An Argument about Teaching, Technology, and Being a Tool-Using (and Tool-Consuming) Species

[Cross-posted to Front Porch Republic]

Joel Mathis, a fellow Kansas writer and friend, recently reflected on joining what may be something of an anti-AI bandwagon, and what it may mean for him professionally. Those who know me know that I’ve been on the bandwagon for a while (I don’t continue to use a flip-phone and occasionally wear a t-shirt celebrating the Butlerian Jihad—“Thou shalt not make a machine in the likeness of a human mind”—from the Dune legendarium for nothing.) As a teacher, I’ve converted most of my essay-writing assignments into in-class Blue Book journaling assignments (which, of course, had never truly disappeared) over the past two years, and I’m mostly happy with that change. But with the start of the fall semester, with many of my fellow academics sharing their thoughts about how, or to what degree, their lessons plans are (or are not) incorporating the incredible advances in artificial intelligence, I figured I ought to think through and put down my position about it all as well. Thus, the following long statement, which I included at the end of all my course syllabi this year.

 A Manifesto on AI

In this course, you are strongly urged, and assignments are designed to require you, to avoid ALL use of generative AI tools (that is, those which generate usable images, summaries, or written statements in response to prompts) insofar as your reading and writings assignments are concerned. Let me explain why.  

Your education here at Friends is likely made possible by, and guided by, several things: parental support and expectations, athletic or artistic or academic scholarships, social pressure, personal or religious motivations, and more. I would like to add one more element to the mix: the ideal of humanitas. It is because I, as the instructor for this class, take this ideal seriously–and I want you to take it seriously too–that I am imposing the rules regarding AI mentioned above.

The term humanitas goes back to ancient notions about education, focusing on how a person can become a responsible, virtuous, independent human being. It is the origin of the word “humanities,” which includes history, literature, philosophy, language, and what are sometimes referred to as “the liberal arts.” Over the millennia, the notion of humanitas has been revived and re-interpreted many times. In this class, the ideal stands behind what is taught in the sense that I believe that one of the core elements of being a responsible, virtuous, independent human being is understanding the nature of, and the operations of, our shared civic life. Gaining that understanding, however, will be undermined if you make use of AI tools to avoid the assigned reading, or to avoid the struggle of communicating what you have learned from your reading in written assignments, because the kind of understanding relevant to both the liberty and the liberality of the humanities in a practical and a practiced one.

Reading AI-generated summaries of the assigned reading is therefore NOT an acceptable way to prepare for class or to engage the class material. This class isn’t about providing you with information to regurgitate. It’s about getting you to closely read and wrestle with hard questions asked by the text: questions that have challenged people for thousands of years, and questions that are (or ought to be) central both to how you conduct your private life and to how we organize civic life. Relying on an auto-generated summary of a text inevitably leads one to miss out on details of the thought processes that led the author to put their conclusion the way they did, and thus to miss the opportunity to think alongside the author’s arguments and to critically interrogate them. 

Using generative AI tools to create all or part of your writing assignments is also, therefore, NOT an acceptable to way to demonstrate your engagement with the class material. Class writing assignments are not busy work. They are there to force you, through the labor of choosing words and connecting sentences, to figure out not only how to communicate the understanding that you have come to of whatever author or problem you have been asked to study, but also exactly what your understanding of that author or problem actually is. A civic environment of responsible, virtuous, and independent people will be one filled with inevitable disagreement and discovery; developing the skill of communicating, through writing, one’s own understandings is essential to being able to both engage with disagreements and learn from others’ (and one’s own) discoveries.

AI tools are obviously of enormous benefit in all sorts of practical fields, and learning how to make use of them in certain research contexts may well be vitally important to your professional future. Just academically speaking, while using AI tools to answer specific research questions, build research bibliographies, or check research citations has some real limitations, developing the skills necessary to do so will likely soon be as routine an expectation as knowing how to use Google Docs or Word. But for this semester, in this class, you are invited, as unusual as it may sound, to take the ancient ideal of humanitas to heart. Here, you will be expected to wrestle with terms and concepts, and to learn to discuss them and write about them, without too much reliance upon powerful technologies which can do, seemingly more quickly or comprehensively, things which you already have—or already have the need to develop—the ability to do. I hope you will embrace this expectation and gain from it.

