What makes one team of people smarter than another? A new field of research finds surprising answers.
For a century, people have been devising tests that aim to capture a person’s mental abilities in a score, whether it is an IQ test or the SAT. In just an hour or an afternoon, a slate of multiple choice questions or visual puzzles helps sift out the superstars — people whose critical thinking skills suggest they have potent intellectual abilities that could one day help solve real-world problems.
But separating the spectacularly bright from the merely average may not be quite as important as everyone believes. A striking study led by an MIT Sloan School of Management professor shows that teams of people display a collective intelligence that has surprisingly little to do with the intelligence of the team’s individual members. Group intelligence, the researchers discovered, is not strongly tied to either the average intelligence of the members or the team’s smartest member. And this collective intelligence was more than just an arbitrary score: When the group grappled with a complex task, the researchers found it was an excellent predictor of how well the team performed.
The new work is part of a growing body of research that focuses on understanding collective behavior and intelligence — an increasingly relevant topic of research in an age where everything from scientific progress to entrepreneurial success hinges on collaboration. Embedded in a century’s worth of Broadway shows, the interactions of online communities, or the path a ball travels between soccer players, researchers are finding hints about how individual people contribute to make a group creative and successful.
The interest is fueled in part by the Internet, which provides an unprecedented opportunity for people to join and leave groups, unbounded by geography. In the digital age, interactions between people are also creating a huge stream of data, giving scientists new ways to glean precise insights about how complex, aggregated behaviors arise. What they are finding is that groups, as entities, have characteristics that are more than just a summing up or averaging of those of its members.
“Intuitively, we still attribute too much to individuals and not enough to groups. Part of that may just be that it’s simpler; it’s simpler to say the success of a company depended on the CEO for good or bad, but in reality the success of a company depends on a whole lot more,” said Thomas W. Malone, director of the MIT Center for Collective Intelligence and senior author of the recent study, published in the journal Science. “Essentially what’s happening as our society becomes more advanced and more developed is that more things are done by groups of people than by individuals. In a certain sense, our intuitions about how that works haven’t caught up with the reality of modern life.”
As the mechanics of how groups work emerge, such insights are forming the basis of a scientific approach to engineering better groups, with experiments already unfolding in sports arenas and scientific laboratories. The best-selling book “Moneyball” told the story of how the Oakland Athletics used an unconventional statistical approach to build a winning baseball team without a big budget. The new research suggests it may one day be possible to give a test to a sales team and predict how well it will sell in the following year, or to pick a management team with a good sense of exactly how it is likely to respond to an array of challenges.
“It’s kind of staggering, it’s 2010 and we’re only beginning to realize what look in this paper to be very strong effects,” said Iain Couzin, an assistant professor of ecology and evolutionary biology at Princeton University who studies collective behavior in animals. “I run a relatively large lab, and I was thinking reading this paper about how I could make my lab more effective.”
People have been studying group dynamics for decades, seeing crowds variously as sources of madness and wisdom. Theories have arisen about people acting in plural, from the “groupthink” decision-making in the Bay of Pigs invasion to the “collective mind” of the flight operations on an aircraft carrier. But despite that long history, Malone and colleagues could not find an example in which people had asked the relatively simple question of whether groups had intelligence, the same way individual people do.
The field of measuring and ranking people’s mental aptitudes has been rife with controversy, but the finding that something called “general intelligence” exists has persevered. By giving people a set of tests, researchers can calculate a factor that predicts how a person will perform on a variety of cognitive tasks — as well as their performance in school and work. The MIT and Carnegie Mellon University researchers decided to see if the same concept applied to groups. While people have measured group performance on specific tasks, what Malone sought to understand was whether there was such a thing as general group intelligence.
In two studies, researchers divided 699 people into groups of two to five people. They measured each team member’s intelligence individually, but then gave the teams intelligence-testing tasks to solve — figuring out the next pattern in a sequence, brainstorming the different potential uses of a brick. Then, the group performed a more complex “criterion” task, such as playing checkers against a computer or completing a complicated architectural task with Legos, which was used to understand whether the collective intelligence researchers measured in the initial tasks correctly predicted the group’s abilities.
What the researchers found was that groups’ collective intelligence strongly predicted how well they did in the computer checkers game and on the Legos task — evidence that something called “collective intelligence” did in fact exist. What was more surprising, however, was that neither the average intelligence of the group members nor the person with the greatest intelligence strongly predicted how well the group did.
