[.article__paragraph--cap][.small-caps]In the fall of 2024,[.small-caps] the Office of the Chaplain at the university where I serve organized a reception for campus religious leaders with the Dean of the School of Engineering. The dean is a pioneer in the field of artificial intelligence, and the topic of discussion was the role of AI in higher education. He began with opening remarks that remain some of the most sober, balanced, and thoughtful reflections on the nature and limitations of AI that I’ve heard from someone in the field. The dean clarified that AI doesn’t actually have internal, conscious thoughts or judgments. If you ask an AI model, “What does coffee taste like?” it can tell you, “Coffee tastes bitter.” But it cannot experience coffee’s bitterness. Consequently, the dean argued that AI cannot by its very nature make value judgments – which is why, he very generously concluded, it can never replace people like us, the campus religious leaders who help students navigate questions of right and wrong. [.article__paragraph--cap]

You might be nodding your head in agreement right now; I was at the time. But then in the Q&A we inevitably came to the question of AI’s use in the classroom, and this is where the conversation started to go off the rails. To my shock, after such a measured approach to AI in his opening address, the dean said something that was, to my mind, a mix of naivete and insanity. Why shouldn’t we, he excitedly proposed, let students use AI to write their papers and assignments for them, at least in rough draft form? Imagine the time and energy that could be saved, which in turn would allow students to focus on higher-order tasks, like creative thinking! Moreover, the dean continued, he saw no reason why very soon all peer-reviewed academic papers couldn’t be purely AI-generated for similar reasons.

I share this anecdote not because I hold this dean in contempt but precisely because his fundamental approach to AI was so thoughtful and nuanced – and so markedly different from the braggadocious and philosophically naive tech entrepreneurs of Silicon Valley – and yet he nevertheless wound up advocating for a future that I instinctively regard with horror. And this is why I left that meeting profoundly unsettled. It was the first time I was really forced to reckon with whether we are truly headed for a future where writing, all the way up to and including published academic papers, will be AI-generated. And if so, what does this mean for the future of education? Or for the very nature of human knowledge itself?

My struggle was heightened in the months that followed by speaking to people who use the latest AI models in very effective ways. The economist in my community group at church finds AI to be indispensable in her research; a scientist I know doing cancer research has found that AI can discover new forms of protein folding; a fellow dad uses AI to turbocharge his work in data analytics; my own wife, a consultant, now uses the latest Claude model on a daily basis for her work. I could add more. These stories made it impossible for me, despite my serious reservations about this technology, to be a full-on AI Luddite. It is undeniable that AI is going to be a permanent and integral part of work and research going forward. The question I felt compelled to answer was: Does my gut reaction that AI threatens something fundamental about being human have any rational basis?

What is it really that is lost by letting AI compose assignments, essays, and academic papers for us? Are those with reservations about AI just a twenty-first century version of the curmudgeons from the last century who resisted giving up their typewriters for word processors? If those with instinctive unease related to AI are to be taken seriously, we need to carefully define what is uniquely human about creative and intellectual endeavors – and therefore, persuasively demonstrate what important things we will lose if we surrender these things to the thinking machines.

[.article__paragraph--cap][.small-caps]The first place I turned[.small-caps] to evaluate the merits of my intuition was to one of the most celebrated essays on the relationship between writing and thinking, George Orwell’s seminal “Politics and the English Language.” First published in 1946, “Politics and the English Language” is eerily prescient about the dangers of severing writing and conscious human thought from each other.[.article__paragraph--cap]

In it, Orwell critiques a style of writing that emerged from mid-twentieth-century mass media culture, especially mass political writing, a style Orwell says “at its worst does not consist in picking out words for the sake of their meaning and inventing images in order to make the meaning clearer [but] consists in gumming together long strips of words which have already been set in order by someone else, and making the results presentable by sheer humbug.” That final sentence sounds like a description of AI-generated writing. And as with using AI, Orwell says that the chief attraction in this way of writing “is that it is easy.”

He gives several examples of the kind of writing he means, one of which comes from Professor Lancelot Hogben (Interglossia):

Above all, we cannot play ducks and drakes with a native battery of idioms which prescribes egregious collocations of vocables as the Basic put up with for tolerate, or put at a loss for bewilder.

