Theory 219/Peer review/Ratio lens
Value Is Signal Against Noise
The claim
The predictive value of review is set by the ratio of true quality signal to reviewer noise, so adding reviewers helps only while noise is independent.
01Mechanism
Averaging reduces independent noise but not shared bias; the ratio determines the gain from more reviews.
02Prediction
Gain should diminish as shared bias rises.
03Refutation
Gain increasing indefinitely with reviewers would remove shared bias.