Yujian Zhou, Illinois Institute of Technology; Reva Freedman, University of Pittsburgh; Michael Glass, Illinois Institute of Technology; Joel A. Michael and Allen A. Rovick, Rush Medical College; Martha W. Evens, Illinois Institute of Technology
Hinting is an important tutoring tactic in one-on-one tutoring, used when the tutor needs to respond to an unexpected answer from the student. To issue a follow up hint that is pedagogically helpful and conversationally smooth, the tutor needs to suit the hinting strategy to the student’s need while making the strategy fit in with the high level tutoring plan and the tutoring context. This paper describes a study of the hinting strategies in a corpus of human tutoring transcripts and the implementation of these strategies in a dialogue-based intelligent tutoring system, Circsim-Tutor v. 2. We isolated a set of hinting strategies from human tutoring transcripts. We describe our analysis of these strategies and a model for choosing among them based on domain knowledge, the type of error made by the student, the focus of the tutor’s question, and the conversational history. We have tested our model with two classes totaling 74 medical students. Use of this extended model of hinting increases the percentage of questions that students are able to answer for themselves rather than needing to be told.