Dialog Strategies for Adaptive Help Systems

Abstract

Many help systems provide assistance independent of their user’s current situation. However, there are good reasons that two users in different situations should be assisted differently while dealing with the same problem. We describe a decision-theoretic planning approach by which an adaptive help system can determine its dialog strategy in accordance to the user’s current needs. These needs, e.g. the need for a rather comprehensive or a rather concise sequence of instructions, can be derived from a user model. We show how information from the user model can be used to parameterize the decision-theoretic planning process and therewith to generate dialog strategies adaptively.