If you have a question about what’s going to happen next in Syria or North Korea, you might get more accurate predictions by asking a group of ordinary people than from foreign policy experts or even, possibly, CIA agents with classified information. Over the last few years, the Good Judgment Project has proven that crowdsourcing predictions is a surprisingly accurate way to forecast the future.
The project, sponsored by the U.S. Director of National Intelligence office, is currently working with 3,000 people to test their ability to predict outcomes in everything from world politics to the economy. They aren’t experts, just people who are interested in the news.
“We just needed lots of people; we had very few restrictions,” says Don Moore, an associate professor at University of California-Berkeley, who co-led the project. “We wanted people who were interested, and curious, who were moderately well-educated and at least aware enough of the world around them that they listened to the news.”
The group has tackled 250 questions in the experiment so far. None of them have been simple; current questions include whether Turkey will get a new constitution and whether the U.S. and the E.U. will reach a trade deal. But the group consistently got answers right more often than individual experts, just through some simple online research and, in some cases, discussions with each other.
The crowdsourced predictions are even reportedly more accurate than those from intelligence agents. One report says that when “superpredictors,” the people who are right most often, are grouped together in teams, they can outperform agents with classified information by as much as 30%. (The researchers can’t confirm this fact, since the accuracy of spies is, unsurprisingly, classified).
“To the extent we have an advantage over those highly informed insiders, it is that our forecasters really don’t care about anything other than being accurate,” says Moore. “If you’re inside–say in the CIA or at the State Department–of course there are benefits to being accurate, being able to accurately forecast things. But if you’re inside any organization, there are more complicated political winds that blow.”
Crowdsourcing could be useful for any type of prediction, Moore says, not only what’s happening in world politics. “Every major decision depends on a forecast of the future,” he explains. “A company deciding to launch a new product has to figure out what sales might be like. A candidate trying to decide whether to run for office has to forecast how they’ll do in the election. In trying to decide whom to marry, you have to decide what your future looks like together.”
“The way corporations do forecasting now is an embarrassment,” he adds. “Many of the tools we’re developing would be enormously helpful.”
The project is currently recruiting new citizen predictors here.