The Wisdom of Crowds13
Committees are notorious for making bad decisions. Ask a committee to design a horse and the result will look like a camel or so goes the old joke. General Motors’s famous engineer Charles Kettering said that the best way to kill a good idea was to give it to a committee.
However, in four conditions collective intelligence can be superior to that of individuals in solving three types of problem.The first condition for success is diversity in attitude toward risk, information, and point of view. Jehn (1999, 2001) found that “For a team to be effective, members should have high information diversity and low value diversity.” There is some evidence from studies of corporate boards and national cabinets that groups of 12-15 people can meet these criteria without becoming unmanageably large.
Second, each member must remain independent despite pressures to conform and the threat of information cascades (Chapter 5) in which everyone simply goes along with an influential or dominant person. If a group is to find the best possible solution, it is vital for the leader to maintain an atmosphere in which every team member is comfortable raising and responding to questions.14
Third, each member of the group must be a specialist on some aspect of the problem. There is no evidence of general decision- or policy-making skills independent of specific, relevant knowledge. Groups are good at choosing among solutions, but only individuals are good at suggesting them.
Fourth, there must be some way to aggregate the information and opinions. Failures to do so gave us 9/11, the North Korean invasion of South Korea, the Battle of the Bulge, and Pearl Harbor.15 Possible methods are Bayes's Theorem (Chapter 2), consensus (Chapter 7), decision by a leader (Chapter 9), voting (Chapter 11), and free markets.
Under these four conditions, groups usual come up with better solutions than individuals do to three types of problem.
The first are “cognition problems”—those with a specific correct answer. The simplest example is guessing the number of objects in a jar, with the mean of all guesses usually bettering the guesses of at most one or two in a large group. An impressive instance involved the disappearance of the submarine Scorpion in 1968. The only things known were when it left port and its destination. No one knew which route it took, how fast it was going, or why it sank. John Craven divided the Atlantic into grid squares and asked a group of mathematicians, oceanographers, submariners, and salvage men to guess independently in which square each thought the wreck was. The search began in the grid with the highest combined probability with the probabilities continually readjusted using Bayes's Theorem, quickly locating the Scorpion.The second type is the “coordination problem," in which individuals act independently in their self-interest. Bookmakers do not have to figure out themselves what odds to give on a horse, fighter, presidential candidate, or team. Instead, they constantly adjust the odds (or point spread in games such as football) so that the same amount of money is bet on each possible outcome. Using the same approach, the Iowa Electronic Market predicts election winners more accurately than the national polls and the Hollywood Stock Exchange usually has the best predictions for Oscar winners and box office returns. Edward Chamberlain (1962) demonstrated free markets work better than alternatives even when people aren’t sure how to maximize their own self-interest, do not have perfect knowledge, and are willing to settle for less than perfect outcomes. The idea is not new: Adam Smith called it the “Invisible Hand” in 1776.
The third type of problem in which crowds make better decisions than individuals are “cooperation problems,” those that require individual sacrifice for social benefit. Solving them (e.g., setting tax rates, reducing pollution, encouraging donations to charity) requires a broader definition of self-interest and trust in other group members without which pure self-interest is the only rational strategy. Adam Smith addressed this problem in Theory of Moral Sentiments (1759). Robert Axelrod (1985) argued that people who deal with one another on a continuing basis recognize the advantage of cooperation. Markets work best when people believe that the long-term benefits of fair dealing outweigh the short-term benefits of sharp dealing. Scammers still exist, so “trust but verify” remains useful but, as we have seen elsewhere, people or even monkeys and dogs treated unfairly (Chapter 2) prefer nothing to being cheated. Overall, the best strategy is to be nice, retaliate if injured, and forgive if reform results—the tit-for-tat solution to Prisoners' Dilemma (Chapter 2).