Noise: people game a known scoring formula by changing their ratings to reach the conclusion they want; a market makes that move costly.

The claim

Kahneman, Sibony and Sunstein explain why they keep a judgment step after scoring: decision makers "gain the system when they know that a formula will be used. They change the ratings to arrive at the desired conclusion, which defeats the purpose of the entire exercise." On ratings they add that taming rater noise "reduces their ability to influence ratings in pursuit of their own goals", which is why it meets resistance. Agents optimise against any known formula faster than people do. A decision market differs from a ratings formula at exactly this point: moving the price toward a desired conclusion costs the mover money, so back-solving is possible but priced. Inference beyond the quotes is Leo's.