The phenomenon known as “collective intelligence” (in their various appellations: “intelligence of crowds”, “crowdsourcing”,….) has received a great attention in recent years, especially since the emergence of the Internet and social networks. And that is, we are told, more interaction and communication could lead to the crowds to create knowledge, or find answers, that would be out of the reach of particular individuals. Well: thanks to the blog Not exactly rocket science, Ed Yong, I have discovered the publication of a couple of investigations which appear to show that the social influence can reduce the intelligence of the crowds.
The first experiment was carried out by Andrew King, a researcher at the Royal Veterinary College, and his co-workers. King was asked to 82 individuals, in isolation, to try to guess how many sweets there were in a boat. The average of the estimates of the 82 subjects was 751. Surprisingly, the actual number of sweets was 752.
In a second movement, King was made known to the various sub-groups of those 82 individuals, what was the estimate which had been made by other colleagues (both in the form of an estimate chosen by King random, as in the form of the average of estimates made before by other individuals). With this information, the intelligence of the collective is affected: the average of estimates passed to range from 882 to 1109.
The second experiment is very similar to that of King, and was carried out by Jan Lorenz and his co-workers. This time, the researchers used 144 students francs, to which, in separate cubicles, they respond to questions very specific, as for example: “what is the population density in Switzerland in inhabitants per square meter?”, or “how Many new people has earned Zurich?”.
As in the experiment of King, when students did not have any information of what was the estimates of their peers, the average of estimates collectively for each question closer to the answer (in the case of the new inhabitants of Zurich, for example, the average of the estimates was 10,000, and the actual value 10.067).
But when Lorenz was made known to the students what had been some of the estimates of his colleagues, collective intelligence impaired: the range of estimates was severely limited around to values that were far from the solution. Interestingly, in addition, the students increased their level of confidence with regard to the possible certainty of his answer, even when what they were doing considerably worse than before.
But this descent into the collective intelligence seems to be reversible. Tells Ed Yong that when King, in a third movement in his study, made known to the individuals what had been the best estimate up to the moment (without revealing the solution), the average of estimates of the collective recovered to a value closer to real (795 estimated, compared to 752 real). In the words of Yong:
So the crowd loses stis wisdom when it gets random pieces of information about what sti members think, but it regains sti wisdom if it finds out what the most successful individual said.
This latter finding gives rise to Yong to make a reflection that fully endorse. The experiments of King and Lorenz are contrived situations, in which the experimenters know the solution to the question. And that rarely happens in real life. Thus, the results of King and Lorenz have more to do with the importance of choosing well who to follow and who to trust when we make our decisions with the intelligence of the crowds itself.
This of course leaves us with a new problem, that is repeated again and again in life (especially in digital): who do you follow?, how can we determine which opinions are reliable and which are not?
These issues are important due to their potential negative effects, and are receiving more and more attention, although still in a relatively unstructured way. As an example, we have a review of Jonah Lehrer’s study of Lorenz, or the recent article by Javier Celaya, where the author reports a paradox that already pointed in the previous post of this blog: although the information does not stop growing, it seems that is not translating into a deeper understanding and more structured; on the contrary, Celaya maintains that it is producing a copy going on between the users of the Internet, with the impoverishment cognitive that this entails.
In the end, the question is: do you really take advantage of the crowds to create and share new knowledge, or individuals are limited to copy and assimilate content more popular? Perhaps some day see how there is a new field of study, trans-disciplinary, that is able to answer this question, and others related to: a true epistemology of the Internet. For the moment, what do you think?
Image of arribalasqueluchan!