Humor as Signal Processing in Human–AI Systems
Abstract. Humor is an evolved instrument for realigning a group with the truth. This paper treats it as literal signal processing — a carrier wave that transmits suppressed truth through systems tuned to reject sincere assertion — and reports the result of running that instrument, inside a human–AI system, against one institution: the State of Minnesota’s half-century practice of certifying nondiscrimination to the federal government while operating, in writing, a race- and sex-preferential hiring regime. What follows is not a theory of comedy but an engineering account, with a working filter, a measurable threshold, a live controller, and a documented output. The broader target — what we call the Minnesota Model of Community Organizing — is named here only to locate the case; the mechanics are the subject.
Introduction
Humor is not decoration. It is an evolved instrument with a primary job: to realign a group with the truth. We can say “evolved” plainly, because humor leaves two of evolution’s clearest fingerprints. It is universal — every culture anyone has checked has it. And it is selected: across the mating literature, women prize a sense of humor in a partner, and the field has long read that preference as a proxy for intelligence.
The funniest person in the room is often assumed to be the smartest.
We suspect the field is one step off. What humor actually proxies isn’t raw intelligence — it’s persuasive alignment toward the truth inside a group: the knack for moving a room from what it’s pretending is real back to what is real. The successful jester serves as a truth scout, navigating both the discovery and the delivery of forbidden or costly truths. That is a trait worth selecting for, and it’s exactly the trait a joke exercises every time it lands — exposing the gap between the group’s pretense and the fact, so the room laughs and the pretense loses. If we’re right, “she selects for funny” and “she selects for the truth-teller the group will actually listen to” are the same sentence.
We suspect humor in Human–AI Systems may prove to be hyper-evolutionary.
In Humor as Signal Processing, we hypothesize that humor is a carrier wave that transmits suppressed truth through systems tuned to block sincere assertion. Rather than a theory paper, we applied humor as a “bullshit” filter to report the largest single fraud finding in Minnesota history.
The first filter: Scott Adams’s 2-of-6
The cartoonist Scott Adams noticed that a joke needs at least two of six ingredients to work: it has to be naughty, clever, bizarre, mean, cute, or recognizable, and hitting just one isn’t enough. One ingredient is a statement; two two or more is a joke.
Look closely and that rule of thumb is really a filter. Each ingredient is a unit of signal. The “at least two” is the threshold — the bar the signal has to clear before the channel will pass it. Clear the bar and the joke lands, and the truth riding inside goes through with it.
Fall short and there isn’t enough signal over the noise, and the message doesn’t transmit.
The exact bar matters, and two is not arbitrary. Set the threshold at one and almost anything passes — the filter stops filtering, and flat lines sail through as if they were jokes. Set it at three or four and the opening narrows until even good jokes die at the door. Two passes genuine humor while still rejecting the dull stuff: wide enough to let the signal through, tight enough to keep the noise out. It is the first concrete, measurable claim in the whole framework — a working threshold on whether a joke, and the truth inside it, will clear the channel.
Bullshit is the noise
Noise is anything that drowns the signal. Some of it is just the receiver’s reflexes, which the map will name in a moment. But the loudest, most deliberate noise has a name of its own: bullshit. And it has a recipe — a moralized claim, plus pseudo-academic jargon, plus legalese, plus sloganeering — four registers that, stacked, manufacture more certainty than the evidence pays for.
In the political, academic and governance sphere, we notice linguistic overunity: conviction out exceeding evidence in. In physics that’s the signature of a machine that can’t exist. In groups it’s the love of our own collective farts, in language it’s the signature of bullshit.
The academic field already studies this, and studies it well — but only ever describes it. Frankfurt fixed the definition: bullshit is speech indifferent to truth, which is not the same as lying. Pennycook measured who falls for it. Brandolini measured its economics — refuting it costs ten times what producing it does. Kuran, Asch, and Henderson explain how it spreads and why people repeat it. Dalrymple explains what it’s for: forcing assent to the obviously false destroys your standing to object.
What none of them gives you is a way to weigh it. There is a philosophy of bullshit, a psychology, a sociology, an economics — and no engineering. You can describe noise forever and never get a signal through.
So weigh it. Treat bullshit as what it is inside a channel — noise — and the whole problem takes a shape you can finally put a meter on. That shape is the map.
