Colleges Created the AI Cheating Problem

Like many who’ve spent any degree of time instructing undergraduates, I’m all for complaining about the youth, and I have accumulated substantial material to work with. In many ways, the discussion around AI is no exception. I detest simply being told, “AI can just do this for me,” I resent reading obviously AI-generated writing, and I especially abhor being asked to justify my expertise against someone’s LLM du jour. 

However, despite my curmudgeonly disposition, much of the recent discourse around AI cheating gives me pause. Yes, it’s definitely a problem: college students are using AI to cheat, they’re having LLMs do their homework, and their test scores reveal that they’re not learning as effectively as they were before. More than three years after teaching my last undergraduate course, I find the idea of dealing with any of this terrifying and outrageous, and I understand the harsh responses I see from my friends and colleagues. However, amidst such responses—from calls to ban AI to demands to universalize AI fluency across campus—we seem to have forgotten some important things. 

Yes, AI is here, and it is not going anywhere. Yes, its disruptive power has implications for higher education. But AI is, ultimately, a tool, and it can facilitate learning just as easily as impede it. That said, none of the skills required to use AI responsibly are innate, nor does simply attending college guarantee that students will develop them. Thus, if we’re going to complain about students’ behavior, we ought to ask ourselves why they’re behaving this way in the first place.

Nearly a century ago, social psychologist Kurt Lewin proposed a mathematical equation to describe human behavior: B = f(P,E). In plain English, behavior is a function of the person and their environment. This equation has become one of the most important foundations of behavioral science. Higher education is currently failing both sides of this equation by creating and maintaining environments that encourage precisely the behaviors we’re complaining about, while simultaneously failing to develop the personal characteristics that would help students behave responsibly within those environments.

There does certainly appear to have been a cultural shift among today’s undergrads, with honesty and integrity being devalued in favor of accomplishment. This culture is problematic, and students must be held accountable for misusing AI in academic settings. 

However, cultures don’t spontaneously generate: they evolve as adaptive responses to particular environments. This means that, in some way, cheating—including the misuse of AI—has proven uniquely advantageous to these students. This is because higher education has allowed, and even actively created, an environment in which this is true.

Neither students nor faculty are incentivized, in the current state of higher education, to place students’ education or development above mere accomplishment. 

For students, grades matter enormously, determining scholarships, graduate school admissions, academic standing, and opportunities after graduation. Prestigious credentials matter, and students are understandably concerned about obtaining them.

Yet when assigning such credentials, much of what is assessed is the finished product: essays, take-home exams, or completed assignments. When students can earn good grades without demonstrating that they’ve actually mastered the material, we’ve created an environment wherein the appearance of accomplishment can be just as rewarding as accomplishment itself. Add AI, and the incentives become painfully obvious.

Students have competing priorities, and they’re not going to suddenly work their absolute hardest in every single class. If technology allows them to obtain the same grade with substantially less effort, some students are inevitably going to take advantage of that opportunity: that’s just how humans work. This doesn’t make cheating acceptable—but it does make it predictable.

Now consider the environments that instructors operate in.

Professors are often overworked, underpaid, and undersupported, with the majority lacking tenure or even a place on the tenure track. However, such faculty are still expected to manage substantial course loads, often while facing considerable pressure to maintain favorable student evaluations. Thus, while it can be tempting to just tell professors to “teach better” in order to AI-proof their classes, such recommendations ring hollow. 

Faculty who demand more of their students in today’s universities can face serious professional repercussions, particularly when student evaluations influence decisions about tenure, promotion, or reappointment. The result is less effective teaching, more grade inflation, and numerous other educationally undesirable outcomes. 

Even in some ideal world where every assignment is AI-proof, the higher education environment would still have these problems. We would still have students incentivized to prioritize grades over learning and professors whose professional futures can depend, at least in part, on keeping those students happy.

Higher education has also dropped the ball when it comes to the other half of Lewin’s equation: the person.

