In the late 2000s, the CIA conducted a research project with Harvard University called Project Looking Glass, designed to understand why the intelligence community had failed to foil the September 11 attacks. The project consisted of a spy-game simulation of a terrorist attack: a team of scientists was tasked with planning an attack, and a group of intelligence officers asked to prevent it.
During the simulations, the “terrorists” consistently beat the “spies”. Researchers noticed that the experts struggled to collaborate effectively because they were not making use of individual team members’ different strengths. It was only when they forced them to communicate properly – instructing members to talk to each other about their areas of expertise before moving forward – that they managed to be successful.
The experiment led one of the researchers, Anita Woolley, to rethink the intelligence of teams. Traditionally, a group’s intelligence was assumed to be the aggregate of the intelligence of its individual members, measured by a metric such as IQ. But Woolley wondered if groups of people actually possessed a collective intelligence that could transcend the sum of its parts.
In collaboration with a team from MIT, Woolley brought nearly 700 people to her lab at the Carnegie Mellon Tepper School of Business in Pittsburgh, Pennsylvania. The participants were given individual intelligence tests and then randomly assigned to teams to work on tasks that included solving puzzles, making moral judgments and negotiation.
Woolley observed that if a group performed well on one task, they tended to perform well on the others. This wasn’t predicted by the maximum nor the average intelligence of the team members. Instead, Woolley found a collective intelligence score, “c”, with predictive power: when the teams were brought back to the lab to play a video-game simulation, their performance was correlated to their c factor. The study, published in 2010 in the journal Science, was one of the first to suggest a metric for collective intelligence.
In subsequent experiments, Woolley and her team began dissecting what actually contributed to collective team intelligence. Factors like group satisfaction, motivation or psychological safety did not contribute. Competition within a team actually lowered its intelligence. One finding was that teams with more women outperformed male-dominated ones. “You have a benefit to having a majority of women, but you still need some men,” she says. “The teams that are consistently more intelligent are gender diverse.”
Woolley also found that teams that communicated a lot, with plenty of conversation between members, also tended to be more intelligent than those where a few people monopolised the conversation.
These two observations could perhaps be explained by a third – that teams whose members were more socially perceptive were also more collectively intelligent. Social perceptiveness was based on a test called Reading the Mind in the Eyes, which measures how well people infer emotional states from images of other people’s eyes. Women tend to score higher than men on this test, and socially perceptive people also tend to be better communicators. “That doesn’t mean that individual ability doesn’t matter,” Woolley says. “What we’re saying is that what matters is both that individual ability and co-ordinating it effectively.”
Woolley was curious to see to what extent her findings on collective intelligence would translate to teams working online. Her team built a web-based software for online groups to chat and collaborate on tasks. Much like working face-to-face, the best teams chatted more and participated more equally. What surprised her, however, was how much social perceptiveness remained a strong factor for collective intelligence – even though team members couldn’t see each other’s facial expressions. “This is related to something called the theory of mind, which is the ability to know what the other person is thinking or feeling, and it generalises to all sorts of inputs,” she says. “Turns out the inputs can be as subtle as text chat. It’s a skill that transcends different kinds of communication.”
During the coronavirus pandemic, Woolley received many emails from companies unsure about how to navigate the world of remote work. “I was probably one of the few people cheering when people started working from home,” she says. “A lot of companies are recognising that there’s much more that can be done. They feel like they need to bring people together out of laziness and poor management practices. Given the right tools you can foster co-ordination and collaboration online.”
Building such tools, however, is not straightforward, and they can easily have a detrimental effect. When Woolley trialled an AI chatbot manager that helped divide and assign tasks she found that “a number of teams got preoccupied with planning and spent very little time actually working on the tasks.” On the other hand, an AI facilitator that helped group members talk about their skills and expertise – similar to in Project Looking Glass – had a positive effect.
For Woolley, this represents the first sketch of what the productivity software of the future might look like: a facilitator that’s running in the background, picking up on the fact that people are good at different things and prompting them when they are available. It’s about managing individual skills and the allocation of effort, she says: “The tools that help prompt that conversation.”
More great stories from WIRED
🧠 Can’t focus? Here’s how to concentrate when working from home
🕺 Across London havoc is being caused by illegal Airbnb nightclubs
👟 If you started running during lockdown these are the best running shoes in 2020
🔊 Listen to The WIRED Podcast, the week in science, technology and culture, delivered every Friday
👉 Follow WIRED on Twitter, Instagram, Facebook and LinkedIn