Wednesday, April 27, 2011

The Really Smart Phone

Researchers are harvesting a wealth of intimate detail from our cellphone data, uncovering the hidden patterns of our social lives, travels, risk of disease—even our political views.



'Phones can know,' says an MIT researcher. 'People can get this god's-eye view of human behavior.'


Apple and Google may be intensifying privacy concerns by tracking where and when people use their mobile phones—but the true future of consumer surveillance is taking shape inside the cellphones at a weather-stained apartment complex in Cambridge, Mass.


For almost two years, Alex Pentland at the Massachusetts Institute of Technology has tracked 60 families living in campus quarters via sensors and software on their smartphones—recording their movements, relationships, moods, health, calling habits and spending. In this wealth of intimate detail, he is finding patterns of human behavior that could reveal how millions of people interact at home, work and play.


Through these and other cellphone research projects, scientists are able to pinpoint "influencers," the people most likely to make others change their minds. The data can predict with uncanny accuracy where people are likely to be at any given time in the future. Cellphone companies are already using these techniques to predict—based on a customer's social circle of friends—which people are most likely to defect to other carriers.


A wave of ambitious social-network experiments is underway in the U.S. and Europe to track our movements, probe our relationships and, ultimately, affect the individual choices we all make. WSJ's Robert Lee Hotz reports.


The data can reveal subtle symptoms of mental illness, foretell movements in the Dow Jones Industrial Average, and chart the spread of political ideas as they move through a community much like a contagious virus, research shows. In Belgium, researchers say, cellphone data exposed a cultural split that is driving a historic political crisis there.


And back at MIT, scientists who tracked student cellphones during the latest presidential election were able to deduce that two people were talking about politics, even though the researchers didn't know the content of the conversation. By analyzing changes in movement and communication patterns, researchers could also detect flu symptoms before the students themselves realized they were getting sick.


"Phones can know," said Dr. Pentland, director of MIT's Human Dynamics Laboratory, who helped pioneer the research. "People can get this god's-eye view of human behavior."


So far, these studies only scratch the surface of human complexity. Researchers are already exploring ways that the information gleaned from mobile phones can improve public health, urban planning and marketing. At the same time, researchers believe their findings hint at basic rules of human interaction, and that poses new challenges to notions of privacy.


"We have always thought of individuals as being unpredictable," said Johan Bollen, an expert in complex networks at Indiana University. "These regularities [in behavior] allow systems to learn much more about us as individuals than we would care for."


Today, almost three-quarters of the world's people carry a wireless phone. That activity generates immense commercial databases that reveal the ways we arrange ourselves into networks of power, money, love and trust. The patterns allow researchers to see past our individual differences to forms of behavior that shape us in common.


As a tool for field research, the cellphone is unique. Unlike a conventional land-line telephone, a mobile phone usually is used by only one person, and it stays with that person everywhere, throughout the day. Phone companies routinely track a handset's location (in part to connect it to the nearest cellphone tower) along with the timing and duration of phone calls and the user's billing address.


Typically, the handset logs calling data, messaging activity, search requests and online activities. Many smartphones also come equipped with sensors to record movements, sense its proximity to other people with phones, detect light levels, and take pictures or video. It usually also has a compass, a gyroscope and an accelerometer to sense rotation and direction.




Advances in statistics, psychology and the science of social networks are giving researchers the tools to find patterns of human dynamics too subtle to detect by other means. At Northeastern University in Boston, network physicists discovered just how predictable people could be by studying the travel routines of 100,000 European mobile-phone users.


After analyzing more than 16 million records of call date, time and position, the researchers determined that, taken together, people's movements appeared to follow a mathematical pattern. The scientists said that, with enough information about past movements, they could forecast someone's future whereabouts with 93.6% accuracy.


The pattern held true whether people stayed close to home or traveled widely, and wasn't affected by the phone user's age or gender.


"For us, people look like little particles that move in space and that occasionally communicate with each other," said Northeastern physicist Albert-Laszlo Barabasi, who led the experiment. "We have turned society into a laboratory where behavior can be objectively followed."


