. This is a preliminary paper advocating a particular approach, and demonstrating its feasibility. Fundamentally, we are concerned with the manner in which machine learning techniques can be used to learn formal models of knowledge and belief. We suggest that this kind of advance will be important for future applications of AI.