The Despair of the Professor in the Age of A.I.

In my writing, and in my idle thoughts, I often return to Arjuna’s lament upon surveying the battlefield of Kurukshetra and finding that he must kill his friends and family. What is his duty in that extraordinary moment? What is my duty, then, in far more ordinary and less harrowing circumstances?

This question, which unfolds throughout a conversation with Krishna, appears in the first chapter of the Bhagavad Gita, a book I somehow did not know existed until I took a Hinduism class in college. I was not a good student, routinely skipping class, but, worried that there would be a price to pay for my truancy and bad grades, I tried, sometimes. A few books got read, including the Gita, which I didn’t really understand until my professor—the sort of charming, gray-haired stoic found in religion departments of liberal-arts colleges across New England—explained, in a capacious and friendly way, that one should do their duty without considering the outcome. And, even though I don’t remember doing particularly well in that class, I have spent the last twenty-five years thinking about Arjuna’s despair and Krishna’s command.

When I think about whether my nine-year-old will need to go to college, I mostly hope she won’t, because I don’t think this country should rely so heavily on a credentialling system that’s far too expensive, inaccessible, and time-consuming to be worth it. But I do worry that she will miss out on experiences like mine, when a nineteen-year-old was forced to read something he wouldn’t have otherwise and was guided toward a revelation, however banal or vain, by a patient professor. How do you place a value on something like that?

There will always be idealistic, ink-stained people who want to devote their lives to scholarly pursuits—their role to inspire young people to love ideas as they do. But this transfer, more than anything else in the academy, has been increasingly blocked by A.I. in the classroom. This past April, Jane Sloan Peters, a professor of religious studies, wrote a stirring Substack post in which she described a course she had designed, some years ago, about what people throughout history have been willing to endure for their faith. The class, called “Letters from Prison,” typically culminated in students trying to synthesize an overriding theme about what they had read. “When I began teaching this course four years ago, students struggled to come up with their own themes,” Peters wrote. But, through brainstorming and revision, the students would ultimately land on some understanding that both felt personal to them and proved they had grappled with the assigned texts.

Last year, the struggle ended—or, at least, got subverted. “Not one of my sixty students in ‘Letters from Prison’ struggled with this task,” she wrote. “I received tidy summaries of the text—the kind of compelling reviews you’d find on a book jacket—as well as perfectly vapid course themes that somehow took account of everything while not saying much.” What Peters suspected was that many of the students had asked A.I. to help. Like so many professors who have been confronted with the dispiriting new reality of student work, Peters adjusted, adding some handwritten brainstorming processes to her course, in the hope of making it A.I.-proof. But when she presented these new expectations to her students, something unexpected happened. “A wave of sadness washed over me, and I actually got choked up in front of the class.” Peters writes. “ ‘Before AI,’ I told them, ‘Students used to work hard to come up with their own ideas. I’d help, and they’d struggle, but they’d come to something that was their own. That doesn’t happen anymore and I grieve that.’ ”

In the past few years, I’ve spoken to a number of academics and instructors at the college and high-school level who have said similar things. They talk about a sense of loss and of despair, because the one thing that brought them meaning has been erased, or blotted out, by the arrival of A.I. Most, like Peters, do not blame the students, nor do they believe all students welcome the changes wrought by the new technology. “I’ve seen students respond with this disdain for teachers who just let A.I. use happen,” Peters said. “There’s this indignance, like, ‘Why don’t you want more from us than this?’ So, even if they’re using it, they’re still wanting us to hold them to a higher standard.”

“This is an exacerbation of a transactional model of education that has lasted for a long time,” Peters told me. Students are told that they’re in school to get a degree, one that comes with a high price tag and, for many, a debt burden. They are told that they will be assessed by the work they turn in. And, because A.I. allows them to turn in what Peters admitted was superficially “pretty good quality material,” they might not see why it’s such a big deal when they can’t explain what they have generated. “There are these waves of relief that wash over me when I see misspellings and poor grammatical structure in sentences,” Peters said. “When I can tell that they’re really working through it themselves.”

The teachers and professors I’ve spoken to have varying perspectives on what A.I. is doing and what it may yet do. But common concerns emerge. What follows are testimonials from Peters and eleven other faculty members at colleges across the country on how A.I. has changed their work.

Susanna F. Boxall

Lecturer in philosophy, California State University, Chico

I am very grim about the outlook of my career. I am about to be forty-five; if my job still exists by the time I am fifty, I will consider myself lucky.

