An article from New York Magazine has been making the rounds this week.
It touches on several aspects of the problem of student use of ChatGPT and other AIs in their coursework, including its prevalence, the underestimation of the phenomenon by faculty, the growing acceptance of its use (see Cluely), and the possible consequences of all of this. For some, it will be an eye-opener. If you’ve been reading the posts here about AI over the past five years I don’t think you’ll be too surprised.
That said, there are some noteworthy passages. Here’s one that illustrates what students are doing with AI and how they don’t quite understand what it is they’re doing—including what they’re doing to themselves.
Wendy, a freshman finance major at one of the city’s top universities, told me that she is against using AI. Or, she clarified, “I’m against copy-and-pasting. I’m against cheating and plagiarism. All of that. It’s against the student handbook.” Then she described, step-by-step, how on a recent Friday at 8 a.m., she called up an AI platform to help her write a four-to-five-page essay due two hours later.
Whenever Wendy uses AI to write an essay (which is to say, whenever she writes an essay), she follows three steps. Step one: “I say, ‘I’m a first-year college student. I’m taking this English class.’” Otherwise, Wendy said, “it will give you a very advanced, very complicated writing style, and you don’t want that.” Step two: Wendy provides some background on the class she’s taking before copy-and-pasting her professor’s instructions into the chatbot. Step three: “Then I ask, ‘According to the prompt, can you please provide me an outline or an organization to give me a structure so that I can follow and write my essay?’ It then gives me an outline, introduction, topic sentences, paragraph one, paragraph two, paragraph three.” Sometimes, Wendy asks for a bullet list of ideas to support or refute a given argument: “I have difficulty with organization, and this makes it really easy for me to follow.”
Once the chatbot had outlined Wendy’s essay, providing her with a list of topic sentences and bullet points of ideas, all she had to do was fill it in. Wendy delivered a tidy five-page paper at an acceptably tardy 10:17 a.m. When I asked her how she did on the assignment, she said she got a good grade. “I really like writing,” she said, sounding strangely nostalgic for her high-school English class — the last time she wrote an essay unassisted. “Honestly,” she continued, “I think there is beauty in trying to plan your essay. You learn a lot. You have to think, Oh, what can I write in this paragraph? Or What should my thesis be? ” But she’d rather get good grades. “An essay with ChatGPT, it’s like it just gives you straight up what you have to follow. You just don’t really have to think that much.”
I asked Wendy if I could read the paper she turned in, and when I opened the document, I was surprised to see the topic: critical pedagogy, the philosophy of education pioneered by Paulo Freire. The philosophy examines the influence of social and political forces on learning and classroom dynamics. Her opening line: “To what extent is schooling hindering students’ cognitive ability to think critically?” Later, I asked Wendy if she recognized the irony in using AI to write not just a paper on critical pedagogy but one that argues learning is what “makes us truly human.” She wasn’t sure what to make of the question. “I use AI a lot. Like, every day,” she said. “And I do believe it could take away that critical-thinking part. But it’s just — now that we rely on it, we can’t really imagine living without it.”
The article is “Everyone is Cheating Their Way Through College” by James D. Walsh.
A common (and, to my mind, clearly correct) response I’ve been seeing is:
I think the question why both (1) and (2) are the case is at least as interesting and important as the NYMag piece itself.
Could you explain why (2) is obvious? It is not obvious to me – essay exams have been a standard feature of all the departments where I’ve taught or been taught, though perhaps not the general rule.
I’d be surprised if it has been a standard feature to use essay exams instead of out-of-class writing assignments, though.
I’m really sorry for being obtuse about this, but that’s just the situation I’m thinking about. Faculty have exams or out-of-class writing assignments like essays. They don’t have both. I don’t think admin care. The writing requirement that many philosophy classes fulfill is often met by the in-class essay exam.
Again, not everyone does that. There’s a great diversity in the assessment-oriented assignments. But I don’t think people assign essays because upper-admin wants them to. I’ve mostly been at large public schools, though, maybe that’s the difference?
Ah, then it’s I who was being obtuse! The places at which I’ve studied and taught have been all over the map, but on the basis of your comment I suspect I’ve been overgeneralizing.
Well, I’m still interested in hearing more about the other side, if you care to elaborate on it.
It’s different in the US, but here in the UK, admin has a lot more influence over how assessment takes place, and do care. And admin in many UK universities are blocking faculty from moving to in-person exams. There are two kind of financial reasons for this. One is that in-person exams (especially if implemented more widely than currently) required administrative resources for scheduling and physical spaces to host all the exams. Building space costs money. Second, exams are unpopular with students. As of today, 40% of UK universities are projected to be in financial deficit by this summer. They can’t afford to put off students, whose fees they need to survive.
