Towards a Pedagogical Conversational Agent for Collaborative Learning
- Yugo Hayashi, Ritsumeikan University, Ibaraki, Osaka, Japan
AbstractThis study focuses on collaborative learning involving a knowledge integration activity, whereby learner dyads explain each other's expert knowledge. It was hypothesized that learning gain can be determined by the degree to which learners synchronize their gaze (gaze recurrence) and use overlapping language (information overlap) during their interaction. Thirty-four learners participated in a laboratory-based eye-tracking experiment, wherein learners' gazes and oral dialogs were analyzed. Multiple regression analysis was conducted, wherein learning performance was regressed on the two independent variables. Then, a simulation was conducted to view how the model predicts performance based on the collaborative process. The results showed that both gaze recurrence and lexical overlap significantly predicted learning performance in the current task. Furthermore, the suggested model successfully predicted learning performance in the simulation. These results indicate that the two variables might be useful for developing detection modules that enable a better understanding of learner-learner collaborative learning．