Wisdom of Crowds

In 1906, the British polymath Francis Galton attended a livestock fair in Plymouth, England, where roughly 800 people paid sixpence each to guess the weight of a slaughtered ox. The contestants included farmers, butchers, and ordinary townspeople with no particular expertise in livestock. Galton, who was deeply interested in questions of human competence and had strong views about the superiority of expert judgment, collected the entry tickets after the contest and analyzed the data. He expected the results to confirm his belief that the average person was not qualified to make such estimates. Instead, he found something that disturbed his assumptions: the median guess of the crowd was 1,207 pounds. The actual weight of the ox was 1,198 pounds. The crowd had been off by less than one percent — a level of accuracy that no individual contestant, including the experts, had matched. Galton, to his credit, published the finding. He titled the paper "Vox Populi" — the voice of the people.

The wisdom of crowds is the phenomenon in which a group of diverse, independent individuals collectively produces judgments, estimates, or decisions that are more accurate than those of any single member, including the most expert member. This is not the same as consensus, which implies that the group has deliberated and agreed. The wisdom of crowds operates through aggregation of independent signals, not through agreement. In fact, as we will see, the process of reaching agreement is often precisely what destroys the wisdom.

Why It Works: The Mathematics of Error Cancellation

The mechanism behind crowd wisdom is statistical, and understanding it requires seeing each person's judgment as a combination of signal (information that points toward the truth) and noise (random error, biased assumptions, gaps in knowledge). When you aggregate many independent estimates, the signals — which all point roughly toward the same reality — accumulate and reinforce each other. The noise — which is random and points in different directions for different people — cancels out. The result is a collective estimate that retains the signal while shedding the noise.

James Surowiecki, who popularized the concept in his 2004 book The Wisdom of Crowds, identified four conditions that must hold for this aggregation to work. First, diversity of opinion: each person must bring some private information, perspective, or interpretive framework, even if it is only a hunch. Homogeneous groups fail because they share the same blind spots, meaning their errors are correlated rather than random and do not cancel out. Second, independence: people must form their judgments without being influenced by those around them. The moment individuals start copying each other, the number of independent signals drops, and the statistical power of the crowd collapses. Third, decentralization: knowledge and expertise should be distributed across the group rather than concentrated in a single authority. Fourth, aggregation: there must be a mechanism — an average, a vote, a market price — for combining individual judgments into a collective answer.

The psychologist Philip Tetlock's landmark research on expert prediction, published in Expert Political Judgment (2005) and later expanded in Superforecasting (2015), provided dramatic empirical support for the wisdom of crowds. Tetlock tracked over 28,000 predictions by 284 experts across multiple domains and found that the average expert performed barely better than random chance. But aggregated forecasts from groups of informed non-experts — particularly those trained in probabilistic reasoning — consistently outperformed individual experts. The Good Judgment Project, which emerged from Tetlock's work, demonstrated that a structured crowd could outperform professional intelligence analysts with access to classified information.

Two Examples: Market and Meeting

Prediction markets are perhaps the most striking large-scale demonstration of crowd wisdom. In these markets, participants buy and sell contracts whose value is tied to real-world outcomes — election results, product launch dates, economic indicators. The market price at any moment represents the crowd's aggregated probability estimate. The Iowa Electronic Markets, run by the University of Iowa since 1988, have consistently outperformed major polls in predicting U.S. presidential election outcomes. They succeed not because any individual trader is unusually insightful, but because the market aggregates thousands of independent assessments, each informed by different local knowledge, personal observation, and analytical frameworks. The price is the wisdom.

At a personal scale, the wisdom of crowds operates in any situation where you seek multiple independent opinions before making a decision. A project manager who asks each team member to independently estimate how long a task will take — before any group discussion — and then averages those estimates will almost always arrive at a more accurate timeline than if she had asked the most senior person for a single estimate or let the team discuss and converge on a number. The key is the word "independently." When team members estimate after hearing each other's numbers, anchoring bias and social pressure distort the results. The independent aggregation is what produces the wisdom; the group discussion is what destroys it.

When Crowds Become Mobs: The Failure Modes

The wisdom of crowds is not a law of nature. It is a conditional phenomenon that evaporates when its preconditions are violated — and in practice, they are violated constantly.

The most common failure mode is the collapse of independence. When people observe each other's judgments before forming their own, information cascades replace independent assessment. The economist Sushil Bikhchandani and his colleagues described this phenomenon in a 1992 paper: when early signals in a sequence point in one direction — even if those signals are weak or based on limited information — subsequent actors rationally ignore their own private information and follow the herd. Each person's imitation makes the cascade stronger, and each new follower provides additional social proof that reinforces the cascade further. The result is a crowd that appears to have reached a strong consensus but has actually amplified the judgment of the first few movers while suppressing all subsequent independent information. Financial bubbles — from Dutch tulips in 1637 to the dot-com collapse of 2000 to the U.S. housing crisis of 2008 — follow this pattern precisely. The crowd was not wise. It was an echo chamber where independence had been replaced by mimicry.

Groupthink is the organizational version of this failure. Irving Janis coined the term in 1972 after studying catastrophic U.S. foreign policy decisions, including the Bay of Pigs invasion and the escalation of the Vietnam War. In each case, a group of highly intelligent, well-informed advisors converged on a disastrous decision because the social dynamics of the group suppressed dissent. Members self-censored to preserve group cohesion, dominant voices anchored the discussion, and the illusion of unanimity replaced genuine independent assessment. Groupthink does not require that members are stupid or uninformed. It requires only that the conditions for crowd wisdom — diversity and independence — are replaced by the conditions for crowd foolishness: conformity and deference.

