True to Things

Sifting through the bullshit for what's real.

Draft

Bullshit Without a Bullshitter

By Matthew Emmett Brown · 7 min read

Brief

The Brief below is compressed from priv/notes/deeper-into-bullshit-cohen.md, which already works the argument out in full; treat it as a transcription, not a proposal.

Frankfurt was wrong to treat indifference to truth as the essence of bullshit. That is the essence of only one kind — bullshit as an activity, defined by the state of mind of its producer, the kind found in advertising, PR, and politics. There is a second kind, distinct and in the academy more important: bullshit as output, defined by a property of the text itself — its unclarifiable unclarity — regardless of anyone’s intentions. For this kind the producer’s mental state is neither necessary nor sufficient: an honest lover of truth can produce it by accident, a sincere person can pass it on without any indifference to truth at all, and a would-be bullshitter can try to produce it and fail. So bullshit has no single essence. Frankfurt took for granted that the bull wears the trousers; sometimes it is the shit that wears them. The test is not psychological but textual — discourse that cannot be rendered unobscure, where adding or subtracting a negation makes no difference to its plausibility, and where any successful clarification produces something unrecognizable as a version of what was said. Frankfurt’s lying/bullshitting contrast is separately malconstructed, because he identifies the liar by his goal (to mislead) but the bullshitter by his tactic (what he asserts) — a category mismatch. The stakes are practical: because output-bullshit need not issue from a bullshitter, the proper target is the product, which is visible, rather than the process or the producer, which are invisible and where accusations of bad faith come too cheaply. This is the essay Bullshit hands to — it closes by naming Cohen as where the next essay starts — and it is the move that makes the machine case tractable, since on this account a text with no mind behind it stands or falls on its own evidence. Frankfurt’s can be stretched to reach machines — drop the intent to mislead, keep only the indifference — but the stretch must still locate an attitude somewhere. Cohen’s need not.

Outline

The doorway is the 2008 structured-credit collapse: instruments whose documents became unintelligible to the desks trading them, a reckoning that could reach the product but never the intent, and sincere participants — quants who believed the copula, funds deceiving no one — who produced bullshit all the same. The reportage for that scene (Salmon, Li, FCIC, Levin–Coburn, House Oversight, Rakoff, Eisinger) is already assembled with URLs in priv/notes/deeper-into-bullshit-cohen.md and is deliberately not duplicated onto the Reading List below, which carries the argument rather than the case. The Sokal affair is held in reserve as the academic mirror-image, but not on the legend’s reading. The editors say they could tell: they found the piece “a little hokey” from the first, they ran a non-refereed journal of political opinion, and they asked Sokal for revisions he refused. What they concede is better than what the legend asserts — they published without ever forming a judgment on whether the reasoning held, because their criteria were documentary and symptomatic rather than argumentative. The product was not misjudged; it was never assessed. And Sokal supplies the principle himself, defending the hoax’s ethics: “it’s true that the author doesn’t believe his own argument. But why should that matter?” That is this essay’s move in the specimen author’s own mouth, and it sits oddly beside his other complaint — that the editors should have detected a parody, which is a demand that they read an intention. Cohen’s distinction is what diagnoses the wobble.

Plagiarism carries the sidecar. The same widening — intent-shaped verb to textually-shaped noun — has already run once in living memory. Traditional plagiarism was an act defined by the plagiarist’s mind: knowing copying meant to deceive. Turnitin made it a property of the text, scored on n-gram overlap regardless of intent, and universities followed; cryptomnesia, sloppy notes, and now unattributed LLM paraphrase produce plagiarism-the-noun without plagiarism-the-verb-in-the-guilty-sense. Same structural move as Cohen’s, forced by the same kind of technological pressure. Placement: one paragraph after the Cohen promotion and before the machine payoff, so the reader meets the widening once in a domain they already accept, and the machine case lands as the second instance rather than the first. Randall and Howard earn the citations — Randall because she argues plagiarism is a property conferred by reception rather than an act of the author, which is Cohen’s move made independently in the other domain, and Howard because patchwriting names the noun without the guilty verb. One sentence here is still unsourced: the claim that Turnitin scores n-gram overlap regardless of intent needs its own citation or should be softened to what the company’s own documentation says.

Source ownership: this essay owns Cohen, Carson, and Pennycook as the spine — Cohen for the account, Carson for the counterexamples that break Frankfurt’s activity version on its own terms, Pennycook because his stimuli were machine-generated and rated profound anyway, which is the output kind demonstrated with no producer at all, nine years before ChatGPT. Cite Cohen from “Complete Bullshit” rather than “Deeper into Bullshit”: §I is the 2002 essay and §II is new. Take §II’s Balibar footnote as the specimen — honest thinker, bullshit sentence, in one line — but leave §II’s thesis about Francophone philosophy alone; it is the part Cohen least stands behind, and following it drags the essay into a culture war the 2008 doorway was chosen to avoid. Moberger is owned here too but as the objection, not the spine: he holds that nonsense cannot be identified independently of the nonsense-talker, and answering him is what earns the central claim. Levy carries the practical stakes, where “target the product” becomes an institutional question with costs. The Sokal affair is three documents — hoax, reveal, editors’ answer — and the third is not optional, because it denies the legend. Frankfurt is owned by Bullshit and read here only as the position being corrected, though his “Reply to G. A. Cohen” partly concedes the ground. Randall and Howard carry the plagiarism sidecar. Carnap is optional, as the ancestor of the textual charge of nonsense that Cohen revives without the verification criterion that made it enforceable. Cut: Hicks et al. (owned by The Work That Remains, and an opponent here rather than support — they keep the Frankfurtian frame and say outright they do not adopt Cohen’s account); Posner (leans on deception and reliance, i.e. the intent side, and cannot carry the Turnitin claim); Sokal & Bricmont (the defect they document is mostly clarifiable-and-false rather than unclarifiable, so it is the wrong diagnosis — reinstate only if §II’s thesis goes in, since Cohen rests it on them at n.30); Maes & Schaubroeck (survey; the extension work is carried by Carson and Moberger); Fredal (rhetoric register, redundant against the spine); Bowles (critique that does not advance the arc); Cassam (political register — belongs to The Answer Machine / Making Sense). Stokke is cut but conditionally: if the tactic/goal malconstruction stays in the argument, Stokke & Fallis (Ergo, 2017) is the literature that works it out and should be acquired.

Reading List

Print / PDF

Letters

Write to the author

Have a considered response? I publish letters that add something — a correction, an experience, a sharper question. You'll get one email to confirm it's you; nothing appears until I've read it.