proving the wisdom of crowds -- 10/2/19
Today's selection -- from Superminds by Thomas W. Malone. The wisdom of crowds:
"When lots of nonexperts vote on possible answers to a question, they often arrive at an answer that is just as good as one that an expert would give. In the Galaxy Zoo project, for instance, hundreds of thousands of online volunteers are helping astronomers by classifying the shapes and other characteristics of a million galaxies in distant parts of the universe that astronomers have previously observed through telescopes. Even though a single volunteer might mistakenly classify an astronomical object, when many volunteers look at that same object and vote on how to classify it, the results of the group's votes are extremely accurate, allowing the classification to happen much faster than if it were being done by a handful of experts.
"In the Eyewire project, led by my former MIT colleague Sebastian Seung (now at Princeton), online volunteers use a similar voting approach to help neuroscientists map the neural connections in detailed images of the brain taken with an electron microscope. Interestingly, the human volunteers work together with AI algorithms, indicating connections the algorithms have missed and letting the algorithms complete the marking of new connections the humans notice. This approach makes it possible to map far more parts of the brain far more rapidly than would be possible if all the mapping had to be done by professional scientists and their paid assistants.
|A dense crowd in London in 1908.|
"Counting votes is a fairly simple way of combining people's opinions, but some of the most interesting examples of what you might call democratic truth finding involve more than just voting; they rely on more sophisticated ways of combining opinions. For instance, the Good Judgment Project, led by Philip Tetlock at the University of Pennsylvania, was part of a competition organized by the US intelligence community's Intelligence Advanced Research Projects Activity (IARPA). Teams led by different universities competed to develop innovative methods for predicting the answers to a wide range of questions about geopolitical events, such as:
· Will Serbia be officially granted European Union candidacy by December 31, 2011?
· Will the six-party talks on the Korean peninsula resume before January 1, 2014?
· Will the London Gold Market Fixing price of gold (USO per ounce) exceed $1,850 on September 30, 2011?
"In each case, the groups didn't make simple yes-or-no predictions. Instead they tried to estimate the probability of the events occurring. New questions usually had end dates several months in the future, and groups were able to update their predictions every day until the end date.
"The Good Judgment team was the hands-down winner of the IARPA competition."