Our Deceptive Memory of Pentecost

On Risk, Reading, and ChatGPT

This essay appeared in the zine vol. iii. Order a copy here.

An AI generated image of Pentecost; courtesy of Up-To-Date AI Webdesign

The humanities, those swimmy-bordered fields outside the sciences, are shrinking both within the academy and without. College enrollment’s down and diving. Major schools are cutting or combining departments: Last year, West Virginia University proposed cutting over two dozen majors, mostly in the humanities. In North Carolina, humanities professors at public universities are now legally prohibited from receiving a (better-paying) “distinguished” professorship. And so on and on. Closer to the ground, I know buddies teaching literature speak of extreme absenteeism: Even when plenty of students are enrolled, lots are habitually not there.

Outside the academy, publishers still sell books, but the market has shifted hard to nonfiction; any fiction’s lucky to push a thousand copies. As far as criticism goes, the academic stuff is read only by those who have to cite it to pass the ecumenical circle-jerk of peer-review, and magazine criticism works as gossip-fodder for those in the industry, or as grant-app-padding for well-reviewed but under-read authors (aside from rare exceptions like, say, Christian Lorentzen or Leo Robson, whose incisive humor opens their essays to less specialized audiences).

I don’t say these things pessimistically. I don’t think they’re necessarily bad, nor do I think they’re particularly bad for me. But they need some explanation that isn’t just “folks don’t read like they used to” or “film and TV have taken over entertainment because people are lazy.” Those explanations might be fine. I don’t know. I don’t think about them much, because they don’t match the world my eyes and ears take in.

What my sensors do pick up is that there are fields, like literature and philosophy and history, that once occupied much more real estate on the collective farm of thought because they didn’t provide new worlds, like film or TV; or prescribe authoritative notions of how our world is, like the sciences in the era of “believe the science”; but they offered descriptions of worlds full of holes that readers could fill with their own, very personal imaginations—that is, worlds for their readers to create themselves. Reading a novel is playing God if you do it right. Same goes for a good philosophical book, an honest history. It’s intoxicating, that power, and addictive. But that drug’s street peddlers have seen a drop in clientele, and that’s the mystery here. How does a proven good drug become less addictive?

Teaching in the humanities often feels cruel: After knowledge of the canon, the profession’s most important skill is confident domination. That sounds harsh. I’ll explain. If I’m, say, a subcontractor with a saw, clearing a lot for construction, all truths involved in the question of whether I’ve done the task allotted me are clear-cut: If the lot is clear-cut, my contract is fulfilled. If I’m in a chemical lab, and the desired product has been produced, and it smells right and looks right and reacts right with the other reagents with which it’s to be mixed, God’s justice reigns on this here plot of lab. There are no questions. Truth is absolute.

Such professions are not affected, in the base-tasks that construe them, by the postmodern anxieties the humanities have wrestled for decades, now. The deconstruction of a tree changes only when a new type of saw comes to market, and the tree’s deconstructionists don’t give a chopped and quartered shit about the philosophy that subgirds the marketing of that new saw: If it cuts wood better—that is to say quicker, or with less waste of useful material, or with lower risk of injury—progress has been made.

But in the humanities—last week, my students turned in a rough draft of their midterm essay on Shakespeare’s Richard II, and I told one that I could tell he had used AI software to write it, and if he didn’t rewrite it, he would surely fail. “That’s not fair,” he said. Said it almost like a hiccup, and looked away from me like he was ashamed for having said it. And this particular guy was too smart not to be ashamed of that.

But I knew what his shameful hiccup meant. The class I teach is required for every undergraduate, so most of my students major in either psychology, biology, or computer science. In those fields, if there’s a tool that attains some desired result more quickly and with less effort, it simply is used. (Or so the thinking goes. Psychology still, somehow, thinks itself a science with ends as definite as cutting logs. Biology…) In any case, in those fields’ exams for undergrads, truth is generally Boolean: There are right and wrong answers. Ones and zeroes. And, as my students have always told me, in their high-school English classes, constrained by standardized state curricula if not standardized tests, the same sort of standard reigned.

So this student of mine, he was right, by the norms of the fields he and most of his classmates primarily studied and, maybe, those of his previous literature classes. To use a tool that does the job more quickly and reliably and with less effort than the older tool should be praised, not penalized.

I will henceforth ignore the fact that current AI-generated writing is simply bad, except on subjects so specialized as to exclude low-quality writing from the training corpus an LLM might use to counterfeit them, and that all the tech journalists who have written of the difficulty of distinguishing AI writing from human writing in the humanities have not thereby said anything about AI-generated writing but have very simply outed themselves as bad readers. These things are, compared to what I’d rather talk about here, painfully obvious, and to explain them further would insult the Cracks in PoMo readers and myself. I won’t do that to you or to me.

