Understanding student engagement based on multimodal data in different types of agent-based collaborative learning contexts

Abstract

Computer-supported collaborative learning is one of the major learning styles for learners. However, students' unequal engagement in CSCL leads to free-riding. This paper designs a novel agent called Virtual Human Agent (VHA) and Text Agent (TA), which can be triggered through real-time analysis to give students dynamic support and address the problem of unequal engagement. A total of 24 undergraduate students were divided into eight groups and asked to write the English literature review collaboratively, and we investigated the effect of different types of agents on student engagement based on multimodal data. Students' classroom videos, screencasts, learning logs, and heart rates were collected. The results show the emotional and social engagement of the experimental group with VHA was significantly higher than that of the control group with TA. However, there was no significant effect of different types of agents on behavioral engagement and cognitive engagement. This finding was supported by the video data in that more interactions among the group members were observed in the VHA group. The limitations of this study are discussed, and future research directions are provided.