The birth of AI has turbocharged student cheating. No, that doesn’t say it. It is a massive dose of anabolic steroids straight to the heart. No… that’s still underselling it. AI is more like fluorine. Fluorine is the most reactive element, hoovering up electrons from almost everything it gets near. If you add two oxygen atoms to two fluorine atoms, you get dioxygen difluoride. That stuff detonates things at -180C. It sets ice on fire and blows up if you add water vapor. It’s too aggressive to be used as rocket fuel, even though rockets are propelled by explosions. Chemists are terrified of it.

AI is a tanker truck full of dioxygen difluoride firehosed straight onto the academic quad.1

Within the professoriate, there have been two main reactions:

1. Denial and wishful thinking.

2. Despair, panic, and existential dread.

In the denial camp are those stubbornly convinced that they now and forever can tell the difference between the AI-made and the human-made.

It’s already not looking too good for that assumption. Take 2D art for example. In one experiment, 11,000 people examined 50 pieces of art and tried to tell which were AI and which were by human artists. They did slightly better than chance. I’m sure that proves nothing. How about poetry? A recent article found that AI-generated poetry is indistinguishable from human-written poetry and is rated more favorably. Of course, that research was about non-expert poetry readers. No doubt the experts will fare better, right? Right?

Some faculty think their super-specific writing prompts and questions will winnow out AI; others think they have magic insight into the AI “voice.” Some believe they can tell the difference between AI hallucinations and plain old students bullshitting. A few place all their faith in AI-detection software, confident the robot police will capture the robot criminals. Really the best test is whether the prose shows excellent command of spelling, grammar, structure, and those mysterious syntactical sprinkles known as apostrophes. If it does, it’s probably AI.

I think those in the denialist camp are living in a fool’s paradise. Whatever success they imagine they have today in spotting computer-generated work will disappear with the next generation of AI, or the one after that. If anything, even the most pessimistic estimates of student dishonesty aren’t pessimistic enough.

Many faculty are abandoning out-of-class essay writing altogether, simply because everything they get back is AI-written. Instead they are moving to more in-class essay exams. Not even that strategy is foolproof, though. One of my colleagues caught a student using AI to cheat in class on an essay test. The student was entering the exam questions into her smartwatch, getting AI to answer them, and then simply writing down the results off her watch.

I know of another professor who, after repeatedly telling his students that he did not offer any extra credit in the course, received an email begging for exactly that. Typical, right? Yes, except the following line was at the end of the email after the student’s signature: “This version conveys urgency and accountability while still being polite and respectful. It acknowledges the timing and emphasizes your commitment to improvement.” That’s right, students are even using AI to grade-grub.

Lots of faculty, especially those still feeling sympathy with Defund the Police, resent being classroom cops. They don’t want to spend all their time suspicious of students, trying to root out the cheaters, and delivering punishment. I get it. There’s a lot about my job I really don’t want to do either, and didn’t exactly sign on for. Filling out bureaucratic paperwork is near the top of the list. Grading too, although I suppose I did sign on for that.

The problem is that we have to give grades, and grades are a kind of testimony: we are testifying that this student has mastered the material to X degree. Or at least, this student has successfully done X amount of work. If there’s a pretty good chance that student cheated and we didn’t try to figure it out, then we don’t know that our testimony is true. We’re not even justified in making the assertion. That is epistemically bad.2

So why? Why the tsunami of academic dishonesty? First we need to ask more broadly:

Why do students cheat?

I can think of five answers: panic, hyperbolic discounting, akrasia, ease, and The Experience Machine. Let’s take a look.

Panic. There’s lots of reasons for panic—a generic panic attack, forgetting that the test is today, or just extreme insecurity about your own abilities. So you look over at your neighbor’s work on the exam. Or you scribble the Periodic Table on your forearm in advance. Or you lightly disguise your plagiarism of that encyclopedia article. Maybe you’re compelled to take a class that you have zero aptitude for, and there’s no way you could pass under your own power. I know someone with excellent writing skills and a master’s degree who cheated her way to a D in mandatory high school chemistry. Chemistry was all Greek to her. Panic cheating is going to be uncommon, just because panic itself is not routine.

Hyperbolic discounting. This is when your decisions are biased to the present, and you forgo larger payoffs later in order to gain some small benefit now. It’s spending all your disposable income instead of saving for retirement, or eating cookies instead of salads. With hyperbolic discounting you view the desires and preferences of your future self as less valuable than the desires and preferences of your present self. Your future self will be pleased that your past self saved for retirement, but your present self doesn’t care about what some future geezer wants when that boat is on sale now.