Beyond its “avoidable ugliness,” this passage has two principal vices according to Orwell: stale imagery and imprecision. “The writer either has a meaning and cannot express it, or he inadvertently says something else, or he is almost indifferent as to whether his words mean anything or not.” The word one might use to describe Hogben’s writing is slop.  “[T]he concrete melts into the abstract and no one seems able to think of turns of speech that are not hackneyed: prose consists less and less of words chosen for the sake of their meaning, and more and more of phrases tacked together like the sections of a prefabricated henhouse.” Which again, sounds eerily LLM-like.

But, says Orwell:

A scrupulous writer, in every sentence that he writes, will ask himself at least four questions, thus:

1. What am I trying to say?
2. What words will express it?
3. What image or idiom will make it clearer?

And he will probably ask himself two more:

1. Could I put it more shortly?
2. Have I said anything that is avoidably ugly?

The scrupulous writer, in other words, far from being “indifferent to whether his words mean anything or not,” relentlessly interrogates what he means to express: each word, phrase, metaphor, and sentence structure is intentionally chosen to communicate, in the clearest and most cogent manner possible, what the writer wants to say. And this, in turn, forces the writer to think carefully about what it is he actually means. When attempting to clearly set down his train of thought in precise terms, the scrupulous writer will be confronted by where his thinking has been vague, muddled, or contradictory, and the task of composition will compel him to make his thinking more cogent as a result.

But the modern writer is not obliged to go to the trouble of writing clearly and thoughtfully, Orwell says. You can surrender to the ease of “throwing your mind open and letting the ready-made phrases come crowding in. They will construct your sentences for you – even think your thoughts for you, to a certain extent – and at need they will perform the important service of partially concealing your meaning even from yourself.” If this was true of the age of mass media, it is even truer in an age where AI can sever writing from conscious thought altogether.

And the danger of such slovenly and careless writing, Orwell argues, is that it inevitably leads to slovenly and careless thinking. Just as a “man may take to drink because he feels himself to be a failure, and then fail all the more completely because he drinks,” so too writing “becomes ugly and inaccurate because our thoughts are foolish, but the slovenliness of our language makes it easier for us to have foolish thoughts.” And this, Orwell warns, has political ramifications, as the title of his essay suggests. The mass production of slop is connected to what Orwell would later describe in Nineteen Eighty-Four as “doublethink” and “newspeak” – smooth, inane, and fundamentally manipulative forms of language that conceal more than they reveal.

[.article__paragraph--cap][.small-caps]The conclusion we should draw[.small-caps] from “Politics and the English Language” is that human cognition (and therefore human learning) is an organic process, in which the higher-order capacities (creativity, conceptualization, sound judgment) are inexorably erected on the foundation of lower-order capacities (like reading, researching, and writing). Remove the lower-order capacities and the whole edifice collapses.[.article__paragraph--cap]

Here’s an example of what I mean. Last year my eldest son began learning to play violin. The initial lessons involved tasks like practicing proper posture, proper bow hold, and how to hold the violin. His first music lessons involved clapping his hands to different rhythms, and his first “recital” involved no actual music at all. He thought that all of this was, in his exact words, “boring.” But of course, all these seemingly pointless tasks are not pointless at all. Learning proper posture, bow hold, and rhythm are foundational for being able to accomplish the “higher-order” task of playing Beethoven’s Violin Concerto. In fields as diverse as music or sports, mastering seemingly tedious fundamentals is crucial to developing higher capabilities. The same thing, I would argue, is true of writing and scholarship.

And this led me to discern the fundamental error that afflicts the enthusiasts of AI-generated writing, which is a basic confusion about the very nature of education and human intellectual activity itself. Such people believe that education and intelligence are properly concerned with “information” and “data.” In this theory, AI is a tool that augments human information processing. But this is based on a flawed theory of the human mind. My son could never “download” his elementary Suzuki lessons to his brain, much as he might like that. The neural connections necessary for playing violin well can only be learned from habit-forming practice.

