Résumé:
CSCL (Computer-Supported Collaborative Learning) environments are transforming collaborative learning by facilitating interactions and knowledge co-construction. However, their full potential remains untapped without fine-tuned adaptation of learner groups and individual learning paths. To address this challenge, our work proposes three complementary algorithmic strategies. First, dynamic learner profiling characterizes each individual based on their competencies and learning style. Second, real-time analysis of collaborative interactions assesses engagement and active participation within existing groups. Finally, a genetic algorithm optimizes group composition by cross-referencing this data with pedagogical criteria, ensuring a balance between cognitive diversity and interpersonal compatibility. This multidimensional approach aims to maximize both academic performance and learner satisfaction.