The introduction of A.I., plus the demographic cliff, has had a devastating effect on higher ed. I think that big research schools will weather the storm, but lower-tier universities like mine will shrink or go away entirely. Already before COVID, there was a big push for online education; post-COVID, many programs switched to online to survive. However, with the introduction of A.I., all of those programs have become diploma mills. I taught online before and after A.I. In the pre-A.I. era, online education was qualitatively inferior to the in-person experience, but it was not a joke. Now, online classes are a simulacrum of education: the students pretend to learn, and I have to pretend that I am teaching them something.

In-person classes can still maintain some degree of rigor, and cheating can be reduced to zero as long as all assignments are done in the classroom. The problem is that this is not a solution to the enshittification of education—I can no longer assign papers because seventy to a hundred per cent of the students will use A.I. This term, I was able to do a comprehensive oral final with a single class because it was a small seminar with eleven students. Even then, I had to book a room for a six-hour slot in order to have a meaningful conversation with my students—scaling this when you have a hundred and fifty-plus students is impossible. Furthermore, because not all faculty are as concerned as I am about A.I. use, and because the students are using A.I. in online classes, the students are much less cognitively capable than they used to be.

Kevin Sun

Teaching assistant professor of computer science, University of North Carolina, Chapel Hill

I’m quite pessimistic about the impact of A.I. on both education broadly and my career personally, given the recent decline in C.S. enrollment.

The most obvious change in my teaching has been the elimination of difficult homework problems, which used to be a major component of my course. I’ve been trying to lean into social pressure as a source of motivation for students to learn, with group quizzes and in-class presentations, but there’s only so much an individual instructor can do given the systemic forces of A.I., grade inflation, the job market, student evaluations, etc. I’m worried that these forces allow many students to coast through school without learning as much as they used to. I acknowledge Bryan Caplan’s point that college is mostly about signalling, not learning, so it’ll survive as long as it’s a useful signalling device to employers. But, as college gets easier, the signal gets weaker, so who knows how things will change.

On a positive note, A.I. has helped me write course syllabi, lecture plans, exams, etc. It’s possible to use A.I. to grade and/or provide feedback to students—though I haven’t done so yet. I have also used A.I. to help me create in-class assignments in which students evaluate A.I.-generated code/content. In C.S., A.I. has shifted the emphasis from writing code to evaluating code. To train students in this, I present them with A.I.-generated code that is either correct or subtly incorrect, and I ask them to evaluate it.

I have a colleague who has completely embraced A.I. My understanding is that his course is much harder than it used to be, but students are allowed to use A.I. during exams. I see the motivation—A.I. is supposed to enhance our skills/productivity, so students should be expected to produce more—but I don’t want to create a situation where students are helplessly dependent on A.I. because they don’t have a solid grasp of the fundamentals.

Daniel Silver

Professor of sociology, University of Toronto, Scarborough

A.I. has fundamentally changed how I teach, and it demands basic reflection about what we are trying to accomplish. It has added a huge amount to my workload this year, since I spent a lot of that time trying to create new types of sociology assignments. The idea, basically, is to create multi-agent simulations where students create representations of the theories of writers like Adam Smith or Max Weber, and then they experiment with them. This was a huge commitment from me, the students, and the T.A.s, but it was worth it. The best final projects showed far more creativity and intellectual work than the typical second-year essay would have.

Beyond that, students still would use A.I. in a thoughtless way, as a replacement for their thought and judgment. So I made a point to just call them on it, and make them meet with me personally. I saw dozens of students, often for thirty- to forty-five-minute conversations. I wanted to understand where they were coming from. I would give them a zero on the assignment and allow them to redo it, after we talked about intelligent use. They usually improved, but not always. I felt that the act of meeting them was the most important part, so they felt that somebody, especially a professor, was paying attention to them and what they produced, which, alas, is rare in larger universities.

I also show them “replacement-level work,” on the model of the sports analytics concept of “wins above replacement.” These are basically variations of A.I.-generated versions of assignments. The students can see clearly how they all kind of look the same. Those are C-level answers, at best, and the students know they need to produce work that is better than what the replacement would be.

Over all, A.I. did really knock me out of a fairly comfortable set of teaching habits, which is producing a lot of emotional upheaval. But I do feel we all, including the students, are learning how to live with it, and we’ll come out better on the other side.

