The Wisdom of Crowds

A highly interesting book about how groups of normal people can be smarter than smart individuals.

Under the right circumstances, groups can be smarter than the smartest person in them. What are these circumstances? How can there be collective wisdom in individual ignorance?

Necessary for the crowd to be wise are diversity of opinion, independence, and an aggregation or weighing scheme. Maybe also a particular kind of decentralization, or rules. The best collective decisions are the product of disagreement and contest; not of consensus or compromise. Voting systems or markets work to consolidate these conflicting views.

Diversity of opinion

Individual errors tend to cancel out when individuals have at least some limited knowledge about the true state. Private vs public info: private is what only you know or think, public is what you observe in others. Mixed skill levels in a group help good decisions by giving more perspectives. Groups of only the strong spend too much time exploiting, and not enough exploring.

Independence

Information feedback loops happen when people base their decision on that of others. Optimizing for yourself, it may be better to follow the crowd then trust your own judgment, but optimizing for all, it is better to decide by yourself, add different opinions and take out the sequen
tial feedback loop. Herding, that is copying the behavior or opinions of others, makes groups dumber. This is called "groupthink", and seen in stock bubbles. Speculation, that is basing you decisions on the assumed decisions of others, who do the same, makes groups unstable, unpredictable and also can lead to bubbles.

Aggregation

The goal of aggregation must be to avoid dependence fallacies, and to weed out or weigh out mistakes. Systems that generate many alternatives, recognize losers and kill them quickly can be successful. Sequential decision making encourages herd behavior, because you observe others' decisions and they influence your own. This is unavoidable for markets and other large groups, but can be avoided with care in small teams.

Cognition

Fact guessing - bigger groups are better. Decision markets forecast future developments better per participant than costly market research studies or polls, and that is polls of the people themselves in case of votes!