Like many other academics, I’m a bit romantic about the idea of learning at the same time that I nurse a cynicism about educational institutions. The question of cheating, to my mind, has always been a marginal one — if students cheat, that’s their choice, just as it’s my choice to detect it and fail them, ça va? Of course, that was before cheating scaled up, before my employer rolled out our university-branded edition of “ChatGPT for Education.” My free choice as a teacher means little in the face of institutional endorsement and ever-blurrier lines between cheating and the American virtue of resourcefulness. 

Last fall, I assigned my Early American History students Incidents in the Life of a Slave Girl by Harriet Jacobs, a text that explores not only the violence and depredations of chattel slavery, but also what it meant to emancipate oneself and ultimately, purchase one’s own freedom. Jacobs considers the complexity of securing liberty while also providing financial capital that will sustain the slaving system. One of my students — a struggling freshman — came to my office hours and asked why he had received such a low score on his reading response. I reread it quickly and remembered: at the end of the book, my student wrote, Jacobs is sold back into slavery, showing the engulfing power of the slave system.

This was, I said to him, simply incorrect — it wasn’t clear to me, on the basis of his response, that he had actually read the book. He mumbled, “Oh, okay,” and sloped off. Naively, I had expected a little more shame. 

Since then, more of my students have submitted stilted, inaccurate responses to their assigned readings — enslaved people are recaptured, Native peoples lose their land, America triumphs — even as the texts themselves suggest complexity, resistance, and marginal victories against the manifest destiny of American expansion and exploitation. In a moment of curiosity, I loaded a prompt into ChatGPT and discovered these were the themes it suggested when a user asked about the books I’d assigned. If I were a certain kind of reader, I might try to infer something about ChatGPT’s attitude about American triumphalism, but fundamentally, I don’t see A.I. models as thinking or communicating anything at all: one might as well psychologize a refrigerator. 

When I was in college, one of my favorite essays was “Confessions of an American Pumpkin Eater,” by Lauren O’Neil. “When students help each other cheat, they may not exactly be promoting learning, but they are choosing comradeship over competition,” O’Neil writes. For cheaters of old — including myself, once upon a time — cheating meant talking to each other, taking and sharing notes, and discussing what was on the test. Now, like so much else, cheating has become a solitary activity, sans effort, sans camaraderie. I can’t say what will become of my students. As with so much else, ChatGPT and its ilk have de-skilled the once-venerated art of cheating.