So students procrastinate, valuing their current time as more important than their future time. Then assignments pile up, there’s no legit way out, and they cheat.

Akrasia. Weakness of the will is not the same as hyperbolic discounting. Akrasia is why New Year’s resolutions are so hard to keep. You’re serious about getting in shape, cutting back on the booze, reading more literature, and finally painting the garage, but then, well, things come up and those resolutions just aren’t very fun. Unfortunately there’s no necessary connection between what you tell yourself on January 1 and what you do on March 1.

A student might have every intention of knuckling down and hitting the books this weekend, but when the weekend comes and all her friends are heading over to the frat house, it turns out she just can’t stick with her original plan. In Being and Nothingness, Sartre says this is the problem with freedom: we’re always free to reject our previous commitments and promises to ourselves. Past vows of action mean nothing, as much as we wish they did. We have to remake our projects and intentions at every moment.

Ease. Back in ye olden times, cheating on a term paper meant going to a paper mill, trying to recycle one your friend wrote a couple of years back, or hiring someone to write it for you bespoke. Those things took effort and possibly money. AI has made cheating low effort and low cost. It’s tempting to think that the ease of cheating explains the whole shebang.

It may be part of the explanation, but it’s not the whole thing. When other vices were liberalized, that didn’t lead immediately to crisis. After the end of Prohibition in the US, the number of alcoholics went up, but not everyone became a drunk. Same with the legalization of marijuana—there may be more pot smokers, but not everyone turned into Snoop Dogg.3 Same with gambling. Casual drinking, occasional toking, and modest gambling are far more the norm than excess. You’d expect academic dishonesty to be more of the same, that making it super easy pushes a few more people over the line but doesn’t lead to a stampede.

We should expect cheating due to panic, hyperbolic discounting, akrasia, and ease to be stable. The percentage of students falling prey to each of these causes is, if not a constant exactly, probably fluctuates pretty closely around some set point. There’s no reason to think that AI somehow increases the number of students panicking, procrastinating, with weak wills, or seeking an easy way out.

The best explanation has to do with the experience machine.

The Experience Machine

In Anarchy, State, and Utopia, Robert Nozick4 writes, “Suppose there were an experience machine that would give you any experience you desired. Superduper neuropsychologists could stimulate your brain so that you would think and feel you were writing a great novel, or making a friend, or reading an interesting book. All the time you would be floating in the tank, with electrodes attached to your brain.” Nozick asks whether you would plug into such a machine for life, with all your experiences no more than preprogrammed illusions. Hedonists would surely reply, “heck yes, the sooner the better!”

Nozick was anti-hedonism. He argued that as appealing as the experience machine initially seems, we would not want to spend our lives attached to it. There are two key reasons why: first, we want to do things with our lives, to accomplish things. It is not merely the artificial feeling of accomplishment that we crave, but the actual fact of it, the truth that we have done something with our time in the world. It is better to write the Great American Novel than just have the false conviction that you have written it, it is better to be loved and have real friendships than buy the machine’s lie otherwise.

Second, the experience machine not only strips us of a connection to real accomplishment, but prevents us from becoming authentic persons. Are you courageous, kind, intelligent, witty, loving, generous? Plugged into the experience machine of phony pleasures, you aren’t anything. Instead, you are simply a passive blob, a receptacle of pleasurable (but fraudulent) sensations. If you prefer to do and have done things and to become a real and genuine person, then you should reject the allure of the experience machine.

That’s Nozick’s argument. Nozick is wrong. AI is the birth of the experience machine, and students prefer it. That’s why they cheat.

Writing is not a tool to express our thoughts—the act of writing is the act of thinking. Sitting down and making words, pondering not merely le mot juste, but how the argument works, what metaphors are appropriate, what really is the conclusion you want to reach, that is the creation of your own point of view. A sophomore telling ChatGPT or Gemini or Claude to write a critique of euthanasia is not developing her own ideas about euthanasia. But it does give the illusion.

Finally, having cheated all the way through college, letting AI do the work, students can have the feeling of accomplishment walking across the stage at graduation, pretending to be an educated person with skills and knowledge that the machines actually have. Pretending to have earned a degree. If Nozick were right then AI would not lead to an explosion of cheating, because students would want the knowledge and understanding that college aims to provide. But in fact many just want the credential. They are hedonists abjuring the development of the self and the forging of their own souls.

So why is AI destroying academic integrity? Because hedonism is true and we have built the experience machine.