Handling negative assumptions in a generic user modeling framework

Abstract

Traditionally, there have been two approaches to powerful logic-based user modeling: First, in the modal logic approach, there is one knowledge base that consists of formulas of one (modal) logic formalism. Second, the partition approach divides the user model into partial knowl-edge bases, mainly to distinguish between different types of assumptions about the user. For the user modeling shell system BGP-MS an approach to integrate partitions with modal logic was developed and later refined to become the user model representation and reasoning framework As-TRa, which is also applicable in the more general case of agent modeling. In this framework, however, there is a representational gap between partitions and modal logic. A specific kind of user model contents, which we call negative assumptions, falls into this gap. Since negative assumptions have been quite frequently used with BGP-MS, we developed specialized mechanisms for dealing with them. This paper gives a brief overview of AsTRa, and formally presents the above-mentioned mechanisms. Like the whole AsTRa framework, they are semantically related to modal logic, which is proven.