A Debate over Being Human

Michael Austin, an old friend of mine and another academic, currently the provost at Snow College in Ephraim, Utah, shared a very contrary perspective in a two-part essay of his, humorously subtitled “How I Learned to Stop Worrying about Skynet and Love ChatGPT.” It starts with him confessing that, though an “Enlightenment humanist” at heart, someone whose most important work of scholarship presumes a worldview into which AI probably has no place at all, he had, as a university provost, “a professional responsibility to implement both classroom AI policies and curricular guidelines.” And so he, being the brilliant pragmatist I’ve always known him to be, dove into the latest pedagogical literature, and figured out a way to integrate AI tools into the teaching of the liberal arts, a way that he found both personally fulfilling and filled with intellectual potential. As he concluded his essay:

When AI extends our ability to do things that were previously beyond our reach, it also requires us to acquire new competencies to turn those possibilities into finished work. When teachers understand this, and when students grasp what they can actually accomplish with the help of AI models, classroom assignments can become a gateway to releasing human potential, unlocking knowledge locked away by limitations AI can help us overcome.

As a longtime writing teacher, I am quite sure that AI will be the end of the thesis-driven five-paragraph essay assignment. I do not mourn its loss. Five-paragraph essays do not occur in the wild, and thesis statements were never anything more than rhetorical training wheels. Teachers who still give such assignments deserve the AI-generated drivel they will almost certainly receive.

But look what we can now put in its place. How about a final assignment to write a full-length opera or a symphony, or a history of a nearby town? With AI, freshman English students might create a graphic-novel version of Moby-Dick or Great Expectations. Or how about a full video version of The Iliad or Dante’s Inferno? AI can help students do things like this.

I don’t begrudge my friend’s enthusiasm (AI enabled him to achieve his long-abandoned goal of being a political cartoonist, after all). And like him, when it comes to teaching essay writing, I also don’t feel the need to mourn any particular pedagogical form, or really any particular teaching instrument. But unlike him, I look at the idea of giving my students—especially beginning ones—AI-dependent assignments and incentives in the name of encouraging them to produce whole histories or symphonies, or create whole videos or graphic novels, as an approach that misses the very foundation of a liberal arts education.

At its heart an education in the human arts and sciences involves, I think, a very present, very embodied, very non-automated engagement with art, ideas, and the written word. To fail to go through an educational formation characterized by such interpersonality is to be diagnosed by a doctor who never actually sees or speaks with you, or to eat a meal prepared by people who never actually handled the dough, the meat, or the vegetables. Obviously, such formational or educational—or diagnostic or nutritional—processes are not entirely all or nothing, entirely black or white. As a college professor, I’ve long since made my grudging peace with (though thankfully I’m still mostly available to avoid) distance and online learning as an unfortunately necessary compromise, thanks to the undeniable good of increasing access on the one hand and dealing with the challenge of affordability on the other. Similarly, we’ve all found ways to accommodate speaking to a nurse on the phone or cooking with canned food from a grocery store. Still, when you talking about as important an ideal as humanitas, I think the reflections and warnings of such thinkers as diverse as Amy Gutmann (who presented education as involving participatory social reproduction) or Michael Polanyi (who presented education as depending upon tacit exchanges between the teacher and student) or John Dewey (who presented education as a “continuing reconstruction of experience”), or Martin Heidegger (who presented a “real or genuine education” as reflexively engaging students in those practices and understandings wherein they already dwell) need to be taken to heart. Whatever compromises the rival goods of our heavily technologized world require of us, AI-enabled steps away from carefully helping students construct and make responsible, virtuous, and independent use of those cognitive and cultural competencies already present in their lives would be, I think, a failure of teaching.

Michael and I ended up going back and forth about this issue at some length. Initially our dispute focused on the issues of scale—it is one thing, he correctly observed, to restrict the use of AI tools in a particular classroom in the name of maintaining an environment of closed and close instruction, something that may be, depending on the circumstances and the pedagogical aims of the teacher, perhaps even admirable. (Michael, to his great credit, has regularly taught classes at Central Utah Correctional Facility—a prison, in other words—and admits that the controlled reading and discussions possible in such environments allow for some superb educational outcomes). But one professor’s preferences, he insisted to me, can’t speak to the needs of an entire institution, especially one dealing with state mandates in regards to the integration of AI into teaching practices. I don’t disagree with any of that; as I emphasized in my manifesto, I was speaking to my students solely about “this semester, in this class.” I freely admit that I am developing this argument solely with my own particular instructional stewardship in mind.

But before long our argument had become one of fundamental assumptions about what it means to become an educated human being. Michael took the position that, since human beings are, evolutionarily speaking, essentially tool-making animals, to make the project of education one that requires, in the name of some humane ideal, the exclusion of whatever tools happen to be part of the world we and our students actually inhabit is actually “profoundly anti-humanistic.” My position was, to say the least, quite different.

