Investigating Clues for Estimating Near-Future Collaborative Work Execution State Based on Learners’ Behavioural Data During Collaborative Learning

Abstract

Our final goal is to realize an Intelligent Tutoring System using cognitive tutor agent(s) who provide appropriate adaptive feedback according to the learner’s state. We considered that estimating how learners would behave in the near future would be an important cue for providing appropriate support. In this study, we attempted to estimate whether learners will collaborate in the near future during task execution from their behavioral data. We conducted an experiment of cooperative learning between humans, and obtained data on facial expressions, gaze, and voice during the experiment. As a result, we suggest that the participant’s state of execution of collaborative tasks in the near future can be inferred to some extent from multimodal information. On the other hand, it was not possible to successfully estimate whether or not participants would be actively collaborating in the near future. Extraction of features that can estimate the activity state of learners regardless of individual differences is a major task for the future.