I’m at a State Department event today, giving a talk about technology and activism – the event is under “Chatham House Rules“, which means that the discussion can be reported, but can’t be attributed. But I’ve asked Dr. Nathan Eagle if I can offer a summary of his talk, and he was gracious enough to agree.
Eagle is an MIT Media Lab researcher who’s now basing himself in Kilifi, on the northern coast of Kenya. His research at MIT focused on mining social data from mobile phone networks, a technique he calls “reality mining”. His team gave a set of phones to MIT students which had been specially enabled to track their behavior with highly transparent spyware. (The students were aware that the phones were monitored, and their identities were protected.) The phones logged the location of a user via cell-tower ID, the proximity of the user to other users by bluetooth scans, and all calls and SMS from the phone (but no content of those communications.) The resulting set of data – available on his research site – is the largest set of human behavoral data available in this space.
He shows an amazing visualization of a slice of this data, watching a hundred individuals move through a map of Cambridge. Watching the interaction between inviduals gives you some fascinating insights into their relationships – two people who are proximate in the lab during the work day doesn’t tell you much about their relationship; two people proximate in downtown Boston on a Saturday evening tells you lots more. The patterns of these interactions can give you some guesses about what events are taking place – finals week at MIT is quite apparent, as was the fact that the Red Sox were in the playoffs and then the World Series in 2004.
One way that Eagle got participation in the study is by answering useful questions for mobile users – How far did I travel last week? When did I last have lunch with John? And Eagle got very good at predicting the locations of people… or at least of some people. Low-entropy individuals – people who follow close routines – were predictable with 90-95% accuracy. Unpredictable, high-entropy individuals include MIT freshmen, who might be out partying at 3am rather than home in bed. In these high-entropy sets, prediction rates were closer to 60%.
The network was also extremely useful in identifying friendships and connections – Eagle compared people’s reported social networks (collected via survey) with the friendships the data seemed to reveal – the data analysis revealed 96% of the friendships and gave a very accurate picture of the topology of the network. The graph produced from the reality mining data was somewhat richer than the reported data as it showed the strength of connections as well – a friend you spend lots of time with versus one you rarely interact with is apparent within the dataset.
One application that emerges from this sort of tool is the possibility of matchmaking – business students participating in the project expressed a strong interest in meeting MIT geeks. The system allowed users to put their phone into a “socially promiscuous” mode, meaning they were willing to accept introductions to other users who were nearby.
One of Eagle’s long-term dreams is to model much larger networks – he points to the possibility of monitoring 250 million mobile users and 12 billion calls in a European nation. Watching how ideas move through a network like that might be similar to watching models of pathogens move through populations of people. Enthusiasm for a particular product is a contagion, not entirely dissimilar to an airborne pathogen.
Eagle points out that 59% of mobile phone users are in the developing world. In Kilifi, he’s able to pay for his cab with his mobile, something he can’t do in the US. Africa is the fastest growing mobile phone market in the world. While there are only 200,000 households with electricity, there are 7 million mobile phone users. He tells us about a trip to “cellphone alley” in Nairobi, where he picked out the innards, a colored case, a keypad and had the phone soldered together, giving him an unlocked GSM phone for $15.
The pervasiveness of these mobiles is having economic impact in Kenya – day laborers no longer have to gather on a particular street corner to seek labor – SMS could disperse the day laborers and make it possible for people to broker their own labor. Who’s going to build these new sorts of applications for the South? Probably not a Finn shivering the winter away at Nokia.
Eagle’s new project – EPROM (entrepreneurial programming and research on mobles) – is trying to encourage people in developing nations to learn how to build applications for mobile phones. This involves building a community of mobile developers and providing curiculum for students to learn how to build applications in this space. EPROM is running an “SMS bootcamp”, encouraging developers to build tools around SMS. There’s a real challenge in teaching this course in Ethiopia, where the local telephone company ETC is blocking most SMS traffic. Teaching in Addis Ababa, he managed to convince ETC to provide a small supply of unlocked SIM cards, which has let students try applications like movie listings, weather information, craig’s list-type applications, and “crush lists” for automated online flirting.
Kilifi, where Eagle lives, has the highest endemic rate of malaria in the world. To try to figure out why the population is so suceptible, there’s an ongoing survey of the population. These surveys are moving from pen and paper to using mobiles to report results to a central server. Phones are computers, he points out, letting you do everything from price fish, to send money to friends, to gain insight into how people live in societies. Excitement over this realization – phones as computers – is sparking international investment in Africa, like the multiple consortia attempting to get fiber into Kenya. “Putting fiber into every Ugandan town might be more effective than driving Landrovers around,” he notes, taking a swipe at the development aid industry.
I talked with a friend a few days ago about the potential to use the rise of mobiles as an opportunity to teach programming in Africa. I was somewhat wary, pointing out that many mobile applications are really just client/server CGI applications, and that development on the handsets themselves is quite tricky, especially in terms of crossplatform issues. But Eagle’s enthusiasm is quite infectious, and I’m looking forward to seeing what sorts of curiculum his project starts putting together.