In my department (and I think in just about every department in every college), the number one discussion in meetings and email discussions for the last three years has been what to do about AI. The main question–sometimes it seems like the only question in my department–has been “how do we AI-proof our classes?”

I get it: students can have ChatGPT cook up a paper for them on any subject in a few seconds. The paper can be well-written enough to get an A if the student asks for that. If the student is too worried about getting caught, they can have ChatGPT serve up a B- or C+ paper instead. While most of us teaching ENGL101 in America have some nose for papers that don’t quite smell like student-written work, any teacher who says they can unfailingly sniff out AI-written prose is lying, at least to themselves if not to you.

So yeah, our teaching lives are different now. Almost everyone who liked being a teacher before, say, 2023 doesn’t like what’s happening now. It occurred to me not long ago that if I had begun my teaching career in 1923 or even 1933, I could have completed a thirty year teaching career without having to live through many (or even any) cataclysmic technological changes. There would have been major social changes to navigate–the Great Depression, WWII, the GI Bill, widespread entry of women in colleges, desegregation–but the technology of teaching and classroom learning wasn’t radically different between 1933 and 1963. Had I started in 1933, I would not have been forced by technological change to reinvent my teaching practice every few years.

When I really did start teaching, though, was 1993. The technological changes we’ve seen since then have been massive. Not all of my students were even using word processors in those first few years–I still took in typewritten papers every once in a while. For that matter, I still distributed handouts that I had made on a mimeograph machine from 1993-95. From then to now, I’ve taught through the total hegemony of the word processor, the internet classroom, YouTube, Khan Academy, social media, learning management systems, the smart phone (as well as the tablet and the ubiquitous Chromebook) before I had ever heard of ChatGPT. And all of those developments have had deep implications for the way I do my work.

But ChatGPT and all its logorrheic LLM siblings have deeper implications still. They are cataclysmic for the work I do.

My colleagues are intelligent and sweet-natured, and I am lucky to be working with them. But, despite their voluble commitment to political progressivism, we all can be some of the most emotionally conservative people around, at least when we all get in a room together. Is there a way we can, you know, find a way to keep teaching the way we’ve been teaching? Let’s just do that! our department seems to be saying, at least if you read our meeting minutes.

I can bitch, and have bitched, about the fact that I have to upend my entire teaching practice to accommodate a tool that will write competent prose and summarize any reading in a matter of seconds. It’s all the more galling that the tool comes to us by way of Elon Musk, Mark Zuckerberg, and the rest of their techbro robber baron buddies and their shareholders. But this is the way creative destruction works: in an open market, entire systems of wealth and production are continually being destroyed by new technology. And if I can’t see ways to use LLMs to support my teaching practice, I’m going to get chewed up and spit out all the more quickly in the coming years.

Sooner or later, AI will be teaching everybody. In the long run, there is no AI-proofing the classroom. A computer that can write competent prose and read anything can also, sooner or later, teach people to read and write. It’s already being used by many teachers as the vaunted “papergrader.com” that some of my waggish colleagues used to pine for 20 years ago. However, I remain optimistic that for at least a little while longer, a human teacher who knows what they’re doing–and who cares about students–can offer something a computer is not yet able to.

So for now, until the computers kick me out of the classroom, here are some of the ways that I’ll be trying to deal with the new regime: taking advantage of the many blessings of AI where I can, minimizing its malign influence whenever possible. I offer these as a starting point for conversation with my colleagues and my students.

