Abstract
Purpose Traditional collaborative learning group models have two key drawbacks. For one thing, it relies solely on grades for a simple ability assessment, ignoring differences in students' knowledge structures. For another, it over-relies on fixed homogeneous or heterogeneous grouping strategies and cannot adapt to diverse teaching scenarios. To address these limitations, we propose a mixed grouping model. Design/methodology/approach First, students' knowledge structures are constructed via concept maps. And we adopt a novel quantitative method to measure knowledge differences. This method integrates concept difficulty, importance, conceptual relationships and students' accuracy. It can avoid the crudeness of traditional grade-based assessment methods. Second, we develop a quantitative formula for student friendships based on bidirectional ratings and propose a mixed grouping model. This model integrates knowledge structure, learning style, and gender and adopts a differentiated weight strategy adapted to teaching tasks. Knowledge structure includes two refined indicators: intra-group knowledge complementarity level (KCL) and inter-group knowledge balance level (KBL). Finally, we conduct experiments on real-world datasets to compare the grouping results generated by the algorithm with those initiated by the students themselves. Findings The experimental results demonstrate that the proposed model highlights the heterogeneity of knowledge complementarity and the rational balance of overall knowledge distribution across groups. Meanwhile, the model maintains appropriate homogeneity in students' learning styles during group allocation. Such a reasonable collocation greatly enhances the adaptability of collaborative groups to diverse teaching demands. In addition, this method effectively avoids the limitation of student-initiated grouping, which overly relies on interpersonal friendship as the main grouping basis. Therefore, the mixed grouping scheme can flexibly adapt to various teaching scenarios, including knowledge integration tasks and creative collaborative discussions and better serve the actual needs of classroom collaborative learning. Originality/value We first apply concept map to represent knowledge structure in grouping and achieved a refined quantification of knowledge differences through two indicators. Then, we propose a friendship formula based on bidirectional ratings, addressing the subjectivity issue in friendship evaluation. Finally, we construct a multi-dimensional collaborative grouping model that integrates dual knowledge structure objectives, learning style, gender and task-adaptive weights, breaking the limitations of fixed strategies and single-objective grouping.
Showing the abstract — retrieve the full paper via the Exa API.