One person choosing alone can usually be trusted to land somewhere near the middle. Put twenty of those people in a room, let them talk, and ask for one shared answer, and something odd happens to that middle. It doesn’t settle toward safety, and it doesn’t flatten into the average of everyone’s private picks. It slides outward, past where anyone was standing before they opened their mouth. A graduate student noticed this in 1961 while working on a completely different question, and the effect he stumbled into is one of the few findings in social psychology that has held up for more than sixty years.
🧠 The Student Who Wanted to Study Risk
James Stoner was a master’s student at MIT when he ran a set of experiments for his 1961 thesis. He handed people a questionnaire of life dilemmas. Each item asked how low a probability of success you’d accept before choosing the risky option: take the secure job with modest pay, or the venture that might pay off big and might leave you with nothing. People answered alone first. Then Stoner put them in small groups, had them discuss the same dilemmas, and asked each group for one joint answer.
The groups chose more risk than the average of their own members had chosen minutes earlier. Nobody in the room had pushed them toward it. The discussion did that by itself. Stoner called the result the risky shift.
Eight years later, in 1969, Serge Moscovici and Marisa Zavalloni ran a version with French students evaluating Charles de Gaulle and Americans. The students who started out favorable came out more favorable; the ones who started out critical came out more critical. The direction wasn’t toward risk at all. It was outward, away from the center. That study gave the effect its real name: group polarization.
⚙️ How It Works
Everything starts with one condition. The group is not evenly split. Most people already lean the same way, even if by different amounts. If the room is genuinely divided, you get something closer to a tug-of-war instead of a slide.
Then two engines switch on at the same time.
The first is arguments. Everyone carries around a handful of reasons for their view, and no two people carry quite the same set. Some of those reasons are shared, meaning the whole room has already heard them. Others are unshared, known to just one person. Because the group leans one way, most of those unshared reasons point the same way too. As people talk, each person hears new reasons they hadn’t weighed before, and nearly all of them support the side they already favored. The pile of evidence gets taller. You don’t feel persuaded. You feel confirmed.
The second is comparison. At the same moment, everyone is quietly reading the room to work out where the acceptable position sits. Nobody wants to be the one lagging behind, so each person stakes out a spot a little past what they think the group’s average is, the same reflex that makes you want to sound slightly more committed than the person next to you. But everyone is doing this at once, so the average moves, and the whole cluster drifts a few degrees further out.
The two engines feed each other. More arguments make you more confident, and more confidence makes it easy to take a bolder public stand, which raises the bar everyone else is measuring against. Stack the effects and the group’s joint answer lands more extreme than any single member’s private answer. A 1986 meta-analysis by Daniel Isenberg settled the long fight over which engine mattered: both run at the same time, in different proportions depending on the situation.
🤔 The Twist
The instinct here is that discussion makes groups wiser, that pooling many minds beats one mind alone. Group polarization says the opposite about a room of like-minded people. Talk doesn’t pull them back to the middle. It carries them past everyone’s starting point.
The reason is independence. The “wisdom of crowds” works only when each judgment is formed separately, because separate errors cancel out in the average. Let people talk first and that independence is gone. The errors stop cancelling and start agreeing, and the crowd’s answer drifts away from the truth instead of toward it. This is why online spaces can be worse than a physical meeting room. In a 2002 study by Sia and colleagues, discussions held online, where people couldn’t see one another and could stay anonymous, polarized more than face-to-face ones. More unexpected arguments surfaced, and so did more of the quiet competition to sound the firmest.
🔗 Why It Matters To You
If you have sat on a small committee that was meant to gather opinions and produce a decision, you have watched this run live. A room of people who all lean the same way tends to walk out more confident than it walked in, and the extra confidence was manufactured by the room, not earned from the evidence.
The same shape appears anywhere a group reinforces itself. An investment committee talks itself into a bolder number. A fan community confirms the same read back to itself every day. A chatbot that agrees with you no matter what isn’t neutral at all; it is a one-person echo chamber, holding the door open while you drift further out. A companion that never disagrees is the most comfortable version of this and the least useful one.
Writing has its own version. A cast of characters who all believe the same thing will read as obviously correct to each other and obviously flat to anyone outside the room. The character who stays in the middle is usually the one the reader trusts.
🎲 A Small Trick
There is a cheap way to disarm the whole thing. Before the group discusses anything, have everyone write their position down privately, and collect the answers. The average of those private judgments is the group’s real center, and it is also where the errors still cancel.
Cass Sunstein’s related point is simpler still: one person willing to say “I’m not sure” out loud can break the spell, because they give permission to the silent half of the room who were already uneasy. Order matters. Write first, add up, then talk about the disagreements. Run it the other way and you mostly amplify.