By Kate Greene|Thursday, December 08, 2011
Bob Evans has spent most of his life obsessing over how to track data. When the Google software engineer was a boy in Louisville, Kentucky, he collected star stickers to show that he had done his chores. In college, where he studied philosophy and classical guitar, Evans logged the hours he spent playing music. Later, as an engineer for a Silicon Valley software company, he defended his dog, Paco, against a neighbor’s noise complaints by logging barks on a spreadsheet (the numbers vindicated Paco, showing he was not the source of the public disturbance). For Evans, collecting data has always been a way to keep tabs on his habits, track his goals, and confirm or dispel hunches about his daily existence.
Last May, Evans reminisced about those early days in data collection as we sat in a large-windowed conference room in Building 47 of the Google campus, near San Jose, California. His personal fixation is shared by a growing number of self-trackers, a movement that is spreading far beyond data-obsessed engineers. Taking advantage of new wearable wireless devices that can measure things like sleep patterns, walking speeds, heart rates, and even calories consumed and expended, more and more people are signing up to download and analyze their personal data. Nearly 10 million such devices will be sold in North America in 2011, according to the market forecasting company ABI Research.
Most self-trackers are extreme fitness buffs or—like Evans—technology pioneers inherently interested in novel software applications. But Evans believes that personal data collecting could have stunning payoffs that go beyond just taking a better measure of everyday behavior. Already, some proponents claim personal benefits from logging their habits—eliminating foods that trigger migraines or upset stomachs, for instance, or saving certain tasks for their most productive time of day. Applied more broadly, data collected by self-trackers could help them find better treatments for diseases and even predict illness before symptoms become obvious.
Evans also sees the potential for individual citizens to pool nonmedical data collected through tracking experiments. Such data sets could have important social benefits. For instance, if members of a community tracked their feelings about safety in their neighborhood and shared their data regularly, crime trends could be detected earlier and addressed more effectively.
As Evans’s history with data collection shows, basic self-tracking is possible with nothing more than a pencil and paper. Still, people have been reluctant to sign on to an activity that has historically required inordinately high levels of self-curiosity and motivation. Now, with the wildfire spread of smartphones and tablet computers, that resistance could be melting away—and Evans plans to capitalize on the change. He has developed a tracking tool, conveniently contained in a mobile phone app, that he thinks can make self-tracking appealing to the masses.
Most self-tracking devices currently on the market measure only a few data points and have their own proprietary software and code limiting how users can analyze their own metrics. Evans’s app is different: It can be set up to track any kind of behavior or event and keeps data in one place, making it possible to analyze it all together. It is also designed to address another major objection to such detailed self-reporting, the fear that our personal data could too easily be leaked, stolen, or simply exposed to the public.
My visit to Google was a chance to understand Evans’s vision and to try out its practical application. I’m not a data obsessive by any means. If Evans could convert me, self-tracking just might be for real.
In 2009, while Evans was working for Google to help create new tools to increase programmers’ efficiency, he realized no one was working on the “soft science” side of the equation to help the programmers become more productive in their personal behavior. In his data-oriented way, he set out to understand everything that happens in a programmer’s work life. He wondered how attitudes toward food, distractions, and work environment—sampled throughout the day —might affect creativity. If a programmer was stressed out or unhappy with a project, could a glance at her daily stats help set her right? Could immediate insight from a survey encourage her to make a change for the better? Evans had a hunch that by gathering the right data sets, he could help people improve their job performance in real time.
To make this process as simple as possible, Evans decided to collect the data through the smart cell phones that Google employees already kept close at hand. He set up an app so a programmer’s phone would chime or buzz a few times throughout the day at random times, as if a text message had arrived. When the employee clicked the message open, the app would ask her if she felt passionate and productive about her project. If not, it asked what she could do to change it.
In addition to gathering data about work habits, Evans set up another survey that asked programmers to outline their work goals. When the app checked in later, it listed those goals and asked which one the programmer was engaged in—the idea being that if a programmer had been distracted, a reminder of what she wanted to accomplish might improve her focus. “I thought it would be cool to build a platform that was not just for collecting data,” Evans says. “It could have the tools and interventions so people could do their own self-improvement.”
The survey was rolled out two years ago to a small number of programmers at the Google campus. Although Evans worried that the app would be too intrusive, he was heartened to see that most programmers continued to use it even after the pilot program officially ended. Since each programmer had different goals, measuring the overall effectiveness of the app was difficult, Evans says, but subjectively, he and his colleagues felt the simple act of observing their behavior through the app led them to change in ways that helped them meet their work goals.
Evans’s daily productivity surveys soon inspired him to create a broader, more flexible mobile platform for self-experimentation that he dubbed PACO—an acronym for Personal Analytics Companion, but also a tribute to the dog that helped inspire his data-tracking ideas. Now PACO is used by thousands of Google employees, and not just for productivity. The app is fully customizable, which means it can track any data point a user dreams up. Some Googlers employ it to log exercise or participation in volunteer programs. Evans tailored his version of PACO to monitor his work tasks and exercise and as a reminder to eat fewer sweets. A colleague uses it to track carbohydrate intake and weight fluctuations and to compare trends across PACO experiments. “I look at the information I track every couple of months and remind myself of the progress I’ve made, or where I need to change my behavior,” Evans says.
