The Wisdom of Crowds

In 1906, at a livestock fair in Plymouth, roughly eight hundred visitors paid sixpence each to guess the weight of an ox that would be slaughtered and dressed. Francis Galton, the scientist who had long doubted the judgment of ordinary people, collected the tickets afterwards, expecting to prove how poor that judgment was.

The tickets refused to cooperate. The median guess, taken from the 787 entries that were usable, came to 1207 pounds. The ox weighed 1198 pounds. A mixed crowd of butchers, farmers and curious onlookers had landed within one percent of the truth, and Galton reported this in the journal Nature in 1907 under the title Vox Populi, admitting that the result was more creditable to democratic judgment than he had anticipated. What stands out in the retelling is the smallness of the error, nine pounds on an animal of nearly twelve hundred.

The story became a founding parable of management thinking. In 2004 the journalist James Surowiecki gave it popular form in The Wisdom of Crowds, arguing that a group can outperform its best members when four conditions hold. These are diversity of opinion, independence of judgment, decentralised knowledge, and a way of combining private guesses into one answer. Firms soon began to run internal prediction markets and staff polls, hoping to capture the same effect.

Independence is the condition that is easiest to lose and hardest to notice. Plymouth worked because each visitor wrote a number alone, without hearing a neighbour. In 2011 a team of researchers led by Jan Lorenz reported in the Proceedings of the National Academy of Sciences that when people in groups were shown one another's estimates, their guesses drifted together, the spread of opinion narrowed, and the crowd grew more confident without growing more accurate.

Plymouth worked because each visitor wrote a number alone, without hearing a neighbour.

Defenders of crowds offer a reply that deserves a hearing. Influence is not always corrupting, they argue, because people who hear others often correct obvious blunders, and a crowd that includes a few genuinely informed members may improve when their view spreads. Financial markets, in which traders watch each other constantly, are sometimes cited as proof that connected crowds can still be sharp. Some economists add that a crowd is also helped when its members err in different directions, since mistakes that pull opposite ways tend to cancel. The reply is not empty. Yet it quietly changes the claim. A crowd that converges on an informed minority is no longer pooling many separate errors that cancel. It is following a few people, and whether those few deserve to be followed is exactly the question the crowd was meant to answer without needing to ask.

The difficulty widens well beyond oxen. Juries, elections, polls, markets and online reviews all promise wisdom from the sum of many judgments, and all of them depend on people forming views apart before those views are combined. Modern life is built to prevent exactly that, since every opinion now arrives already surrounded by the opinions of others. If a crowd is wise only to the extent that its members do not listen to one another, then a society that listens more closely may be learning less, and it remains unclear how a people could ever be at once fully connected and fully independent.