An Optimized Group Formation Scheme Considering Knowledge Level, Learning Roles, and Interaction Relationship for Promoting Collaborative Problem-Based Learning Performance

Group formation is one of the key processes in collaborative learning because having adequate groups enables a good interaction among the members and is fundamental in order to obtain appropriate learning performance. Therefore, this study presents a novel genetic algorithm-based group formation scheme (GAGFS) that considers heterogeneous complementation of students' prior knowledge levels and learning roles, and the homogeneity of social interaction relationship of the members in the same learning group to help students enhance learning performance and facilitate interaction in collaborative problem-based learning (CPBL) environment. The quasi-experimental research method is used to assess the effects of three different group formation schemes including the proposed GAGFS, random group formation scheme, and self-selection group formation scheme on the learning performance and interaction effects in CPBL environment. Eighty-three students in three Grade 6 classes in an elementary school in New Taipei City, Taiwan were invited to participant in the experiment. One class was randomly assigned to the experimental group that adopts the proposed GAGFS for CPBL. Two classes were randomly assigned to the control 1 and control 2 groups, respectively. The control 1 group adopts the random group formation scheme, whereas the control 2 group adopts the self-selection group formation scheme for CPBL. Learners in the three groups all use CPBL system to perform problem-based learning activities associated with the subject of "global warming", but adopting different group formation schemes. Learning outcomes of problem-based learning and interaction evaluation are used to determine the learning performance and interaction effects among learners in the three learning groups with the support of different group formation schemes. Finally, a semi-structured interview is conducted to validate the variation of the three different group formation schemes in learning performance and interaction effects. The result shows that the proposed GAGFS that considers students' prior knowledge levels, learning roles, and the interaction relationship is significantly superior to the random group formation scheme in terms of the learning performance of the "action 2" learning stage during the four CPBL stages. Moreover, analytical results also show that the proposed GAGFS for group formation is significantly superior to the random and self-selected group formation schemes in terms of peer interaction effects assessed by social network measures.