Correlated Error

The wisdom-of-crowds depends on disagreement doing useful work. Independent people encounter different evidence, apply different assumptions, and make errors in different directions; aggregation can then cancel those errors. Influence couples their judgments. Once participants inherit the same anchor or information source, their mistakes become correlated, and the average repeatedly counts the same underlying bias. The diversity-prediction-theorem therefore concerns diversity of error, not merely a diversity of people.

Agreement alone does not prove correlated error: people may converge because the common information is accurate. The warning applies when aggregation is being trusted to cancel mistakes but the judgments are no longer genuinely independent; determining whether influence or truth caused convergence requires separating plausible confounders.