Under the right conditions, a group can outperform an expert. The Internet abounds with examples of the phenomenon called the Wisdom of The Crowds. Today U.S. intelligence threat assessments are based on collective wisdom.
Exactly 10 years ago today two airliners flew into New York's twin towers. While the world looked on with shock and horror, America suffered the terrorist attack of September 11 with special despair. For Lowell Jacoby, Director General of the Defense Intelligence Agency, DIA, the worst case scenario became a reality.
Jacoby was an advocate for better exchange of knowledge of U.S. intelligence. Not only between the CIA and FBI, but also from the network of informants such as foreign embassies and domestic port offices.
With the new terrorist attack, the U.S. security apparatus reached a low point. In less than 10 years four significant attacks against U.S. interests had slipped under the radar. In 1993, the terrorists attempted to destroy World Trade Center. In 1998 the U.S. embassies in East Africa were hit, and in 2000 the destroyer USS Cole was bombed in Yemen.
After 9/11, there was a Congressional inquiry which quickly showed that closed hierarchies and tradition-bound thought were key problems. As DIA Director Lowell Jacoby put it:
"Information considered irrelevant noise by a group of analysts can provide an essential trace or detect a significant relationship when they are thoroughly tested by another group."
The experts' solution was called FutureMAP and Policy Analysis Market (PAM). These were two markets or stock exchanges, where, in simple terms, guesses about the future were bought and sold. The first was reserved for the intelligence service's own circles, preferably in ad hoc groups of under 30 members. The second, PAM, would be placed in the Middle East and be publicly accessible.
The political storm was not long in coming. The project was called "totally inane" and liquidated after a few years. Especially because prominent policy-makers in the U.S. found it indecent to bargain about such things. A strange subject, believes American economic journalist James Surowiecki, who through his book "The Wisdom of Crowds" has gained a significant position in the debate on the possibilities of the modern knowledge society. For is not that good intelligence makes every day?
If the U.S. intelligence functioned optimally, it would have appreciated the information that at flight schools around the U.S. there were some students of Arabic descent who were indifferent to learning how to land an aircraft. Perhaps noise to some, but quite relevant knowledge to others, highlights Surowiecki. When market valuations are combined with openness, it will make for a gigantic sum of knowledge assessed in new ways with fresh eyes, he said.
The word crowd in this context does not mean a mob full of football fans, but a large enough group of people who by common effort evaluate several options and find the best solution. The trick is to find the best conditions for the wise crowd.
Could more effective knowledge-sharing have prevented the terrorist attack on the World Trade Center and the Pentagon a decade ago?
"The most important thing is to balance correctly the extent to which people copy each other, with considering their own individual knowledge," says Professor Yaneer Bar-Yam.
He researches complex systems at the New England Complex Systems Institute, better known as NECSI. In collaboration with MIT and Harvard, NECSI is studying the wise crowd.
"When they are more likely to copy each other, herding behavior dominates. When each person considers their own information and does a more limited amount of copying, the group can work effectively in response to the information that people share," explains Yaneer Bar-Yam.
Under optimal conditions, there is virtually no limit to what the "Wisdom of the Crowds" can be used for. Professor Bar-Yam points to a couple of examples of networks which illustrate this kind of group behavior:
"Even though Wikipedia is not flawless, it is a good example of how people can collaborate in an efficient manner. Another good example is the 'open source' movement that develops software."
Researchers at NECSI also look at how a lot of making his choice from a collective knowledge. Professor Bar-Yam lists three ways:
"(1) Everybody expresses an opinion about a single issue (whether to do A or B). The end result is a decision to do A or B, (2) Everybody does part of the task and together the actions of many individuals combine together, this is the way wikipedia works mostly. (3) Everybody only knows part of the ideas that are relevant to making a decision A or B, and even though they are not expressing an opinion about A or B, together they make a decision."
"This is perhaps the most difficult idea to understand, but it is actually the most powerful. It is like creating a good sports team where each individual does something different but they work together to make the best team action," explains Yaneer Bar-Yam.
A wise mass is created not only by good relationships and decisions, says economics journalist James Surowiecki. Whether a crowd has a chance to act wisely depends, inter alia, on it consisting of sufficiently different people with different opinions comfortable working independently. Both with a boss and each other. The mass must also have a decentralized organization, and be able to collect and disseminate its knowledge.
Together it makes a clever and self-regulating mass. When herd mentality trumps self-regulation, however, the crowd can act stupidly.
In the 1960s the art group Provo tried to eliminate Amsterdam's traffic by placing white painted bicycles free of charge. After one month, bicycles were destroyed, stolen or tossed in one of the city channels. The project failed because those who supported it were fewerthan those who wanted to vandalize it.
Wikipedia is the opposite. In the book "Here Comes Everybody" Internet sociologist Clay Shirky explains how problematic articles, e.g. those about Islam, survive. When the article is hacked or deleted the original is the restored and online within two minutes. Unlike in the bike project, vandals are trumped almost immediately.
Although the American politicians in 2003 rejected the idea, the wise crowd survived as a "soothsayer". Thanks to Internet penetration and increased computing power, collective knowledge can be used to conduct surveys of a given field and calculate likely outcomes.
"In private business financial markets are already used to assess the likelihood of different scenarios. And everything suggests that decentralized and user-generated markets can also be used to for prediction, for example of political events," says Henrik Johannsen Duus, working with Strategic Forecasting as an associate professor at CBS, Copenhagen Business School.
American intelligence abandoned the idea of a joint information center. In 2006 the CIA created a wiki called Intellipedia. A closed network for 3,600 users with the ability to share information in groups and upload pictures and video clips. Among the contributors are the CIA and NSA. Today, Intellipedia has more than 28,000 articles such as threat assessments and terrorist activities.
For Yaneer Bar-Yam this is a step in the right direction but not the perfect solution: "It would be better to create an open network where information flows by itself. For example, using Twitter. To expect that people could find information stored in a closed wiki is not good enough."
The hope for FutureMap and PAM was that the two markets could act as a wise crowd. The idea is tantalizing, but involves a number of problems. First and foremost, an assessment can affect the outcome. You must be a stupid terrorist if you do not change your plans when they are on the list of probable future events. Next, it is impossible to determine whether an incident was likely when U.S. intelligence comes into action to prevent it. Last but not least a "guess" about the future needs to be concrete.
"The main problem with this idea is that to make a prediction, you have to be able to ask the right questions," points out Yaneer Bar-Yam.
Had FutureMAP or PAM, despite the theoretical problems, been running before 9/11, they would probably have hoisted a few red flags before the terrorist attack, says Henrik Johannsen Duus. Especially if they were paired with other types of analysis software.
"The result would not necessarily show what was wrong, but that something was wrong. There is no doubt that when we analyze different situations in different ways, there are remarkable things," says Henrik Johannsen Duus.
In the autumn of 1906 the British scientist Francis Galton was at the cattle show. Here he came upon a competition where you had to guess the weight of a slaughtered ox. Galton was convinced that the crowd was the same as an ignorant mob. To prove his theory Galton asked about the 800 ballots. The results showed the opposite. Many guesses were entirely in the woods, but their average value was a pound under the ox's processed weight of 1,198 pounds.
In 2005 Nature magazine did a comparison between Wikipedia and the Encyclopedia Britannica. Nature asked experts in various fields of research to assess 50 articles in both encyclopedias. The result showed that there were 2.9 errors per article in the Encyclopedia Britannica, while Wikipedia had 3.9 errors.
Phone: 617-547-4100 | Fax: 617-661-7711 | Email: office at necsi.edu
277 Broadway Cambridge, MA USA