Elizabeth Strom

Associate professor, School of Public Affairs, University of South Florida

I teach a lot of fully online classes. There is really no way you can prevent use of A.I. in a fully online class. There are a few situations where A.I. responses are so off the wall I can give the student an F on the assignment, but most of the time it can be difficult to figure out the provenance of a short essay written by a student I’ve never met. They don’t know me or, often, the other students in the class, so there are fewer social norms that reinforce doing their own work. Usually a handful of students will be very interested in the topic and take advantage of the many opportunities to meet with me in person or on MS Teams, and make an effort to complete the readings and discuss assignments with me, but they are the minority. For others, it’s just an easy way to get three credits. I have tried to frame assignments to make it harder to avoid doing the reading—I ask for citations, for page numbers, for their opinion. I try to come up with engaging assignments: debate this topic! Role play this scenario! Turn this into a meme! But it is possible to game those requirements and hard for me to discern the originality of the work.

The university doesn’t have any sort of policy on the use of A.I. other than “follow the guidance of your instructor,” so I don’t think students are even getting a consistent message. And while some faculty will claim they can always recognize A.I.-generated assignments, I don’t think that’s the case when you are reading fifty short essays. There have been times when students have handed in gibberish that is clearly A.I.-generated work that they didn’t bother to check against the assignment, and other times when the work is so generic and lacking in detail that it’s clear they did not do the reading. But there is no definitive way to check, and with fifty students I don’t want to spend my time playing “CSI: Who Wrote This Paper?”

Neal Hebert

Assistant professor, department of visual and performing arts, Grambling State University

I’m a theatre professor, and when I teach plays I’m not looking at them as literature—instead, I ask my students to imagine these scripts are like the bones of a human body, and the way you get a full human is when you use those bones to “realize” a play onstage. The students have to write papers, but these aren’t research papers. Instead, I ask them, for example, to list two words that they feel describe the physical world of the play, and then write a paragraph explaining why they picked those words. The idea is to get students thinking like possible collaborators on a show: designers, actors, and directors.

I was excited the first time I got to read my students’ papers on August Wilson’s “Fences.” I selected that play because students could both read the script and watch the film if they want to experience it like that before writing the paper. In an intro class, I have to assume that, for some students, this might be the first play they’ve ever read, and they might need help to visualize it.

Out of forty students, the vast majority chose similar words, phrasing, and concepts, and most papers were written in that inimitable ChatGPT style: “This isn’t a simple story about injustice—it’s a clarion call for a positive understanding of justice.” If you’re a prof like me who also has ten or so years of public-high-school teaching in English and theatre fairly recently, the pattern of words spit out by LLMs is really easy to spot. It’s like elevator muzak, but in words.

I have since revised the assignment. I scour lesser-known anthologies for plays that have been rarely produced or published, and are too obscure for ChatGPT to learn about. If ChatGPT is used on these assignments now, it hallucinates characters, plotlines—it just makes shit up, since it has nothing to go on. I tell students that ChatGPT is disallowed from their writing process, that I can immediately tell when ChatGPT has been used, and that I will fail the student on this assignment if it is used—and, potentially, for the entire course, if we go through a formal appeals process. I’ve stopped being a collaborator in these intro courses and started being a plagiarism cop, and I do resent that a bit. I wanted to be the kind of professor my professors were for me.

Even in my upper division courses, I still sometimes get fake papers. I tell my theatre majors, “I get paid the same whether I pass you or fail you. But what you’ve just done is told me and everyone else in our department that you are so lazy you would rather outsource your collaboration to an app than risk being an artist.” Sometimes that gets them to write the paper honestly. Sometimes it doesn’t—and I can’t prove it in a historical drama class, because so many of the plays I have to teach are canonical, and were used to train the A.I. models.

I worry that if we start to produce students who can’t be bothered to read and think about the plays they are performing in—or may one day perform in—then the next generation of theatre professionals will only be able to meet and critique the world in which they live in the most pedestrian and boring ways possible. Can you imagine A.I. Performing Arts Slop? The theatrical equivalent of the images ChatGPT and its competitors spit out, soulless and inert, arriving on stage stillborn? I can.

Lauren Aulet

Assistant professor, department of psychological and brain sciences, the University of Massachusetts, Amherst

The tension around A.I. in higher education feels less like a simple administration-vs.-faculty divide and more like a mismatch between the speed of institutional response and the slower, unresolved pedagogical questions faculty are facing in the classroom. Universities are understandably trying to respond to A.I. quickly, but faculty are often the ones dealing with the hardest implementation questions: what counts as student work, how assessment should change, and how to preserve the kinds of struggle and independent thinking that learning depends on.

At the same time, I do think A.I. has expanded what I can do scientifically. For me, the biggest effect is that it lowers the cost of trying things computationally (i.e., with code). Coding is a fairly ubiquitous part of my job and I can now move more quickly from an idea to an analysis script or proof of concept. That does not replace scientific judgment, but it does make certain ideas feel more testable than they would have before.