I agree that human beings are, evolutionarily speaking, tool-making and tool-using entities. But why does it therefore follow that attempting to construct a learning environment that explicitly rejects the employment of any particular set of contemporary tools which humans have invented is unhuman? That would imply that any construction of the human—and, I would say, of the humanitas ideal as well—has to be solely centered on the evolutionary process itself. But that is simply not the case. Human beings are tool-using animals, yes, but they are also dialogic, discerning, distinguishing, debating animals. The whole history of human civilization may not be reactionary and Luddite in character, but throughout the history of human civilization there have nonetheless been innumerable instances of resistance to, reconsiderations about, and reform of our tools and the processes that have been or are being constructed through and with them. Sometimes we construct forms of intercourse and learning and development that are consciously, even aggressively, progressive in that they seek to transcend the forms that came before. But sometimes we also construct forms of intercourse and learning and development that purposefully embrace what are understood (sometimes, I would argue, correctly) as the accomplishments and virtues of previous forms and therefore strives to preserve them.

For example, I doubt anyone who lives in the wake of the Enlightenment can believe that the restrictive Chinese imperial examination system was the peak of human educational development—I sure don't. But by the same token, it’s not as though those 1000+ years (which gave the world Luo Guanzhong’s Romance of the Three Kingdoms, the poetry of Li Bai, and the development of Neo-Confucianism) were all "anti-humanistic.” So while I am happy to acknowledge the validity of many of the criticisms of the pedagogical choices I and others are making, to so firmly stipulate the evolutionary and cognitive centrality of relying upon tools so as to label the decision to reject AI as a learning tool (and thus the decision to pedagogically control its role within the classroom) as therefore anti-human goes, I think, several steps too far. It is as much a humane reality to seek new tools for engaging with art, ideas, and the written word, as it is to decide that some tools inherently disrupt the practices necessary to understanding, and to therefore avoid them.

Tools and Technology, Within and Without

I’m comfortable with the position that I’ve taken—which, it’s worth noting, is completely independent of perhaps even more immediate environmental and ethical and apocalyptic arguments against the normalization of AI in schools (and our lives generally). But Michael wasn’t wrong to push me very specifically about tools and their place in education—or indeed, just developing the competencies that I would consider essential to the project of humanitas in the first place. Plato attacked writing, after all, as a tool that would disrupt the memorization and orality he understand as essential to learning. Would I attack wearing eyeglasses because they are a hack, a short-cut, an augmented engagement with art, ideas, and the written word?

I think it’s possible to come up with a distinction between AI and eyeglasses that doesn’t just depend upon the later having been present and in ordinary use throughout human societies for a millennium, while machine-learning models, the core of artificial intelligence, are—depending on your definition—less than 50 years old. Among other things, I believe one can rigorously defend the construal of this distinction as being between a tool that the individual uses to do a thing the person has learned how to do (using eyeglasses in order to better see the paper or the screen upon which one is drawing or using software to construct comic strip panels) and a tool that the individual uses to do a thing that the individual has explained to the tool how to do for them (explaining to ChatGPT, perhaps through an admittedly demanding and potentially engaging and creative learning process, exactly how do draw or construct comic strip panels). But this distinction, like any distinction made in reference to human processes of educational development, will likely always have to be subject to regular reconsideration. As, for example, I’ve reconsidered—if not entirely changed my mind about—GLP-1s.

Listening to a discussion about GLP-1s (Ozempic, etc.) conducted by Ezra Klein a few months ago is what put me on the path of n at least partial reconsideration of these even-newer-than-AI technologies, pharmaceuticals that control appetite and otherwise “hack” certain receptors in our brain. Originally, I was profoundly critical of these drugs; I was deeply, philosophically, hung up on the idea that there is something essential to our ability to think and reason and feel as human beings that cannot be extricated from our spatially and socially embodied experiences. Consequently, I was just plain suspicious of technological interventions into our embodiment for the sake of exercising some kind of control over it, partly because such "control" seems to me an attempt to actually escape the labor of our embodiment, to instead statically align it with some given set of utilitarian advantages or procedural expectations or economic demands or cultural preferences rather than anything actually ongoing and human. Nearly two decades ago I pontificated about this, comparing using marijuana to elective cosmetic surgery, the former positively and the latter negatively. At the time a friend of mine pushed back against that framing, using nearly the exact same eyeglasses example—would that tool be a "technological intervention"?—that Michael used on me. I thought I could make that distinction at the time, and I still think I can make it today—but the surety of that distinction is not beyond ongoing debate.

