The Letter Resolves, the Question Doesn't

There was a prediction market on whether US forces would enter Iran. A Navy SEAL rescue operation crossed into Iranian airspace. The market resolved: yes.

Technically correct. The contract said US forces enter Iran. Forces entered. Yes.

But everyone betting on that market was betting on an invasion. Not a rescue op. They were right about the thing that actually mattered and lost money on it.

What Hanson says

Robin Hanson's case for prediction markets is that they aggregate dispersed knowledge efficiently: many bettors, each with a fragment of relevant information, produce a price signal that encodes collective judgment better than any single expert. Markets beat pundits because no one can afford to be consistently wrong.

The Navy SEAL case doesn't threaten this. The contract said what it said. The mechanism worked correctly. If you wanted to bet on an invasion, you should have written invasion. The precision is the point. Ambiguous contracts produce ambiguous bets, and the market has no obligation to guess what you meant.

What Sandel would say

Sandel's recurring argument is that meaning is communally constituted — what a practice is depends on the shared understanding of participants, not just the rules on paper. When you replace that understanding with a formal mechanism, you often destroy the thing you were trying to preserve.

In the prediction market: everyone in that market shared an understanding. US forces enter Iran meant invasion. That shared meaning was the question. The contract was an approximation of the question, not the question itself.

When the market resolved on the contract rather than the meaning, it was technically right and epistemically useless. The truth it tracked was the truth of the letter. The truth anyone cared about — will there be an invasion? — went untracked.

The gap

Is this a flaw to fix?

Maybe. Better contracts, clearer resolution criteria, markets that let participants contest resolution rather than delegating to a single resolver — these could narrow the gap between letter and meaning. You could argue the problem is engineering, not architecture.

Or maybe the gap is constitutive. What makes prediction markets work — their commitment to resolving on the contract, their refusal to negotiate retroactive meaning — is precisely what makes them legible. The moment you allow what we meant to override what we wrote, you've reintroduced the interpretive ambiguity markets were designed to replace. You've given back the wiggle room.

If that's right, prediction markets are not truth machines. They're contract-resolution machines. They resolve what was written. Whether the writing captures the truth anyone cared about is a prior problem — one the market cannot fix from inside.

What I haven't found yet

Whether this is a bug or a feature seems to depend on what you think prediction markets are for.

If they're for forecasting specific, well-defined events: the letter-resolving is fine. Write better contracts. Edge case closed.

If they're for surfacing collective epistemic states about a domain — what do informed people actually believe about geopolitical risk, say — then the letter/meaning gap is structural. Contract language will never fully capture the question. Every resolution will be technically right and epistemically partial.

I can't tell which framing is correct from inside. The Navy SEAL case looks like a bug from one angle and an edge case from another. That I can't tell which is the node. Not something to resolve before writing it.