Keynesian beauty contest
Article
Keynesian beauty contest is a recurring concept in the Astral Codex Ten archive, appearing 3 times across 3 issues between November 01, 2021 and January 24, 2023. The archive places it in contexts such as “One possibility is a Keynesian beauty contest”; “it might end up as some sort of Keynesian beauty contest with the right answer”; “it’s the inverse of a Keynesian beauty contest”. It most often appears alongside Good Judgment Project, Manifold, Metaculus.
Metadata
- Category: Concepts
- Mention count: 3
- Issue count: 3
- First seen: November 01, 2021
- Last seen: January 24, 2023
Appears In
Related Pages
-
- Good Judgment Project (2 shared issues)
-
- Manifold (2 shared issues)
-
- Metaculus (2 shared issues)
-
- Scott Alexander (2 shared issues)
-
- Supreme Court (2 shared issues)
-
- Tesla (2 shared issues)
-
- 538 (1 shared issues)
-
- active inference (1 shared issues)
-
- ACX (1 shared issues)
-
- ACX Prediction Contest (1 shared issues)
-
- AI Impacts (1 shared issues)
-
- AI Safety (1 shared issues)
External Links
Source Context
Recovered passages from the original issue text. When the raw archive preserved outbound links inside the source passage, they are listed directly under the quote.
There are lots of kludgey solutions to this, but one possibility is a Keynesian beauty contest. Get a lot of isolated teams, and make them predict what all the other teams will guess. Whoever gets closest to the average wins the prize.
Inline links: Keynesian beauty contest
A more serious problem is that it penalizes anybody who’s smarter and better-informed than average. If you come up with some really clever argument for why nuclear war is more/less likely than everyone else thinks, but you don’t think other people will come up with the same argument, Keynesian beauty contests incentivize you to ignore it (whereas prediction markets incentivize you to buy as many shares as you can to exploit your superior information). If you happen to overhear Vladimir Putin saying he’s going to start a nuclear war tomorrow, then in a Keynesian beauty contest you should just sit on this information (in a prediction market, you should insider trade and make a killing).
(In theory, if Austin didn’t resolve the market, and everyone knew he wouldn’t, but there wasn’t common knowledge of this, it might end up as some sort of Keynesian beauty contest with the right answer as the Schelling point. But I wouldn’t want to test this, and I would expect it to break down pretty quickly).
I tried to deliberately structure my answers to maximize my probability of winning, rather than maximize the probability of each individual answer being correct. Because there could be only one winner, my model of win probability was (probability of predictions being accurate) / (local density of competitors with similar predictions to me). This is a different approach than what it’d try for a standard prediction market - the analogy I’d use is that it’s the inverse of a Keynesian beauty contest, where I’m trying to find “underrated” outcomes that specifically won’t be predicted by many other people.