Favoring Collaborative Learning in PBL: An Automated Solution for Semantic Group Formation

This research-to-practice full paper proposes an automated solution for forming groups of students in computing higher education with the Problem-Based Learning (PBL) approach. PBL is a pedagogical model that promotes the development of professional skills (knowledge, skills, and attitudes) for problem-solving through collaborative learning. In this context, effective team building is essential to the success of PBL, as it can enhance student learning and development. The formation of teams using the PBL approach involves effort and observation of several aspects, which makes manual grouping inefficient in formatting balanced groups. In addition, both the instructor and the student can participate in the grouping decision process, considering the constraints defined by the instructor and the student's satisfaction with the recommendations made. In this sense, when using an automated approach that considers these criteria and possibilities for forming groups, it is possible to streamline the process of structuring teams and mitigate the formation of groups with low potential for success. This study proposes an automated solution for forming groups of students using the DSR (Design Science Research) method applied in evolutionary cycles. It considers the following research question: RQ) How can balanced groups in computing higher education with PBL be formed automatically, considering predefined criteria, attributes, and affinities between their members? This solution considers multiple attributes of the individuals involved and the organizer's criteria for dividing groups, using a semantic structure for group formation. The solution also allows students to negotiate based on affinity with peers and satisfaction with the group. System prototypes were created and evaluated throughout the DSR cycles to evaluate the proposal, demonstrating compliance with the defined restrictions and indication of balanced teams. As primary limitations, we point out challenges in student data collection that hinder group formation. The DSR method iteratively improved solutions, but more application cycles will be necessary to ensure the solution's robustness. In future work, further analysis should link team performance to group formation, focusing on criteria for balanced groups in PBL courses.