Less Wrong is a community blog devoted to refining the art of human rationality. Please visit our About page for more information.

HalFinney comments on "Statistical Bias" - Less Wrong

13 Post author: Eliezer_Yudkowsky 30 March 2007 06:55PM

You are viewing a comment permalink. View the original post to see all comments and the full post content.

Comments (8)

Sort By: Old

You are viewing a single comment's thread.

Comment author: HalFinney 31 March 2007 11:46:17PM 0 points [-]

I think this same basic formula is behind the argument for majoritarianism: the crowd's consensus (average) squared-error, plus the variance in the crowd, equals the expected squared-error for a random person in the crowd. Therefore the crowd consensus view has a lower expected squared-error than the average squared error for individuals in the crowd. Hence a random participant will do better to substitute the crowd consensus for his own estimate.

I'm reading a book, "The Difference", by Scott E. Page, which discusses how and when crowds do well, and he calls it the Diversity Prediction Theorem: Given a crowd of predictive models, Collective Error = Average Individual Error - Prediction Diversity.