So the tension for me is that the same tools that can be genuinely useful for research can also be destabilizing for education. “A.I. helps me code” does not straightforwardly translate into “A.I. is good for universities.”

Auyon Siddiq

Associate professor of decisions, operations, and technology management, Anderson School of Management, University of California, Los Angeles

Our administration has definitely encouraged faculty to be A.I.-forward, but adoption across faculty varies. The default policy is that A.I. is permitted, and it is up to the faculty member to specify when it is prohibited.

I teach the core statistics class for first-year M.B.A.s, and we encourage use of A.I. for the gritty details (e.g., coding) so students can focus on concepts. We actually made our exam a hundred per cent A.I.-friendly, as an experiment, with the only constraint being you can’t use your phone to take photos of the exam. The class average was still only seventy-five per cent, because the students who were truly lost apparently weren’t saved by A.I. But this could change in the future.

Interestingly, voluntary uptake of these tools is also mixed across students. In many cases, students haven’t really used them regularly until they are required to do so in my class. But this is also changing rapidly.

A.I. is definitely making me rethink teaching. Honestly it has freed up mental bandwidth to think more about my lecture structure, pacing, anecdotes and case studies, and so on, rather than being consumed by the mechanics of making PowerPoint slides and homework assignments. So, personally, I find it very exciting. Of course, there’s the lingering anxiety that we might be replaced entirely by A.I., although I don’t think that is likely to happen anytime soon. I think there will always be demand for human relationships, even (and maybe especially?) in higher education.

David Roach

Assistant professor of history, Campbellsville University

I have seen a shocking and discouraging amount of A.I. in the classroom. I have to teach a few online classes every year, and I would estimate that half of my students have used A.I. on these assignments. In two in-person history surveys I taught recently, I believe that a little over half of the students used A.I. on their last short paper.

Initially, I responded, for my in-person classes, by deëmphasizing out-of-class writing, and, for my online classes, by devising ways of policing the assignments: prompts that might trigger incorrect answers from A.I., hidden words that had to do with the assignment but might get the A.I. talking about a different speech or document. But both of these techniques have negatively affected student learning and my experience teaching. I believe that students need to practice putting their thoughts in an essay—they need the friction and difficulty of putting words together and learning how to think. So I’m frustrated because I cannot teach how I want to teach.

I have also come to believe that students now know how difficult it is for faculty to prove A.I. cheating. Early on, a couple of months after ChatGPT hit the scene, and before many faculty became aware of it, I had a typo in a prompt that helped me catch A.I. cheating. I met with eight students, about ten per cent of the class, and all but one of them admitted to using A.I. This last year, in many meetings with students about A.I., only one of them admitted to using it.

My emotional response to all this is hard to describe, something between disgust and despair. Was it always the case that half of our students would cheat if it were easy enough? If they knew that it would be hard to prove? It’s hard to consider that and not despair.

David Song

Professor of Asian American studies, East Los Angeles College

In my classes I allow some A.I. utilization but with appropriate attribution and all the usual academic-honesty provisions. Of course, no one bothers with that, even with the content that’s blatantly generated.

With community-college students, many are just trying to complete their general-education requirements, so there’s generally less investment in actual learning and more focus on checking off all the boxes before they transfer out. I suppose that would be the case in any institution with grad requirements. The most rampant abuse is generally with the asynchronous online courses where students are faceless.

This brings up another major problem that seems more specific to community colleges. A lot of financial-aid fraud has apparently taken place, with fake students enrolling in classes and using A.I. to actually complete various assignments. That started a few years back. I remember sensing something was off in my intro to Asian American history class when I would read student introductions and there would be, like, three to four students with generic Wasp-sounding names—our student body is predominantly Latino, with a smaller percentage of Asians—writing introductory posts about how they have already taken other classes in Asian American issues and literature, and that just sounded like bullshit.

Beth Ritter-Conn

Assistant professor of religion, Belmont University

The erosion of trust between faculty and students has been the hardest part for me. the magical thing about teaching, especially teaching undergraduates, is that, at its best, it is this experience of discovery and curiosity, people trying ideas on for size and seeing what still fits them from their upbringing and what they need to leave behind so they can grow. There is trial and error, there are mistakes that lead to revelations, and the classroom is this safe laboratory where you get to do that in conversation with your peers who are all doing the same thing. These days, I feel like I can’t trust my students to be willing to make that mess with me in quite the same way.

The tipping point was last year when I had Honors students—Honors students!—using A.I. to write reflection journals. Literally the only task there is “tell me what you are thinking inside your own head.” There is no right or wrong answer. It’s just, Give me your thoughts on this thing. And I had students who outsourced that task to the robots. I can’t give honest feedback when it is not honest work. I can’t help you work out how you want to think about something, how you want to be in the world, if you are not using your own brain to tell me where you are. We also can’t have an honest and meaningful class discussion about a specific text if people are using A.I. to summarize the assigned texts.

