Research on multi-agent assisted collaborative argumentation design based on regulated learning

Abstract

Argumentation is an effective way for students to engage in collaborative learning, but students are often faced with a lack of regulating skills such as a lack of specific expertise and knowledge of argumentation, and low levels of motivation and engagement. Based on regulation learning theory, the study designed a collaborative argumentation activity incorporating the support of large language model agents and examined its effects on the quality of students' argumentation maps and argumentation discourse patterns. It was found that in terms of argumentation maps, large language model agents can help students search for multiple evidence to support claims and can provide a stable environment. In terms of argumentation discourse patterns, large language model agents can help students reorganize their knowledge, optimize the collaborative process, and facilitate the change of argumentation patterns from individual argumentation to critical co-construction. The findings are informative for the design and development of large language model agents in education and the development of collaborative learning activities.