MIT Technology reports new research on the “wisdom of the confident”:
It turns out that if a crowd offers a wide range of independent estimates, then it is more likely to be wise. But if members of the crowd are influenced in the same way, for example by each other or by some external factor, then they tend to converge on a biased estimate. In this case, the crowd is likely to be stupid.
Today, Gabriel Madirolas and Gonzalo De Polavieja at the Cajal Institute in Madrid, Spain, say they found a way to analyze the answers from a crowd which allows them to remove this kind of bias and so settle on a wiser answer.
… Their idea is that some people are more strongly influenced by additional information than others who are confident in their own opinion. So identifying these more strongly influenced people and separating them from the independent thinkers creates two different groups. The group of independent thinkers is then more likely to give a wise estimate. Or put another way, ignore the wisdom of the crowd in favor of the wisdom of the confident.
To test this result, they eliminated those who updated their estimates based on that of the crowd:
Madirolas and De Polavieja began by studying the data from an earlier set of experiments in which groups of people were given tasks such as to estimate the length of the border between Switzerland and Italy, the correct answer being 734 kilometers.
After one task, some groups were shown the combined estimates of other groups before beginning their second task. These experiments clearly showed how this information biased the answers from these groups in their second tasks. …
That allows them to divide the groups into independent thinkers and biased thinkers. Taking the collective opinion of the independent thinkers then gives a much more accurate estimate of the length of the border.
The funny thing about this research is that anyone who believes in the wisdom of crowds and updates their belief based on that collective wisdom is then excluded from the collective estimate. The wisdom of crowds needs someone who trusts their own opinion more than that of the crowd. It is similar to the efficient markets hypothesis relying on those who don’t believe in it – if everyone believed markets were efficient, no one would invest effort in finding and acting on information that might affect market prices. That effort is what allows prices to reflect this information.
So who are the confident people who form this more accurate estimate? The Dunning-Kruger effect tells us that the unskilled will be overconfident as they don’t have the cognitive skills to recognise their ineptitude. But despite this effect, the more skilled do tend to be more confident than the unskilled – just not by as much as the skill gap warrants. As a result, eliminating the less confident can still cut the least skilled.