The question is interesting. Students are going to be a mix. Sure, they want to learn. But they also want to party. Or even, just have good grades b/c good grades > bad grades.
Therefore, even in a benign environment, incentives and trades off guarantee some interest in short cuts.
But, in a cut throat environment, where better grades from a better school can alter your entire career arc... then yes of course cheating is absolutely a strategy some players (sorry, students) will consider.
And if said careers carry risks to the outside world (medicine, avionics engineers), it seems fair to me that companies relying on the credentials provided by the academic system not be deceived when a student says he's in the top 5% of his Harvard class...
I came here to write something similar. Obviously cheating doesn't matter as much from a faculty perspective, but from a student perspective, grades can make a big difference toward getting a good job and having a good life.
I'd argue that MOST of the time, medicine and engineering notwithstanding, the requirements of the job DON'T actually depend that much on the academic foundation. Even in medicine and engineering, there are many pre-req undergrad courses that have little bearing on the individual's ability to be successful at the job. So long as a student has some aptitude to learn and DO the job, they just need to GET the job with strong intellectual fitness signals from their academic records (e.g. GPA, school reputation etc.). The situation is even further exacerbated by courses graded on a force-ranked curve in this competition for high GPA, class rank, and post-grad opportunities.
So the real victims of LLM based cheating are not the deceived faculty, nor the unlearned cheaters themselves, but the honest students who want to compete fairly by studying the material and using their own brains. They are most at risk of doing the right thing but getting outcompeted for fitness signals. The faculty should make efforts to enforce fairness and root out cheaters for the benefit of these good students that they actually want to serve.
I am not without sympathy to this case, but two things:
One is that (from what I remember reading) most cheating isn't an A- student trying to jump up to an A+, but a D student trying to squeak out a C+. People turning in LLM-generated papers mostly aren't beating out honest students for competitive internships, they're just staying off academic probation.
Second, this is very much part of the tension I see between denouncing credentialism while also embracing it. If the reason why LLMs (or samizdat solution manuals) are a scourge that we must go to great lengths to fight is that it's essential to preserve the credential as a career signal, can we then turn around and denounce students for giving top priority to the credential over genuine academic interests?
I don’t really have a problem with credentialism. If universities were only for people interested in learning for learning’s sake, they’d have a couple of thousand students…
I love lots of subjects (economics, history, history of sciences, related humanities like sociology) but even there, without the carrot/stick of getting a career, I’m not sure how hard I would have pushed myself.
Maybe when the robots do all the work and we’re all effectively indolent aristocrats can universities return to being purely about eager to learn students…
I think students give top priority to the credential because they know that employer hiring and higher ed admissions practices rely on these credentials. So long as they continue to make decisions based on coarse but easy to rank signals like GPA and academic reputation, they sustain the incentive to win the competition by any means necessary.
In tech, I've seen firsthand candidates who pay to get ahold of the interview bank questions ahead of time. This becomes self evident when they start answering specific parts of the question that haven't been asked yet. Very recently a student was expelled from Columbia for building an LLM based program to help students cheat on live coding exercises. So while it may not be the majority of cheating, there does still appear to be willingness to cheat at the top.
We may also be talking about cheating with LLMs as more black and white than it actually is. In reality, I think many students probably do try to learn the material and then supplement that learning with LLM tools to make their work better, or to check that it's right. Sure it takes away from mastery and complete self-reliancy, but maybe they're just preparing for a world where this is kind of how we do everything. There may be an unfair analogy to calculators and word processors making mental math and handwriting obsolete.
LLMs are really good tools. We should be teaching students how to make best use of them, not trying to prevent their use. We should also teach how to get things done when the internet is not available and not pretending that isn't always the case.
"It’s probably impossible to force disinterested students into putting in the effort to actually learn (but then, it never was) ..."
Did you mean "always"?
Sorry to start off with a suggested edit. I really did enjoy this essay, and I think you make a solid point, your self-deprecating "inconclusive noodling" ending notwithstanding.
Never was possible, yeah. I think I probably edited "It's probably not possible" into "it's probably impossible" and forgot that there was a callback in the parenthetical that should have parallel construction. That's the kind of thing I do a lot.
“The ones who are interested in the opportunity to learn will still do the work, if they know that’s what it takes.” This is going to take a big sell job on our part, or at least bigger one than it used to be when I was only trying to challenge physics majors to not Google “Griffiths hw solutions,” at least not every week. I’m teaching premeds this fall, so presumably will be on the front lines of this issue…
I've had the advantage the last few years of teaching quantum off a colleague's lecture notes rather than a named book whose solution sets are Out There, but even when I used Griffiths, I was surprisingly successful at getting them to not just Google the solutions. Or if they did Google them, they were staggeringly bad at transcribing them...
Pre-meds may very well be a different deal, though. Engineers don't seem to be, interestingly.
The question is interesting. Students are going to be a mix. Sure, they want to learn. But they also want to party. Or even, just have good grades b/c good grades > bad grades.
