Intelligent Interface Learning with Uncertainty

This paper presents an intelligent user interface agent architecture based on Bayesian networks. Using a Bayesian network knowledge representation not only dynamically captures and models user behavior, but it also dynamically captures and models uncertainty in the interface’s reasoning process. Bayesian networks’ sound semantics and mathematical basis enhances it’s ability to make correct, intelligent inferences as to the user’s needs. We show explicit examples of our agent’s reasoning using our Bayesian network and present results showing the utility of Bayesian networks in the domain of user interfaces. Content Areas: user interfaces, agent architecture, cognitive reasoning, expert systems, reinforcement learning, probabilistic reasoning