The third failure mode is homogeneity. A 2004 study by Scott Page and Lu Hong, published in the Proceedings of the National Academy of Sciences, demonstrated mathematically that a group of diverse problem-solvers will outperform a group of high-ability problem-solvers when the high-ability group is homogeneous. The reason is that homogeneous groups — people who trained at the same institutions, read the same sources, share the same assumptions — make correlated errors. Their mistakes all point in the same direction, so averaging does not help. Diversity is not a moral nicety in the context of crowd wisdom. It is a mathematical requirement.

Limitations

The wisdom of crowds has clear boundaries that are frequently overlooked by enthusiasts.

First, it works best for problems with a definite factual answer — the weight of an ox, the probability of an election outcome, the completion date of a project. For problems that involve values, priorities, or ethical trade-offs, aggregation is far less useful. Averaging a crowd's moral judgments does not produce moral wisdom any more reliably than it produces factual accuracy. The crowd that estimated the ox's weight with stunning precision is the same crowd that, throughout history, has endorsed slavery, witch-burning, and genocidal wars. Factual wisdom and moral wisdom are different capacities.

Second, the mechanism requires that errors be uncorrelated — that different people are wrong in different ways. When systematic biases affect the entire crowd, aggregation amplifies the bias rather than canceling it. If every member of a prediction market overestimates the probability of a dramatic event (a well-documented tendency called the availability bias), the market price will reflect the shared bias, not the truth. Crowd wisdom is only as good as the independence and diversity of the crowd.

Third, the concept is easily distorted into a justification for populism or anti-expertise. Saying "the crowd outperforms the expert" is true under specific conditions but profoundly misleading as a general principle. The crowd outperforms the individual expert when the crowd is large, diverse, and independent. A small, homogeneous, mutually influenced group is not a wise crowd — it is a mob. And in domains that require deep specialized knowledge (surgery, structural engineering, molecular biology), the crowd of non-experts does not outperform the specialist, because the non-experts lack even the baseline information needed to generate useful signals.

Fourth, aggregation mechanisms matter enormously and are rarely neutral. How you combine individual judgments — simple average, weighted average, median, vote, market price — changes the outcome. Each method has different properties and different vulnerabilities. A simple average is sensitive to outliers. A vote is vulnerable to the tyranny of the majority. A market price reflects the views of those who choose to participate, which may be a biased subset of the population. The "wisdom" that emerges is always filtered through the aggregation mechanism, and that mechanism is a design choice, not a natural law.

Connections to Other Concepts

Social proof is both the close cousin and the mortal enemy of crowd wisdom. When social proof operates on independently formed judgments — when you check the reviews written by people who each evaluated the product separately — it channels genuine crowd wisdom toward individual decisions. But when social proof operates as a substitute for independent judgment — when you buy a stock because the price is going up and you assume the crowd must know something — it destroys the independence that crowd wisdom requires. The difference between wisdom and folly often comes down to whether the crowd's members influenced each other before or after forming their views.

The Overton Window constrains which conclusions a crowd can reach. If certain ideas or possibilities fall outside the window of socially acceptable positions, crowd members will self-censor to avoid them, and the crowd's aggregated judgment will reflect the boundaries of the window rather than the full range of available evidence. A crowd asked to estimate the probability of a politically sensitive event will produce different answers depending on whether the answers are anonymous or public — because public answers are constrained by the window, and anonymity removes the constraint.

Groupthink is the specific failure mode that emerges when crowd wisdom conditions are violated within organizations. Janis's eight symptoms of groupthink — illusion of invulnerability, collective rationalization, belief in inherent morality, stereotyping of out-groups, pressure on dissenters, self-censorship, illusion of unanimity, and self-appointed mindguards — are essentially a point-by-point description of how diversity, independence, and decentralization collapse within a cohesive group.

Tall poppy syndrome suppresses the diverse viewpoints that crowd wisdom requires. When individuals who hold contrarian positions know that voicing those positions will result in social punishment — being labeled difficult, being excluded, having their other contributions scrutinized more harshly — they self-censor. The crowd loses access to precisely the independent signals that would make its collective judgment valuable.

The Independence Check: A Self-Test

The practical application of crowd wisdom is a discipline: before you aggregate, verify the independence. The self-test is the independence check — when you find yourself about to form an opinion on something, ask whether you are drawing on your own information and analysis or whether you are recycling the conclusions of others.

The internal experience is a subtle gravitational pull. You read a headline, see a trending topic, notice what your peers seem to believe, and feel your own assessment bending toward alignment with the perceived consensus. The pull feels like understanding — it feels like you have "figured out" what is going on — but what has actually happened is that social information has overwritten private information. The trigger situation is any moment when your opinion on a question forms unusually quickly and with unusual confidence, and you notice that you have recently been exposed to strong social signals pointing in that direction. Speed and confidence in the presence of social influence are the warning signs that independence has been compromised.

Galton went to the Plymouth livestock fair expecting to prove that the average person was not competent to make informed judgments. What he discovered instead was that the average of all those "incompetent" judgments was more accurate than any expert's. But the discovery came with a condition that Galton's one-paragraph paper did not emphasize and that a century of subsequent research has confirmed: the crowd's accuracy depended entirely on the fact that each person estimated independently, without seeing the others' guesses. Had they called out their numbers one by one, each influenced by the last, the wisdom would have evaporated, and the crowd would have been no wiser than the loudest voice in it. The magic is in the independence. Protect it, and the crowd is a precision instrument. Destroy it, and the crowd is a stampede.