Doctor of Philosophy
Arthur C Graesser
Scotty D Craig
Philip I Pavlik
Intelligent tutoring system (ITS) developers often assume that the system’s effectiveness is driven by its interactivity, and for good reason; research regularly shows that interactivity is a critical component for learning. However, some evidence suggests that students with low prior knowledge learn best in vicarious settings, or in settings where they can model a peer agent’s positive learning behaviors. The expertise reversal effect suggests that high prior knowledge students may not benefit from the increased granularity and interactivity that a conversation-based ITS can provide. In the current study, participants were randomly placed into an interactive condition where they interacted with AutoTutor, an ITS for critical thinking, or in a vicarious condition where they observed another user interact with AutoTutor. Students who are observing a tutoring session can model positive learning behaviors of the tutee. However, the results indicated that low prior knowledge participants who watched high prior knowledge participants (LH) interact with AutoTutor was the worst performing subgroup. This suggests that observing more accurate and positive learning behaviors in AutoTutor was not enough to promote learning for the LH group. High prior knowledge participants who observed low prior knowledge participants (HL), and both the high and low prior knowledge interactive participants significantly outperformed the LH group on posttest scores. There was no support for the expertise reversal hypothesis. While the results were not significant, high prior knowledge participants who viewed other high prior knowledge participants (HH) had lower posttest scores on average than high prior knowledge participants who viewed low prior knowledge participants (HL). There was also no significant difference between the high prior knowledge interactive group and the low prior knowledge interactive group, which runs counter to what the expertise reversal hypothesis would predict. These findings suggest that the learning of high prior knowledge participants was not inhibited by the increased amount of feedback and potentially redundant information provided by low prior knowledge participants in the vicarious condition. Instead, the trend suggests that the increased amount of negative feedback, conflict episodes, and potential contradictions observed in vicarious settings may have benefitted high prior knowledge participants in vicarious settings.
Dissertation or thesis originally submitted to ProQuest.
Shubeck, Keith Thomas, "Student Prior Knowledge and Learning in an Intelligent Tutoring System: Comparing the Effectiveness of Vicarious and Interactive Dialogues" (2022). Electronic Theses and Dissertations. 3276.