The Midterm That Broke the System
Roberto Serrano didn’t set out to start a fire. He just wanted to give his students a break.
After the December 2025 campus shootings—two dead, nine injured—he made a humane call: let them take the midterm at home. Closed-book, yes, but unlimited time. He’d done it once before, twenty years ago, and it worked fine. This time, he thought, maybe they’d finally relax enough to think.
They didn’t.
The average score? 96%. Not 96% for the top 5%. Not 96% for the 10% who’d been crushing problem sets. 96% across the board. In a class where the historical average hovered between 65 and 80.
He didn’t panic. He waited. He’s been teaching economics at Brown for nearly two decades. He knows talent. He knows when students are good. And this? This felt wrong.
So he made a promise: if the final, taken under proctoring, didn’t match, he’d void the midterm. He didn’t accuse anyone. He just said: let’s see who you really are.
The final? 48.6%. A record low. Three students scored zero.
That’s not a curve. That’s a collapse.
"We cannot afford to have a society in which a significant fraction of our best young minds think that cheating is OK," he told them. "We cannot choose to become idiots."
And he was right.
The Numbers Don’t Lie (But They Do Hurt)
The GAITL committee’s report didn’t surprise anyone who’d been paying attention. It just put numbers to the quiet dread we all felt.
56% of undergrads use AI daily or weekly. 67% of grad students. 85% of master’s candidates.
And yet—88% of undergrads and 73% of grad students say they’re afraid AI is making them dumber.
That’s not hypocrisy. That’s grief.
They know what they’re doing. They know they’re outsourcing their thinking. And they hate it.
Faculty? 95% fear for their students’ long-term cognitive health. Four out of five expect a measurable decline in critical reasoning over the next five years.
And here’s the kicker: 75% of professors believe AI will increase cheating. Not might. Believe. Because they’ve seen it.
The American Association of Colleges and Universities found the same pattern nationwide: 25% of assignments submitted with AI help. And that number’s climbing every semester.
This isn’t a Brown problem. It’s a generational one.
The Lie We’re Telling Ourselves
We keep saying "AI is a tool." Like a hammer. Like a calculator.
But a calculator doesn’t write your proof. It doesn’t explain the Keynesian multiplier in your own voice. It doesn’t rewrite your thesis paragraph until it sounds like you wrote it.
AI isn’t a tool. It’s a crutch that pretends to be a brain.
And we’re letting students use it to bypass the hard work of learning.
The GAITL report says students use AI most often for "explaining complex problems" and "debugging code." That sounds noble. But here’s the truth: they’re not using it to learn. They’re using it to skip.
One student told me, "I just want to understand the concept." But then handed in a 12-page paper that read like a GPT-4 essay with a few personal pronouns swapped in.
We’re not teaching them to use AI. We’re teaching them to fake using it.
What Brown’s Doing (And What It’s Not)
Brown’s response? Guidelines. Literacy training. Departmental policies.
That’s fine. But it’s not enough.
Guidelines won’t stop a student who’s drowning in deadlines and anxiety. Literacy training won’t help if the professor doesn’t know how to spot AI-generated prose.
We need to stop pretending this is a technical problem. It’s a pedagogical one.
We need to redesign assignments so AI can’t be a shortcut. We need to make exams about process, not product. We need to ask students to annotate their drafts, to show their thinking, to fail publicly.
And we need to stop grading the final output and start grading the struggle.
The Real Crisis Isn’t Cheating. It’s Apathy.
The most terrifying statistic in the GAITL report? Not the 96%. Not the 48.6%.
It’s this: students know they’re being dumbed down. And they’re still doing it.
That’s not rebellion. That’s resignation.
They’ve internalized the message: it doesn’t matter if you learn it, as long as you get the grade.
And if that’s true—then we’ve already lost.
Serrano didn’t just expose cheating. He exposed a quiet surrender.
We’re training students to be efficient, not thoughtful. To perform, not to understand.
And if we don’t fix that, no policy, no AI literacy course, no proctored exam will matter.
Because the real AI threat isn’t the algorithm.
It’s the idea that thinking doesn’t matter anymore.
