Stalking Aristotle in AI

Part I: The New Cave Dwellers

There’s no denying that the generative AI epistemic period we’ve entered is the most intoxicating cave shadow show yet. Our search for answers online yields truths, facts, and findings dazzling in their split-second amassing, in their plentitude and seeming certitude, in their ransacked-Borgesian-library unassimilability. But for many who’ve acquired the habit of interpretation, in my case, from lifelong bibliophilia and doctoral study, we must plumb the underlying conditions of this strange new knowledge ecology—the invisible and possibly intractable rules of online rhetoric. The new AI overviews in particular can have a kind of foreclosed-forlorn quality. Small symbolic sarcophagi confront the truth-seeker and fact-finder. Someone, rather, something has already done the scouring of what is called, spookily, “corpora” (machine-readable bodies). Computational heuristics have already found the good-enough (which is sometimes very good indeed). The human-to-human roads—online’s Blue Ridge Parkway, Devil’s Highway, Routes des Grand Crus, Manali-Leh Highway—have lost their signposts, and most are rerouted to the AI superhighway, far less scenic and meandering, requiring less gas. And the things-in-themselves to which the instantaneous “truths” correspond are receding further now, and with synthetic data, will be “increasingly refined approximations of approximations.” Inside of our flickering cave is AI’s own shadow show.

High Human Style

The question of how we know truth, how we construct it, interrogate it, come to believe it, and persuade others to as well, is an ancient one. Today’s truth claims have long been buffeted by a ceaselessly circulating (late capitalist) logic, and yet, in digital terms, we’ve witnessed the sudden, confounding, illogical damming of an entire content-publishing ecosystem of .com flows and .edu eddies—Anna Kornbluh’s blockade principle. If truth eluded us in the unending flow of feeds and streams, links and likes, it now elides us in the terminal summary. Even before the arrival of the search(ing) zero-click cul-de-sacs of 2024, we witnessed social media’s reversion from its early conversational days to the deculturalizing ideology of broadcast mode. These search and social media dynamics have furthered the dilapidating dialogic surrounding knowledge production and dissemination online—cognitive, social, rhetorical, institutional.

As the online symbolic economy slides into an ever more totalizing partnership with computation, often in lieu of human deliberation, and with AI as the ultimate arbiter of credibility and influence, some of us are searching for ways to share our ideas with other human beings in—let’s call it—High Human Style. A tautological challenge definitionally, it merely wants to be recognizably human. To that end, some have suggested typos should be retained, but this seems to me too much like the dinner guest who deems his dyspepsia the best means of communicating appreciation to the chef. Others have recommended the strenuous exercise of taste in our rhetorical and creative expressions, but this could be said of, say, the High Renaissance period of the early sixteenth century, so hardly uniquely ascribable to our current period, and begging the question: what is tasteful? Do we mean practical judgment or aesthetic discernment or both?

In a fine bit of irony, High Human Style’s compulinguistic features are immediately recognizable by AI: high entropy—unpredictability in information theory, and short stylometric distance—authorial singularity (of the “lower case” variety) across multiple texts. Along the same lines but in more quotidian human-literary terms, there is universalized cliché avoidance, the urbane use of metaphor, and a mastery of syntactic exoticism: our composition-class lessons writ large as noncomputational defense. Yet—and here’s where the plot thickens considerably—we must accommodate the AI overlords in some manner, by some method, lest our communicative artifacts suffer the ignominy of digital invisibility.

To find HHS, there are books, plural still, not (yet) Kevin Kelly’s liquefied omnibook, and a single author’s entire oeuvre, read exhaustively because the persona and conversation with that mind is largely the point. There are paywalled articles (always already scraped) whose disciplinary rhetorical features remain indigestible roughage for the AI maw. Podcasts present to generative AI the most vexingly indexical of human traits: the actual human voice, in a room somewhere. And there are the disintermediated freemium spaces—like this one, where voluminously developed paragraphs are, for better and for worse, the slow boat to citation success. An archipelago of rhetorical refuges afloat in a deteriorating rhetorical commons, where persuasion, judgment, authority, credibility, truth, and, hopefully less successfully, High Human Style, are adjudicated by computational systems.

