Vierailuluento: Yang Shi "Understanding Learning with AI: A Human-Centered, Data-Driven Approach to Programming Education"

Understanding Learning with AI: A Human-Centered, Data-Driven Approach to Programming Education

Yang Shi
Utah State University

Abstract: AI offers new opportunities to understand how people learn and to develop more effective learning support. However, computational insights must be grounded in evidence of how students actually think and work, and translated into support that meaningfully benefits learners. My research investigates how computational methods can help us understand student knowledge and reasoning, using programming education as a setting. In this talk, I will begin with student knowledge modeling, examining how approaches developed in other STEM domains translate to programming and why learning context matters. I will then explore complementary perspectives on representing and discovering student knowledge, from data-driven knowledge component discovery and program semantics to theory-guided analyses of students’ reasoning. These investigations reveal both the potential and limitations of computational approaches to understanding learning. I will next examine how these insights can inform the design of learning tools, closing the loop between understanding learners and supporting their learning. Ultimately, I will outline a vision for human-centered AI that uses computational understanding not as an end in itself, but as a foundation for designing learning environments that respond to learners’ needs, respect their agency, and support their reasoning and development.

Bio: Yang Shi is an Assistant Professor in the School of Computing at Utah State University. His research lies at the intersection of human-centered AI, computing education, and learning analytics. He develops data-driven and AI-based methods to model student knowledge and reasoning from programming behavior, investigates how program representations and learning theory can inform computational models of learning, and translates these insights into AI-supported learning tools. His work has appeared in venues including CHI, SIGCSE, EDM, LAK, and UMAP, receiving the EDM Best Short Paper Award (2026), EDM Best Full Paper Award (2021), and ICER Best Full Paper Runner-up (2024). He serves as an Associate Editor of the Journal of Educational Data Mining and will serve as Program Chair for EDM 2027 and Publication Chair for SIGCSE TS 2027. His research has been supported by the NSF, the Gates Foundation, and Digital Promise.

This guest talk is hosted by Assistant Professor Juho Leinonen, Department of Computer Science.