Don't Ban the Car. Build the Seat Belt.

When cars were first invented and accidents started happening, people didn’t destroy the vehicles and go back to riding horses. Instead, we created traffic laws and invented seat belts. The car itself was not a problem to be eliminated, it was a power that needed to be properly regulated. AI in education is no different. What we need is not a ban to turn the engine off, but proper safety measures. The real question should not be “should we allow AI?” but rather “how do we build seat belts?”

Last October, a massive cheating scandal occurred during an online midterm exam at Korea University, where many students used AI to solve problems. In a student community poll, 190 out of 353 respondents admitted to cheating. It is completely natural and responsible for educators to worry about AI abuse and try to keep grading fair. But their instinct to ban AI from classrooms is the wrong response to the right concern. AI is not a threat to be blocked — it is a reality to be navigated, and universities that fail to recognize this are already falling behind.

AI is the same kind of shift. And research shows it brings real gains: in productivity and in equality. When BCG embedded AI into actual consulting work in late 2023, professionals using it outperformed those who didn't across every measure, and the weakest performers improved the most. A separate study from Cornell found the same pattern among academics. Those using AI published significantly more, and the gains were largest for nonnative English speakers — writers whose ideas had always been there. For students like me, AI doesn't replace ideas. It finally lets them be heard.

But this power comes with a cost that is easy to miss. Researchers from the University of Washington found in a recent study that heavy AI users ended up with writing that was more neutral, more formal, and less their own — and reported similar satisfaction with it as those who had barely used AI at all. They could not tell what they had lost. In late 2023, MIT researchers found that regular AI use over four months measurably weakened brain engagement. And Anthropic's own study found that when people used AI to learn something new, they finished at roughly the same speed. However, they retained far less with the steepest drop in debugging, the exact skill needed to catch what AI gets wrong.

This is the problem with a tool that is this easy to use. The productivity gains are real. But used without intention, AI does not just help you work faster — it quietly does the thinking for you. You feel fine. You might even feel productive. And that is exactly where education falls short.

So what does a seat belt look like in education? There are two places to start: how we assess students, and how we assign work. On assessment, oral exams are the most AI-proof tool we have. AI can write an essay, but it cannot sit across from a professor and defend one. This is already happening. In a machine learning course at Penn this semester, every homework submission is followed by a twenty-minute session with a TA — no notes, no AI. Students walk through a randomly selected problem end-to-end, defending their reasoning in real time. The base credit comes from submitting. The real grade comes from understanding. A documentary from Harvard found that even Ivy League graduates could not explain basic science concepts they had studied for years — because they had never been asked to out loud. AI makes this problem worse. Oral exams can help fix it. On assignments, the answer is not less AI — it is harder problems. My professors at KAIST understood this. Instead of banning AI, they provided the tools and assigned projects that AI alone could not solve. We had to learn how to guide it, push it, and think beyond it. That is what AI literacy actually looks like.

Universities that ban AI are not protecting their students, they are leaving them unprepared for a world that has already moved on. The goal is not to turn the engine off. It is to teach the next generation how to fasten their seat belts and take control of the steering wheel.

American Studies 355
Calderwood Seminar in Public Writing: Critiquing American Popular Culture
Paul Fisher
Wellesley College, Spring 2026