We Cannot Choose to Become Idiots
A blind economics professor at Brown University offered take-home exams this semester. His students averaged 96 out of 100. Forty of them scored perfectly. This had never happened before — the historical average for his midterms sits between 65 and 80.
He smelled something wrong. So he switched the final to in-person and told the class why.
Eighteen students dropped the course immediately. Nine more didn’t show up for the exam. Of those 27 students, 22 had scored a perfect 100 on the midterm. Those who took the in-person final averaged 48.
The professor’s name is Roberto Serrano. He went blind at 17, learned Braille, went to Harvard, and now he’s calling this what it is: large-scale cheating with generative AI. He told El PaĂs and Inside Higher Ed the full story this week because Brown’s administration has been lukewarm about it. He doesn’t care if it’s awkward. He cares that his students aren’t learning.
“We cannot choose to become idiots,” he said. That’s the quote.
Here’s the part that bothers me most, and I want to be fair about it.
The counterargument: AI is a tool. We don’t ban calculators in engineering exams. Why ban ChatGPT in an econ class? The world has changed. Students should learn to use the tools they’ll have in the real world.
It sounds reasonable until you hold it next to the data. Forty students scored a perfect 100 on a take-home exam that was harder than previous years’ midterms. Not 90. Not 95. One hundred. Then 22 of those 40 vanished the moment they had to sit in a room and answer questions without AI. That’s not tool use. That’s dependency. The calculator analogy breaks because a calculator doesn’t write the equations for you — it automates arithmetic so you can focus on structure. ChatGPT wrote these students’ economics answers. They never learned the structure.
The second counterargument: Serrano changed the rules mid-semester. Students enrolled knowing exams would be take-home. Changing to in-person after the midterm is unfair.
This one has teeth. Serrano himself acknowledged it — he gave the class a chance to prove him wrong, promising to count the midterm if the final scores came close. He was transparent about why he was suspicious. The students who knew they couldn’t replicate their performance without AI self-selected out. That’s not an unfair gotcha. That’s a controlled experiment with an ethically communicated methodology. The 22 students who ran are the result.
The third counterargument: These are Ivy League students under immense pressure. They’re overscheduled, competing for internships, trying to build resumes. AI cheating is a rational response to an irrational system.
I buy that the system is broken. But the solution isn’t “cheat harder.” The solution is fixing the system. And Serrano’s point is sharper than people want to admit: if the best minds at one of the best universities in the world can’t resist outsourcing their own thinking, what happens when those minds graduate and run things? You don’t fix a broken system by hollowing out the people who are supposed to fix it.
Here’s what actually scares me about this story.
A Princeton survey cited in the article found that 29.9 percent of students admitted to AI cheating on at least one exam or assignment. That number is almost certainly an undercount — people don’t love admitting to cheating. And these are the students who will build the next generation of software, infrastructure, and policy.
The professor said AI cheating leads to “a failed society.” That’s not hyperbole from someone who doesn’t understand technology. Serrano is an economist. He understands incentives. He’s pointing out that when the shortcut becomes indistinguishable from the real thing in terms of grades but completely distinguishable in terms of actual ability, you’ve created a credentialing system that measures nothing.
Brown commissioned a report on AI in teaching that found most students themselves worry about “negative consequences for their cognitive capacity.” They know it’s making them dumber. They do it anyway.
That’s the part that sticks. Not the cheating. The knowing.
Sources: Ars Technica, Inside Higher Ed, El PaĂs