Inside The Weird, Profitable Study Of “Social Physics”

An MIT professor studies how counterintuitive cultural practices can make companies work better.

Inside The Weird, Profitable Study Of “Social Physics”
[Image: Wikipedia]

MIT professor Alex “Sandy” Pentland studies how ideas flow through groups and the effect that flow has on productivity, something he calls “Social Physics.”


Social physics can predict the creativity and productivity of a team based on their communication patterns alone, determine which apps you will download or which business plans executives will choose. By tracking idea flow within a particular team or company, you can radically improve its performance.

Pentland says his approach is an outgrowth of sociology, with the addition of data. “The name comes from the fellow who created sociology,” says Pentland. “He had this theory about how it was ideas that drove society but of course he had no data and no math.” Pentland’s team has studied communication data from day traders and call centers, German banks, and communities of young parents. The rules which govern how new ideas band together to become innovations and spread through society turned out to be similar in all those groups. Here’s how it works.

How Ideas Flow

The first question is what is an idea anyway? “People use information and ideas interchangeably and I’m trying to tease them apart,” says Pentland. “Information is just facts or possible facts. Ideas are actually like little strategies. In this situation if you do this you will get that result. Ideas are actionable. Idea flow is if I am exposed to ideas, I see people saying things, how likely is it that I am going to pick up that as a habit. So it’s really the flow of habits through a community.”

Your habits are often things you do and acquire almost unconsciously, what economist Daniel Kahneman calls “fast thinking.” Habits are also hard to break so people tend to be conservative about adopting new ones. The collective habits of a particular company add up to what is often called its culture.

“What changes behavior is seeing people experimenting with the same behavior,” says Pentland. “I mean really physically seeing it. It’s a sort of cooperative group behavior where people explore new ways of doing things without even being aware that it’s cooperative. It’s what the community is thinking about as a collective and that’s very different from the newest topic of conversation. Those sometimes turn into behaviors but very rarely.”

In Pentland’s view only rich channels of communication like face-to-face and to a lesser extent video conferencing and telephone lead to changes in behavior and therefore in culture. Email and text are particularly ineffective when it comes to the adoption of ideas. So effectively electronic channels can help maintain culture but they rarely create it.


Building Productive Teams

One of the most important mathematical measures Pentland’s team has found to predict productivity in teams is engagement: the frequency and type of interaction between team members. The team often uses badges which measure not only a person’s location but also use accelerometers to determine whether he is sitting, standing, or walking and to detect big body gestures. A built-in microphone samples volume and pitch in conversations rather than the content itself.

“Are you screeching at someone? Are you being distracted?” says Pentland. “In people who are negotiating a salary, it turns out that you can predict the negotiation pretty accurately from body language and the tone of voice. In speed dating, you can tell if the people are going to trade information or not by essentially how engaged they get.”

Groups where everyone talks to everyone else, in other words engagement is high, tend to be highly productive no matter what kind of work they are doing. “And it’s not official meetings. It’s stuff around the coffee pot or at lunch or or just in the hall.” Pentland persuaded the manager of a Bank of America call center to schedule coffee breaks for team members simultaneously (usually only one person has a break at a time). Productivity in the worst-performing teams jumped by 20%. More generally, Pentland has found that social time can account for up to 50% of improvements in communication

“A typical mistake which big companies make is to put a teleconference unit down the hall that you have to sign up for. That’s very formal. It has to be more casual. What I like about, for instance Skype, is that you just say ‘I’ve got to talk to X’ and go bang, bang, bang, bang and there he is. It’s just like popping around the corner and sticking your head into somebody’s office. Those kind of spontaneous, dynamic, peer-to-peer tools are really important together with opportunities to work together physically.”

One software company tried without success to improve engagement by organizing events like beer nights, but simply making lunch tables longer so that people who didn’t know each other interacted more, increased productivity by 5%. The number of opportunities a workplace provides for social learning via engagement is often the largest single factor in its productivity.

Contrarians And Creativity

While engagement is a good predictor of productivity, creative teams exhibit a behavior Pentland calls exploration. “Creative groups are distinguished by communication outside of their group,” says Pentland, “They talk to lots and lots of different people and usually this is something that their boss would consider a waste of time or not in their job description.”


It’s hard for a team to maintain high levels of both engagement and exploration since the time that its members can spend on communication is limited. So creative teams often oscillate between periods of high exploration and high engagement. Team members bring back new ideas from their wider network and then integrate and act on them during the high-engagement phase.

Not all external sources of ideas are equally useful, however. “In a lot of the tech world we are just embedded in echo chambers,” says Pentland. “One of the very best things you can do is find contrarians,” says Pentland. “You look for people who ought to think one way based on who they spend time with but actually they have a different opinion. What that tells you is they have some other information, strategy, other community that they touch that is different than the ones that you see.”

At a certain point people become so interconnected that the flow of ideas is dominated by feedback loops. Groups of people whose opinions are very regularly correlated, which can be the case in highly engaged teams, may be operating in an echo chamber, which is bad for business. “The heuristic is does everybody think the same thing all the time? Or do you find yourself going to the same sources all the time? That’s an indication that you are not diverse enough.”

Pentland’s group studied day traders on the trading website eToro. Traders can follow other traders, copy an individual trade or an entire trading history, operate independently, or combine several strategies. “What you find is that if you plot the amount of this exploration,” he says, “how diverse are the set of people that they are talking to, that predicts how much money they’ll make almost perfectly.” Traders who had a balance of diversity and social learning earned 30% more than those at either extreme.

Reward the Social Fabric

Within companies, managers often use economic incentives to try and change behavior. “What I’ve been able to show with this rich data is the reason that economic incentives don’t work the way they should is that everybody also has a set of valuable exchanges of ideas and favors and so forth with their peers,” says Pentland, “Economic incentives that apply to the individual ignore the social fabric and the social fabric often wins in that struggle. What you can also do is reward the social fabric.”

Pentland performed an experiment where half of a test group got an economic reward if they lost weight and the other half were assigned a weight-loss buddie within the group and were rewarded if their buddie lost weight. The second group were much more successful.


“We have been able to set up rewards like that to get enormous changes in behavior,” says Pentland. “That sort of thing where it’s essentially a social incentive or sometimes it looks like a group incentive seems to work much better in many situations than individual incentives. The mathematics of it says that it will be at least twice as efficient, in other words twice the bang for a given budget as economic incentives. What we find in real experiments is that we often get factors like four times more efficient or eight times more efficient.”