A Method for Promoting Vicarious Learning in an Online Conversation-Based Intelligent Tutoring System

Keith Shubeck and Xiangen Hu

Conversation-based intelligent tutoring systems (ITSs), like AutoTutor, are highly effective at promoting learning across a wide variety of domains. However, not all students benefit equally from these ITSs. Some evidence suggests vicarious learning and learning alongside a peer agent (i.e., trialogues) are most appropriate for learners with lower levels of domain knowledge. Other studies suggest that high domain knowledge learners benefit more by teaching other student agents. There is a need for careful observations of interactions between learner aptitude and various conversation frameworks. In an ongoing online study, participants are assigned to either an interactive condition where they hold a conversation with a tutor agent, or a yoked-vicarious condition where participants observe the interaction. The results of this experiment will shed light on the varying effectiveness of vicarious learning on students with different levels of domain knowledge.