A couple semesters ago—after I assigned a memorization, then presentation of a speech from Macbeth—another student stayed after all the others had left, looking nervous, and asked me, “But Banquo’s name’s on this speech. You mean I should say his words?” This guy wasn’t dumb. I explained it to him, and he did a fine job, and—like the student I mentioned before—he wrote a fine paper on the play, one that achieved real personalization of thought.

But the same crisis had infected both of them: They weren’t used to being asked to express their own agency, and they weren’t used to the expressions of their respective agencies being valued more highly than the precision of a cutting-edge tool. They thought of themselves—at least in the context in which I met them—not as agents with valuable experiences and idiosyncratic wills, but as more or less competent operators of tools.

And the teacher’s job—my job—in this case, is simple hierarchical domination: To assert that there are standards. And those standards aren’t universal, they’re my decisions. And when the principal standard is personal, original thought, the humanities’ friction with the sciences (at least at that undergraduate level) becomes glaringly, at times painfully, obtrusive.

And all this, it’s the same mechanism that makes reading The Brothers Karamazov, or Difference and Repetition, or Shooting Niagara too risky for most folks of the social class that, back in the day, would have read those things: Getting into a novel or a good piece of philosophy doesn’t allow anything like consensus or the precision of a mechanical tool. The thoughts that result from it can’t be predicted. That’s part of a good book’s suspense-apparatus. The same way you can’t predict where all the characters and ideas will wind up at the end, you can’t predict what you’ll be at the end. That’s always been a risk. But, somehow, it’s become a more demanding one.

At the most basic level, the risk of reading and writing well has been intensified by economics: In a world in which it is no longer impolite to discuss politics and religion in public, but is increasingly impolite to refuse to do so, and (in the humanities, especially) one’s ultra-competitive livelihood may well depend upon doing so sensitively, we’ve got to guard our hearts and minds from ideas that, by some slip of the tongue, may escape over drinks with friends or colleagues and damage our cultural capital—hence the much-talked-about tendency, shared by folks across whatever political spectrum, to read and socialize within bubbles of similar-minded people. When every word may be a shibboleth, reading or writing things that give idiosyncratic meaning to words—like any great literature—becomes a terrifically dangerous pastime.

And that explanation works fine within a hegemonic cosmopolitan culture in which the various shibboleths and the cultural bubbles they defend are well-defined—that is, in the world of those already initiated. Nobody who has passed the social filters to be at a PEN America gala is going to mispronounce one of that institution’s shibboleths by accident—and, for the same reason, they’re generally not going to trust an LLM to generate their essays. But the incentive structures governing speech, and the reading and writing that inform it, are profoundly different in a classroom, and especially in the highly diverse classrooms of a major public university.

By diverse, I don’t mean ethnically, but culturally. There might be ten distinct races, even countries of origin, present in a room at a prestigious private college, but the incentive structures and selection filters that have landed someone in that room have generally ensured that there is precisely one culture present: Everyone knows the relevant shibboleths, though some may have learned them more thoroughly than others. Below the top tier of universities, however, the economic problem of professional shibboleth-maneuvering becomes a theological and linguistic problem traceable from the story of the Pentecost in the Acts of the Apostles to modern sociolinguistic research.

After Jesus returned and ascended to heaven, his remaining Apostles were in Jerusalem celebrating the Jewish Feast of Weeks. The others around them spoke a wide variety of languages, and the Apostles had an impossible but burning desire to share the news of Jesus’ return and ascension. In answer to their wish, the Holy Spirit descended on the wind and turned their tongues to fire, and they preached “in tongues,” so that everyone present understood them, regardless of their linguistic backgrounds.

The Pentecostal ideal, that there is some language of God, or of self-evident spiritual truth, has run through modern thought from John Dee’s claims to have rediscovered Enoch’s angelic language to the terminological obsessions of Kantian and post-Kantian philosophy to our contemporary saturation with ever-shifting political shibboleths. Back in my Evangelical church days, every discussion of apologetics began with some variant of the idea that if I spoke pure, Godly truth to a nonbeliever, the Holy Spirit itself would take over and make that truth apparent to them. In my academic career, I’ve heard the same idea repeated in both the explicit form of the French Realists’ fixation on finding le mot juste and the implicit form of the academy’s professional and political shibboleth-matrix, often derided by those outside the academy as either “jargon” or “political correctness” or “wokeness.”