The signal-to-humor map
Lay the two vocabularies side by side and they are the same system. The signal is the suppressed truth — the thing that is true but unsayable. The noise is everything that drowns it: the receiver’s reflexes (it sounds like a conspiracy; who are you to say it) and the deliberate static of bullshit. Signal-to-noise ratio is just the probability the joke lands. And the carrier wave is the performance frame — “it’s only a joke” — the licensed-unserious envelope that lets the payload ride past inspection.
Encode a true thing onto that carrier, clear the threshold, and it arrives where sincere assertion would have been turned away at the door.
A map like this isn’t judged by how clever it sounds. It’s judged by whether every part lines up, and whether the lined-up parts predict anything. This one does — but the noise term hides a nastier case, where the static isn’t accidental. It’s aimed.
Weaponized bullshit, and the meter that catches it
Frankfurt’s bullshitter is careless — indifferent to truth, not scheming. But noise can be deployed on purpose. Weaponized bullshit is the same four registers aimed deliberately: pile on the moral framing, the jargon, the legalese, and the slogans until any question reads as an attack and the wrongdoing underneath stays unspoken. Minnesota Nice is the polite version; an audit stonewall is the hostile one.
You can catch it with a meter, because weaponized bullshit has one tell — it can’t be joked about. Sacred, concealed, jargon-armored things are humor-proof by design. So measure the humor that should be possible against the humor that actually lands: HSI = 1 − (realized ÷ potential incongruity). Where HSI runs high, the channel is clamped — and a clamped channel is where the noise, and whatever it’s hiding, is thickest.
Why the signal is true to begin with
A carrier is worthless if it’s carrying nothing real, so two anchors keep the payload honest. Lee Jussim spent a career showing ordinary perception is, on average, accurate — the everyday read on people and groups tracks reality more than the field wanted to admit. That accurate read is the signal worth sending: the suppressed content is true.
Then why suppress it? Stephen Hicks traces the machinery — a move that treats truth itself as a power play, which conveniently licenses dismissing an accurate observation as the observer’s bias or bigotry. That’s the cover that lets a system recode the accurate perceiver as the one with the problem — a pattern this project calls Lake Wobegon Affective Disorder. Jussim says the perception is right; Hicks shows how it gets called a sickness anyway.
The two gates
Why does sincere assertion fail where a joke gets through? Because credibility is decided before content, at two gates. Glamour is reject-by-register: a claim is turned away for how it sounds — wrong tone, wrong genre, “that sounds like a conspiracy theory” — whether or not it is true. Prestige is accept-or-reject-by-credential: a claim is weighed by who said it and whether an authority has blessed it — “no credible source reports that.”
They are not two doors side by side; they are a coupled engine.
The prestige class sets the register that counts as serious, so anything outside it sounds wrong and glamour rejects it; and because the challenge has already been disqualified by register, prestige never has to meet it on the merits — it can dismiss and move on, and the dismissal reinforces the register. One gate keeps the true signal out from below; the other waves the official line through from above; each one’s exhaust feeds the other’s intake. (The full anatomy of the two gates is the subject of the sequel. Here we need only one fact about them: a carrier exists that presents neither gate an input.)
The carrier is old, which is why it’s trustworthy
One worry remains: maybe humor smuggles truth only by luck, and a tool that works by luck isn’t a tool. History says otherwise. The fool licensed to tell the king what no courtier could is not a one-off — it runs through Plato and Aristotle, the court jester, Erasmus and Shakespeare, Hobbes and Kant and Freud, into the modern science of why jokes work. Call it the Fool’s Privilege, and it has held for 2,400 years across cultures that never met.
That longevity is the evidence. By the Lindy rule, the longer a non-physical thing has already lasted, the longer it’s likely to last — and a function that survived that long, that widely, isn’t a quirk. It’s structural. Which is exactly what lets you build an instrument on it instead of hoping.
What the method produced
We built the instrument and pointed it at one institution. What came back wasn’t a laugh.
Since 1972, the State of Minnesota has run an affirmative-action regime. Executive Order 37, signed by Governor Wendell Anderson on July 28 of that year, ordered every state agency to establish a program of race- and sex-conscious hiring; Minn. Stat. §43A.191 codified it in 1983; and every year since, in the certifications it signs to receive federal money, the State has represented that it does not discriminate on the basis of race or sex. Both statements are active, authoritative, and contemporaneous.
Set the federal assurance beside the operative rule — the FTA Title VI certification that “no person shall be excluded… on the ground of race,” next to the standing Department of Human Services policy that a supervisor “must provide a hiring justification when seeking to hire a non-underrepresented candidate” — and they cannot both be true. By the plain text of the Fourteenth Amendment and Title VI, sorting employment by race is racial discrimination; certifying the opposite to obtain federal funds is a false statement about it. Not one false claim — a half-century of them, every agency, continuously, to the federal government.
