Responding to Unexpected Student Utterances in CIRCSIM-Tutor v.3: Analysis of Transcripts

Abstract

CIRCSIM-Tutor is a dialogue-based intelligent tutoring system that conducts dialogues with medical students about blood pressure regulation. To obtain models for computer-generated dialogues, we analyzed dialogues involving expert human tutors. In this paper we describe some of the interesting and complex patterns we isolated from the human tutorial dialogues in cases where the student gave erroneous or otherwise unexpected results.