The Wisdom of Crowds and Its Limits

When Crowds Are Wise

In 1906, Francis Galton attended a county fair where 787 people guessed the weight of an ox. The average of all guesses was 1,197 pounds — just one pound off the actual weight of 1,198. No individual guess was as accurate as the crowd's collective average. James Surowiecki's influential 2004 book 'The Wisdom of Crowds' argued that groups outperform individuals under specific conditions: diversity of opinion, independence of judgment, decentralization, and a mechanism for aggregating individual views.

This has direct implications for direct democracy. If citizens' individual judgments are independent and diverse, the majority vote will tend to produce good outcomes — even if no individual voter is an expert. The Condorcet Jury Theorem formalizes this: if each voter is more likely than not to choose the correct answer, the probability that the majority is correct approaches 100% as the number of voters increases.

Implications for Direct Democracy

The science of collective intelligence suggests that direct democracy can produce good outcomes — but only when the conditions are right. Citizens need access to accurate information (which requires a functioning media ecosystem and resistance to misinformation). They need to form judgments independently (which is undermined by partisan media bubbles and social pressure). And the questions put to them need to be ones where collective judgment adds value (factual or value questions) rather than ones requiring technical expertise.

This analysis suggests that direct democracy is not inherently good or bad — its quality depends on the institutional design around it. Citizen assemblies, which provide information and enable independent deliberation, satisfy the crowd wisdom conditions better than referendums, which often feature misinformation campaigns and social pressure. The question is not whether citizens can decide well, but whether the institutions around them are designed to help them do so.

Worked Example: The Condorcet Math — and Where It Breaks

The Condorcet Jury Theorem is worth seeing in numbers. Suppose each voter has just a 51% chance of picking the correct answer on a binary factual question — barely better than a coin flip. The theorem says that as you add independent voters, the majority's accuracy climbs sharply: with 1,000 such voters the majority is correct roughly 73% of the time, and with 10,000 it exceeds 97%. Tiny individual competence, aggregated across many independent minds, becomes near-certainty.

But the theorem has a fatal dependency: the word 'independent.' The moment voters copy each other — following a charismatic leader, a viral post, or a partisan outlet — their errors correlate and the math collapses. A crowd of 10,000 who all take their cue from the same source is no wiser than one person. This is why game-show audiences (independent, anonymous button-presses) are famously reliable, while stock-market crowds (watching each other and chasing the trend) inflate bubbles.

The lesson for direct democracy is precise: the goal of institutional design is to protect voter independence and information, because that — not sheer numbers — is what makes a crowd wise.