ABSTRACT
We show that users of question-answering systems expect those systems to be responsive to their unstated plans and goals. Techniques needed to accomplish this should be special cases of more general abilities, in particular, the ability to recognize the user’s plan and to plan a helpful response. We propose and justify a new system architecture embodying this framework, and we illustrate how that architecture is applied in two implemented systems. The first is a question-answering system, and the second is a simple decision-support system, for which both graphic and linguistic means of communication are available.