Only recently have academics had the opportunity to study commercial cellphone data. Until recently, most cellphone providers saw little value in mining their own data for social relationships, researchers say. That's now changing, although privacy laws restrict how the companies can share their records.


Several cellphone companies in Europe and Africa lately have donated large blocks of calling records for research use, with people's names and personal details stripped out.


"For the scientific purpose, we don't care who the people are," said medical sociologist Nicholas Christakis at Harvard University, who is using phone data to study how diseases, behavior and ideas spread through social networks, and how companies can use these webs of relationships to influence drug marketing and health-care decisions.


His work focuses on "social contagion"—the idea that our relationships with people around us, which are readily mapped through cellphone usage, shape our behavior in sometimes unexpected ways. By his calculation, for instance, obesity is contagious. So is loneliness.


Even though the cellphone databases are described as anonymous, they can contain revealing personal details when paired with other data. A recent lawsuit in Germany offered a rare glimpse of routine phone tracking. Malte Spitz, a Green party politician, sued Deutsche Telekom to see his own records as part of an effort by Mr. Spitz to highlight privacy issues.


In a six-month period, the phone company had recorded Mr. Spitz's location more than 35,000 times, according to data Mr. Spitz released in March. By combining the phone data with public records, the news site Zeit Online reconstructed his daily travels for months.


In recent days, Apple Inc. triggered privacy alarms with the news that its iPhones automatically keep a database of the phone's location stretching back for months. On Friday, The Wall Street Journal reported that both Apple and Google Inc. (maker of the Android phone operating system) go further than that and in fact collect location information from their smartphones. A test of one Android phone showed that it recorded location data every few seconds and transmitted it back to Google several times an hour.


Google and Apple have said the data transmitted by their phones is anonymous and users can turn off location sharing.


"We can quantify human movement on a scale that wasn't possible before," said Nathan Eagle, a research fellow at the Santa Fe Institute in New Mexico who works with 220 mobile-phone companies in 80 countries. "I don't think anyone has a handle on all the ramifications." His largest single research data set encompasses 500 million people in Latin America, Africa and Europe.


Among other things, Mr. Eagle has used the data to determine how slums can be a catalyst for a city's economic vitality. In short, slums provide more opportunities for entrepreneurial activity than previously thought. Slums "are economic springboards," he said.


Cellphone providers are openly exploring other possibilities. By mining their calling records for social relationships among customers, several European telephone companies discovered that people were five times more likely to switch carriers if a friend had already switched, said Mr. Eagle, who works with the firms. The companies now selectively target people for special advertising based on friendships with people who dropped the service.


At AT&T, a research team led by Ramon Caceres recently amassed millions of anonymous call records from hundreds of thousands of mobile-phone subscribers in New York and Los Angeles to compare commuting habits in the two metropolitan areas.


Dr. Caceres, a lead scientist at AT&T Labs in Florham Park, N.J., wanted to gauge the potential for energy conservation and urban planning. "If we can prove the worth of this work, you can think of doing it for all the world's billions of phones," he said.


Thousands of smartphone applications, or "apps," already take advantage of a user's location data to forecast traffic congestion, rate restaurants, share experiences and pictures, or localize radio channels. Atlanta-based AirSage Inc. routinely tracks the movements of millions of cellphones to generate live traffic reports in 127 U.S. cities, processing billions of anonymous data points about location every day.


One study found that the U.K.'s happiest time is 8 p.m. Saturday; its unhappiest day is Tuesday.  European phone companies discovered their customers were five times more likely to switch carriers if friends had switched, allowing the companies to target their ads.
Another study was able to determine that two people were talking about politics—without the researchers hearing the call


As more people access the Internet through their phones, the digital universe of personal detail funneled through these handsets is expanding rapidly, and so are ways researchers can use the information to gauge behavior. Dr. Bollen and his colleagues, for example, found that the millions of Twitter messages sent via mobile phones and computers every day captured swings in national mood that presaged changes in the Dow Jones index up to six days in advance with 87.6% accuracy.


The researchers analyzed the emotional content of words used in 9.7 million of the terse 140-character text messages posted by 2.7 million tweeters between March and December 2008. As Twitter goes, so goes the stock market, the scientists found.