So there’s just so much that feels like this new barrier to the whole process of a liberal-arts education. I’ve made some adjustments to my pedagogical strategies since last year. We’re doing more in-class writing; I’m doing in-class, pen-and-paper exams rather than using our learning management system. I’m being kind of a stickler about requiring hard copies of textbooks and requiring in-class notes to be taken by hand, and requiring phones and laptops to be put away. Things have gone better this academic year. But it still just feels like this extra thing I have to police (or else just give up and decide not to).

Jane Sloan Peters

Assistant professor of religious studies, University of Mount Saint Vincent

In her Nobel Prize address from 1993, Toni Morrison takes a familiar fable and transforms it into a commentary on the power and precarity of language. A group of young people ask an old, sage woman whether the bird they hide behind their backs is living or dead. She replies, “It is in your hands.” The bird is language.

In the original fable, the youths are depicted as foolish and mocking, the elder depicted as measured and wise. But in her retelling, Morrison casts the relationship between young and old as one of mutual incomprehension, and the elder is not wise, but selfish and afraid. She highlights the failure of the old woman to respectfully consider the youths’ question. Veiled beneath their trick with the bird is a plea that she take them seriously, that she acknowledge the way her generation has failed theirs. “Is there no speech, they ask her, no words you can give us that helps us break through your dossier of failures? Through the education you have just given us that is no education at all . . . ?”

I think about Morrison’s address when I think about how artificial intelligence is shaping the student-teacher relationship. Students use A.I. even if they are apprehensive about what the technology will mean for their life. And at a certain level I think they’re looking to their professors for guidance on how to use it, and even more, for affirmation that their words still matter. That they still matter.

Jeremiah Croster

Instructor in English, Houston City College

Houston City College is a large, urban community college with open admission. It’s one of the most international community colleges in the United States—we get a lot of students from Africa, Asia, Mexico, and Central America, a few from South America, and even Europe (for some reason that I can’t figure out, we get a lot of Kazakhstanis). We also have a lot of students from the city. I could easily have a class with thirty per cent immigrants, thirty per cent Houston natives, and thirty per cent people who’ve moved here from other parts of the U.S. A significant majority of our students are either Hispanic or Black, and, again, within those populations, there’s a good mix of immigrants and native Houstonians. Students are pretty diverse in age, too. The majority of my students are in their early twenties, but an average class has three or four students older than thirty and I often have a student or two older than me. Economically, our students are generally poor or working-class. They are often negotiating difficult home lives. Once in a blue moon you get someone from the middle-class whose parents are just opting to save money by sending them to community college, but it’s rare.

The first year after ChatGPT was introduced wasn’t too bad. It was the spring semester of 2src24 when the shit really hit the fan and everyone was using it. There just isn’t much unique voice in the writing anymore. I’d say in that first year of heavy use, probably fifty to sixty per cent of my in-person students were using it and had that homogenized voice. In the online classes that I teach, it’s easily at eighty to ninety per cent.

I do a few things. For in-person classes, I use blue book essays now. It does help, but it also prevents students from developing their thinking on their own time. I used to tell students to take a walk for inspiration if they were struggling with what to write, but that isn’t possible with blue books. For shorter assignments, I’ve switched from written responses to video responses—students upload video of themselves talking. Some of them still read off a screen, but most of them just say what they’re thinking. Finally, I now deduct points from essays for the common tropes of A.I. writing: lists of three, overuse of adjectives, etc. In my experience, ChatGPT is still a pretty bad writer and can’t hack the existential risk involved in spitting out a compelling thesis.

When it comes to my own future, I’m not too worried: I’m forty-nine and I think I can hang on until my retirement plan kicks in. When it comes to the future of teaching college writing in particular, I’m less optimistic. Faculty are mostly opposed to using A.I. in classrooms, but some embrace it, and the administration is convinced it’s the greatest invention since sliced bread—they’re actively trying to get resistant faculty on board. I just haven’t seen any good use for it in the classroom, at all. I know there are fields like biomedical engineering where A.I. might help speed up research processes that will yield medical advances, but that’s very different from the work we do in the liberal arts and especially in the humanities.

A buddy of mine used to say that there are two approaches you can take to college learning: one approach is to help students get an education and the other is to help them get a degree. The get-a-degree approach was already winning even before A.I., but now that it’s here, the education part is starting to feel like something someone will write about in a history book. Or maybe A.I. will do it.