- 17 Apr 2015
A Summary of Alex "Sandy" Pentland's
SFI Community Lecture
By Mary Hoff
What each of us does over the course of a day — where we go, whom we encounter, what we buy — may seem rather mundane and inconsequential. When aggregated with activities of others, however, our seemingly simple comings and goings form amazing and instructive patterns.
The remarkable pictures that emerge when we map social networks — and the beneficial use to which we might put them — were the focus of a Santa Fe Institute’s Community Lecture, presented on March 11, 2015, by Alex “Sandy” Pentland, Toshiba Professor of Media, Arts, and Sciences at MIT and director of MIT’s Media Lab.
Listed as one of Forbes’ seven most powerful data scientists in 2011, Pentland is co-leader of the World Economic Forum’s Big Data and Personal Data initiatives and author of Social Physics: How Good Ideas Spread – The Lessons From a New Science. He took the opportunity of the lecture, sponsored by Thornburg Investment Management, to introduce the rapidly growing science of social physics, which uses computer science, machine learning, mathematics and more to collect, compile and analyze massive amounts of data about human activity in ways that allow us to discern emergent patterns and apply them to improve human well-being.
The term “social physics” was first used, Pentland said, some two centuries ago by scientists who suggested that we apply the objective principles of physics to humans in hopes of better understanding patterns of behavior. That science took on a new life in the 1990s with the development of wearable technology. Pentland was there, as creator of the Cyborg Collective, a pre-Internet, pre-cellphone attempt to imagine the future and understand the implications of the nascent technology revolution. Some of the devices and capabilities envisioned by the collective were remarkably prescient — Pentland shared two-decade-old student sketches of people wearing Google-glass-type devices and carrying what looked for all the world like iPhones — but the big breakthrough for the project, he said, was that participants began to wear sensors. “This was the first time you could observe quantitatively how people behave over long periods of time,” Pentland recounted. That meant instead of relying on averages, researchers could bring together actual records of individuals’ activities, like threads in a tapestry, to create a larger, emergent picture of human behavior.
Since then, Pentland said, the ability to observe quantitative aspects of social interactions — through sensors and through the use of Big Data sources such as de-identified credit card records — has shed remarkable light on what we do and why we do it, yielding insights that can be used to understand and improve communities and social well-being.
Pentland went on to describe a number of studies and discoveries in the realm of social physics:
Nonverbal signals. Aggregation of data collected from many individuals with the use of sensors has shown that nonverbal signals indicating states such as excitement and attention can predict success with uncanny accuracy of everything from job interviews to speed dates to pitching business plans — paying no attention at all to the actual words involved.
Exploration vs. exploitation. Observations of the human equivalent of foraging behavior in nonhuman animals — seeking out known and novel parts of our environments — have shown that the two states, “exploration” and “exploitation,” activate different parts of our brains. Using data from 100 million credit card records, Pentland and colleagues discovered the patterns differ for people in different circumstances. For instance, people without a lot of discretionary money explore to find deals, then exploit those deals by changing their purchasing habits when they find something cheaper. Wealthier people, on the other hand, just explore — they don’t necessarily change their habits based on what they find.
The important thing, though, is having a healthy balance. Banks have used Big Data to discover that an atypical balance between “explore” and “exploit” modes can be used to predict who is most likely to have credit card troubles.
By giving people smartphones and asking them to report how they feel each morning, Pentland and colleagues have also found they could use a shift in the balance between the two modes as an indicator of when individuals become depressed or sick. These findings have since been used by a prominent healthcare system to help medical professionals detect and treat health issues such as congestive heart failure early on. “It’s like a check engine light,” Pentland said. “As a consequence, they get to people before they get really sick. And that’s the kind of thing that will change the health care system.”
The pulse of a city. Pentland has also applied social physics to take what he calls “the pulse of a city.” Using cellphones with GPS to observe patterns of movement and interaction among large numbers of people, Pentland’s team discovered surprising patterns in how people within a particular community behave. It turns out that individuals in the same local community tend to exhibit similar patterns in where and when they shop, work, socialize and play. “Even though these people didn’t know each other, they’re part of the same tribe and they learn from each other,” he says — things like what to wear, what to eat, what to drink and more become norms created by individuals who don’t necessarily interact on a personal basis, creating clusters of communities within urban areas.
Fascinating, yes — but this “social learning” phenomenon turns out to be insightful as well. Exploring further, Pentland’s team discovered that the different “tribes” that self-assembled often had characteristics that made them ripe for intervention — members of one had five times the average risk of having diabetes, and those of another were an order of magnitude more likely to have problems with alcohol.
“What that means is that you can look at the pattern of interactions within a city, and look at how many people you interact with, what tribes you interact with, how far you go to interact, and you can see how new habits develop and spread throughout the city,” Pentland said. “What we’re doing is using the people as a sensor.”
A look at patterns in face-to-face communication can also be used to identify pockets of poverty and crime within communities. If patterns change, chances are there’s a new stressor in the neighborhood. In a study in London, for example, a lack of exploration and engagement was correlated with a higher crime rate.
Shots on goal. Pentland noted that such social physics findings can be used to develop incentives toward social goals that really work. Traditionally, we think of ourselves as individuals, so interventions intended to create change revolve around establishing rewards and penalties intended to motivate individuals. But far too often they don’t work. Why not, Pentland asked, move into game theory and use social incentives instead? In yet another study, he and his team gave participants smartphones with pedometers and split them into two groups: In one, individuals received rewards when they walked more. In the second, participants were able to monitor each other’s walking habits — and when people upped their activity, their friends, not they themselves, received the reward. The buddy system turned out to be eight times more effective in motivating walking — and buddy system participants were more likely to keep up their activity level when the study was over. These findings, Pentland suggested, could readily be applied to efforts to motivate behavior change in other areas such as energy use.
The bottom line, Pentland said, is that data about how people behave can provide valuable insights that can be used to improve quality of life. What social scientists need to do is gather and analyze these data in a way that allows patterns to emerge. One example would be to collect statistics about how people move about large cities in order to address challenges such as improving mass transit and reducing disease and poverty. Pentland concluded with a note of caution, however, pointing out that it’s not easy to gather behavioral data or to ensure complete anonymity, and alerting audience members to the need to reclaim our digital data, which currently are being gathered and used by companies and governments, a practice he compared to serfdom.
Pentland invited those interested in learning more about these topics to check out his new book, Social Physics: How Social Networks Can Make Us Smarter.
The video of Alex “Sandy” Pentland’s public lecture can be viewed at https://www.youtube.com/watch?v=Uxhq2Llat2o .
