Why Harvard’s Dean Encourages Guardrailed AI Use in College

Summary

Harvard College undergraduate education dean David Deming argues that artificial intelligence is pushing universities from an era centered on expertise toward one centered on learning communities, habits of mind, and shared intellectual life. He says AI models may make expertise more widely accessible and provide personalized tutoring and feedback, but they cannot replace the discipline, teamwork, criticism, and moral development learned in a community. Deming’s position is not that students should use AI indiscriminately. He favors a “barbell” approach: enforce bans where they can be enforced, and where they cannot, acknowledge usage and establish clear guardrails rather than create rules that reward secret cheating. For writing-intensive courses, he proposes controlled initial drafting, supervised AI feedback and research, multiple revisions, oral presentations, and device-free discussions so instructors can observe how students form and defend ideas. He also supports oral defenses as an accountability mechanism, while noting that they are costly and difficult to scale in large classes. Deming says he plans to test these methods in a course on the economics of AI, but the interview ends before reporting reactions from faculty and students or results from the experiment.