Groupthink Can Actually Be Good, If It’s Done Right

The wisdom of crowds has been taking a battering lately. Last November, every major news outlet and pollster said Donald Trump would lose the presidential election, and just last week there was similarly unanimous consensus that Britain’s Conservative Party was headed to a big win in its own election. Both seem like perfect illustrations of the dangers of groupthink, of the massive errors that can creep in when people all bring their opinions in line with one another.

Not so fast, says University of Pennsylvania researcher Damon Centola. His research, published in this week’s issue of the Proceedings of National Academy of Sciences, suggests that the classic way of looking at groupthink — in which collaborative decision-making gets worse the more people know what others are thinking — is wrong. People can actually benefit from knowing what others are guessing in uncertain situations. It’s only when some people have too much influence over the crowd that the really big errors creep in.

Centola’s study had more than 1,300 people take part in online surveys where they were given three tries to estimate things like how many calories were in a given meal. The participants were put in one of three scenarios. Some were just given three tries, with no further information about what others were guessing. Some were put in social networks where everyone’s input was given equal weight, and the participants saw their fellow group members’ estimates after each round of guessing. The rest were in groups where certain people were singled out as important, and others in those groups could only see those influential people’s guesses.

Over three rounds of guessing, only those in that second group, where everyone’s guesses were equally important, got more accurate. The other two sets either showed no difference in accuracy or actually got worse the more opportunities they had to guess. Even in cases where the chosen opinion leaders were good at guessing one estimate, they generally proved inaccurate for all the other tasks.

The key then to preventing groupthink might not be to keep everybody wholly independent in their decision-making — which Centola says is what scientists and engineers often strive for when making big collaborative decisions out of fear of groupthink — but rather to ensure that nobody holds too much sway when people compare notes. His own work has focused on how this approach could prove useful in medicine, with doctors potentially using social networks to work together on tricky cases and make better judgments.

The finding also makes a lot of sense in contact with the recent run of poorly predicted elections: Maybe the problem wasn’t that everyone agreed Hillary Clinton and Theresa May were cruising to big victories, but rather a lot of the people most certain of that outcome were also some of the most powerful, influential voices in the political and media establishment. That’s exactly the kind of outsize influence Centola’s research suggests is at the heart of groupthink at its most error-prone and wrongheaded.