Also confused about this–especially given that in my experience, higher ed admin is swamped with people who are much more familiar with practices in STEM than in the humanities and social sciences. And it is very common to basically only hold in class exams (often in conjunction with some low-stakes problem set type homework assignments) in much of STEM. Further, this might not last depending on whether we can recover from Trumpism sooner rather than later, but at least right now, very many of us–not all, but many–teaching in higher ed in the US have complete control or very near complete control over things like what sorts of assessments we choose to give. At my institution this would never even come up, because there is absolutely no oversight surrounding what I do in my own courses. I think some version of this is relatively common, or at least that there is extremely minimal oversight.
We do this as a matter of department policy now. Including for otherwise online courses. We’ve yet to hear a peep of pushback from admin, who are otherwise all on board for having us “teach” students how to “use” AI.
It doesn’t really solve all the problems, but it reduces the incidence of cheating to manageable proportions. Pedagogically, it’s no substitute for taking work home, but the reality is that at the lower level, virtually nobody here is doing their own work at home any more. (A few are, because there are always outliers. But it’s really not very many of them.)
Pedagogically, it’s seemed to me almost from the beginning that the clear solution is to shift to a broadly Oxford-style tutorial system (ungraded essays discussed one-on-one with the instructor), but of course this would require a massive change in how resources were allocated.
That doesn’t sound like a “solution” to me – that sounds like a massive change in how resources are allocated, with a small number of students getting a much more intensive education, and a large number of current students being eliminated from the university!
Think of it, on the contrary, as allocating far more resources to teaching (either by way of more hiring or by way of reducing research expectations but increasing time spent teaching). Instead of large classes with 100 or more students, imagine 10-20 student size courses in which one can have students write essays that are presented individually or perhaps in pairs in front of the instructor who discusses them with the students, proving instead feedback, follow up questions, etc.
aka “an obviously good thing”
The real challenge, I think, is asynchronous online classes. In such classes, closed book, closed notes, closed-AI exams are difficult to achieve. There are online proctoring tools, but they’re kind of a pain to set up, are intrusive, and can fail. It’d be great if I could send all my online students to a testing center on campus for in-person proctoring, but that’s not feasible for various reasons. I truly *hate* teaching asynchronous online for these reasons.
We just book a room–usually in the evening, so that more are available and to minimize scheduling conflicts for students–and invigilate ourselves.
(You do have to tell students from the get-go, though, because it’s confusing for an online course to have in-person exams.)
All my discussions are audio/video exchanges now about personal experience that relate to the topic. Its obvious when they are really scripted thus their grade reflects that. I’ve been having them become content creators ironically. I think some of the most engaging content right now outside of academia is really well done youtube video essays.
Not following you. I have switched to blue books and the administration couldn’t care less (nor would they have the power to do anything about it if they did care).
I have switched to blue books and the administration couldn’t care less (nor would they have the power to do anything about it if they did care).
This is one way the US is different from some other places (and, having expereinced both systems, clearly better, I’d say.) Where I work (in law, in Australia) “assessments” have to be approved by a committee more than a year in advance, and there is a limited range on what you can do. There is a fetish of “authentic assessment” (although what that even means in many cases is unclear), but it limits what you can do. The basic form is set at least by the program level, and often by external regulators. Tons of things that would be thought to be basic matters of academic freedom in the US are set top-down here. It’s not that you couldn’t have these sort of tests (we have in-person, proctored exams on lock-down software at the end of the term for most of our law classes), but that any change has to be approved from on-high, and will be looked at skeptically.
Same
Regarding 1 – I think it’s obvious that we can prevent *this* behavior by these means. The question is whether it’s obvious we can teach the skills that we want to teach. Maybe we can!
But it’s notable that the kind of writing that is assessed by in-class blue book exams is off-the-cuff one-hour writing, rather than the longer-format writing that most of us have been producing in our research lives and have been aiming to teach. Has that been a good aim? If so, maybe the multi-day in-class lockdown-browser essay assignment proposed in an earlier post here is better than blue book exams, though it’s really not at all obvious to me!
https://dailynous.com/2025/04/02/the-multi-day-in-class-lockdown-browser-essay-assignment-guest/
Thanks for the shout-out, Kenny! I’ve really liked the multi-day in-class lockdown browser essay format, and I plan to use it again next fall. I also keep learning how to do new things with the format (I recently added access to Google Translate in the assignment in such a way where students didn’t have access to the rest of Google–this was super helpful for a few students for whom English is a second language).
And, regarding point 2 above about how our institutions are responding to the idea of in-class writing assignments. My institution’s (Indiana University) messaging has largely centered around “cool ways to incorporate AI,” but IU’s Center for Innovative Teaching & Learning was enthusiastic enough about my idea to have me write a guest blog post about it for their blog (it’s an adapted piece from my original Daily Nous piece, with a shout out to Daily Nous!).