I would argue that the mental habits we cultivate are, ultimately, far more important to education and intellection than the “information” we acquire. The habits necessary for the achievement of true human knowledge can be called intellectual virtues. We typically think of virtue as pertaining to moral actions, like acting justly or charitably. But a virtue is broadly any habit that leads us toward the obtainment of some good, and intellectual virtues are habits of mind that lead us toward the obtainment of truth. So, the next authority I turned to in my quest was the man who almost single-handedly resurrected the concept of virtue in modern Anglophone philosophy.

Widely considered one of the most influential English philosophers of the twentieth century, Alasdair MacIntyre published his most acclaimed book, After Virtue, in 1981. MacIntyre defines a virtue as an acquired human quality, the possession of which enables us to achieve certain goods but the lack of which effectively prevents us from achieving those same goods. To give a concrete example, the good of human friendship can only be realized when two people possess virtues like loyalty, love, generosity, and honesty. People who lack these virtues by contrast – who are mean spirited, selfish, and untrustworthy – cannot, by definition, achieve friendship in the truest sense. That’s because a virtue is not only a means ordered to some end, but one in which the end is in some way intrinsic to the means itself – just as friendship is intrinsic to the act of being loyal or loving or generous to another person.

Like friendship, truth and knowledge are goods that are essential to human flourishing. And I would argue that writing is a practice through which we acquire those intellectual virtues that are internal to and partly definitive of the practice of writing itself – virtues like patience, discernment, sound judgment, and careful thought – and which in turn are necessary for acquiring the goods of truth and knowledge. (There are other such practices, like reading and research, though these are activities perilously threatened by AI as well.) And although this language is absent from his essay, the cultivation of intellectual virtue through good writing (and the cultivation of vice through bad) is essentially what Orwell describes in “Politics and the English Language.”

For even the greatest author, the art of writing well is a perpetual challenge that refines our ability to think. In his poem “East Coker,” the great Anglo-American poet T. S. Eliot laments that he has spent the last twenty years “trying to use words, and every attempt / is a wholly new start, and a different kind of failure.” He captures the agonistic aspect of the art of writing by describing it as “a raid on the inarticulate / With shabby equipment always deteriorating / In the general mess of imprecision of feeling.” This may sound frustrating, but it captures the important truth that writing is an essentially challenging activity. Essay, after all, originally meant “to make an attempt.”

And shirking the challenge to “essay” by employing AI is not “efficient”; it is the vice of intellectual laziness. Circumventing that challenge by having an LLM write an essay for you is just as detrimental to the cultivation of human knowledge as it would be to friendship if you circumvented a challenging conversation with a friend by giving him an AI chatbot to talk to instead.

And all this is why the engineering school dean’s proposal is so seriously flawed. The high schooler, university student, or young professional who uses AI to compose essays and reports so he can skip ahead to higher order tasks will fail to cultivate the intellectual habits that are necessary if he is to become a good conceptual thinker in the first place.

[.article__paragraph--cap][.small-caps]I have anecdotal evidence[.small-caps] of this from my wife’s consulting work. When she uses Claude, it produces spectacular results. But when her younger analysts try to use it, it produces slop. The difference is that my wife spent much of her career without AI, so she was forced to hone the requisite mental habits for her job, while her analysts were not.[.article__paragraph--cap]

Certainly, I am not advocating for a Butlerian Jihad against data centers. I have no objection in principle to the widespread use of AI to perform many tasks, even if such uses displace certain skills (though care must be exercised even here). But I am convinced that using AI infringes on something essential to our humanity when it encroaches on those practices that are essential to cultivating intellectual virtue and thus promotes intellectual vice instead.

Unfortunately, our culture is so dominated by what MacIntyre elsewhere described as the “goods of effectiveness” – goods like power, wealth, and yes, efficiency – that these are now the only goods we seem to recognize. Lost are the goods of excellence, which are less about “productivity” and “optimization,” and more about the cultivation of virtue and the formation of character. And it is the latter that most truly and fully actualize our humanity. Our blindness to such goods, and denigration of the practices associated with them, will prove catastrophic if left to continue. If we outsource a significant share of human writing to AI, then I predict that over the next several decades the already perilously poor writing, reading comprehension, and critical thinking capacities of the American public collapse completely – and, as Orwell warned, the political implications for our society will be dire.