Ninety-Six To Forty-Eight

On March 5, an advanced mathematical economics class at Brown University sat for a take-home midterm.¹ When Professor Roberto Serrano graded the eighty-six exams that came back, forty of them had earned perfect scores.² The class average was ninety-six, an unusual number for a course whose average had ranged between roughly sixty-five and eighty in prior years.¹ Serrano ran the questions through ChatGPT and found the tool had produced an oddly convoluted argument for a problem whose real answer was short. That same convoluted argument was appearing, in one form or another, across dozens of submissions.²

He warned the class. He told them the final would be proctored, in person, and would count for half the grade. He also told them that if the final's pattern did not confirm the midterm's, only the final would count. The final was held in late June. On the same students, in the same course, on the same material, the class average dropped from ninety-six to forty-eight.³ Late last week, the story ran under headlines calling it the largest AI-enabled cheating case in Ivy League history.⁴

What The Second Room Proved

The reason the story landed heavily this past week is not, at bottom, the fraud. Some version of the fraud has been reported every month for two years. It is that the proof of it required a physical room. A locked door, a proctor, a class period, an exam the professor could watch the student taking. That is what let him watch the ninety-six become forty-eight. The mathematics were the same. The people were the same. The only variable that changed was whether the room could be seen into.

Anyone reading the story understands quickly why he had to do it. He had no other record. The take-home midterm arrived from the students as a finished document. What happened in the hours between the download and the upload was invisible to him. He was left with two options, and both cost something. Trust the visible artifact, which he now had reason to doubt. Or replicate the day in a room he could watch, and count only the room he could watch. He picked the second, and the class taught him what he already suspected.

The Room That Used To Leave Marks

Every school in the country reading the story on Wednesday morning is facing a version of the same choice. Roll the take-homes back into proctored rooms, and swallow whatever else the take-home was for. Or leave them alone, and try to catch what happened after the fact. Neither is a good option. Both are downstream of a room that used to be visible and quietly is not anymore.

The take-home was never only about the completion of an artifact. In the years the average was between sixty-five and eighty, it was also a record. Of the student who worked in one long sitting and finished tired. Of the student who worked for three hours, gave up, came back, and finished at midnight. Of the student who tried a wrong approach for an hour and switched. That record used to leave marks. Erasures on paper. Timestamps in a document. Draft after draft in a submission portal. A professor who cared could read those marks and know something about the day. Now, more and more, the artifact arrives as if it was born fully formed. There is no way, from the finished document alone, to tell whether the hours were spent with the material or with a machine.

What The Ninety-Six Was Actually Made Of

The ninety-six was not, in a strict sense, a lie about the answer. In a course whose questions are checkable, the answer was probably right. The lie was about the process that produced the answer. And it is the process that a course in advanced mathematical economics is really trying to grow. The finding of a convoluted argument for a straightforward problem, repeated across dozens of exams, is a very particular kind of tell.⁵ It is the trace of a class where the tool did the reasoning and the students copied it, faithfully, without noticing that the reasoning was worse than the reasoning they would have done themselves in a room he could see. The forty-eight, taken in that room, is what those same students could do with the material they had claimed the ninety-six with. That gap is not a discipline problem. It is a learning problem. Half a semester of a course that was supposed to teach a way of thinking left no trace of the thinking having happened.

A Different Room To Build

Serrano's response is understandable. In the short run, if the record cannot be seen, the professor moves work into a room where the process can be. But that is a defensive move, and everyone in the room feels it. Take-homes were not a lax convenience. They were the way schools taught the shape of doing hard work under one's own supervision. They were the way a student practiced the version of thinking that the rest of adult life would require of her. Losing them, in every course, in every school that reads about Brown this month and reacts, is a real loss. It is the closing of the room a school was designed around.

The more interesting question is not whether the take-home survives. It is whether the record does. Whether a course can be designed so the process of a take-home leaves the same kind of marks that pencil and paper used to leave. Whether the pauses, the revisions, the moment the student asked the tool a question and the moment she pushed back on its answer, the false starts and the rewrites and the second thoughts, can be seen the way the exam room lets a professor see a student's hand hover over a question. The instruments to build that room exist. What is often missing is the decision that the record, and not the artifact, is what the course is for.

The Brown story, read this way, is not a warning about students. It is a nudge to schools. The ninety-six was invisible until the room was moved. The next room does not have to be a proctored one. It has to be a legible one.

If the take-home stopped counting on your campus tomorrow, what would you have lost that you could not build back in a different room?