If you’d like to know why companies like Google and Yahoo are so intent on gaining access to your mobile phone, take a look at the research on “reality mining” that’s being undertaken by an MIT team led by Sandy Pentland. In an interview with Technology Review, Pentland explains how modern cell phones provide a uniquely rich record not only of people’s locations but of their actions, social behavior, and even social roles:

Just look at a cell phone. It knows where you are, and this is obviously sort of useful. But the generalization is that maybe it can know lots of things about you. Take your Facebook friends as an example. The phone could know which ones you socialize with in person, which ones are your work friends, and which friends you’ve never seen in your life. That’s an interesting distinction, and reality mining can make it automatic. It’s about making the “dumb” information-technology infrastructure know something about your social life. All this sort-of Web 2.0 stuff is nice, but you have to type stuff in …

Today’s cell phones are on us all the time, and they come with hardware that can act as sensors for your environment. For instance, if Bluetooth is turned on, then the phone can see and be seen by other Bluetooth devices. You can start to make a record of the Bluetooth-enabled devices you encounter throughout the day. Then you can figure out, based on the frequency [with which] you encounter other people’s Bluetooth phones, what sort of relationship you have with them.

The iPhone also has an accelerometer that could tell if you are sitting and walking. You don’t have to explicitly type stuff in; it’s just measured. And all phones have built-in microphones that can be used to analyze your tone of voice, how long you talk, how often you interrupt people. These patterns can tell you what roles people play in groups: you can figure out who the leader is and who the followers are.

Pentland argues that this kind of machine awareness may have useful applications, such as tracking the spread of epidemics or even monitoring the “social health of communities.” But, as the interviewer points out, it “all gets very creepy very fast.” Replies Pentland:

That’s not a trivial thing. Do you really want your government to know about you to that level? It could stop SARS, but there’s a big trade-off there. You could make this a much more transparent world where that’s available to everybody. But we definitely need to talk about it and figure out a new deal for privacy – to use this data and not be abused.

Beyond the potential use of cell phone data by governments, though, it’s easy to see the vast commercial value of automatically harvesting a continuous stream of data on a person’s location, activities, relationships, and social roles and using it to personalize services and advertisements or, in the extreme, manipulate behavior for profit-making ends.

In a paper entitled “Inferring Social Network Structure Using Mobile Phone Data,” Pentland and two coauthors explain that one of the great benefits of the cell phone as a data mining tool is that it provides raw, unfiltered information, which ends up being more reliable than information “self-reported” by people. People’s reports on their own behavior are subject to a great deal of distortion due to memory lapses, cognitive biases, embarrassment, and other factors. Cell-phone reality mining, by contrast, provides “a new method for precise measurements of large-scale human behavior.” Our cell phone know us better than we know ourselves.

To illustrate the power of the technique, the authors conducted a reality mining experiment that involved “ninety-four subjects using mobile phones pre-installed with several pieces of software that record and send the researcher data on call logs, Bluetooth devices in proximity, cell tower IDs, application usage, and phone status. These subjects were observed via mobile phones over the course of nine months, representing over 330,000 person-hours of data (about 35 years worth of observations).” The data provided a remarkably intimate view of the subjects’ lives. The researchers were, for instance, able to “identify characteristic behavioral signatures of relationships that allowed us to accurately predict 95% of the reciprocated friendships in the study. Using these behavioral signatures we can predict, in turn, individual-level outcomes such as job satisfaction.”

Reality mining, the authors write, could revolutionize social research. It provides “a new approach to studying collaboration and communication within organizations – allowing the examination of the evolution of relationships over time. More dramatically, these methods allow for an inspection of the dynamics of macro networks that were heretofore unobservable. There is no technical reason why data cannot be collected from hundreds of millions of people throughout the course of their lives.” But it’s unlikely that the data will only be analyzed in the aggregate by academic researchers. The commercial value of reality mining is far too great to restrict the technique to the ivory tower. The resulting intrusions into personal privacy could well be dramatic, and as Pentland notes in the interview, we can’t assume that our interests will be protected: “The people making policies don’t know what is [technologically] possible, and they don’t necessarily make policies that are in our best interest … These capabilities are coming, but we have to come to a new deal. It doesn’t do any good to stick your head in the sand about it.”

NOTE: Technology Review provides a link to the paper, but it no longer seems to work. I found a copy in Google’s cache. And if you’d like to do a little reality mining yourself, you can download all the data from the researchers’ experiment here.