Refreshments are available starting at 10:30 a.m. The seminar will begin at 10:45 a.m.
Abstract
Today, generative AI can produce polished essays, code, and diagrams. However, helping students produce artifacts is not the same as building their understanding. Hariharan Subramonyam argues that most AI-in-education tools commit a fundamental design error: AI automates students’ knowledge access while bypassing the cognitive processes they need to construct understanding. In this talk, Subramonyam presents his research program that locates cognitive work within AI-based education as a first-class design variable, focusing on the process of understanding in the learner. Script&Shift structures AI-assisted writing around source transformation rather than text generation. DraftMarks provides teachers with real-time instrumentation of student writing behavior, repositioning them as designers of learning environments. SimStep provides task-level abstractions that compile plain-language causal descriptions into interactive science simulations. Finally, SPIRE operationalizes dynamic and interactive learning conversations rather than defaulting to generic Socratic dialogue, grounding AI scaffolding in pedagogical theory. Together, these projects demonstrate how AI can be designed to make the process of understanding visible, tractable, and learner-centered. Subramonyam closes with a broader research agenda for educational AI that develops human expertise rather than substituting for it.
Speaker Biography
Hariharan “Hari” Subramonyam is an assistant research professor at the Stanford University Graduate School of Education and, by courtesy, in its Computer Science Department. He is a core faculty member of the Stanford HCI Group and is a Ram and Vijay Shriram Faculty Fellow at the Institute for Human-Centered AI. His research challenges the automation trend in educational AI by centering cognitive work within the learner, drawing on constructivist learning theory to help students bridge abstract knowledge with practical application through purposeful creation. As part of the NSF-funded National AI Institute for Exceptional Education, Subramonyam develops accessible human-AI tools for children with speech and language impairments. His work has received multiple Best Paper Awards at premier human-computer interaction conferences. Subramonyam earned his PhD in information from the University of Michigan School of Information.