Inducing conceptual user models

Usually, performance is the primary objective in systems that make use of user modeling (Um) techniques. But since machine learning (Ml) in user modeling addresses several issues in the context of human computer interaction (Hci), the requirements on \performance" are manifold. Thus, machine learning for user model-ing (Ml4Um) has to meet several demands in order to satisfy the aims of involved disciplines. In this article we describe the application of a rather unusual Ml method to Um, namely inductive logic programming (Ilp). Though not primarily associated with eecient learning methods, we motivate the use of Ilp by showing translucency of derived user models and the explanatory potential of such models during a user adapted ltering process. In course of the OySTER project our goal is to induce conceptual user models that allow for a transparent query reenement and information ltering in the domain of Www meta{search.