Collective intelligence in small teams
October 1, 2010
A new study co-authored by MIT researchers documents the existence of collective intelligence among groups of people who cooperate well, showing that such intelligence extends beyond the cognitive abilities of the groups’ individual members, and that the tendency to cooperate effectively is linked to the number of women in a group.
Many social scientists have long contended that the ability of individuals to fare well on diverse cognitive tasks demonstrates the existence of a measurable level of intelligence in each person. In a study published Thursday, Sept. 30, in the advance online issue of the journal Science, the researchers applied a similar principle to small teams of people. They discovered that groups featuring the right kind of internal dynamics perform well on a wide range of assignments, a finding with potential applications for businesses and other organizations.
“We did not know if groups would show a general cognitive ability across tasks,” said Thomas W. Malone, the Patrick J. McGovern Professor of Management at the MIT Sloan School of Management, one of the authors of the paper. “But we found that there is a general effectiveness, a group collective intelligence, which predicts a group’s performance in a lot of situations.”
That effectiveness, the researchers believe, stems from how well the group works together. Groups whose members had higher levels of “social sensitivity” — the willingness of the group to let all its members take turns and apply their skills to a given challenge — were more collectively intelligent. “Social sensitivity has to do with how well group members perceive each other’s emotions,” said Malone. “In groups where one person dominated, the group was less intelligent than in groups where the conversational turns were more evenly distributed.” Teams containing more women demonstrated greater social sensitivity and in turn collective intelligence, compared to teams containing fewer women.
When ‘groupthink’ is good
To arrive at their conclusions, the researchers conducted two studies in which 699 people were placed in groups of two to five and worked on tasks that ranged from visual puzzles to negotiations, brainstorming, games and complex rule-based design assignments. The researchers concluded that a group’s collective intelligence accounted for about 30 to 40 percent of the variation in performance.
Moreover, the researchers found that the performances of groups were not primarily due to the individual abilities of the group members. To determine this, many of the participants also performed similar tasks individually. The average and maximum intelligence of individuals did not significantly predict the performance of their groups.
The paper’s lead author was Anita Woolley, an assistant professor at Carnegie Mellon University’s Tepper School of Business. The other researchers in the study were Christopher Chabris, an assistant professor of psychology at Union College in New York; Malone; Alexander Pentland, the Toshiba Professor of Media Arts & Science at the MIT Media Lab; and Nada Hashmi, a doctoral candidate at MIT Sloan. The study received funding from the National Science Foundation, the Army Research Office and Cisco Systems.
To record the interactions of people, the researchers equipped study participants with wearable electronic badges — designed by Pentland’s Media Lab group — that provided a complete record of a group’s conversational patterns and revealed a group’s propensity to take turns. “When you do that, it’s possible to get patterns you’ve never seen before,” said Pentland.
Only when analyzing the data did the co-authors suspect that the number of women in a group had significant predictive power. “We didn’t design this study to focus on the gender effect,” Malone said. “That was a surprise to us.” One implication is that the level of collective intelligence should keep rising along with the proportion of women in a group. To be sure, as Malone said, that gender effect is a generalization. “Of course some males have more social skill or social sensitivity than females,” Malone acknowledged. “What our results indicate is that people with social skills are good for a group — whether they are male or female.”
Malone said he believes the study applies to many kinds of organizations. “Imagine if you could give a one-hour test to a top management team that would allow you to predict how flexibly that group of people would respond to a wide range of problems that might arise,” he said. “That would be a pretty interesting application. We also think it’s possible to improve the intelligence of a group, by either changing the members of a group, or teaching them better ways of interacting.”
How universal is it?
Colleagues in the field found the results intriguing. Jeremy Gray, an associate professor of psychology at Yale University, said the study “was very well done,” adding that “the key point is great, that features of the group can be more important than features of the individuals that make up the group, for determining outcomes.”
However, Gray, responding to questions by e-mail, noted that the study raises additional questions for further investigation. Beyond the relatively routine tasks used in the study, he wrote, “high-stakes or high-risk situations would also be very important to understand. There is no guarantee that the same pattern of results would hold, for example, for a jury deliberating a death-penalty case, a corporate board facing a hostile-takeover bid, criminal gangs battling a rival gang, and so on. We just don’t know yet.” Moreover, he added, “clarifying the conditions under which the proportion of women makes a difference would be interesting.”
Malone said the co-authors “definitely intend to continue research on this topic,” including studies on the ways groups interact online, and are “considering further studies on the gender question.” He added that “collective stupidity,” the failure of a group to perform to the abilities of its members, exists along with collective intelligence. “Part of the research agenda for this field is to understand better the conditions that lead to one rather than the other,” Malone explained. “Many factors can affect a group’s intelligence, including the social sensitivity, norms, and motivations of group members, as well as the composition of the group.” For now, Malone said his group has identified a general principle indicating how the whole really can be greater than the sum of the parts.
“Having a bunch of smart people in a group doesn’t necessarily make the group smart,” concluded Malone.