Cheating Traps

The recent IHE piece about the professor who set traps to catch cheating students—by posting an exam with incorrect answers on a commonly used site—has been rattling around in my head. It’s a new twist on a long-standing issue.

I see a few separate issues at hand.

The first is whether it’s OK, in the age of AI, to insist on original work at all. Call me old-school, but I’m still a solid yes on that one. The point of assignments is to push students to learn by doing. If they aren’t the ones doing the assignment, they aren’t learning. But the damage from cheating isn’t just to the students who shortchange themselves. In competitive settings—whether individual classes graded on curves, or subsequent admissions to law school or med school—grades are essentially positional. If Amber cheats her way to a better grade than Billie, both are harmed. Cutting slack for Amber amounts to condoning harm to Billie. One could argue that grades shouldn’t be competitive at all, but a change like that can only work at scale. In the world in which we actually live and work, ignoring the positional nature of grades and shrugging at cheating effectively rewards those who cheat and punishes those who don’t.

The second is around websites that publish exams without professors’ permission. Doing so is an obvious copyright violation, but they operate at such scale—and there are so many of them—that the game of legalistic whack-a-mole feels rearguard at best. Why these sites are legal is entirely beyond me, but they’re everywhere. Given the manifest ineffectiveness of trying to cease-and-desist-letter them into oblivion, I understand the appeal of salting the earth instead.

But that strategy, which is what the professor did in this case, strikes me as fraught. It’s tricky for the same reason that these sites are hard to stop: information leaks. If the poison were confined to the cheating site, I’d be much less bothered. But it could easily make its way from that site to others—especially social media and/or other AI sites—at which point a well-meaning student could mistake it for truth.

My sense of student cheating is that it’s frequently opportunistic and/or defensive. Take the case of Amber and Billie above. If Billie has reason to believe that the majority of the class is like Amber, then she faces a dilemma: either cheat, too, in order to keep up, or refuse on moral grounds and accept one of the worst grades in the class. It’s a bit like speeding on the highway. When everyone else is going at least 10 to 15 miles per hour above the speed limit, doing the speed limit amounts to making yourself a road hazard. When the official reality and the observed reality are sufficiently far apart, the official reality loses much of its force. In that setting, Billie may perceive her own cheating as defensive. We can condemn her for that, but at base, it’s a system failure.

From a system perspective, I could imagine a few solutions, but only one of them has a realistic chance of working.

The professor could use moral suasion and appeal to students’ sense of civic virtue. The problem with that is that the only students on whom that would work are the ones who don’t need it.

An elaborate crackdown could work, sort of, for a while. But it would be costly, it would turn off students, it would establish a terrible learning environment, it would result in a host of false positives and, over time, through a sort of natural selection, it would reward the cleverest cheaters. I don’t see this as a serious answer.

Professors could flood the zone with bad information, thereby trapping those who look in the wrong place. But as the past several years have taught us, bad information escapes its initial boundaries. It’s also the exact opposite of what we want professors to do. They’re supposed to pursue truth—that’s the point of tenure, for example—rather than make up disinformation. I don’t see this strategy ending well.

The strategy likeliest to work, I think, is to restructure assignments and grading. That can entail a good bit of work, and in some cases, may require institution-level course redesigns. (Going with oral exams, for instance, would put limits on class sizes.) But as long as students are evaluated on few, high-stakes tests, the incentive to cheat will be powerful.

Wise and worldly readers, is there a better way?