Christopher Dede of Harvard warns that higher education overemphasizes testable "reckoning" skills while neglecting human judgment — the quality that best complements AI. He urges institutions to move beyond automating instruction and instead adopt active, project-based learning and immersive simulations that let students rehearse real-world decisions. Dede also stresses teaching students how to evaluate AI outputs to avoid overreliance on "hallucinations." He predicts many non-elite colleges will adapt curricula to certify graduates' judgment for AI-rich workplaces.
Harvard Fellow: Teach Judgment — Not Just Testable Skills — To Prepare Students For The AI Era
Harvard Senior Research Fellow Christopher Dede says higher education institutions need to focus on helping students learn to work with AI, rather than how to prevent them from using technology to cheat.
Higher education stands at a crossroads as campuses nationwide confront how to integrate artificial intelligence into teaching and assessment. Rather than centering policies on preventing misuse, Harvard Senior Research Fellow Christopher Dede argues colleges should redesign curricula to cultivate human judgment — the capability that complements AI's strengths — so graduates can thrive in AI-rich workplaces.
Dede's Central Argument
Dede distinguishes between "reckoning" (calculative prediction and pattern recognition, the kind of skills measured by tests like the GRE, SAT and LSAT) and "judgment" (contextual, moral and situational decision-making). He traces this distinction to philosopher Brian Cantwell Smith and warns that current systems emphasize the former to the detriment of the latter.
"Our educational systems are geared to teaching reckoning and their quality is measured by assessing reckoning... So what we're doing is preparing human beings to lose to AI as opposed to preparing human beings with skills that complement AI." — Christopher Dede
Why Reckoning Alone Is Not Enough
Dede notes that much of the discourse about AI in education focuses on automating existing practices — "doing things better." But he cautions that if institutions merely teach the same calculative skills more efficiently with AI, they will train students in skills that workplaces will soon automate. Anything AI can reliably teach is likely to be replaced by workplace tools that perform those tasks directly.
As generative AI becomes more commonplace, higher education institutions are grappling with how to approach the subject with students.
Teach Judgment Through New Designs
Instead of banning AI or reverting to paper-only exams, Dede advocates for educational designs that develop judgment: active, collaborative and project-based learning, plus immersive simulations where students must apply skills in realistic, high-stakes contexts. These formats require decision-making under ambiguity and therefore cultivate capabilities machines struggle to replicate.
For example, in a negotiations course, students can rehearse with AI acting as a landlord, used-car salesperson, or boss. Such simulations allow learners to practice strategies and deliver skilled performances in realistic exchanges — and generative AI can analyze those performances to reveal what students truly retained.
Teach Students To Interpret AI
Dede also emphasizes the importance of teaching students how to reason about AI outputs so tools do not become "magic." Students must learn to spot when an AI is producing a hallucinatory or spurious result versus a reliable insight, and to combine AI's analytical reach with human values, empathy, and moral judgment.
While some professors have focused on getting AI out of the classroom entirely, Harvard Senior Research Fellow Christopher Dede warns they are making a mistake.
Concrete Stakes And Anecdote
He highlights an anecdote in which a colleague received roughly 200 internship applications, many of which appeared identical because students had used generative AI to draft cover letters. The colleague quickly narrowed the candidate pool, illustrating that students who cannot distinguish themselves from AI-generated output will struggle to stand out in the job market.
Looking Ahead
Dede is skeptical that elite universities will rapidly overhaul curricula, but he predicts many other institutions will adapt to survive — shifting toward programs that demonstrate graduates possess judgment and an informed understanding of reckoning so AI is treated as a scrutinizable tool, not inscrutable magic.
Implication: Rather than policing devices or reverting to analog tests, higher education should redesign learning experiences so graduates can pair AI's computational strengths with distinctly human judgment.
Help us improve.
