Improving controllability and predictability of an interactive user model driven search interface

In this extended abstract, we discuss user controllability and system predictability in the context of exploratory search. When a user is directing a search engine by interactively refining a probabilistic user model using uncertain relevance feedback, usability problems regarding controllability and predictability may arise. By interpreting user’s actions as setting a goal for an optimization problem instead of passive relevance feedback, and by allowing the user to see the predicted effects of an action before committing to it, these problems can be reduced. A preliminary user study with an exploratory search interface indicates small improvements in user acceptance, perceived usefulness and task performance.