User Models, Intelligent Interface Agents and Expert Systems

The proliferation of user modeling as a means of accurately capturing the beliefs, desires, and intent of users is apparent. We present a dynamic, uncertainty-based user model knowledge representation for use in an intelligent interface agent called GESIA. GESIA’s numerical uncertainty management representation, which has its roots in Bayesian networks, not only dynamically captures and models user behavior, but it also dynamically captures and models uncertainty in the agent’s reasoning process. We critically discuss the construction of user models, relating their construction with user intent and the relevancy of observable properties in the user’s “world.” Our agent’s usefulness is demonstrated in the domain of a generic expert system shell called PESKI.