User models and user interfaces: A case for domain models, task models, and tailorability

My thesis is that the case for user models has little empirical support, necessarily must confront currently intractable problems, and is motivated by aspects of a metaphor that does not provide, at least presently, an effective base for the design of user interfaces. More positively, I argue that some of what is sought in the name of user modeling can be accomplished by basing interface design on models of application domains, user tasks, and by providing tailorable systems. I am led to this position primarily from experiences with the design of ICAI systems and construction of mul-timodal interfaces to complex systems. In the limited space of these proceedings I am confined to sketching a portion of my case against user modeling. I focus on the metaphor that I think motivates most user modelin