Development of Group Formation System for Collaborative Learning Using Genetic Algorithm with Multimodal Affect Estimation

Haruka Tada, Chun Xie, Itaru Kitahara

Abstract: This research develops a group formulation method to maximize the effectiveness of online collaborative learning by incorporating “emotions” as a primary metric. Affects are multidimensionally recognized from learners’ voice and facial expressions during a greeting session immediately prior to group work. The multidimensional data are compressed into two dimensions to facilitate efficient integration with other metrics in the group formation process. These two affective metrics, derived from voice and facial expressions, are then combined with traditional metrics such as knowledge level, communication skills, and leadership abilities, resulting in five metrics that ensure diversity within groups during the grouping process. In a validation experiment involving 34 university students, the validity of recognizing affects from both voice and facial expressions and the extent of information loss during dimensionality reduction are examined. The proposed method is one of the few group formation approaches that take emotional diversity into account, significantly contributing to the field of Computer-Supported Collaborative Learning.

External IDs:doi:10.1007/978-3-031-93567-1_27