AutoTutor: A human tutoring simulation with an animated pedagogical interface

This paper presents an overview AutoTutor, an intelligent tutoring system that engages in conversationally smooth dialog with students. AutoTutor simulates the discourse patterns and dialog moves of human tutors with modest tutoring expertise. In order to concretize the situation we begin with two short snapshots of AutoTutor in action. In one snapshot, involving an articulate and knowledgeable student, AutoTutor begins with a how question and then simply pumps for further information. In the other, with an inarticulate and less knowledgeable student, AutoTutor begins with a why question and follows this with numerous hints and prompts until the topic is covered. The remainder of the paper describes the system's architecture, which is comprised of seven modules: a curriculum script, language extraction, speech act classification, latent semantic analysis, topic selection, dialog move generation, and animated agent. AutoTutor responds to the learner in real time and runs of a single Pentium processor.