Transparency, Explanation and Trust in User-Adaptive, Personalized Interaction
Lead: Cristina Conati
Transparency, explanation and trust in user-adaptive, personalized interaction. A promising approach to personalization is mixed-initiative adaptation, whereby the interactive system has the ability to suggest suitable personalizations, but the process is mediated by an on-going dialogue with the user(s). A grand challenge in this approach is how to enable the user-adaptive system to explain the reasons for its suggestions in a way that can foster an effective and unobtrusive mixed-initiative interaction, conducive to building the user’s trust in the process. This in itself is an exercise in personalization, as different users, communities and applications are likely to generate different answers to the questions of when and how a mixed initiative dialogue should happen. We are currently investigating these questions in the context of explaining hints during interaction with educational software, as well as the suggestions of a music recommender system.