In 1906, a British scientist named Francis Galton went to a country fair where people were paying to guess the weight of an ox. After the contest ended, he gathered the eight hundred guesses, expecting most of them to be wildly inaccurate. The crowd was made up of regular people—farmers, merchants, kids—not livestock experts. When he averaged all the guesses, he got a number startlingly close to the actual weight of the ox. Not just close. Within one pound. The average guess was more accurate than nearly every individual guess in the crowd. For decades, this finding was treated as a curious anecdote. But in recent years, researchers have discovered that Galton stumbled onto a deep statistical principle—one that quietly shapes everything from prediction markets to scientific consensus to AI systems.

It turns out that crowds, under the right conditions, really do think better than the individuals inside them. Not because the people are smart, but because their errors cancel each other out. What feels like collective intelligence is actually math. And once you understand the mechanics, the wisdom of crowds stops looking like magic—and starts revealing both why groups can be brilliant and why they sometimes fail catastrophically.

‍

Why Groups Sometimes Outperform Experts

If you've ever wondered how a random group of strangers can outperform a single expert, the answer comes down to one statistical fact: averaging reduces error.

When individuals make independent guesses about a quantity, their errors tend to be scattered. Some guess too high, some too low, some land in the middle. The errors aren't perfectly symmetrical, but they're often close enough that, when averaged, they cancel out.

The result is a group estimate that's frequently more accurate than the typical individual estimate—and sometimes more accurate than the best expert in the room.

This effect has been replicated across thousands of contexts. Estimating jellybean counts. Predicting stock prices. Guessing election outcomes. Forecasting weather. Even diagnosing rare medical conditions when groups of doctors share their independent assessments.

When the conditions are right, the crowd is genuinely smarter than its smartest member.

The Four Conditions That Make It Work

Crowds aren't automatically wise. They're only wise when four specific conditions are met.

The first is diversity. The members of the crowd need different perspectives, knowledge bases, and ways of approaching the problem. A homogeneous crowd just amplifies the same errors.

The second is independence. Each person needs to form their estimate without being influenced by the others. The moment people start looking at each other's answers, the math breaks down.

The third is decentralization. People need to be drawing on their own local knowledge and experience. A crowd where everyone is reading from the same source isn't really a crowd—it's an echo.

The fourth is aggregation. There needs to be a way to combine the individual answers into a single estimate, whether through averaging, voting, or some other mechanism.

When all four are present, crowds become remarkably accurate. When even one is missing, the wisdom collapses.

Why Independence Is the Most Fragile Condition

Of the four conditions, independence is by far the easiest to lose.

Humans are deeply social. We pay attention to what others think. We adjust our opinions toward the group, often without realizing it. This is sometimes called information cascade—the process where each person's decision is influenced by the people who decided before them.

A famous experiment showed how easily this happens. When people were shown a clearly correct answer in a group setting, but watched several confederates pretend that an obviously wrong answer was correct, a substantial number of participants started agreeing with the wrong answer. Not because they didn't see the truth—but because the social pressure outweighed the perception.

This is the failure mode at the heart of every viral misinformation event. Once enough people appear to believe something, the pressure to align with the group can override individual judgment. The crowd's wisdom turns into the crowd's certainty—and certainty is much harder to correct.

How Diversity Beats Expertise

One of the most counterintuitive findings is that diverse groups often outperform expert groups, even when the experts are individually more knowledgeable.

The reason is mathematical. Expert groups tend to share the same training, the same assumptions, and the same blind spots. Their errors line up. When you average their guesses, the errors don't cancel—they reinforce.

A diverse group, by contrast, makes errors in different directions. The errors fight each other into the noise, leaving the signal behind.

This is why decision-making teams that mix backgrounds, experiences, and viewpoints often produce better outcomes than teams of similar specialists. It's not because the diverse team is individually smarter. It's because the team's collective error map is more spread out.

When Crowds Get Catastrophically Wrong

The same mechanisms that make crowds wise can make them disastrous when the conditions break.

Stock market bubbles are a classic example. When investors stop making independent decisions and start following each other, the wisdom of the crowd disappears. Prices stop reflecting underlying value and start reflecting collective expectation. Everyone is buying because everyone else is buying.

The same dynamic plays out in panics. Bank runs. Fashion fads. Political mobs. Social media pile-ons. The crowd isn't aggregating independent information anymore—it's amplifying a single shared signal until the signal overwhelms reality.

In moments like these, individuals who hold independent views often see the truth more clearly than the crowd. But they also pay a social price for going against it.

The wisdom of crowds and the madness of crowds are the same machinery, running with different inputs.

How AI Quietly Uses This Principle

The wisdom of crowds isn't just a human phenomenon. It's also one of the foundational ideas in modern AI.

Many machine learning systems use a technique called ensemble learning, which combines the predictions of many different models to produce a single more accurate prediction. This is essentially Galton's ox-weighing strategy, applied to algorithms.

Each individual model has its own errors. By averaging or voting across the ensemble, the errors cancel out and the accurate signal rises to the top.

Even large language models benefit from a version of this. When researchers ask a model the same question multiple times and aggregate the answers, the result is often more accurate than any single response. The same statistical principle that helped Galton predict an ox's weight helps AI systems predict almost everything.

Why Some Questions Aren't Crowd-Solvable

Crowds aren't always useful. There are entire categories of questions where the wisdom of crowds doesn't apply.

Highly technical questions where most people lack the foundational knowledge to make a meaningful guess.

Questions involving rare or specialized expertise, where one person's opinion is genuinely worth more than a thousand others.

Questions that require sustained reasoning rather than quick estimation.

Questions about novel situations with no historical pattern to draw on.

For these, individual experts—or small groups of experts working carefully—often outperform the crowd. The wisdom of crowds is real, but it's a tool with a specific use case, not a universal answer machine.

The Democracy Connection

The wisdom of crowds is also the implicit logic behind democratic systems.

The reasoning isn't that voters are individually well-informed. They often aren't. The reasoning is that aggregating millions of independent judgments tends to produce better outcomes, on average, than relying on a single ruler—because the errors of voters cancel each other out while the errors of a single decision-maker don't.

This logic only holds, however, when the four conditions are met. If voters lose independence—because they're all consuming the same media, following the same influencers, or pressured by the same social environment—the wisdom collapses.

Modern democracy faces this challenge directly. Information cascades, algorithmic amplification, and social media pressure all push voters toward correlated rather than independent judgments. The math that makes democracy work depends on the diversity that makes it hard.

A Smarter Way to Think About Group Decisions

When you understand how the wisdom of crowds actually works, you can use it more deliberately.

Focus on:

  • Gathering opinions independently before discussing them as a group
  • Including genuinely different perspectives, not just different people
  • Aggregating answers mathematically, not just emotionally
  • Being skeptical of group consensus that formed quickly
  • Recognizing that confidence in a crowd is not the same as accuracy

The instinct to assume crowds are either always wise or always foolish misses what's actually happening. A condition-based view treats crowds as systems that succeed or fail based on structure.

Intelligence Lives in the Spaces Between People

For all the focus on individual brilliance, the deeper truth is that some of the most reliable forms of intelligence don't live in any single brain. They live in the structure of how multiple brains combine.

A jellybean jar. An ox at a fair. A market. An election. A scientific consensus. A neural network ensemble. Each one is a system where the right kind of combination produces accuracy that no individual could match alone.

And that's the shift: from seeing intelligence as something contained inside a person to seeing it as something that can emerge between people, when the conditions are right.

Because once you understand how collective wisdom actually forms, you stop being surprised that crowds can be brilliant—and start being careful about when they're not.

‍