A Brown professor switched his exams from take-home to in-class. The grade distribution collapsed. He was shocked.
Everyone else wasn’t.
Let me be clear about what happened, because the details matter. A professor—presumably thoughtful, presumably experienced—decided to move his exams back into the classroom. No laptops. No ChatGPT. Just students, a desk, and whatever they’d actually retained. The result was a bloodbath. Grades plummeted. The gap between what students were producing at home and what they could produce in a room was so wide it could only mean one thing.
Most of them had been cheating.
We didn’t discover cheating. We discovered that we’d been trusting a locked door that was never locked.
The professor framed this as a revelation. An alarming insight into the state of academic integrity in the age of AI. The comments section, predictably, was less impressed.
“All tests should be in-person. This is an easy problem to solve.”
“The collective always taking the path of least resistance is not the profound finding the article chocks it up to be.”
Ouch. But also: correct.
Here’s what nobody wants to say out loud: the cheating wasn’t a moral failure. It was a systems failure. And the system that failed wasn’t AI—it was the assumption that you can hand someone a take-home exam, give them access to the most powerful answer-generating tool in human history, and trust that intrinsic motivation would win.
That assumption was always delusional.
The scandal isn’t that students used AI. The scandal is that we built a system that assumed they wouldn’t.
Think about the incentive structure. You’re a college student. You’re carrying five classes. You’re sleep-deprived, stressed, and staring at a take-home essay exam. On one side of the screen is the exam question. On the other side is ChatGPT, which can produce a B+ answer in twelve seconds. Nobody is watching. The stakes feel high. The risk feels zero.
What exactly did we think was going to happen?
This is the part where educators get defensive. “Students should have integrity.” “Learning is about the process, not the grade.” “If they cheat, they’re only hurting themselves.”
All true. All irrelevant.
Human nature doesn’t break because a new tool arrives. It just finds a faster path.
The path of least resistance isn’t a character flaw. It’s a law of physics applied to behavior. Water flows downhill. Students use the easiest available tool to complete a task they don’t intrinsically value. This isn’t cynicism—it’s observation backed by every behavioral economics study ever conducted.
The professor at Brown didn’t uncover a crisis. He uncovered a mirror. And what it reflected was an assessment system designed for a world that no longer exists.
Take-home exams made sense when the only resources available were textbooks, notes, and maybe a roommate’s advice. They tested your ability to synthesize information over time. They assumed a relatively level playing field of temptation.
That world is gone. ChatGPT didn’t destroy it—it just made the destruction undeniable.
You can’t design a system that fights human nature and then act betrayed when human nature wins.
The solution isn’t better AI detection. It’s not honor codes. It’s not moral panic or sermons about academic integrity. The solution is so obvious that the top comment on the article said it in one sentence:
“All tests should be in-person. This is an easy problem to solve.”
Is it really that simple? Mostly, yes. In-person exams remove the opportunity. They don’t eliminate all cheating—nothing does—but they eliminate the specific category of cheating that AI enables: the effortless, invisible, zero-risk kind.
But here’s the uncomfortable follow-up: if in-person exams are the answer, what were take-home exams ever testing? And what does it say about higher education that we’ve spent years grading work that may have been largely AI-generated without knowing it?
Every grade given on a take-home exam in the last two years is a data point in an experiment we didn’t know we were running.
Some students were honest. Some weren’t. We couldn’t tell the difference. The transcripts don’t distinguish between earned knowledge and generated knowledge. And now we’re discovering that the gap between the two is enormous.
This isn’t just about Brown. It’s about every institution that kept handing out take-home assignments after November 2022, pretending that nothing had changed. It’s about a profession that responded to the most significant disruption to assessment in centuries with a shrug and a plagiarism checker.
The professors who moved to in-person exams early aren’t heroes. They’re just people who read the room. The ones who didn’t aren’t villains. They’re just people who trusted a system that was already broken.
The real lesson from Brown isn’t that students cheat. It’s that educators forgot to ask whether their tests were testing anything at all.
So here’s the choice. Universities can keep performing the ritual of take-home assessments, adding AI-detection software and honor code reminders like band-aids on a wound that requires stitches. Or they can admit what the Brown professor discovered the hard way: the easiest way to stop AI cheating is to stop creating the conditions where it’s possible.
In-person exams. Oral defenses. Live presentations. Project-based assessments with checkpoints. None of these are new ideas. They’re just the ideas we abandoned because take-home exams were easier to administer, easier to grade, and easier to scale.
We optimized for convenience. We got exactly what we optimized for.
The path of least resistance runs in both directions. Students took it with AI. Administrators took it with take-home exams. Nobody has clean hands here.
The grade distribution at Brown isn’t a scandal. It’s a reckoning. And every school that hasn’t done the same experiment yet is just delaying the inevitable.
The question isn’t whether your students are cheating. The question is whether you’ve built a system where cheating is the rational choice.
If the answer is yes—and for most of higher education, it is—then the problem isn’t the students. It’s you.
FAQ
Q: Isn't this just blaming educators for students' lack of integrity?
A: No—it's recognizing that system design beats moral instruction every time. You can lecture about integrity for four years, or you can remove the opportunity to cheat in four seconds. One of those actually works.
Q: What's the practical implication for universities right now?
A: Move high-stakes assessments in-person immediately. Use take-home work for practice and feedback, not for grades. If an exam can be aced by AI without the student learning anything, it was never measuring learning in the first place.
Q: Isn't in-person testing outdated in a remote-first, AI-augmented world?
A: If your assessment can't survive contact with a classroom, it wasn't assessing anything meaningful to begin with. The problem isn't the format—it's the fantasy that convenience and integrity can coexist without structural guardrails.