When everything went online in 2020, there were legitimate worries that educators could be replaced by once-recorded lectures and cheap TA labor for assessment.
With the prevalence of AI, now, there’s a good case to be made that real learning (and, thus, credentialing that is worth a dang) can only happen in a live, interactive setting, with in-person supervised graded work etc. I think that’s an opportunity for to reclaim some of the value we bring as educators.
I’m confident that at least half of the students enrolled in my online asynchronous classes this semester used generative AI on almost every assignment. I believe (though I am not very confident) that the students who earned A’s and A-‘s did not use this technology much, but plenty of students who earned B’s did. This was one reason I decided to move jobs to a university with less demand for these sorts of classes. I can’t see myself staying in this profession if this becomes the norm, though.
My own view is that there’s a first best response and a second best response.
The first best response is to lean in to the human meaning of teaching. The class is a conversation–no tech allowed, you have to bring the physical books or course packet, everyone participates. All writing assignments are iterative and require meeting with the instructor to brainstorm, etc. Papers are the result of a dialogue between prof and student with real feedback that grow out of class conversations. Students like it because it’s very supportive and they feel like they’re actually accomplishing things. The problem with the first best solution is that it requires small classes and light teaching loads. I can pull it off at a fancy R1 but just barely.
The second best response is to just go back to the old ways. Do bluebook handwritten exams, with no tech in class allowed. Everything is done in analog. You barely even use Canvas/Blackboard. No PDFs, nothing, to encourage actual reading as opposed to AI summarizing. The problem with the second best solution is that people may not want to take those classes because they’re harder than classes they can cheat at without compensating advantages, so enrollment may drop. But I think this is the best that most people can do.
I sure wish there was an option that was as good as the first best, but less resource intensive.
The problems of the second option can be fixed through collective action (like oh so many things). If everything everywhere is done in analog, this goes away.
One of the biggest issues I see is that it is chipping away at their confidence to write without assistance and realize it is ok to struggle with concepts. All the testing has made them afraid if imperfections. Microsoft’s recent study of ai usage in the work place is applicable to us as well. Trust in AI is high when the task is perceived as inconsequential, stress is high, deadlines loom. All the same factors that contribute to normal cheating.
From what I understand, at Oxford (and perhaps other places in the UK) assessment is almost exclusively done by way of final examinations, many of which are administered in person through some centralized system. Does anyone know how much of a problem this is at places like that?
Although Oxford is not specifically mentioned, a recent study, mentioned in the THE here:
https://www.timeshighereducation.com/news/most-universities-still-use-online-exams-despite-cheating-fears,
would seem to indicate not.
The first paragraph:
“UK universities’ continued use of remote online examinations without supervision poses a grave threat to the integrity of undergraduate assessment, argues a new study which urges institutions to abandon exam practices introduced during the pandemic.”
For Oxford specifically, according to data published in the linked study, 13% of exams at Oxford are remote online, while 0.4% are remote online and supervised.
That would imply that at Oxford, 12.6% of exams are unsupervised, online examinations.
I suppose that one sould still claim that ‘many’ are administered in person, though.
Genuine question: why the scare quotes around ‘many’ (and the “I suppose”)? You make it sound as if inferring “many are administered in person” from “87% are administered in person” would be something of a stretch.
I meant that though while the figures quoted don’t undermine anything in your post, it is still interesting to know that Oxford is not unaffected by these changes, to an extent that some may find surprising.
Serious question – what if we asked students to quote, explicate, and then critically evaluate the responses of two or three LLM’s to a posted prompt of philosophical importance? Students would be required to input exactly the prompt provided, and by having them cite the response, instructors would have a direct comparison between the writing/compositional styles of the different LLM and that of the student.
You can tell the LLM to do that and then to do the comments but in a different voice.
LLMs are arbitrary text generators. There is no possible assignment that requires the students to generate a text that they cannot have an LLM do. No amount of recursion, meta-textuality, or cleverness can change that.
(Except that LLMs still can’t put together a half decent bibliography.)
You can tell LLMs to write, say, the plot of a Toni Morrison book in the voice of a slightly drunk Scottish flute player afraid of rabbits relating the plot to an alien from Venus who needs everything to be told to them in limericks. They’ll do their best. Will it be great? Not often. Will it be a clearly distinguishable voice from (e.g.) any other voice you told it to use? Absolutely. So it’s child’s play for an LLM to generate a response distinct from another LLM’s response, especially if you prompt the LLM properly.
One might argue students won’t know how to prompt it properly if you conceal the purpose of the assignment from them. But in addition to ethical concerns about deceiving our students, there are easier ways to catch bad LLM users than coming up with cockamamie assignments like this that force students to do what all of us I am sure already have come to hate doing, which is engage with the output of these goddamn LLMs one iota more than is absolutely required in order to survive.