The collective intelligence of the group has long been extolled by companies as a good thing -- mitigating individual...
biases of all sorts. But how much smarter is the group relative to the individual and can companies take steps to make their collective intelligence better?
Research by Tom Malone, the Patrick J. McGovern professor of management at the MIT Sloan School of Management and founding director of the MIT Center for Collective Intelligence, is pulling back the covers on the importance of group dynamics and the power of organizational design in developing the collective intelligence of the enterprise. And he has some good news for CIOs: Stretching a company's collective intelligence muscle has never been easier, thanks to technology.
Measuring collective intelligence
It's easy to assume that smart individuals create a smart group, but that's not necessarily the case. Malone's research found that beyond individual intelligence, three additional factors are significantly correlated with collective intelligence:
- social perceptiveness or the ability to pick up on non-verbal cues;
- equal participation among group members; and
- the number of women, who on average tend to have more social perceptiveness than men, within the group.
These three factors have an impact on the collective intelligence of a group, regardless of group size. But, Malone continued, as the number of individuals increases, there is an additional factor that can shape the collective intelligence of the group -- and that's organizational design.
"Even if you've got a bunch of very smart, interpersonally skilled people but they're stuck in the middle of a huge organization that's designed really badly, there's still not a lot they can do to help make the organization intelligent," he said during his presentation at the recent Real Business Intelligence conference in Cambridge, Mass.
The five cognitive processes
But technology is beginning to strip away barriers created by organizational design. Networks that enable the sharing of information globally, digital platforms that crowdsource talent beyond the company payroll, and machines that teach themselves are the foundation for a new organizational design -- one that will ultimately have a transformative effect on collective intelligence in the enterprise, according to Malone.
"Cheap communication now allows us to move information that's needed for knowledge work almost anywhere on the planet, almost instantly, and almost free," he said. "That means we can get global economies of scale for every specialized task being done by people almost anywhere in the world."
To help understand exactly how machines and humans, together, can ratchet up the collective intelligence of the enterprise, Malone provided a framework of five types of cognitive thinking an intelligent entity must be able to perform -- be it as an individual or as a group. They are as follows: to create possibilities for action, to decide which of those possibilities to do, to sense things that are going on in the world, to remember and to learn.
He also provided examples of how machines can -- and are -- helping humans tap each type of cognitive thinking to produce a collective intelligence that exceeds the abilities of humans collaborating alone.
Create possibilities for action
How can companies plug technology in to fulfill the first of the five cognitive processes? Malone suggested they create possibilities for action by embracing hyperspecialization of work.
"It basically means specializing work into much smaller pieces than is done in today's jobs," he said. It's not a new concept, but, in the past, hyperspecialization has been applied to physical tasks. Thanks to technology, companies are now able to hyperspecialize knowledge work.
He pointed to Topcoder, a software development platform with more than 600,000 registered freelance programmers, and Foldit, a video game for folding protein molecules important to biological and pharmacological research, as examples of hyperspecialized knowledge work.
Another example is the Amazon Mechanical Turk, a name with origins that date back to 18th century Europe and a robot chess player that turned out to be controlled by a human hidden in a cabinet under the chess board. The Amazon Mechanical Turk works the same way: Humans complete tasks that computers can't yet do, but the humans work behind an API.
Decide what to do
To boost the group's cognitive process of deciding what to do, Malone pointed to the Good Judgment Project, run by researchers at the University of Pennsylvania and funded by the U.S. intelligence community. The project aims to predict geopolitical events before they happen. The machine, in this instance, is a crowdsourcing platform designed to bring together a wide swath of talent. Those who signed up to participate didn't need special qualifications.
The top 2% of participants who correctly predicted events were labeled "super forecasters" and placed into groups of 10 to 15 where their predictions were mined for insight. As it turned out, the super forecasters were so accurate, they bested predictions made by the intelligence community on average, according to Malone.
"That's a pretty interesting lesson," he said, "that more or less, ordinary people carefully selected and trained can make more accurate predications than people within the whole apparatus of the multibillion-dollar U.S. intelligence community."
Another example of using technology to boost the collective IQ is prediction markets, where bets are placed on how accurate a prediction might be. Malone experimented with prediction markets to see if humans and machines making a prediction together produced more accurate results than humans or machines predicting alone. Augmented intelligence, in this case, won out.
Sense, remember and learn
Finally, machines can now be used to help improve a company's ability to sense, remember and even learn. Malone talked about the last three cognitive processes in relation to the Human Diagnosis Project, a platform that crowdsources clinical expertise to create a bank of medical knowledge.
"For instance, if you're a nurse in a remote village in Africa, you might use a system like this to get opinions from remote specialists that would help treat your patients better than you alone could," he said.
But the platform isn't just a place to collect, store and share information; it is also outfitted with a machine learning algorithm that uses the data to help clinicians make future diagnoses.
"When they've done studies of this system, they found that when you have a bunch of people predicting, you do better than a single doctor," Malone said. "And they found that predictions made from machine learning based on the knowledge base that is already here for common cases [are] almost as good as the predictions made by human physicians."
Humans are good at diagnosing cancer; machines are better
Robots will rival humans in the enterprise. Are CIOs ready?
Artificial intelligence is better when a human is in the loop