MIT’s Alex Pentland can predict whether you’re getting the flu – or whether you’ll gain weight – from measurable patterns of social interaction. (More)

Social Physics, Part I: Exploration, Engagement, and Idea Flow

This week Morning Feature considers MIT professor Alex Pentland’s new book Social Physics: How Ideas Spread – The Lessons from a New Science. Today we begin with the measurable patterns of social interaction that comprise what he calls “idea flow.” Tomorrow we’ll see how idea flow and social intelligence dwarf individual factors like intelligence, training, and experience in predicting the success of teams, companies, and communities. Saturday we’ll conclude with how his New Deal for Data can both protect individual privacy and enable flexible, robust, and sustainable data-driven societies.

Alex Pentland is the Toshiba Professor of Media Arts and Sciences and Director of the Media Lab Entrepreneurship Program at MIT. His research focus is on information flows and incentives in social networks, the big data revolution, and converting this technology into real-world ventures. He is the founder and director of the Human Dynamics group, and the Media Lab Entrepreneurship Program, and the World Economic Forum’s lead academic for its big data and personal data initiatives.

Once upon a virus….

It’s Tuesday, the middle of the work week, and you’re buzzing around the office. You have a few sniffles – allergies, you suspect – but you feel fine. You banter with your coworkers in the corridors and around the coffee machine. At the end of the workday you surprise them by joining them at a local coffee shop. Once home you pause to chat with a neighbor on the sidewalk before going inside, where you hop on Twitter and Facebook to connect with other friends. You feel bubbly and effervescent.

Then comes Thursday, when the sneezing and sore throat and body aches set in. You stay in your office or cubicle most of the day. You nod to colleagues as you pass, but you really don’t feel like talking. You tell your boss how sick you feel, and she sends you home to get some rest. That evening, you call family and friends to commiserate or at least distract yourself from how icky you feel.

Had you been wearing a sociometric badge at work on Tuesday, or had the funf app on your cell phone, your boss might have sent you home at noon on Tuesday … because an algorithm could have detected that you were coming down with the flu. That bubbly effervescence was actually the flu virus prompting you to spread it to as many people as possible. And Thursday, when you started to feel sick, you followed another very predictable pattern: withdrawing from casual social contact during the daytime, but reaching out to friends and family in the evenings.

Neither the badge nor the funf app needed a blood sample. They didn’t record your body temperature, or count your sniffles or sneezes. The algorithm analyzing your data didn’t need a doctor’s review. The pattern of your social interactions – perhaps reinforced by Google searches in your community – were enough to calculate the probability that you were carrying the flu virus.

Once upon an idea….

We’re all familiar with the phrase “going viral” to describe a video or news story that’s ‘suddenly’ everywhere. And that’s a good phrase, because it turns out that ideas do spread much like viruses.

But unlike the flu, the spread of ideas helps us. In fact, Dr. Pentland argues that idea flow is the primary source of innovation and effective problem-solving. Most ‘new’ ideas are really new links between familiar concepts. Early astronomers struggled to explain the motions of the planets, until Isaac Newton linked those distant motions to that of objects moving here on earth. George de Mestral looked at how burrs stuck to his clothing and his dog’s fur during a hunting trip, and saw the idea for Velcro.

We usually tell such stories as examples of individual genius, but those geniuses rarely worked in isolation. They talked with colleagues, and non-colleagues. They bounced ideas off of friends and passersby. And, often, the ideas were refined by teams who worked together, sharing and enhancing best practices and other social norms that collectively shape each group’s culture.

And Dr. Pentland’s research shows that idea flow – into and within the group – was a better predictor of their success than the individual members’ intelligence, training, or experience … combined.

Exploration and “weak ties”

Idea flow hinges on two patterns behavior, the first of which is exploration. Effective groups look for good ideas, and not only within the group. In fact, looking only within the group reduces the idea flow, because people in groups tend to already know what each other knows, and already look at problems and solutions through similar lenses. Dr. Pentland found that insular groups are prone to bubble-and-bust patterns because they create echo chambers and grow too confident in the group consensus.

To get new perspectives, then, groups must explore outside the group, from what sociologist Mark Graovetter called “weak ties.” Sales staff need to talk to customer service. Professional athletes need to talk to athletes on other teams, or who play other sports. Legislators need to talk to janitors. The kernel of a great idea – and a life-changing friendship – might begin with by chatting with a homeless person.

Engagement and “strong ties”

But new ideas coming into a group only become useful as they are adopted and adapted within the group, and that is a function of “strong ties.” And as that study found, those within-group ties need not be formal meetings. Ideas also spread in informal interactions – in corridors or at the coffee machine – where group members talk about common problems or simply affirm trust relationships.

Indeed Dr. Pentland’s equations don’t factor in what group members discuss. Rather, they use measurements of who talks with whom, how often, for how long, and how the conversation moves from person-to-person. Effective groups have balanced interactions, with everyone participating but each typically saying only a few words before listening to the others. Sociometric badges don’t record the conversations, but they do measure gestures and tones of voice, and how well each person picks up on and responds to the others’ nonverbal cues … and that data very accurately predicts the strength of ties and flow of ideas within a group.

It also predicts how well groups will establish norms. For example, Dr. Pentland can predict which college freshmen will gain weight by looking at the behaviors of their peers, not just their friends but also the other students in their dorms and at nearby tables in the dining hall. When ‘you’ decide to ask for an extra serving of fries, there’s a good chance you’re mirroring the prevailing actions – the social norms – of the people around you.

Indeed research shows that our brains release feel-good endorphins when we engage in synchronized activities like rowing, dancing, or singing. Studies suggest we get the same feel-good boost when we fall into step as we walk alongside a friend, or even a stranger who happens to be going the same way. As a social species, we are wired to copy each other, and we usually do it without even thinking about it, as in this famous Candid Camera skit:

Predictions are hard….

But you’re an individual. You think for yourself and make up your own mind. That’s the story we’ve all been told since childhood, and it fits our inner experience of puzzling out problems and making choices.

And that’s true, to a point. Maybe you’re not getting the flu. Maybe you’re just having a really great Tuesday. Maybe you’re not mimicking others in the cafeteria. Maybe you’ve begun an exercise plan and you’re really hungry for those extra fries … all on your own.

Social physics acknowledges that. Dr. Pentland’s equations are probabilistic. The algorithm may say there’s an 83% probability that you have the flu, but that leaves 1-in-6 chance that you’re just having a great Tuesday.

But with groups, the Strong Law of Large Numbers kicks in. You may just be having a great Tuesday, but if 60 people in your office building are behaving the same way, such that the algorithm projects an 83% chance that each has the flu …

… it’s extremely likely that about 50 of you have and are spreading the flu … and you should all go home before you infect anyone else. If only your boss could know that….


Happy Thursday!