Answers are grounded in this paper and cite their sources.
Understanding how teams evolve over time is crucial for improving collaboration and performance in both educational and organizational contexts. Existing models of team development, such as Tuckman’s five-stage framework (Forming, Storming, Norming, Performing, Adjourning), remain largely qualitative and lack scalable tools for real-time analysis. This paper presents a novel generative AI-based system that integrates zero-shot emotion detection and large language models to analyze team conversations and map them into Tuckman’s stages. The system processes textual interactions, detects emotional cues using natural language processing, and aggregates individual and group-level emotion profiles to provide stage-specific feedback. Through a layered architecture, the tool enables both live interaction ("chat mode") and retrospective analysis ("log mode"), offering real-time insights into team dynamics. The model is validated using both synthetic team dialogues, carefully crafted to reflect distinct developmental stages, and real-world data from student project communications. Results show that the system reliably identifies emotional patterns and accurately classifies team stages, providing personalized and collective feedback to support team development. This research bridges computational science and social systems modeling, demonstrating the potential of AI to transform how we assess and support group performance in dynamic, collaborative environments.
No takes yet. Share an insight, caveat, or question.
Barrios-Fleitas et al. (2025) studied this question.