Impact of AI-agent-supported collaborative learning on the learning outcomes of University programming courses

With the rapid development of artificial intelligence technology in the field of education, AI-Agents have shown tremendous potential in collaborative learning. However, traditional Computer-Supported Collaborative Learning (CSCL) methods still have limitations in addressing the unique demands of programming education. This study proposes an innovative AI-Agent-supported Collaborative Learning (AI-CL) approach, aimed at enhancing students' programming learning experience and effectiveness through intelligent assistance from AI-Agents. The research developed an AI-Agent system based on large language models and employed a quasi-experimental design to investigate the effects of the AI-CL method on students' learning achievement, self-efficacy, cognitive load, and learning interest. The experiment was conducted at a university in Shanghai, recruiting 45 undergraduate students who were randomly assigned to the AI-CL experimental group (n = 24) and the CSCL control group (n = 21). The experimental context was set in an ACM programming competition teaching environment, lasting for 6 weeks with 70-min sessions each week. Results showed that the AI-CL group significantly outperformed the CSCL group in learning achievement, self-efficacy, and learning interest. Regarding cognitive load, the AI-CL group demonstrated significantly lower mental effort compared to the CSCL group, while there was no significant difference in mental load. These findings not only provide new theoretical perspectives for the application of cognitive load theory and self-efficacy theory in AI-assisted learning environments but also offer strong practical guidance for higher education institutions to introduce AI-Agent-assisted collaborative learning models in programming courses.