But back in 2022, a group of linguists led by J. Michael Terry found that, even in such a seemingly culture-neutral field as math, students performed markedly worse when given questions that included grammatical constructions they hadn’t grown up with—even if they’d grown up speaking English, and even when their performance on other aspects of the test was equal. When the kids saw some grammar that they knew but hadn’t known as their own, their brains sent off error signals like the test had made a grammatical mistake.

In a humanities classroom, then—how many students grew up with Shakespeare’s idiosyncratic English? The test the classroom gives them, daily, is then, first, to switch off their brain’s error-alarms just enough to comprehend the canonical texts they’re given, many of which are canonical precisely because they developed new ways of using the English language. Second, they’ve got to keep other error-alarms running at compensatory high-alert so as to communicate to the institution’s representative (me) about that unfamiliar language I’ve handed them. So, often, what I hear from my students is the perfect opposite of Jordan Castro’s “unbearable rightness of professor-speak”: Instead of ending each statement with a consensus-hungry “right?”, they begin their sentences with “I don’t know if this is right, but…”

Since universities have treated accessibility to ever-greater numbers of students as an unqualified good in recent decades, and thus ever-more students from outside the culturally dominant groups raised with the operant shibboleths are sitting in classrooms where they need to know those shibboleths, at the same time that tuition increases have outpaced wage growth to a hideous degree, the idea of their future grades hovers over most students like some kind of Red Pen of Damocles. They need to make their transcript worth the debt they’re accruing for it, but their upbringing, including their primary and secondary education, has not given them the tools necessary to do so. So, like the Apostles, they seek a universalizer—either in AI-powered editors or text-producers or in pay-per-job essay services where a more advanced student makes some side-cash to write for them.

So that student who told me it wasn’t “fair” to penalize him for using AI to write his midterm was doubly right. By the standards of the Boolean-truth fields of his other classes and by the standards of a dominant linguistic ideology based upon the Pentecost and its precision-obsessed philosophical and religious offshoots, none of which accounts or allows for dialectic or idiolectic difference, he was right. Which makes my penalizing him for using an LLM to write his essay doubly cruel.

Now, none of this is to say that the standards—especially stringent—required of a humanities classroom are unjustified. Nor is it to say that universities should adopt more restrictive admissions policies (though such a response would certainly save lots of people lots of unpayable debt).

Ultimately, there is no “should” here. I’ve simply described what I see and hear: The humanities are shrinking at universities and, in other ways, suffering in the book market. Students are fearful, for economic reasons subtended by cultural and, at a still more fundamental level, theological reasons, to read and write risky works. Unlike the professors’ self-sure (but still pathologically insecure) reliance on the tag-question, “right?”, students’ habitually prefatory “I don’t know if I’m right, but…” belies the minefield of shibboleths—political, cultural, and epistemic—that hides beneath all the humanities’ surface-proclamations about “academic freedom” and the like.

In the end, and pointedly less precisely, there is a measure of good faith—the assumption that the folks hearing your voice will make an earnest effort to understand your words as though they were spoken by a bright, well-meaning equal—sorely wanting in the humanities, which the sciences don’t have to address to the same degree. In fields with no Boolean truth but, rather, rhetorical truths, free expression of ideas often runs counter to those same fields’ desperate degrees of economic competitiveness and reliance on institutional patronage, especially for students new to those fields and fretting day and night about whether they’ll ever be anything but new to them.

And so what we’re left with is a problem of scale that I can’t change. Likely one that none of us can change, but which we can only hope to describe faithfully, then respond to usefully, but personally, not universally: The incentive systems and cultural filters of the humanities—be they academic, literary, or whatever—are simply as they are. But, though I can sketch them at wide-scale and tell my own experiences, every region’s—and every room’s—cultural mélange will offer subtly different problems, thus different solutions. So I’ve done my damnedest to describe, but not to prescribe any means of changing anything, since such a prescription would reduce me to thinking at the scale of a state or some other abstract power. A futile pursuit, in the best case. But that next-to-last paragraph’s notion of “good faith” comes close to a prescription and approaches the level of solution in many cases.

At the most extreme length of interpretation, that exhortation to “good faith” entails, not just a trite reminder to listen openly, but that every speaking, reading, and writing problem is a translation problem, not just metaphorically but as seriously and completely as any rendering from Portuguese or Tlingit to English, except that, within a language, the necessities of translation are much subtler, more difficult to satisfy, more a mystic than a scientific exercise, more like the sudden insight of the Apostles at Pentecost than anything registerable by Google Translate or DeepL.

Jonah Howell writes and teaches literature in New York. His play Lillita and the Tramp will open in May at the Hudson Guild Theatre.