Epistemically Active Adaptive User Interfaces

An accurate model of the user is one of the most important factors impacting the success of adaptive user interfaces. However, the input information from the user may not be sufficient for construction of accurate user models. We propose epistemically active adaptive user interfaces that initiate subtle interactions with the user in order to collect new information about the user’s state based on his responses (as in mixed initiative systems) or reactions to the probe. These epistemic actions can increase accuracy and decrease the computational cost of user modeling while requiring only low cost responses from the user. ACM Classification: H5.2 [Information interfaces and presentation]: User Interfaces. Graphical user interfaces. General terms: Design, Human Factors