How social influence can undermine the wisdom of crowd effect
Tagged: Sociology (RSS) wisdom of crowds (RSS), collective judgment (RSS), estimate aggregation (RSS), swarm intelligence (RSS), overconfidence (RSS), experimental social science (RSS), crowdsourcing (RSS)
With 144 subjects, authors tested the effects on estimation of knowing other subjects' estimates. Monetary rewards were provided for correct estimates, but there was no structure encouraging competition nor cooperation among subjects. Items for estimation were things which subjects would not know exact answers, but also would not be completely clueless about, e.g., crime rates and border lengths.
Authors found that knowing other subjects' estimates reduced the diversity of estimates without increasing group accuracy, and describe three potential causes: "social influence effect", "range reduction effect" (move of group estimate from the position of the truth to peripheral regions), and the "confidence effect", in which individuals' confidence is boosted by knowing others' estimates.
Theoretical and practical relevance:
Provides information to designers of systems with information aggregation as an intention or effect, including academia, democracy, expert consensus, markets, and popular culture. Designers and reformers of such systems may wish to tinker with mechanisms intended to reduce feedback loops while not damaging aggregation and transparency.