Other tenets of group success also seemed to fall by the wayside: A group’s motivation, satisfaction, and unity were unimportant. Instead, the researchers found that when a group had a high level of collective intelligence, the members tended to score well on a test that measured how good they were at reading other people’s emotions. They also found that groups with overbearing leaders who were reluctant to cede the floor and let the others talk did worse than those in which participation was better distributed and people took turns speaking. And they also found that the proportion of women in the group was a predictor of collective intelligence — a factor they believe was likely influenced by women’s generally superior social sensitivity.
Though intriguing, this work is just a first step. What Malone and colleagues are ultimately interested in is how to predict a group’s abilities in real-life scenarios — how they handle an environmental cleanup or design a blockbuster product. Legos and checkers are a surrogate for complicated tasks, but the ultimate test will be in determining whether collective intelligence truly predicts how teams, of all sizes, work on everyday tasks. Since groups and situations in the real world have fluidity and complexity whose individual components can be difficult to break down and measure, however, other research is focusing on dissecting the dynamics that build to group behavior.
Take the game of soccer: A player who never scores a goal may play an integral part in the team’s success. Baseball has its RBIs, ERAs, and OBPs, but, as with soccer, most real-life activities do not come with discrete statistical measures of how people interact or work together toward a goal. In a paper published in the journal PLoS ONE this summer, Luis A. Nunes Amaral, a professor of chemical and biological engineering at Northwestern University, worked with colleagues to see if they could quantify individual players’ performance by modeling the game of soccer — creating a network in which each player in each match of the 2008 Euro Cup is a node, and tracing the ball flow between them. Each time the ball flowed toward a possible shot, it passed through players on a team, and measuring just how central any one player was in that team flow allowed the researchers to develop a quantitative measure of the individual players’ contributions to the team.
Amaral has also studied how team members contribute to creativity by analyzing 113 years of Broadway musicals. In a 2005 paper in Science, he studied the changing rosters of librettists, lyricists, producers, composers, choreographers, and directors and found that success and creativity seemed to depend on groups that do not become stale, using the same slate of collaborators each time. The same thing held true for teams of scientists conducting research in various scientific disciplines over 50 years.
Questions about how to make groups better have taken on new urgency as evidence has accrued that teams are usurping the central spot once occupied by solo contributors. A 2007 Science study found that in science and engineering, patents, social sciences, and even to some extent in the arts and humanities, there is a shift at work — new knowledge is increasingly being produced by teams.
“This is a matter that is of national interest,” Amaral said. “We have limited resources to spend on any activity — including scientific research — so we would want to get highest possible benefit from the money we spend.”
The tendency to assign credit to a discrete individual, not a group, runs deep. There are certainly group projects in school and bonuses built on team performance, but there is also a seemingly inescapable impulse to search in a group for the narrative of the individual. How did the president guide the country at a particular time? Who is the scientist that created the lifesaving drug? People gravitate toward stories of individuals who matter, despite the fact that much of human history has been shaped not by one person at a time but by networks of people, whether they are bands of hunter-gatherers or corporations.
“Very rarely are we coming up with something that influences the rest of society as a maverick, on our own,” said Robert Goldstone, a professor of psychological and brain sciences at Indiana University.
Instead of seeing groups as nameless and faceless affiliations that swallow up an individual’s identity, the new work on collective behavior suggests that in company lies opportunity. The field of intelligence testing has long been controversial, in part because of concerns that such scores were crude and biased, pigeon-holing people as stupid or smart. In contrast, collective intelligence offers a new spectrum of possibilities. Instead of pronouncing a person’s intellectual engine good or bad, the research suggests that group intelligence is highly malleable and that concrete steps — such as mixing newcomers into an established team or not allowing a single leader to dominate — could fundamentally alter the group’s intelligence.
More broadly, groups and the complex social structure of human interactions may help account for how people got “smart” in the first place. The dramatic changes in science, culture, art, language, technology, and music over the past thousand years are not due to the development of brand-new mental or physical capacities. Instead, it is a particular kind of group benefit, Goldstone argues, in which human progress bootstraps upon itself through a collective cultural memory. Knowledge ratchets up in successive generations without our having to reinvent technologies, discover laws of nature anew, or risk tasting all the mushrooms in the forest.
“There’s been a tendency to focus on the negative, the mob psychology, the idea that people can bring out the worst in each other,” Goldstone said. “There’s just as much evidence that people can bring out the best in each other.”
Carolyn Y. Johnson is a Globe reporter. E-mail firstname.lastname@example.org.