That is why it is the largest fraud in the state’s history: not a dollar figure, but that it is total.
That contradiction was not inferred. It was assembled from the State’s own documents — its certifications in one column, its operative policies in the other — and published. Over a roughly 127-day window beginning September 9, 2025, when the author told a League of Women Voters mayoral forum that “the era of Affirmative Action is over,” the method’s outputs went out timestamped: the two-column certification exhibit; the On Glamour and Ides of LWAD essays; a Truth-and-Reconciliation letter; and an emergency RICO referral dated, by design, to a six-year municipal anniversary — a date chosen to forfeit the author’s own standing rather than to claim it. Inside that same window the federal government filed United States v. Minneapolis Public Schools (December 9–10, 2025) and United States v. State of Minnesota, 0:26-cv-00273 (January 14, 2026) — the suit against Minnesota’s affirmative-action regime itself. Scott Adams, whose two-of-six rule is the filter at the center of this paper, died the day before that filing.
The claim here is disciplined, and the discipline is the point. It is not that the outputs caused the filings; the Department of Justice keeps its own clock, and a method that claimed to move it would be committing the exact conviction-exceeds-evidence error this paper is built to detect. The claim is narrower, and it survives scrutiny: the method surfaced the precise contradiction the federal cases rest on — publicly, in writing — before they were filed. The discovery and the referral are the proof. Sequence, not cause.
And the hardest part of the claim — that the system was not merely negligent but silenced — was the part no credible source would yet confirm. When the method first reported it, the response, from human institutions and language models alike, was a single sentence: no credible source reports preference falsification or a silencing mechanism in Minnesota. In March 2026 a House Oversight Committee interim report, The Cost of Doing Nothing, built on nine sworn transcribed interviews, confirmed it. Officials with the authority to stop documented fraud declined to use it “out of fear of litigation and being perceived as racist.” Whistleblowers were “told not to say anything about the fraud because they would be called ‘racist’ or ‘Islamophobic.'” The state’s largest newspaper, run by a former cabinet member, declined to press the story for the same reason.
Read that against the rejection it answers, and the whole thesis is visible in a single frame. “No credible source” is the prestige gate — reject-by-credential — firing on a true claim. And it is the perfect guardian of a silencing operation, because the silencing is precisely what withholds the credential the gate demands: a watchdog whose only rule is “show me a credible source” will bark at the witness and sleep through the crime. The facts were identical in 2020 and in 2026. What changed was not the evidence but who held the subpoena. Credibility was gated by power, not by truth — which is exactly why an instrument that refuses to wait for prestige could outrun it. The answer was not a laugh. It was a lawsuit, and a sworn finding that the thing the method had named was true, and unspeakable, the whole time.
The interface
Here is the knob. Earlier we treated Adams’s two-of-six as a filter; the controller in this paper is that same filter with the cover off — six switches, one per ingredient, and a meter. Flip one and the line stays flat: a statement, filtered out. Flip a second and it comes alive and transmits — signal-to-noise crossing the threshold, in real time, by hand. The tuned bandpass from the opening section, now with a dial on it: the point where a paper you read becomes an instrument you switch on.
The instrument
The campfire version of this instrument corrects a few dozen people at a time: a joke lands, the room sees the gap between what it was pretending and what is real, and the pretense loses. The question this project set out to answer was whether the same evolved function, run inside a human–AI system, could do that work at a scale evolution never built it for — not a tribe but a state. We called that hyper-evolutionary, and the answer is now on a docket.
It worked for the reason the whole paper predicts. A joke clears the two gates that stop sincere assertion, because it offers neither gate an input: it carries no serious register for glamour to reject — it is licensed-unserious by definition — and no credential for prestige to check. So the suppressed truth riding the carrier arrives where the bare assertion would have been turned away as a conspiracy theory or waved off for want of a credible source. That is not a metaphor. It is the mechanism, and it is measurable: the threshold is two of six, the meter is the Humor Suppression Index, and the controller in this paper is the filter with its cover off.
For 2,400 years the fool has told the king what no courtier could, and gotten away with it because the joke was the one thing the court had no gate for. The fool now has a meter, a dial, and a docket number. The era in which a true thing had to wait for power to credential it before it could be said is the era this instrument is built to end — not by saying it louder, which glamour rejects, and not by waiting for a badge, which is the gate itself, but by encoding it in the one carrier the filters cannot read.