Therefore, even in a benign environment, incentives and trades off guarantee some interest in short cuts.
But, in a cut throat environment, where better grades from a better school can alter your entire career arc... then yes of course cheating is absolutely a strategy some players (sorry, students) will consider.
And if said careers carry risks to the outside world (medicine, avionics engineers), it seems fair to me that companies relying on the credentials provided by the academic system not be deceived when a student says he's in the top 5% of his Harvard class...
I came here to write something similar. Obviously cheating doesn't matter as much from a faculty perspective, but from a student perspective, grades can make a big difference toward getting a good job and having a good life.
I'd argue that MOST of the time, medicine and engineering notwithstanding, the requirements of the job DON'T actually depend that much on the academic foundation. Even in medicine and engineering, there are many pre-req undergrad courses that have little bearing on the individual's ability to be successful at the job. So long as a student has some aptitude to learn and DO the job, they just need to GET the job with strong intellectual fitness signals from their academic records (e.g. GPA, school reputation etc.). The situation is even further exacerbated by courses graded on a force-ranked curve in this competition for high GPA, class rank, and post-grad opportunities.
So the real victims of LLM based cheating are not the deceived faculty, nor the unlearned cheaters themselves, but the honest students who want to compete fairly by studying the material and using their own brains. They are most at risk of doing the right thing but getting outcompeted for fitness signals. The faculty should make efforts to enforce fairness and root out cheaters for the benefit of these good students that they actually want to serve.
I am not without sympathy to this case, but two things:
One is that (from what I remember reading) most cheating isn't an A- student trying to jump up to an A+, but a D student trying to squeak out a C+. People turning in LLM-generated papers mostly aren't beating out honest students for competitive internships, they're just staying off academic probation.
Second, this is very much part of the tension I see between denouncing credentialism while also embracing it. If the reason why LLMs (or samizdat solution manuals) are a scourge that we must go to great lengths to fight is that it's essential to preserve the credential as a career signal, can we then turn around and denounce students for giving top priority to the credential over genuine academic interests?
I don’t really have a problem with credentialism. If universities were only for people interested in learning for learning’s sake, they’d have a couple of thousand students…
I love lots of subjects (economics, history, history of sciences, related humanities like sociology) but even there, without the carrot/stick of getting a career, I’m not sure how hard I would have pushed myself.
Maybe when the robots do all the work and we’re all effectively indolent aristocrats can universities return to being purely about eager to learn students…
I think students give top priority to the credential because they know that employer hiring and higher ed admissions practices rely on these credentials. So long as they continue to make decisions based on coarse but easy to rank signals like GPA and academic reputation, they sustain the incentive to win the competition by any means necessary.
In tech, I've seen firsthand candidates who pay to get ahold of the interview bank questions ahead of time. This becomes self evident when they start answering specific parts of the question that haven't been asked yet. Very recently a student was expelled from Columbia for building an LLM based program to help students cheat on live coding exercises. So while it may not be the majority of cheating, there does still appear to be willingness to cheat at the top.
We may also be talking about cheating with LLMs as more black and white than it actually is. In reality, I think many students probably do try to learn the material and then supplement that learning with LLM tools to make their work better, or to check that it's right. Sure it takes away from mastery and complete self-reliancy, but maybe they're just preparing for a world where this is kind of how we do everything. There may be an unfair analogy to calculators and word processors making mental math and handwriting obsolete.
Yep, that’s better expressed.
LLMs are really good tools. We should be teaching students how to make best use of them, not trying to prevent their use. We should also teach how to get things done when the internet is not available and not pretending that isn't always the case.
"It’s probably impossible to force disinterested students into putting in the effort to actually learn (but then, it never was) ..."
Did you mean "always"?
Sorry to start off with a suggested edit. I really did enjoy this essay, and I think you make a solid point, your self-deprecating "inconclusive noodling" ending notwithstanding.
Never was possible, yeah. I think I probably edited "It's probably not possible" into "it's probably impossible" and forgot that there was a callback in the parenthetical that should have parallel construction. That's the kind of thing I do a lot.
Here in Australia the schools have been going very hard on catching AI.
To the point where some business leader are now asking, "How come all these graduates as so unprepared for the real world? They suck at using AI"
“The ones who are interested in the opportunity to learn will still do the work, if they know that’s what it takes.” This is going to take a big sell job on our part, or at least bigger one than it used to be when I was only trying to challenge physics majors to not Google “Griffiths hw solutions,” at least not every week. I’m teaching premeds this fall, so presumably will be on the front lines of this issue…
I've had the advantage the last few years of teaching quantum off a colleague's lecture notes rather than a named book whose solution sets are Out There, but even when I used Griffiths, I was surprisingly successful at getting them to not just Google the solutions. Or if they did Google them, they were staggeringly bad at transcribing them...
Pre-meds may very well be a different deal, though. Engineers don't seem to be, interestingly.