How to Educate for AI

"Which school will best prepare today's students for an AI-infused world?" This question, or some variation of it, comes up in nearly every board retreat or parent meeting we facilitate. AI is invariably seen as threatening mass job extinction, as well as the banishment of whole professions and career paths. Parents and students want to AI-proof their preparation, and trustees quickly spot an opportunity for competitive advantage.

A new report from the Massachusetts Institute of Technology (MIT) Ad Hoc Committee on AI Use in Teaching, Learning, and Research Training almost inadvertently suggests an answer to that question. We say "inadvertently" because MIT ends up making a hundred-year-old progressive education argument, and it is not an institution one usually associates with that pedagogical flavor.

The committee describes learning as a cultural practice through which students learn to make meaning, exercise judgment, form identities, and participate in community. The committee then names the failure mode it fears most: students will internalize a transactional model in which assignments are outputs, teachers are evaluators, peers are optional, and knowledge is a commodity to be acquired as efficiently as possible. It asks instructors to make metacognition a central task of teaching, so that students understand how they learn and why learning matters, and to build reflective practice and personal agency. Knowledge is built through cognitive friction, as in a study group working through a proof and an experiment adjusted until it works. This is constructivist learning theory stated transparently, and it justifies nearly every specific recommendation that follows.

The report says more about assessment, the area where AI seems to pose the biggest challenge — how do you know it is the student taking the exam and not AI? — and it should be required reading for school leaders thinking about strategy in the age of AI.