Simulating team dynamics with generative agents: a case study on data science teams from digital platforms

Abstract Team-based collaboration is central to tackling complex analytics challenges, yet such collaboration involves intricate and dynamic social interactions. Factors such as collaboration duration, power imbalances in communication, trust levels, and instances of dishonesty have been shown to influence team chemistry in varying contexts. Managing team dynamics remains inherently difficult, prompting increasing interest in using generative artificial intelligence (AI) tools to support analytical work. However, the effectiveness of these tools in real-world team settings is still debated. To better understand and evaluate team dynamics in analytics-driven projects, we develop a multi-agent team dynamics model that leverages large language models as generative agents. This approach allows us to simulate team interactions in the context of open innovation challenges, revealing key insights into the relationship between team behaviours and performance outcomes. Our contributions are twofold: (1) we propose a unified model to holistically understand team dynamics, and (2) we develop a novel simulation framework using generative AI to model text-based team interactions. These contributions offer new perspectives for managing collaboration in analytics-intensive environments and inform the evolving role of AI in supporting innovation and decision-making.