Usable adaptive hypermedia systems

Adaptive interfaces have received much criticism because adaptation and automatic assistance generally contradict the principles of direct-manipulation interfaces. In addition, their success depends highly on the ability of user models to capture the goals and needs of the users. As the construction of user models is often based on poor evidence, even the most advanced learning algorithms may fail to infer accurately the user goals. Previous research has put little emphasis on investigating usability problems of adaptive systems and developing interaction techniques that could resolve these problems. This paper examines these problems and presents an interaction model for adaptive hypermedia (AH) that merges adaptive support and direct manipulation. This approach is built upon a new content adaptation technique that derives from fisheye views. This adaptation technique supports incremental and continuous adjustments of the adaptive views of hypermedia documents and balances between focus and context. By combining this technique with visual representations and controllers of user models, we form a twofold interaction model that enables users to move quickly between adaptation and direct control. Two preliminary user studies exhibit the strengths of our proposed interaction model and adaptation technique. Future extensions to our work are outlined based on the weaknesses and limitations that the studies revealed.