Being a student should be about more than completing assignments.
Much ink has been spilled over how to respond to artificial intelligence use in education. Harvard College Dean David J. Deming frames cheating with AI as a “temptation”; students know it’s wrong, but we just can’t help ourselves — so assignments have to be restructured to make cheating impossible.
We think this framing is mistaken. Instead, we should consider AI misuse to stem from the failure to cultivate curiosity in students.
As students, part of our engagement with our education is externally motivated — and reasonably so. Competitive admissions select for students who treat knowledge as capital, something to acquire and exchange for other goods: we pursue grades, summer programs, and extracurriculars to get into selective universities. Once we get to Harvard, these pressures return as the pressures to compete for post-grad opportunities.
Many of us are also motivated, at least in part, by curiosity; the burning drive for understanding. Even among the more extrinsically motivated, cultivating curiosity is well worth the while. Curiosity has been shown to increase creativity, enhance memory, foster a sense of meaning in life, and promote engagement with one’s work. When coupled with effort, it even matches intelligence as a predictor of academic performance.
Teach a man, and he learns for a day. Teach a man to love learning, and he learns for a lifetime.
Suppose we make cultivating curiosity a primary goal of higher education. Then, even in fields one chooses primarily for employment — say, perhaps no longer, computer science — the task of the university is to bestow the love of learning computer science on the student. The development of specific skills and knowledge employed in computer science follows incidentally.
We might be tempted to think that assignments and grades facilitate this curiosity. Take the example of writing an essay. The traditional university could hope that in the process of writing, students become interested in what they are writing about. The assignment should eventually give the student a reason to pursue understanding for its own sake, beyond the instrumental goods like grades.
However, there were signs this method was flawed before large language models — Harvard students have long prioritized external rewards. We search for “gems,” skip readings, bring our laptops to class, and apply for jobs in sections.
LLMs then fully overturn the calculus. The technology lowers the cost of outsourced cheating in terms of time, risk, money, and effort.
When we see that almost 40 percent of student respondents to one survey will admit to “regularly using AI for coursework in ways their instructor may view as inappropriate or against class policy,” we are confronted with just how instrumentally rational we really are. (This is not for lack of clarity, either, since two-thirds of surveyed students agreed they “always” or “often” comprehend their instructors’ policies.)
Completing an A-level essay in prehistoric times required a student either to engage deeply with the subject, or cheat. Pre-AI, cheating was expensive and risky. One had to plagiarize others’ work, or pay someone to do it for you. This means that even the most well-resourced and instrumentally rational students are incentivized to write the essay as normal.
AI doesn’t create an incentive to avoid intellectual work. It reduces the cost of acting on that incentive. Or in other words, LLM use isn’t the siren’s song; cheating is. AI only loosens our bounds.
Students attuned to the goods internal to the pursuit of learning should have little reason to misuse AI: it would cost them their prized learning. Conversely, using AI to bypass your work makes sense when you’ve been taught to value your grade instrumentally above all else.
So while we laud efforts to redirect students back into intellectual engagement, we should also recognize that AI misuse is a mere symptom. The malady? Students have not been made curious. Changes such as implementing oral exams and presentations might realign incentives, but we have no reason to think they will be more effective at cultivating curiosity than pre-AI policies.
In conversations about adapting course design to advances in AI, it’s time we stop thinking solely about how to make misuse less feasible, and start thinking about what it would take to seed students’ curiosity.
We are uncertain what an alternative to this incentive system should be, but we can offer a criterion for judging one: Students should be courted to fall in love with understanding.
Zakiriya H. Gladney ’27, a Crimson Editorial Chair, is a double concentrator in Social Studies and Statistics in Dunster House. Louis-Francois P. Belanger ’28, a Crimson Editorial editor, is a Philosophy concentrator in Dunster House.
Want to keep up with breaking news?Subscribe to our email newsletter
Have a tip for The Crimson?Share it confidentially
