Potential Problems Behind Crowd Wisdom
In class, we discussed potential problems behind information cascades. When individuals follow the crowd, they sacrifice their own private information in favor of crowd wisdom. Usually, this is a smart move, because the collective knowledge tends to be more accurate than one’s own knowledge. However, crowd wisdom is not necessarily very accurate. For instance, in the marbles experiment discussed in lecture, the crowd based their information solely on the first two high signal results, even though their choice might be wrong. A study conducted by researchers from ETH Zurich further expands upon this problem.
In the experiment, the subjects were given a set of questions regarding geographical questions and crime statistics. Each subject initially answered the question without any outside information. They were later given the option to choose between their own answer and the average estimate of everyone else’s answer. The researchers offered monetary payments for good estimates as an incentive for the subjects to answer to the best of their ability. The results of the experiment showed that when the subjects chose the “crowd-sourced answer”, it was a generally accurate estimate. However, this approach was subject to several errors. Averaging everyone’s answers and disregarding one’s “private answer” diminished the diversity of answers without improving the accuracy of the collective answer. Furthermore, averaging the answer also potentially moves the collective answer from the truth (in this case, the actual answer to the question). Lastly, the experiment showed that the subjects generally had more confidence in the collective answer, even if it were more inaccurate than their actual answer.
The experiment exposes the potential problem in relying on crowd-sourced information. If information sourced from the crowd is inaccurate, it could be difficult to switch the flow of information to a more accurate answer. The collective answer could migrate farther from the truth and it would be difficult to change it around because those in the crowd have a high, but erroneous confidence in their crowd sourced information.
source: http://www.pnas.org/content/108/22/9020.full.pdf_html