1. Speak Frankly with Students: If my and my colleagues’ stated feelings about AI are any guide, students are getting mixed messages about use cases for AI. And even if we educators weren’t giving mixed messages, students are certainly receiving mixed messages from the culture at large, from the techno-utopian advertising they see from Google and Apple and Meta to creepy cautionary tales like M3gan. Given that my job as a teacher of rhetoric is to help people understand how arguments work, and given that one of the main functions of LLMs is to confect natural-sounding arguments, part of my job now involves helping students consider LLM use cases. I’m far from an LLM hater, despite some of the obvious losses that LLMs present for my work as a writer and teacher of writing. But I’m also deeply skeptical about any utopia that Google et al. are selling. For now, I expect my students not to use LLMs to create text that they pass off as their own. They can expect me not to use LLMs to grade their work. Only one of these expectations is realistic; I know that as a result of their anxiety, laziness, or cluelessness, some students will be trying to pass LLM content off as their own work. I’ll speak to that issue below.
2. Stop Grading Students; Give Them a Fair Assessment Instead: I’ve been arguing that we should get rid of letter grades since long before I ever heard what an LLM was, but LLMs have only made grade grubbing and credentialism more acute: if it’s so easy to get an A by cheating, why would any student accept a C? And if everyone is getting an A, why do we have grades at all? Replace the anti-educational grading system we have with a straightforward, outcomes-based pass/no pass system based on in-person competency testing. These tests can look like a lot of different things, not just essay tests. But they might especially be essay tests, handwritten in a Blue Book or typed on a computer with a lockout browser. (To that end, by the way, many of my colleagues, especially in the math department, argue that our college needs a proctored testing center. I have no doubt that we will have one sooner or later. But my college has never been a leading-edge institution; we’ll have our testing center only after several other colleges in the state system have started one and the practice becomes an official, shiny Best Practice with our State Board for Community and Technical Colleges).
3. Implement a No Devices Classroom. One of the central goals of education is to help students cultivate cognitive endurance: “the ability to sustain effortful mental activity over a continuous stretch of time.” I have no doubt that this goal is made more difficult when students have ad libitum access to multiple screens and information feeds in the classroom. And while the research is equivocal for students as a whole, for lower performing students–those who are over-represented in open door community colleges–the research suggests that device bans help students to stay focused on their classwork. If someone is listening to Spotify on their earbuds, managing a text thread, checking TikTok every 5-7 minutes, and squeezing in a round of Candy Crush during down time in the class (however a student might define “down time”), should I be surprised that they are having trouble identifying the main idea of the paragraph we’re all supposed to be looking at?
   
   One may reasonably ask what AI in the classroom has to do with this fractured attention economy. It’s related in two ways: first, the companies selling AI as an edtech that students should be using in the classroom are often the same companies that benefit from having students constantly plugged in to multiple streams of data simultaneously. Secondly, I believe there is a benefit to having students at least sometimes exert their minds without the cognitive prosthesis of AI, the same way that you’ll get in shape faster riding an old-school “acoustic” bicycle than riding an e-bike (and much, much faster than riding a motorcycle). I’ll admit that this second claim is more vibes-based, and I’ll be happy to revise it in light of high quality research findings. But for now, common sense tells me that it helps for students sometimes to have only their minds to rely on.
   
   Here’s a very simple example. One of the best ways that a person can prove to someone else that they understand something they’ve read is to summarize that reading. For that matter, summarizing is one of the best ways to prove to yourself that you understand what you’re reading. It’s a foundational tool for managing information, as well as a vital step in making a rhetorical analysis, an academic response, a literary analysis, a research paper, and a whole bunch of other academic assignments. It’s also one of the more difficult skills for a person to learn, especially with readings that are challenging. If I assign students to summarize a tough article, it’s a lot to ask that they struggle for an hour or more with a task that a computer could do for them in ten seconds. I can hardly blame some of them for having ChatGPT serve up a summary for them if I assign it as homework. However, if we write the summary together in the classroom–which has the advantage of our being able to puzzle out together the writer’s organizational schema and the main ideas of paragraphs–we might actually write a true human summary together. That only works when there is one part of our lives where AI is not a constant background (or foreground) presence.
   
4. Use LLMs Outside of the Classroom. I’m not ready yet to require that students use LLMs outside of the class–lots of students, especially the more thoughtful ones, are deeply skeptical of LLMs for a lot of reasons. However, I am starting to look for parts of my teaching that I think can be safely off-loaded to AI and which I can recommend to students. One of the big use cases is grammar and punctuation instruction, a part of my teaching that I used to love but which has gotten steadily crowded out by changes to our department’s approach to curriculum.
   
   ChatGPT is a potentially awesome teacher of sentence grammar. As I tell my students, beyond all the debates in lefty spaces about “Standard Edited English” being a tool of colonialism and white supremacy, there’s great value in being able to understand how sentences are put together, how parts of sentences like phrases and clauses interact. One can say a great deal with nothing more than simple declarative sentences. However, understanding how an appositive or an absolute phrase works (whether or not you know the names for those structures) will make it possible to say and write–and think–ideas that are much, much more subtle, as well as much harder to formulate with only declarative sentences. Explaining grammar and punctuation is one of the few areas of life where I claim to have real expertise; nevertheless, I think that ChatGPT is better than I am at it, and it’s certainly more tireless at it.
   