The fact is that even if we somehow managed to AI-proof every classroom tomorrow, that wouldn’t be enough. AI is evolving at an astronomical pace, and even if an AI-proof environment could be constructed, who knows how long we could maintain it? Besides, cheating is a hallowed academic tradition. It didn’t begin with AI, and it certainly won’t end with AI.

So while we absolutely must address academic environments that make cheating easy or desirable, we must also develop people capable of exercising good judgment in difficult situations, like the one brought about by the advent and wide accessibility of AI.

Now yes, research has identified numerous traits linked to responsible, ethical use of AI: humility, self-efficacy, self-control, and critical thinking, to name but a few. However, it’s absurdly naive to assume that college students naturally possess these characteristics at levels sufficient to ensure responsible AI use. Such characteristics must be intentionally cultivated and reinforced—and yet higher education seems unwilling to acknowledge this.

For instance, at far too many institutions, “ethical development” seems to boil down to vague appeals to an honor code, forcing students to read Aristotle, Kant, and Aquinas, and, at least at religious schools, some variation of reminding students that Jesus is watching.

Now, there’s nothing inherently wrong with any of that. However, knowing Aristotelian ethics doesn’t make you an ethical person, which means ethical development has to be about more than just exposing students to information as if ethical development were a mere box to check. If we want to develop people who can responsibly use AI regardless of whether or not we can keep up with its vast potential for either avoiding necessary work or outright cheating, we have to be intentional about it. 

The good news is that higher education has an extraordinary opportunity for this process. Universities are communities of learning. Students spend years learning, working, and living alongside one another and their faculty. That gives us an enormous opportunity to cultivate the characteristics, habits, and norms necessary for responsible, ethical behavior. However, taking advantage of that opportunity will require a lot more intentionality than higher education is currently demonstrating.

As a social scientist, I’m reluctant to offer sweeping policy prescriptions. Universities differ enormously, and what works at one institution, or even in one department, might not work in another. Nevertheless, I think there are some concrete changes that addressing this problem is going to require.

First, higher education must get serious about changing the environments in which students operate. That means establishing clear, meaningful institutional standards regarding acceptable and unacceptable AI use and then actually enforcing them. It also means redesigning educational assessments so that students must demonstrate mastery of the material, focusing less on information regurgitation and more on whether students can understand, apply, and critically evaluate what they’ve learned. And no, simply relying on AI detectors isn’t going to cut it, as these technologies remain unreliable, particularly when used to make high-stakes decisions about students. 

Second, we’re going to have to reconsider how we evaluate and support faculty.

This means that student evaluations will probably need to go, at least as a basis for high-stakes personnel decisions. That doesn’t mean abandoning our assessment of teaching effectiveness—quite the opposite. We need to go beyond asking students whether they enjoyed a class or thought they learned something and actually determine whether effective teaching and learning occurred. This might require more resources, but will also allow for more effective teaching, less easily gamed assignments, and better faculty development. 

Finally, we’re going to have to invest in the intentional development of faculty and students alike, both around and beyond AI. 

This needn’t (mustn’t) be merely a vague aspiration. We know such development is possible, and we know what it looks like. It involves having students work through realistic ethical dilemmas, consider competing perspectives, practice strategies for making difficult decisions, and receive feedback on their judgment. Critically, these experiences can be incorporated into coursework, meaning that ethical development needn’t be relegated to occasional workshops or seminars. And even more importantly, we must assess whether these experiences actually work by measuring changes in students’ ethical judgment, rather than simply asking whether they enjoyed the experience. 

AI isn’t going away, and neither is the temptation to cheat. We can spend the next decade playing technological whack-a-mole with students, implementing ever more elaborate safeguards while students keep finding ways around them, or we can actually fulfill higher education’s mission of creating environments that intentionally develop people capable of exercising ethical judgment even when shortcuts are easily available. 

None of this excuses students who cheat. But universities must also take responsibility for the environments they create and the people they’re supposed to develop.

Lewin gave us the equation nearly a century ago. It’s about time higher education started paying attention to both sides of it.

Aaron Pomerantz is the Assistant Director of Research and Evaluation at Rice University’s Doerr Institute for New Leaders.