AI Reveals the Truth of Education: American Universities Begin to Revive an Ancient Tradition

Oral exams return as AI makes written work less trustworthy

American universities are quietly bringing back oral examinations — not as nostalgia, but as a practical response to generative AI. Cornell University has begun requiring students to defend submitted work in 20‑minute, Socratic-style sit‑downs; the University of Pennsylvania is pairing written assignments with face‑to‑face questioning; and it has been reported that New York University’s Stern School is even trialing voice AI agents to assist in oral assessment. On the surface this looks like an anti‑cheating measure. Deeper down, it is a re‑examination of what a written paper actually proves.

A long arc from oral culture to print and back again

For centuries, teaching and assessment were oral: apprenticeship, debate, and on‑the‑spot interrogation were how mastery was judged. The printing press industrialized education — textbooks, standardized exams and essays became the default signals of learning because producing polished text was costly and thus credible. Now generative AI has collapsed that cost. When a machine can produce a fluent, structured essay in seconds, “being able to write well” ceases to be a reliable proxy for understanding. Professors are asking the question they should have asked long ago: are we evaluating true cognition or just formatted text production?

Why this matters beyond campus

The shift matters both pedagogically and geopolitically. Pedagogically, oral defenses force students to show reasoning in real time — to reveal what remains in the mind when the answer sheet is taken away. Emily Hammer, a Penn professor, has said it isn’t only about cheating; reportedly, instructors see genuine losses in cognitive skills and creativity that polished texts used to mask. Geopolitically, this development unfolds amid a global AI arms race — between the United States, China and other actors — that has rapidly put powerful language models into widespread hands, accelerating the need to rethink assessment and credentialing.

A test of what education is for

Are universities simply retrofitting old rituals, or are they reclaiming a core of education lost to successive media shifts? Marshall McLuhan’s media tetrad would predict this: every medium that reaches an extreme revives what it had displaced. In that sense, the return of Socratic interrogation is less a fad than a structural correction — a move to preserve the one thing AI cannot yet fully counterfeit: the thinking that occurs in the presence of a skeptical interlocutor.