Jean Vanderdonckt - Machine Learning for Improving Adaptive User Interfaces

Although adaptive user interfaces are aimed at optimizing the end user's performance and/or preference, they are known as suffering from a series of shortcomings: user cognitive disruption (the end user is disrupted by the adaptation), lack of predictability (the end user does not know when and how a user interface will be adapted by the system), lack of explanation (the system rarely provides the end user with some explanation on why this adaptive process took place), and the lack of user involvement (the end user is rarely given the opportunity to intervene in the adaptivity process). In order to address these challenges, machine learning techniques, combined with some end-user development, offer a promising opportunity for improving the whole process, but also introduces new challenges. This presentation will review open issues in the domain and demonstrates two software applying machine learning techniques for intelligent widget selection and adaptive layout of graphical user interfaces based on task model.