From Detection to Dialectic

Most of us have been eyeing the sharks circling the shorelines for a while now. Dread of awakening, body-snatched, in the algo-accepting panopticon is everywhere. Rarely a day goes by that I don’t read about how the ubiquity of LLM-predicted prose has produced an uncontrollable detective reflex among many writers, journalists, linguists, educators, and professional POVists, all searching for HAL 9000 twitches and glitches, small lassitudes of reasoning, or—could it be?—a trace in the (possibly…probably AI-generated) essay of the real-world-rounded judgment that comes from, you know, life out here among the embodied.

Provenance detection can be diverting but does not suffice. With a radically new epistemic regime, we need to be far more deep-structural in our analysis of the new persuasion frameworks. We need to seriously size up the new epistemological arbiters and make them less alien to us, or, as many have suggested, make ourselves more unclassifiably and irreducibly human. (Wasn’t this always our task?) Or we need a radically new dialectical synthesis of the two.

Janus-Faced

I live the fraught dialectic of our AI era professionally. As a consultant who helps companies, leaders, and cultural institutions attain and maintain cultural authority, I focus on challenges unique to the thought leadership industrial complex. This has been my way of, among other things, not creating CONTENT, which is where many small crimes against humanity are committed. Thought leadership, to state the obvious, is a digital marketplace phenomenon, not one of the polis or commons. Thus fraught with contradiction, it has two masters, carbon- and bit-based, file- and feeling-founded. This twin-headed predicament manifests in many ways (or, more frequently, goes unrecognized, to dire credibility-eroding effect.) For instance, though it is a rhetorical arena highly aware of the compounding, high-visibility value of original thought, it is teeming with truisms. And though it is a rhetorical space that believes the actionability of its arguments and insights, explicit and implicit, is the ultimate metric of value, economic and symbolic, the bulleted key takeaways so beloved by AI are often missing or of low utility.

I often advise business, individual, and institutional clients that the best thought leadership must first earn the judgment of the audience, a shift in perception, what Aristotle called krisis. How the audience understands their situation has changed, and only then should the thought leadership artifact prompt prohairesis—action or practical decision, which follows organically. In fact, I’ve found that Aristotle’s Rhetoric, a 2,400-year-old framework—and the broader classical tradition it anchors—is more essential than ever for presenting ideas persuasively online, but not simply because of its enduring insights. Mounting curiosity about rhetoric—even now, often taken as a pejorative—is symptomatic of a heightened need to understand the machinery of persuasion, as trust in authorities evaporates and stabilizing metanarratives fray. Our collective epistemic anxiety is further exacerbated by AI’s refraction of humankind’s persuasion at scale: the patterns and interlocking appeals in writing it has learned to replicate and reward. Refracted, not reflected, because Aristotle codified persuasion as the art of the situated human speaker reading a live audience in real time—hardly the conditions of most persuasive prose in the AI era.

Rhetoric, Refracted

Nevertheless, the prismatic effect of AI’s devouring of massive persuasive patterns is to illuminate just how much the credibility and influence of an organization or individual online is the cumulative signal of their entire output. Authority and influence come as well from the sustained coherence of a company, institution, or leader’s reasoning and the overall cogency of their expertise; the consistency of their voice and the company they keep in credible conceptual neighborhoods; the unfailing emotional resonance and skin-in-the-game quality of their ideas; and the timeliness and zeitgeist grounding of their meaning-making. In other words, Logos, Ethos, Pathos, and Kairos.

Aristotle assumed a speaker, an audience, an occasion. We now confront something stranger: a corpus, summarily summarized by the mute black box of algorithmic hierarchization. Part Two begins there, with the four timeless rhetorical appeals, pressurized as they’re pressed into action, and with the question every serious thought leader and institution now faces: what remains of persuasion when judgment is increasingly outsourced before the human encounter can happen?