Mitigating interpersonal uncertainty in collaborative argumentation: Using LLMs as scaffolds for group cohesion, dialogic behavior and learning performance in CSCL

Key Points

• Learning performance improves with the integration of large language models in collaborative argumentation, enhancing group cohesion.
• Participants show enhanced dialogic behavior during discussions when supported by large language models during group tasks.
• Analysis of collaborative processes reveals that interpersonal uncertainty often detracts from effective argumentation and learning outcomes.
• Implementation of large language models as scaffolds suggests promising strategies for strengthening group dynamics in educational settings.