The ChatGPT Cheating Crisis Was Inevitable

Once again I find myself writing about the latest AI outrage cycle, this one sparked by an NYMag article about the rampant abuse of ChatGPT in higher education. It should surprise no one that kids are using AI to cheat constantly and we are about to see a generation of functionally illiterate college graduates hit the workforce. Seems bad!

I feel obligated to point out that the article is a more than a little sensationalist and references a study claiming that 90% of students are using ChatGPT that isn't particularly rigorous. Its narrative also nicely cherry-picks a truly despicable guy to frame its story, "undetectable" cheating app Cluely's founder Roy Lee, who was kicked out of Columbia for creating the tool. But the crisis it's reporting on is a very real one. I think the story fails, though, in unpacking the broader societal failures behind it.

There's a clear and obvious failure in the rollout of these tools here. AI watermarking is already possible and ought to be a legal requirement for any consumer LLM application. If AI ethics or safety were in any way serious fields this wouldn't be a problem at all - instead of unreliable and scammy "AI detector" apps we'd have a built-in, reliable way of detecting AI-generated text. Obviously this hasn't happened because cheating and scamming are among the most profitable uses of LLMs but this would be an extraordinarily easy legislative win if we didn't live in a failed state.

There is a bigger structural failure (or, well, giant rats' nest of structural failures) at play here, though. The thing about a lot of consumer applications of AI is that, while I find them reprehensible, the tools themselves just act as an accelerant for a bigger, more insidious problem. Generally, the problem is the incentive structures behind its use - forcing a tool that isn't ready for prime-time onto the general public, largely for the sake of making number go up, was always going to end poorly. This is especially true in this case, where AI is rapidly putting the lie to everything that's wrong with our university system. The combination of an increasingly precarious professorial workforce, instrumentalization of universities as investment vehicles, and the emphasis on grades over learning have created an environment where cheating is strongly incentivized and professors aren't given the tools, power, or time necessary to combat it.

I've been an adjunct professor before. I really enjoyed the teaching part 0f it, but if you care at all about your work, chances are you're barely making a livable wage. Adjunct professors are only paid for classroom hours. When I was teaching, I was paid $95 an hour, which sounds great, but came out to $285 per week, per class. During this time I was writing my own syllabi, assignments, office hours, lesson materials, and problem sets. Probably if you are less insane than me you would not be doing quite so much work - for a new class, I put in close to 30 hours a week, which put me at less than $10 an hour - but I think even for a person who takes a more reasonable approach to their classes, you're probably putting in at least 8 hours a week for an existing class and 12 hours for a new one, counting office hours, grading time, administrative work, and so on.

Adjuncts who actually want to make a living are forced to either work themselves to the bone or just not give a shit about their classes. And New York has some of the highest adjunct pay rates! So many professors don't have the time or space for the extra overhead of trying to determine whether student work is AI-generated. Add the unreliability of AI detection apps and the extremely mixed messaging university administrators are giving on AI usage and many of them are in a position where they don't feel comfortable failing AI-generated work even if they're sure that's what it is.

I care a lot about education. Hell, the main reason I quit adjuncting was that I felt it didn't give me the resources I needed to do my students justice. But my belief in the importance of good education largely comes from nontraditional environments. Until sixth grade I went to a Montessori school: we had no grades, and half of the school day was dedicated to pursuing the types of study we were naturally interested in. Montessori instilled and encouraged a deep sense of curiosity that has accompanied me ever since, and I'm deeply grateful for that.

For undergraduate, I attended Reed College. Reed is a very unique place: it keeps a 9:1 student-faculty ratio, and the majority of non-lab classes are discussion-based. It rarely ranks highly in the US News and World Report college rankings, partly because it refuses to participate, and partly because its alumni earnings are comparatively low for an expensive liberal arts school. Despite this, it consistently ranks as one of the hardest schools in the US, has the fourth-highest percentage of students who go on to earn a PhD, and has the second-most Rhodes scholars of any liberal arts school. Most non-introductory courses are taught at a graduate level. Reed prides itself on being about the "life of the mind", and that was certainly true of my experience there. I also rarely checked my grades, because Reed discourages doing so - there's zero grade inflation, and you are generally taught to emphasize the process of learning over the results. As a philosophy major, I think I averaged 4-500 pages of reading a week. Reed is still probably the hardest I have ever worked in my life. While the pressure-cooker environment was horrific for my mental health, and I have mixed feelings about its culture, ten years on I am mostly grateful for the experience. It made me smarter, more thoughtful, more critical, and more curious. It also taught me to read incredibly quickly, which is a generally useful skill to have in life.

Most people don't have educational experiences like these ones, though. Having attended a West Virginia public high school and NYU for graduate school, I have a lot of sympathy for the kids using ChatGPT to do their work for them. When I was in high school I would have done the same! I'm very glad I ended up going to Reed in the long run, but I also went there because I didn't have a good enough GPA to get into an Ivy. The incentive in high school was always to up my GPA and up my chances of getting into a good school. Coming from Montessori I think I cared more about learning than the average kid but I absolutely cheated my way through the classes I didn't care about. All of them were just hoops to jump through, not opportunities to learn, and most of my teachers didn't do anything to disabuse me of that notion.

The NYU program I chose was a blend of both models. I went to ITP, a program that also has no grades and is largely project-based. It encouraged people to actually learn - not being a career factory was a big part of why I went there - but a solid 50% of my cohort had had the standard education model beaten into them at that point, and treated the experience like all the ones they'd had up to that point. I was astonished at the number of people who never spoke up in class, didn't do the reading, and generally put in the bare minimum to pass. Hell, some of the more wealthy kids there even paid other students to do their homework. Since the program is a cash cow for NYU, there were no real disincentives for not caring about it, and while I'm glad I went that aspect of it lessened the classroom experience for everyone who was actually there to learn. It didn't really make sense to me that you would go to a nontraditional grad school if you weren't excited about learning, but plenty of people did it either for the sake of a visa or just to get a piece of paper.

The popular model of education is not about learning. It's about grades. This is massively disadvantageous to people who have different learning styles, different abilities, or just can't afford things like tutors and SAT prep. When I was teaching, I always tried to grade based on effort and progress, and I think learning environments that incentivize those things (you know, the things that are required in actual learning) are the only ones that can truly be considered successful. A student who comes in with zero knowledge of a topic and comes away with a decent understanding is more impressive to me than a student who comes in already knowledgeable and coasts on it. But right now our system is designed to reward the latter over the former. It breeds complacency in the already-gifted and punishes people who aren't. This is no way to teach or learn.

I'd wager that the majority of educations people receive in the US are closer to public high school than to Reed or Montessori. The standardized testing mindset is endemic to the population at this point, and college is just a piece of paper you have to get (or, if you're Cluely's founder, a networking opportunity) instead of an opportunity to actually learn. Unless education is actually about learning - learning to love learning, learning to be curious, learning to think, learning for its own sake - of course people will treat it as a chore that can be automated. Why wouldn't they? We haven't made any effort to teach them otherwise.