Polymarket has US alien disclosure at 15%. The Probability Lab says that number is mostly emotion.

We ran the December 2026 disclosure question through Aurelon's Probability Lab, 50 scenarios, full adversarial debate, the usual base-rate and overlap correction. The structural answer comes out around 1%, with the upper plausibility band capped near 18%. Polymarket's 15% is not impossible, but it is a price that lives almost entirely at the top of that band, in worlds where a quiet machinery of caution suddenly decides to stop being cautious.

The full Polymarket question is more demanding than the casual phrasing makes it sound. It does not pay out if a senator says something interesting on a podcast, or if a Pentagon spokesperson allows that "non-attribution" could mean a lot of things. It resolves YES only if the President, a Cabinet member, a member of the Joint Chiefs, or a US federal agency makes a definitive public statement that extraterrestrial life or technology exists by December 31, 2026, 11:59 PM ET. That is a narrow keyhole, and the machinery on the other side of it has spent eighty years getting good at not crawling through.

The market price treats that keyhole as if it were a door. That is the gap the run was designed to measure.

Bureaucracies that have practiced for decades how not to say something do not usually break the habit in seven months.

Across fifty scenarios the model behaved exactly the way it should when the headline number is wrong: the YES-leaning pathways generated a lot of weight in the first pass, and then the correction layers ate most of it. The reason was not skepticism. It was overlap. Almost every plausible YES path is a variant of the same story, something major is discovered, then a US official publicly says so, rebadged with different protagonists.

The engine's job, in moments like that, is to refuse to count the costumes. After overlap correction the YES family contributes far less than its raw aggregate suggests, and the NO family, which is dominated by genuinely distinct mechanisms of inaction, ends up doing most of the lifting.

What jumped out of the scenario distribution was how many different ways the answer can stay NO without anybody having to actively suppress anything. They are mostly mundane, which is exactly why a market priced on vibes underweights them.

• Bureaucratic reframing into safety and airspace risk (65%): the issue gets quietly absorbed into the language of drones, foreign tech, and air-traffic safety.
• Adversary-tech reclassification (68%): the most exotic sensor data ends up labelled as Chinese or Russian capability of unknown configuration.
• Lawfare and information-access gridlock (48%): court-ordered releases produce volume, not resolution.
• Sensor modernization explains away the mystery (58%): better optics and processing keep retroactively turning "anomalous" into "explainable, just later."
• Executive messaging fragmentation and preclearance delays (45%): any statement of this magnitude has to clear DoD, ODNI, State, NASA, and the White House.
• Transparency promises become study commissions (45%): when Congress demands answers, the system answers with a panel.

None of these requires a cover-up, a conspiracy, or even particular bad faith. Each is just an institution doing what institutions trained on caution do when handed a politically radioactive question on a tight clock.

The Pentagon has had decades to find a phrase between 'unidentified' and 'extraterrestrial.' It found one, and uses it every time.

The most useful insight from the debate transcripts was that, even in the favorable scenarios, the resolution hinges on language rather than evidence. The skeptic agent kept making the same point and the judge kept agreeing with it: the gap between "the evidence is overwhelming" and "a federal agency publicly says extraterrestrial life exists" is enormous, and it is exactly the gap the US government's default vocabulary is built to live inside.

Officials have an entire palette for this: unidentified, anomalous, non-attributable, not explainable by known terrestrial systems, consistent with, compelling but not conclusive. None of those phrases resolves the market. All of them are professionally safer than the words that do.

Polymarket prices are not always wrong, but they are not immune to the topic they are pricing. Disclosure markets attract a particular kind of trader, the same kind that drove the price up briefly around every Schumer amendment, every Grusch hearing, every David Fravor podcast cycle, and that pool is structurally bullish in a way an NFL futures pool is not. They are paying for the possibility of being right about a generational story, and they discount the wording problem because the wording problem is boring.

Tetlock's superforecasters consistently outperformed experts on exactly this kind of question, and the reason was not better information. It was an unsexy preference for the outside view: how often does the institution actually say something this big, on this kind of deadline, in language this clean? The base rate is brutal.

Aurelon's central estimate of around 1% lives at the floor of the plausibility band; the upper band touches 18%. Polymarket's 15% is therefore not crazy on its face, it sits inside the band, but the shape of the scenario distribution makes that price hard to defend.

The honest reading is that the price is doing what disclosure-themed prices have always done: tilting toward the wish. The structural number is in the low single digits, and even the optimistic case stops well short of 15%.

15% is not a probability. It is what the question feels like to people who would enjoy the answer being yes, with a decimal point attached.

The disagreement is not about whether the universe is interesting. It is about what the United States government will say, in writing, on a deadline. Those are different bets, and the market is pricing the first one.