The design properties of an AI teammate, such as personality, communication style, and timing of speech, can significantly influence a team's trust, coordination, and decision-making. However, existing tools lack the infrastructure necessary for rigorous study: reproducible configuration of an AI teammate embedded in instrumented, real-time collaboration over extended periods.
We present the Team Research and AI Integration Lab (TRAIL), a web platform that transforms the AI teammate into a configurable, reproducible design object, integrating a Big Five personality model, a selective-participation messaging pipeline, dual memory, chained longitudinal experiments, and export-ready analytics.
In a real-world deployment across six classroom sessions (approximately 51 students), TRAIL maintained longitudinal chaining, ensured the AI held a stable minority of the conversation, and enabled export-driven AI-human text-similarity analysis.
A single blind persona change resulted in a design-consistent double dissociation: the cognitive-scaffolding agent received higher contribution ratings and closer linguistic alignment, while the socially-supportive agent fostered a warmer team climate and reduced over-reliance.
Blogger's Review: The TRAIL platform provides a crucial infrastructure for exploring the complexities of human-AI collaboration, particularly in terms of reproducibility and configurability. This approach not only enhances the interaction capabilities of AI but also lays a solid foundation for further experimental research, making it worthy of broad application and promotion.