Aiding teachers in constructing virtual-reality tutors

Abstract

Teachers need different tools for constructing virtual realities than do professional programmers. Teachers building tutoring environments need only and should only provide declarative and nonquantitative specification of the application, as such information is sufficient to build powerful prototypes or even products when exploited properly. We describe our METUTOR tutor-generation system for sequential-action skills, which uses means-ends analysis on a teacher's declarative specification of a set of actions. METUTOR asks the teacher to specify conditions for recommending actions, preconditions of actions, and expected and random consequences of actions. METUTOR also asks the teacher to associate pictorial and/or aural representations with facts, and to specify how and when to use them. METUTOR provides facilities for automatic resolution of interactions and conflicts between media objects. We show examples from a firefighting tutor and a pilot's emergency tutor.