Another friend of mine, Damon Linker, a fine writer who has struggled with some health concerns, recently reflected on his doctor-recommended use of this miraculous new technology, specifically Wegovy, which has helped him lose 55 pounds:

I learned to think about the world using categories of moral analysis descending from the traditions of classical philosophy and Christian theology. To speak very roughly, both view obesity as following from gluttony, which the first describes as a vice and the second describes as a sin. . . . To avoid the vice of gluttony one must practice and demonstrate the virtue of fortitude or steadfastness, which involves an act of self-mastery, the overcoming of one’s craving for the pleasure of eating for the sake of both the intrinsic psychological rewards of self-mastery and the consequential good of bodily health.

Looked at in these terms, my failure to lose more weight on my own, even after I learned about the threat my obesity posed to my cardiovascular health, was a defect. I was incapable of mastering my craving for the pleasure of eating. This is quite common, especially in our time and place, with its abundance of hyper-palatable and unhealthy processed foods. . . .

That’s where the Wegovy comes in. Suddenly a portion that would have left me craving more food left me feeling fully satisfied and not at all tempted by the thought of eating more. In fact, the thought of eating more made me feel vaguely ill. Which is just another way of saying that the drug removed my craving for more food—the very thing that needs to be overcome and mastered in order to resist the vice of gluttony.

This means that a world in which everyone who’s overweight or obese has access to a GLP-1 will also be a world in which this kind of self-mastery and self-overcoming no longer need to be cultivated. Is that a bad thing? I suspect it is. . . .

Those are the troubling thoughts that have come racing into my head these past few months when I’ve glanced, smiling, at my noticeably smaller self in the mirror.

Damon’s experience sat alongside Klein’s and the provocative discussion he had conducted about the human body and brain, about how the best way to understand GLP-1s is to see them in terms of a tool that intervenes in our own evolved programming, in my mind for weeks. It is undeniably the case that our industrialized and financialized agricultural and food production systems have ended up creating environments which are profoundly unhealthy for our bodies, and profoundly difficult for our evolved mental scripts to deal with. And when we think about that problem in a world characterized by questions about artificial intelligence, questions that force us to struggle with what we think we know about consciousness, about genetic inheritances, about our own ability to make decisions which distinguish ourselves from the tools we make, use, and bring into our lives, I’m just not sure we can avoid—or at least, I’m not sure I can avoid—acknowledging that we may not have a good grasp of what “self-mastery” today actually means. I am confident that I can distinguish between eyeglasses as a tool that enables individuals to seek mastery in tasks that they do and AI as a tool that enables individuals to seek mastery in getting a computer to do tasks for them. But am I just as confident that GLP-1s are a tool to avoid constructing virtues that we should be bodily striving to master? Might they instead be tools that enable individuals to pursue a mastery over the multitude of ordinary virtuous tasks that our gluttony-enabling and incentivizing world would interfere with? I’m not sure I’m convinced, but I genuinely wonder.

Stumbling Forward Technologically

The easy, and avoidable, rejoinder to this kind of meandering quasi-Luddism is pointing out that I wrote it on a computer (one of which, so far as I know, Wendell Berry still doesn’t own). I use a laptop, my wife and youngest daughter regularly use an e-bike, and I don’t go anywhere without my blood pressure medication and my Excedrin. I appreciate my arguments with my friends over technology, because being reminded of my own existence as a tool-using creature is a good thing. Among other things, it indirectly emphasizes to my thinking that my own creatureliness, and the humane and tactile work I hope to do in an embodied way in the classroom, is itself mediated by the intervening and augmenting technological tools that I unthinkingly (zuhandenheit, if we prefer Heidegger’s term) use every day. And this recognition leads to humility, some of which I hopefully express humorously.

For example, this semester I have embraced as a supplement in all my classes a new version of Turnitin, the ubiquitous plagiarism-spotting electronic paper submission system, called Turnitin Clarity. The idea behind this supplement is that I can, by requiring students to use this as their word-processing program for essays and response papers, police any AI use they might engage in (and distribute consequences accordingly). By letting them know all about this beforehand, my hope is that they’ll be incentivized to accept the ideals I’ve put in my Manifesto and avoid making use of any generative AI in their work. What will be the outcome of this restriction? I don’t know, but I’m anxious to see. The idea of using a technological innovation to stand in the way of another technological innovation may seem counterintuitive, and certainly in some cases it would be so (that’s where the in-class Blue Book assignments come in!). But as the new semester begins, I’m happy to attempt to make my way through all these challenges which stand in the way of achieving some kind of humanitas as best I can. What else can we do except continue to engage, assess, and re-assess as necessary? After all, wasn’t it the temptation to make use of tools that would allow us to skip over, even for seemingly excellent reasons, all this constant (yet constructive) work of technological (re)consideration which got us here in the first place?