Applications of human tutorial dialog in AutoTutor: An intelligent tutoring system

At the University of Memphis we have created an intelligent tutoring system, called AutoTutor, that helps students learn by holding a conversation in natural language. Decades of research on human tutoring have guided our creation of AutoTutor, which implements effective tutoring strategies. Several studies have shown that AutoTutor promotes significant learning gains. The current research examines which features of the dialog can account for the learning gains, and assesses AutoTutor's appropriate use of dialog. Specifically, we explored the dialog patterns from natural tutoring interactions with AutoTutor and analyzed how short pedagogical feedback is related to learning. We found that AutoTutor creates an appropriate model of student knowledge and responds to the students in a manner consistent with their overall performance. These results together with previous findings support the conclusion that AutoTutor is an effective intelligent tutoring system that uses pedagogical strategies that are appropriate for individual learners.