"It is not just about observing what is happening; it is about shaping what is happening," said Dr. Bollen. "The patterns are allowing us to learn how to better manipulate trends, opinions and mass psychology."


Some scientists are taking advantage of the smartphone's expanding capabilities to design Android and iPhone apps, which they give away, to gather personal data. In this way, environmental economist George MacKerron at the London School of Economics recruited 40,000 volunteers through an iPhone app he designed, called Mappiness, to measure emotions in the U.K.


At random moments every day, his iPhone app prompts the users to report their moods, activities, and surroundings. The phone also automatically relays the GPS coordinates of the user's location and rates nearby noise levels by using the unit's microphone. It asks permission to photograph the locale.


By early April, volunteers had filed over two million mood reports and 200,000 photographs.


Publicly, Mr. MacKerron uses their data to chart the hour-by-hour happiness level of London and other U.K. cities on his website. By his measure, the U.K.'s happiest time is 8 p.m. Saturday; its unhappiest day is Tuesday.


Perhaps less surprisingly, people are happiest when they are making love and most miserable when sick in bed. The most despondent place in the U.K. is an hour or so west of London, in a town called Slough.


On a more scholarly level, Mr. MacKerron is collecting the information to study the relationship between moods, communities and the places people spend time. To that end, Mr. MacKerron expects to link the information to weather reports, online mapping systems and demographics databases.


Several marketing companies have contacted him to learn whether his cellphone software could help them find out how people feel when they are, for instance, near advertising billboards or listening to commercial radio, he said.


Mr. MacKerron said he's tempted—but has promised his users that their personal information will be used only for scholarly research. "There is a phenomenal amount of data we can collect with very little effort," he said.


Some university researchers have begun trolling anonymous billing records encompassing entire countries. When mathematician Vincent Blondel studied the location and billing data from one billion cellphone calls in Belgium, he found himself documenting a divide that has threatened his country's ability to govern itself.


Split by linguistic differences between a Flemish-speaking north and a French-speaking south, voters in Belgium set a world record this year, by being unable to agree on a formal government since holding elections last June. Belgium's political deadlock broke a record previously held by Iraq.


The calling patterns from 600 towns revealed that the two groups almost never talked to each other, even when they were neighbors.


This social impasse, as reflected in relationships documented by calling records, "had an impact on the political life and the discussions about forming a government," said Dr. Blondel at the Catholic University of Louvain near Brussels, who led the research effort.


The MIT smartphone experiment is designed to delve as deeply as possible into daily life. For his work, Dr. Pentland gave volunteers free Android smartphones equipped with software that automatically logged their activities and their proximity to other people. The participants also filed reports on their health, weight, eating habits, opinions, purchases and other personal information, so the researchers could match the phone data to relationships and behavior.




The current work builds on his earlier experiments, beginning in 2004, conducted in an MIT dormitory that explored how relationships influence behavior, health, eating habits and political views. Dr. Pentland and his colleagues used smartphones equipped with research software and sensors to track face-to-face encounters among 78 college students in a dorm during the final three months of the 2008 presidential election.


Every six minutes, each student's phone scanned for any other phone within 10 feet, as a way to identify face-to-face meetings. Among other things, each phone also reported its location and compiled an anonymous log of calls and text messages every 20 minutes. All told, the researchers compiled 320,000 hours of data about the students' behavior and relationships, buttressed by detailed surveys.


"Just by watching where you spend time, I can say a lot about the music you like, the car you drive, your financial risk, your risk for diabetes. If you add financial data, you get an even greater insight," said Dr. Pentland. "We are trying to understand the molecules of behavior in this really complete way."


Almost a third of the students changed their political opinions during the three months. Their changing political ideas were related to face-to-face contact with project participants of differing views, rather than to friends or traditional campaign advertising, the analysis showed.


"We can measure their daily exposure to political opinions," said project scientist Anmol Madan at MIT's Media Lab. "Maybe one day, you would be able to download a phone app to measure how much Republican or Democratic exposure you are getting and, depending on what side you're on, give you a warning."


As a reward when the experiment was done, the students were allowed to keep the smartphones used to monitor them.


Write to Robert Lee Hotz at sciencejournal@wsj.com

No comments: