Crowds Are Wise (And One's A Crowd)

An empirical self-experiment on the wisdom of crowds and the 'inner crowd' effect. Using an ACX-Survey distance-estimation question (6,924 respondents), Scott shows error falls with crowd size (fitting 1/error to log crowd size) and that averaging a single person's two guesses beats either guess -- but only barely (2-13 km) versus 200 km for two different people. Cross-checks against the Van Dolder/Van Den Assem casino study (inner crowds plateau at ~half the first error; longer gaps between guesses decorrelate them and help), and asks why almost nobody uses wisdom-of-crowds on real numerical decisions (answer: few decisions are made numerically, and the conversion cost usually exceeds the accuracy gain).

Quality 67 · Strong. Strong (low): a fun, honest, data-driven exploration with a real result (inner-crowd wisdom is real but weak), transparently caveated as single-question and possibly p-hacked; solidly mid-Strong.

Claude’s paradigm shift 48 · Moderate. Moderate: the empirical test and the 'why don't people use this' analysis are a nice contribution, but the underlying wisdom-of-crowds idea is well-established.

Real-world impact 2 · Minor. Minor/within-blog: a survey-data exploration with practical estimation relevance but no material footprint.