   One of the assignments I’ve been giving, and which I plan to use even more widely this term, is to have students upload a paragraph of a reading we are studying (or sometimes a paragraph of their own writing) to the LLM of their choice, with the instructions that the LLM quiz the student on how the sentences are constructed. Sometimes I have LLM quiz students on the types of clauses that are appearing in each sentence; at other times I have the students try to classify sentences as simple, complex, compound, or compound-complex; at other times I have the LLM test students on the placement of commas or other punctuation in their writing. I do this not because I want students to memorize the nomenclature of clauses and punctuation but because the activity forces students to pay attention to the way sentences are constructed, the same way that musicians learn to pay attention to chord progressions and photographers learn to study the composition of a shot. And not only does ChatGPT know at least as much as I do about sentence grammar and punctuation, but it’s infinitely patient. There are similar huge gains available to us if we use LLMs as reading comprehension aids, as critical readers for students’ rough drafts, as explainers of historical and sociocultural context. I wrote about this phenomenon of LLMs-as-the-Computer-from-Star Trek here.
   
   In fact, practically the only way I want students not to use LLMs is as creators of content that is to be graded. Of course, that’s one of the only things that some students seem to want to use LLMs for, and that’s one of the main reasons to retire this 18th century grading system we inherited from Yale University. As I tell my students multiple times a term, if they are coming to college because they hope a degree leads to a job, they’re only going to get hired to to one of two things: 1. a job the employer would prefer not to do (e.g. toilet cleaning) or 2. a job the employer is not able to do themselves. And if the student has never developed skills that the employer doesn’t already have, they’re going to get the toilet cleaning job. And why go to college for that? As I tell my students, if what you know how to do at the end of your mystical journey in college is to have ChatGPT write a report for you, no one is going to hire you to do that. Every employer in America already knows how to have ChatGPT write a report for them.
   
5. Teach In Person. Notwithstanding 30-odd years of advertising and boosterism that online classes were the wave of the future, I’ve always been an online learning skeptic. I wasn’t impressed by the online classes I took as a student; in the few online courses I taught before the pandemic, I was troubled by how many students seemed to struggle who in my professional estimation probably would have done ok in a face to face class. And nothing I saw as an online-only teacher during the pandemic disabused me of my original skepticism. On the contrary, I think at our college we’re still adjusting to student populations who were subjected to the tender mercies of all-online education for a year-plus.
   
   At this point of human history, when everything I know or might ever know is available for free through LLMs, I have nothing to offer students beyond a human face. But there is still some value in having a human face: we are highly evolved to interact with actual physical human beings. Face to face classes aren’t the only modality that ever makes sense–I would argue that online learning is appropriate for some students (particularly more experienced and self directed students) and for some classes–but for a general education course like ENGL101 at a community college, I believe there should be a presumption of some in-person learning.
   
   What does this preference have to do with LLMs? While of course it’s easier to ascertain that a student, rather than an LLM bot, is doing the classwork when you can actually see the student doing the classwork, the main reason for preferring face to face learning has nothing to do with enforcing some academic honesty regime. Rather, the main advantage of face to face classes in our current LLM world is that most people still like seeing other people and like being seen by them. It’s shocking and sad how often my students confide in me that what they really hope for out of college is to make a friend. Some of them may already have the supposed companionship of an AI therapist or an AI girlfriend, but what they really want is other human beings: old fashioned sacks of meat with smiles and unexpected phobias who don’t respond to their every question with the words That’s a very perceptive question, Dylan, and it gets to the heart of blah blah blah…
   
   If you’re out of school, think back to your own school days. What specific instruction, principles, or words of wisdom do you remember from your own classes? If you’re like me, you can barely remember anything: I know that school taught me certain habits of mind and an ethos around using inquiry to explore reality, but beyond that, I forgot nearly everything twenty minutes after the final exam. But I bet there are some people from your classes that you remember. Some of them could be your best friends today. You might even have married one or two of them. That doesn’t happen much in an online class, and it doesn’t happen at all with solitary LLM-driven instruction.

Just like most everyone else who works with a computer, I am facing a job that has changed radically. What I tried to communicate to students for the first 25 years of my career was that reading and writing are valuable, salable skills in their own right. I’m not so sure of that anymore: an LLM can write in any genre and on any subject better than a typical college graduate, and it has read–and digested–far more than any single human being could be expected to have read.

But having said all that, I believe a human teacher of reading and writing has something to offer students. Reading and writing are still the training regimen by which a person learns to think. Whether or not anyone ever pays you to write or read an argument, learning to make an argument yourself remains one of the most important things you can learn to do. Argument is the process by which you make your thinking clear to others, but just as importantly, it’s the way you make your thinking clear to yourself. However ChatGPT has changed things in the classroom, and will continue to change things, it hasn’t abolished this essential reality of our lives.