Exploring AI-Supported Disciplinary Mediation in Student Project Teams' Text-Based Communication

Abstract

Interdisciplinary project-based learning requires students to negotiate differences in language, assumptions, priorities, and working practices. These differences are difficult to surface in text-based team communication, where discussions can become fragmented and AI tools are often used as private side channels rather than shared supports for collective sensemaking. We present Spritz, a Discord-based LLM technology probe that explores how AI might mediate disciplinary boundaries in student project teams. Spritz monitors group chat for signals of semantic or pragmatic boundaries, prompts members to articulate their perspectives through private channels, and returns anonymized syntheses to the shared discussion. We conducted a technology probe study and co-design workshop with 12 university students from technical, business, and design backgrounds. Participants experienced Spritz during a simulated interdisciplinary resource-allocation task and reflected on AI's role in collaboration. Findings show that participants valued AI mediation not only as cognitive support for boundary crossing, but also as a relational buffer. Spritz helped organize fragmented discussion, surface implicit expectations, and clarify divergent interpretations, while softening interpersonal pressure around disagreement and concession. Participants further imagined future AI mediators as switchable roles, including strategic advisors, cross-domain translators, and perspective challengers. However, these expanded roles introduced a central design tension: the neutrality that made AI acceptable as a mediator became unstable when AI began to advise, challenge, or influence team decisions. We contribute empirical insights and design considerations for AI systems that mediate interdisciplinary collaboration in text-based communication while preserving human agency, trust, privacy, and accountability.