Exploring AI-Supported Disciplinary Mediation in Student Project Teams' Text-Based Communication: A Probe-Grounded Co-Design Study

Abstract

Interdisciplinary student teams must negotiate differences in terminology, priorities, and practices, yet these differences are difficult to surface in text-based communication. We introduce Spritz, a Discord-based LLM technology probe designed to create a situated experience of AI-mediated collaboration for reflection and co-design. Spritz detects semantic or pragmatic tensions, privately prompts members to articulate their perspectives, and returns anonymized syntheses to group discussion. We conducted an exploratory study and co-design workshop with 12 students from technical, business, and design backgrounds. Preliminary findings suggest that participants perceived AI mediation along cognitive and relational dimensions. Spritz helped organize fragmented discussions, surface implicit reasoning, and reduce interpersonal pressure around disagreement. Participants also envisioned AI as a strategic advisor, cross-domain translator, and perspective challenger, raising tensions around neutrality, authority, and accountability. We discuss implications for negotiable interventions, transparent role transitions, and preserving human ownership of team decisions.