The Responsible AI for Learning Lab · UW–Madison
TRAIL Lab is an interdisciplinary research group at the University of Wisconsin–Madison studying the responsible use of AI in real-world classrooms. We work across the learning sciences, learning analytics, machine learning, and human-centered design — with teachers, with students, and with the messy specificity of where learning actually happens.
PI
Shamya Karumbaiah
Established
2023 · Madison, WI
§ 01 · Disciplines
We treat AI in education as a problem too consequential for any single discipline. Our work braids four together — each one keeping the others honest.
How people actually learn — across cognition, culture, and context. Our work is grounded in this question, not in what is technically possible.
Where, when, and whether AI can serve teaching. We study integration as a question, not a foregone conclusion — and we honor the answer no.
Reading the patterns in how students engage, struggle, and grow. We design analytics that surface signal without flattening the people inside it.
Co-designing with the educators and learners who live with the consequences. Their judgments are not user feedback — they are the research.
§ 02 · Focus areas
Field site
AI does not enter education in the abstract — it enters fifth-period algebra, a multilingual ELA cohort, a fifteen-minute lesson with thirty learners. We study what integration means inside those rooms, in partnership with the teachers who run them.
— fig.02
Inquiry
We audit language models for the biases they carry into student-facing settings, examine consent and surveillance in classroom analytics, and ask whose values get encoded as defaults. Ethics here is not a checklist — it is the research question.
— fig.03
Method
We build the empirical record on what AI actually does for learning — through controlled studies, longitudinal partnerships, and methodological work on how to measure bias reliably for small groups. We resist hype and we resist dismissal in equal measure.
— fig.04
A defining stance
“Should AI be in education at all? That is the question we are willing to answer no to. The work begins there — with consent, evidence, and the people who live with the consequences.”
§ 03 · News
Talks, papers, panels, awards. The work in motion.
Kaycie Barron presented on Translanguaging, Systemic Functional Linguistics, and Artificial Intelligence in K-12 Science Classrooms: A Multiple Case Study at the American Education Research Association conference in Los Angeles, California.
Kaycie Barron was awarded the 2026 Michael Vincent O’Shea and Harriet Frisbie Eastabrooks O’Shea Fellowship for her work on advancing ethical AI applications in bi/multilingual learning contexts.
Kaycie Barron presented on Characterizing Language Model Biases Using Linguistic Variations in Translanguaging at the Society for Research on Educational Effectiveness (SREE) conference in Chicago, Illinois.
Alina Guha presented on Teacher Perceptions on AI Support for Bi/multilingual Learners: Superpowers and Constraints at the Artificial Intelligence in Education conference.
Shamya Karumbaiah was an invited speaker at the "AI and Society" event hosted by the Wisconsin International Resource Consortium.
Kaycie Barron presented on Evaluating Language Models Using Linguistic Variations in Multilingual Learners’ Writing at the International Conference of the Learning Sciences (ICLS) in Helsinki, Finland.
28 entries total
§ 04 · An invitation
We work with educators, students, scholars, and engineers — anyone willing to sit with hard questions about AI in learning and answer them with evidence. Whether you are a prospective PhD student, a teacher curious about co-designing tools, or a researcher looking to collaborate, we want to hear from you.
TRAIL LabThe Responsible AI for Learning Lab — asking whether AI belongs in classrooms, not just how.
University of Wisconsin–Madison · Educational Psychology
Feedback or accessibility issues? shamya.karumbaiah@wisc.edu