Affiliations: University of Ulm, Institute for Information Technology, Ulm, Germany. Email: matthias.bezold@uni-ulm.de, wolfgang.minker@uni-ulm.de | Elektrobit Automotive Software, Erlangen, Germany
Note: [] Corresponding author.
Abstract: One advanced feature of user interfaces in smart environments is adaptation, the ability of the interface to improve itself for an individual user based on an observation of the user's behavior. Adaptation to user behavior comprises two steps: drawing conclusions from the user-system interaction and adapting the interface accordingly. In this article, the user-system interaction is regarded as a sequence of basic events, which are emitted by the interactive system. Different user modeling algorithms use the information from these basic events to extract new knowledge. Finally, adaptations are triggered based on this information. Since adaptations rely on the expressiveness of the underlying data model, a semantic layer is introduced as an abstraction of the interactive system. The system-independent logic of the adaptations is defined on top of this ontology-based layer, whereas the system-specific execution is defined separately. A set of reusable adaptations is defined as design patterns and the adaptations are integrated into the framework. We present an adaptation framework that comprises a user modeling component and an adaptation component. In order to show the technical feasibility of the framework, we created a reference implementation that comprises all presented components. In addition, the implementation serves as a test bed for a user evaluation of the adaptations. For this purpose, test subjects fulfilled tasks with test systems that implement the adaptations.
Keywords: Adaptive interactive systems, adaptations, user modeling, ontologies, adaptation patterns, evaluation of adaptations
