Research
My research investigates what AI is doing to the social fabric of collaboration — and what would have to change in the design of AI systems for that fabric to hold. The work moves across three connected areas of inquiry: how AI shapes the moral character of group deliberation, how it shapes whose perspectives and ways of knowing get heard, and how it redistributes — or fails to redistribute — participation and belonging in collaborative settings. The architecture below is the working map I use to think about how these areas relate, the mechanisms through which AI bends each, and the social contexts in which they unfold.
Across nested contexts
- Small Groups & Teams
- Classrooms
- Workplaces
- Civic Spaces
- Society & Public Life
The architecture I use to think with — three areas of inquiry, threaded by mediating mechanisms, unfolding from teams to society.
Moral Voice & Collaborative Moral Decision-Making
Whether people retain the capacity to deliberate together on hard moral and ethical questions when AI is part of the conversation. Whether AI's presence shifts how groups reason about and prioritize human values, and whether it suppresses or amplifies individual moral voice within a group.
This is the newest area of the program. Active experimental work is underway at UCI (results April 2026) with a cross-cultural replication in Vienna (Summer 2026). Publications forthcoming.
Epistemic Agency & Ways of Knowing
Whether people maintain a sense of themselves as legitimate knowers, and whether AI expands or narrows the diversity of perspectives and ways of knowing that surface in collaborative settings. Who shapes what the group comes to know and believe.
- Redistributing Cognition in AI-Supported Writing. Choi, Wu, Tate, Tseng, Harnick-Shapiro, Yim & Warschauer. ICLS, 2026.
- From Wandering to Collaboration: Discourse Patterns in Middle School Generative AI Use. Ritchie, Nixon, Powell Tate & Warschauer. LAK, 2026.
- The Use of Large Language Models in Education. Xing, Nixon, Crossley, et al. International Journal of Artificial Intelligence in Education, 2025.
Active investigation through the Spencer Foundation Vision Grant (“Learning Together in the Age of AI”) with Laura Allen, examining how algorithmic systems shape which arguments surface, how perspectives are framed, and whose ways of knowing gain legitimacy.
Inclusion & Equitable Participation
Who speaks, whose ideas get taken up, who is heard, and who is marginalized in AI-mediated collaborative settings. Inclusion as a process-level phenomenon that unfolds through moment-to-moment interaction — one that can be actively shaped, or quietly suppressed, by how AI is designed and positioned in a group. Identity (gender, race, background) operates here as a cross-cutting moderator across all three areas of inquiry.
- Measuring Inclusion in Interaction: Inclusion Analytics for Human–AI Collaborative Learning. Choi & Nixon. arXiv, 2026.
- Read the Room or Lead the Room: Understanding Socio-Cognitive Dynamics in Human–AI Teaming. Choi, Samadi, JaQuay, Park & Nixon. LAK, 2026.
- Agentic Men, Communal Women?: Exploring Gender Bias in LLM-based Leadership Identification for Collaboration Analytics. Choi & Nixon. AIED, 2025.
- Catalyzing Equity in STEM Teams: Harnessing Generative AI for Inclusion and Diversity. Nixon, Lin & Snow. Policy Insights from the Behavioral and Brain Sciences, 2024.
Active investigation through the Jacobs Foundation Fellowship (“Interventions to Promote Inclusivity in STEM Team Problem-Solving Environments”). Phase 1 deployed across four TRAIL pilot studies through 2025; Phase 2 design-prescriptive interventions begin in 2026.
Mediating Mechanisms
The mechanisms through which AI bends each area of inquiry are a layer, not a single through-line. Two are currently named — others are on the candidate list.
- Framing & algorithmic mediation. What AI surfaces, filters, and makes salient shapes the conceptual space within which a group can reason. AI bends deliberation without anyone necessarily trusting or deferring to it — which is why this mechanism slips under the radar of most human–AI teaming research, and why it sits at the center of the Spencer Vision Grant work on epistemic agency.
- Trust & deference. Calibrated reliance, over-trust, push-back, authority framing. When groups defer to AI versus when they push back, and what that does to the moral and epistemic character of group decisions.
Methods & Approaches
- Group Communication Analysis (GCA) — computational framework for identifying socio-cognitive roles in multi-party discourse.
- Multilevel + temporal modeling — analyzing processes at individual and group levels and how they shape each other, across time rather than in snapshots.
- TRAIL platform — experimental apparatus for controlled studies of human–AI teaming in real group interactions.
- Interdisciplinary base — computational discourse science, natural language processing, experimental social science.
What AI does to the social fabric of collaboration will be determined by decisions being made right now. This work is in service of an empirical foundation for designing systems that protect moral voice rather than suppress it, expand epistemic access rather than narrow it, and distribute participation rather than concentrate it.
