On the job market! 🔍
I'm seeking research positions in HCI, Responsible AI, and AI evaluation & auditing, in both academia and industry. Get in touch.

I build evaluation infrastructure that puts communities in control of AI: methods and tools that let the people affected by AI systems define, measure, and audit what counts as good performance.

  • Infrastructure & Behavior · quantitative work using causal inference to understand how infrastructures such as broadband and LLMs shape human behavior and knowledge gaps on Wikipedia and OpenStreetMap
  • Stakeholder AI Evaluation · mixed-methods work building tools that let stakeholders audit and evaluate AI in their own contexts

I publish at CHI, CSCW, FAccT, AIED, and CSCL. I'm a member of the Collaborative Computing Group (MadCollab), advised by Dr. Jacob Thebault-Spieker, and I collaborate with Dr. Shamya Karumbaiah in the Responsible AI for Learning (TRAIL) Lab. Before my PhD, I earned an M.S. in Urban Data Science from New York University.

Research interests: human-computer interaction · responsible AI · AI evaluation & auditing · computational social science · peer production

Selected Publications

add figure:
img/pubs/facct26_repair.svg
When AI Breaks, Teachers Repair: Pedagogical Repair Work in Situated Classroom Practice
Yaxuan Yin, Shamya Karumbaiah, Yujin Kim, Devansh Saxena
ACM FAccT 2026
What happens when classroom AI breaks down? Using focus groups and a deliberation tool we developed, we document the pedagogical repair work teachers perform to keep AI-supported instruction running, and argue that this often invisible labor should reshape how educational AI is designed and evaluated.
doi · pdf
add figure:
img/pubs/chi26_broadband.svg
Beyond Access: Contextualizing the Benefits of Broadband through Contributor Dynamics on Wikipedia
Yaxuan Yin, Jacob Thebault-Spieker
ACM CHI 2026
Does broadband access translate into online participation? Linking broadband connectivity to contributor dynamics on Wikipedia, we show that the benefits of access are contextual, shaped by who contributes, and from where.
doi · pdf
add figure:
img/pubs/facct25_priorities.svg
Responsible AI in Education: Understanding Teachers' Priorities and Contextual Challenges
Yaxuan Yin, Shamya Karumbaiah, Shona Acquaye
ACM FAccT 2025
What does responsible AI mean to the teachers expected to use it? Using a value-elicitation survey approach, we examine teachers' priorities for responsible AI and the contextual challenges that determine whether those priorities can be realized in practice.
doi · pdf
add figure:
img/pubs/aied25_aibat.svg
AIBAT: AI Behavior Analysis Tool for Teacher-Driven Contextual Evaluation of Language Models in Education
Shamya Karumbaiah, Yaxuan Yin, Aayush Bharadwaj
AIED 2025 🏅 Best Paper Nomination
Can teachers make sense of the harms and benefits of LLMs in their daily tasks? We developed AIBAT, an audit tool supporting teacher-driven, contextual evaluation of language models in education, putting the criteria for "good" AI behavior in educators' hands.
doi · pdf
add figure:
img/pubs/cscl25_teacher.svg
Toward Teacher-Centered AI Design: Exploring the Role of Pedagogical Values and Contextual Factors in K-12 Teachers' Perceptions of Responsible AI
Yaxuan Yin, Shamya Karumbaiah, Ju Lim, Shona Acquaye, Sara Schwartz
CSCL 2025 🏅 Best Paper Nomination
How do pedagogical values and classroom context shape what K-12 teachers see as responsible AI? Our findings motivate teacher-centered approaches to designing AI for education.
doi · pdf
add figure:
img/pubs/cscw25_density.svg
The Effect of Population Density on Remote Humanitarian Mapping Activities: A Triple-Difference Analysis
Yaxuan Yin, Jacob Thebault-Spieker
ACM CSCW 2025
Where does remote humanitarian mapping actually happen? Using a triple-difference analysis, we examine how the humanitarian OpenStreetMap platform reshapes underlying biases along population density, creating opportunity in some places while deepening gaps in others.
doi · pdf
add figure:
img/pubs/cscw24_tradeoffs.svg
Productivity or Equality? Tradeoffs in Volunteer Microtasking in Humanitarian OpenStreetMap
Yaxuan Yin, Longjie Guo, Jacob Thebault-Spieker
ACM CSCW 2024
Microtasking makes volunteer mapping more productive, but at what cost? Using causal inference, we identify tradeoffs between productivity and equality in humanitarian OpenStreetMap, showing how task design shapes whose places get mapped.
doi · pdf

Awards, Fellowships & Grants