Decision-Theoretic Planning for User-Adaptive Systems: Dealing With Multiple Goals and Resource Limitations

Abstract

While there exists a number of user-adaptive systems that use decision-theoretic methods to make individual decisions, decision-theoretic planning has hardly been exploited in the context of useradaptive systems so far. This thesis focuses on the application of decision-theoretic planning in user-adaptive systems and demonstrates how competing goals and resource limitations of the user can be considered in such an approach. The approach is illustrated with examples from the following domains: user-adaptive assistance for operating a technical device, user-adaptive navigation recommendations in an airport scenario, and finally user-adaptive and location-aware shopping assistance. With the shopping assistant, we have analyzed usability issues of a system based on decision-theoretic planning in two user studies. We describe how hard time constraints, as they are induced, for example, by the boarding of the passenger in an airport navigation scenario, can be considered in a decision-theoretic approach. Moreover, we propose a hierarchical decision-theoretic planning approach based on goal priorization, which keeps the complexity of dealing with realistic problems tractable. Furthermore, we specify the general workflow for the development and application of Markov decision processes to be applied in user-adaptive systems, and we describe possibilities to enhance a user-adaptive system based on decision-theoretic planning by an explanation component.