In the early twentieth century, the British researcher Francis Galton tried to prove how stupid the masses were: at a cattle market, he had people guess the weight of an ox and was amazed to find that the average of the guesses were very close to the animal’s actual weight. Galton had inadvertently proven the opposite: a crowd is more likely to find the correct answer than an individual. This has been referred to in the textbooks as the ‘Wisdom of Crowds’ ever since. Almost a hundred years later, James Surowiecki became a best-selling author when he provided many examples of the wisdom of crowds – from stock markets and elections to quiz shows.

However, a new study at ETH Zurich now reveals that the masses can get it wrong after all and miss the best answer by a long way. This can happen even if they are only mildly influenced from outside, as revealed by an experiment conducted by an interdisciplinary research team including mathematician Jan Lorenz, sociologist Heiko Rauhut and the two ETH-Zurich professors Dirk Helbing and Frank Schweitzer. The researchers asked a total of 144 students to estimate facts such as the population density of Switzerland or the length of the border between Italy and Switzerland. In two different experimental settings, they were informed either of the average estimate or of the complete picture, i.e. all individuals’ estimates. These two conditions were compared with a control group in which the participants did not receive any information on the estimates of others. To stimulate good performance, the students were given a financial incentive: the better the answer, the more money they received.

Influenced in the wrong way

The results of the surveys showed the researchers that even a minor social influence – in this case knowledge of the other participants’ guesses – undermines the ‘Wisdom of Crowds’. The estimates began to converge, resulting in a consensus. At the same time, the collective error grew and the correct answer ultimately shifted to the peripheral regions of the spectrum of estimates. The group as a whole was therefore not smarter. The participants were lulled into a false sense of security due to the closeness of their guesses.

When people answer questions, it seems that they want to keep an eye on the opinions of others, just because these could have some extra piece of information. Furthermore, the desire for harmony might be stronger than the search for the correct answer.

The convergence of the judgements suggests that the group thinks they have already found or will find the correct answer. ‘But this is no guarantee that the answer is correct,’ stresses Rauhut. This can also be observed for Internet users: if numbers of downloads are published with it, a mediocre song can become a hit simply by being in the right place. Assuming that many downloads imply quality, users are more confident that the song must be good.

Problems of the search for consensus

In matters of taste, this might not be a problem. But the participants’ confidence in their answer increased as the majority of others arrived at a similar conclusion, even if the estimates were wrong. ‘It’s a problem of our culture that we search for a consensus,’ says Helbing. The herd instinct, coupled with over-confidence, was also one of the reasons for the financial crisis. Warnings were ignored; financial experts influenced other financial experts; anyone who called the problems by name was forced to shut up. Peer and conformity pressure ultimately promoted wrong judgements and solutions.

‘In general, the less you are influenced, the better the average estimates. The independence of the individual judgements is extremely important,’ stresses Helbing, who is also running the FuturICT project to reduce socio-economic crises. This does not just hold for the stock exchange or politics, but also for science. ‘Pluralism is better than consensus.’

Social influence everywhere

However, the scientists admit that these days it is hardly possible to judge something without being influenced. In today’s society, it is difficult to get independent judgements as most people are embedded in social networks and thus influence each other all the time. ‘It is astonishing that even minor social influence can trigger herd behaviour and other negative side effects,’ concludes Helbing. ‘If peer pressure and cooperative or competitive aspects are added, social influence can bias judgements even more.’

Further reading

Lorenz J, Rauhut H, Schweitzer F, Helbing D. How social influence can undermine the wisdom of crowd effect. PNAS early edition. www.pnas.org, doi:10.1073/pnas.1008636108