Stanford Instructors Use AI Tools to Deepen Student Learning

Stanford University instructors are increasingly experimenting with generative artificial intelligence to deepen learning, while Stanford’s Center for Teaching and Learning and the Office of the Vice Provost for Undergraduate Education have created communities to compare approaches. The Center for Teaching and Learning’s Teaching with AI Community Share-out series, part of AI Meets Education at Stanford, has featured 23 instructors over two academic years from the School of Humanities and Sciences, School of Engineering, School of Medicine, Graduate School of Education, VPUE and Stanford Online High School. Organizers Kenji Ikemoto and Yunshu Fan said the series is meant to foster discussion rather than evaluate individual projects. Presented examples showed AI used as a support layer for teaching rather than an automated replacement. Clinical associate professor Amit Kaushal built a real-time lecture-question system that groups, ranks and summarizes student questions, letting him address common confusion without interrupting the lecture or replacing public discussion. Senior lecturer Waheeda Khalfan and colleagues used AI to analyze years of biology course content, identify assumed prerequisites and create self-diagnostic assessments and the Foundations Navigator tool. With human review, the project compresses course-path guidance from months to weeks while supporting students who enter with different levels of preparation. Assistant professor Shima Salehi and senior chemistry lecturer Jennifer Schwartz Poehlmann developed STEPS, an AI tutor for science and engineering problem solving. Modeled on Schwartz’s teaching, it asks guiding questions instead of giving answers, encouraging students to slow down, plan and practice productive struggle. Student feedback shifted from frustration to appreciation, with some describing it as sounding like their instructor coaching them. Assistant professor Ariel Stilerman introduced classical Japanese tools including BungoBot, Hiki and Sata in a graduate seminar. BungoBot acts as a Socratic tutor by asking questions about classical Japanese sentences rather than translating them outright; Hiki supports reverse lookup for composition, and Sata models conversational classical language without interrupting mistakes. The tools accelerated grammar processing, opened the seminar to non-specialists and allowed students to work with rare calligraphic scrolls and transcribe manuscripts for Stanford’s library. Together, the projects reflect a cautious Stanford approach: AI is used to manage scale, personalize preparation and scaffold reasoning, while faculty remain central to accuracy, pedagogy and judgment.