Transparent User Modeling for a Mobile Personal Assistant

Abstract

User models (UM) – the explicit representation of all relevant aspects of a user's preferences, beliefs, etc. – form the basis of virtually all adaptive systems. While early approaches re- lied on explicit interviews to fill in the details of such a UM, more recent systems unobtrusively observe their users and automatically infer the contents of the UM. This reduces the workload for the user, but it also makes her lose control over the system's knowledge about her and the way this knowledge is used. This paper discusses the necessity for transparent user modeling as well as various techniques for achieving it.