Can your cell phone tell if you're happy or overworked?

Researchers at the Massachusetts Institute of Technology think it can do that and more--separate the rich from the poor, the sick from the healthy, even the outgoing from the introverted. Sandy Pentland, director of MIT's Human Dynamics Research program, has focused his work on that unlikely task: using gadgets as simple as a cell phone to better understand the quirks and patterns of human behavior.

Pentland's experiments began with what he calls a "sociometer," a simple badge-like device that hangs from a subject's neck and records his or her movements, tone of voice, and location. With just those signals collected from large groups of subjects, Pentland says he was able to perform a kind of data analysis he calls "reality mining," finding patterns that reveal a surprising range of information--from how a population breaks down into groups, to which groups are most social and productive, to the personality traits of single individuals, all based on measuring tone of voice and body language.

In his most recent experiments, however, the sociometer hasn't been necessary. Instead, Pentland has tracked his subjects through their cell phones, which are carried by around four out of five Americans. Pentland spoke with Forbes.com about the benefits his tracking experiments could offer to society, the privacy problems they pose and how he hopes to strike a balance between the two.

Forbes.com: What is "reality mining?"

Sandy Pentland: Reality mining is about using sensors to understand human beings. The sensors could be security cameras, they could be devices that you wear on yourself, they could be cell phones. The point is it's about people. Data mining is about finding patterns in digital stuff. I'm more interested specifically in finding patterns in humans. I'm taking data mining out into the real world.

What kind of reality-mining experiments have you actually performed?

We developed this thing called a sociometer, a little badge that you wear around your neck that records your body language, your motion and your tone of voice--the tone, not the words. It gives us a nice little package for reality mining.

We've done all sorts of interesting things with this. Just listening to peoples' tones of voice and how they move, we can measure interest level and attention, factors that account for 40% of the variation in the outcomes of things like salary negotiation, dating scenarios, closing a sale, pitching a business plan.

How can all of that be gleaned from a single device?

Tone of voice and movement reveal a lot more than you might expect. When we did this with the British call center Vertex, we found that just measuring the variability of a salesman's tone of voice and how much he listened versus how much he spoke, we could predict whether the customer was going to buy with 89% accuracy.

Humans have a kind of second language that we're not conscious of, a signaling language. We've evolved to be able to share with each other a lot about our internal state in how active we are, the timing of our interactions, how much we mirror each other. Like when I nod my head, you'll likely start nodding your head, and psychologists have shown that you nodding your head makes you much more likely to believe what I'm saying. Just by looking at where people are, their motion and their tone of voice, you can "x-ray" people in this way.

You've applied this to large groups?

Last year, employees of a German bank used our sociometers while they were working on an ad campaign. We compared their data with survey data to verify it, and we found, for instance, that the sociometers could tell whether people felt overworked, who was happy, who felt their group was well-managed. You could pick out the personality traits in any group, who was an extrovert and an introvert.

We also found that a productive group seems to have to go through a number of configurations. They scatter and reach out for information and then come back together to integrate that information. Groups that go between those two states are the more productive and happier groups.

All of that information came from how subjects carried themselves and who they spent time with. In the end, we could construct the entire org chart--the real org chart, not the one in the front office.

And you can do something similar without a custom-made gadget, but with cell phones?

A cell phone can do almost exactly the same thing as a sociometer, the only difference being that it's not around your neck. It talks to cell towers and can tell your location, it has Bluetooth to scan for other devices, and some even have accelerometers to measure motion.

In our tests in 2006, we distributed phones with software to 91 [MIT] Media Lab and [MIT] Sloan [School of Business] students for nine months, so we could tell where they were and who they were talking to and who were their friends. We used cell-tower signals and Bluetooth, and we'd have them fill out a survey, including questions about how productive and how happy they were, to test our results. It turned out the answers to those questions were highly correlated with the network structures we observed.

Basically, that means that with just a cell phone, you can go into organizations and find out how happy and how productive people are, which is really pretty amazing.

Beyond these tests, what are the larger implications of this kind of data mining?

If you take a map of Sprint customers who have chosen to turn on the geo-location capabilities in their phones, you can map data about larger chunks of people, just like you can tell in an organization which groups are happy and which talk to each other. By looking at how often people go to work, how long they stay and how much overtime they work, you can estimate the financial health of a city block. You can even estimate social indices, how integrated a group is in society, how much they mix with people who are not in their group.

This kind of measurement isn't something you could ever see before. Suddenly, maps have people on them, not just streets.

And what, on that large scale, can you do with data taken from cell phones?

Take something like SARS [Severe Acute Respiratory Syndrome]. It took two weeks for the government to realize that an apartment building was a breeding ground for the disease. If you looked at something like this, it would be a huge signal: suddenly the whole area would stop going to work.

Social indices also correlate closely with how much crime comes out of a community. If they're not plugged in to the rest of society, things aren't going very well there.

So there's a common good for this. Being able to detect diseases like SARS could potentially save millions of lives. Looking at social indices could lead to very interesting conversations.

But doesn't all of this lead to privacy concerns?

Yes. This kind of private information is very different even from banking data. This is who you spend time with and where. Were you able to walk straight after you came out of that bar and got into your car? You can tell that with the accelerometers in your phone. It's a very invasive kind of data.

But it has enormous uses. So we have to come up with a deal that says what can be looked at and what can't. There are a bunch of technology challenges in figuring out how you can look at this data in an anonymous or safe way, but there are also public-policy and government challenges: People need to agree that it's a good idea to be able to stop SARS or make the buses go where the people are, or figure out who needs new social services. That's why we're working on what we're calling the "new deal on data."

Interestingly, companies are also noticing the potential for this data and are also following this, but they're terrified of lawsuits or bad publicity or legislation. So what companies want is rules that tell them what they can do with very private data.

What kinds of rules would make it possible to use this data without violating privacy?

There are opt-in situations, for instance, where you let me look at your location in return for certain services--to know where the nearest restroom is or where your kids are or something.

Some things should still never leave the phone. Whether you were drunk when you left the bar would probably be a good thing to leave completely private. Or it might be like a black box on an airplane, where it's only analyzed after the fact. It could be a model where, if nothing bad happens, the data stays private, but if you screw up, the courts come after your data.

Does that mean that as soon as you were under suspicion, your phone company would give up location data that could potentially incriminate you?

Actually, they already do. Phone companies are required to keep location data for a certain period of time for just that purpose. That fact is, cell phones have to talk to cell towers, and that means they always know where you are.

What about within a company? Do private organization have the right to use this kind of data mining? Should employees be worried?

Even within an organization, I think there should be limits to what kinds of violations of privacy would be acceptable. People own their own data, and you can't just steal it. But you can offer them a lot of incentives to give it up.

In some ways, people may want their data to be shared. You may want your boss's boss to know that one group isn't working, and that your group is. One of the things that people are most interested in is how they compare to other people. When they contribute data, even in an anonymous way, they may get see whether they're working harder than other people, whether their boss is as much of a jerk as everybody else's, and that's very valuable to people. Or you can even pay employees for their data.

The goal is put all the data on the table and have a discussion. We have ways, as in the German bank we studied, of measuring how happy people are. So we could optimize an organization so that everyone was happier. That's pretty cool. But to do that, people have to own their own data, and there has to be recognition of a common good that's created by sharing it. That's the new deal on data.