After hearing him describe all the ways PACO has subtly changed the lives of his colleagues, I was ready for my own plunge into the world of self-tracking.
Logging personal data is probably as old as writing itself, but some modern self-trackers trace its origin to that godfather of American ingenuity, Benjamin Franklin. He was interested in how well he adhered to his famous 13 virtues, including frugality, sincerity, and moderation. Each day for several years he noted the ones he’d violated in a book he kept especially for the purpose.
More recently, Gordon Bell, a computer pioneer and researcher at Microsoft, introduced the concept of “life logging.” From 1998 to 2007, Bell collected his emails and scanned documents, photographs, and even continuous audio and video recordings of his day-to-day life into a searchable online database—an attempt to create a digital record of every thought and experience he’d had for a decade.
Within the past three years, though, self-tracking has grown into a veritable grassroots movement, embodied by an organization called Quantified Self, a community of data-driven types founded in the San Francisco Bay Area by journalists Kevin Kelly and Gary Wolf. Most Quantified Selfers have technology backgrounds, or at the very least a penchant for numbers. They gather in online forums and at face-to-face events to talk about their self-experimental methods, analyses, and conclusions. How does coffee correlate with productivity? What physical activity leads to the best sleep? How does food affect bowel movements? Mood? Headaches? No detail, it seems, is too intimate or banal to share.
The current explosion in self-tracking would not be possible without the mass digitization of personal data. Websites for tracking, graphing, and sharing data about health, exercise, and diet—many of which are linked to phone apps—are on the rise. RunKeeper, a popular data collection app for runners, reports 6 million users, up from 2 million in November 2010. The new small, affordable sensors, like the $100 Fitbit, can wirelessly log all sorts of human metrics: brainwave patterns during sleep, heart rates during exercise, leg power exerted on bike rides, number of steps taken, places visited, sounds heard. And a number of these sensors, such as microphones, GPS locators, and accelerometers, come inside smartphones, making some types of tracking effortless. Research firm eMarketer projects that by the end of 2012, 84.4 million people will use smartphones in the United States, up from 40.4 million in 2009.
A 2011 study by Pew Internet, a project at the Pew Research Center that investigates the impact of the Internet on American society, estimates that 27 percent of Internet users have kept track of their weight, diet, or exercise or monitored health indicators or symptoms online. Still, the Pew report also hints at a limitation inherent in the current self-tracking paradigm. It is still done mainly by conscientious people who are highly motivated to collect specific types of data about specific cases. Of the adults surveyed who own a cell phone, only 9 percent have mobile apps for tracking or managing their health.
“It’s still a relatively new idea that phones are windows into your behavior,” says computer scientist Alex Pentland, director of the Human Dynamics Laboratory at MIT. Most people, he adds, think that “health is the responsibility of your doctor, not you.” But self-tracking tools that give both patient and physician a snapshot of symptoms and lifestyle could become increasingly important to personal health.
Health is exactly what was on the mind of Alberto Savoia, a Google software engineer who supervises Evans, when he joined us in the conference room to discuss which PACO experiments had worked best for his team.
Savoia himself had created an experiment to track the effects of his allergy shots. He’d never had allergies until he moved to America from Italy. “I made fun of Americans,” he says, for sneezing at everything from cats to dust. “But lo and behold, I started to sniffle.” He suspected that his shots were helping, but as an engineer, Savoia knew to be skeptical of his own perceptions. He wanted quantitative proof. “Our brains construct fabulous stories,” he says. The daily reports he logged into PACO indicated that his shots for cat dander and pollen were working well: His symptoms were less severe and less frequent than they had been before the shots.
During the same test period, Evans created an experiment called Food Rules, based on the book of that name by Michael Pollan, a journalist who advocates eating simply and avoiding processed food. After each meal, PACO would ask: Did you eat real food? Was it mostly plants? Evans found that the very act of responding to these questions made him more aware of his eating habits. He started choosing his food in the Google cafeteria more carefully, knowing he would have to answer for it after lunch. Within weeks he stopped running the experiment because every answer was “yes.”
I considered their examples. It occurred to me that I sometimes sneeze fairly aggressively after meals. When I was a teenager, I ribbed my mother for her after-dinner sneezes, but in my early twenties I started sneezing too, with no obvious connection to specific foods. My mother had a hunch that the trigger was sugar, but I had my doubts: Who ever heard of a sugar allergy? I never kept a food log to find the actual culprit, but the question seemed perfect for PACO. In just a couple of minutes, the Google engineers walked me through the steps of creating my own experiment, which I called Sneezy, to track the problem.
I constructed a handful of other experiments as well, including one I dubbed Good Morning, Sunshine! in which PACO was programmed to ask me how well I had slept and what I’d dreamed about; Flossy, in which PACO asked me if I had flossed the day before; and the self-explanatory Call Your Mother, which had PACO pestering me on Sunday evenings to see if I had talked to my mother lately—and if so, what we’d discussed.
I chose to keep these experiments private: No one else could sign up to use them, and my data would be stored, encrypted, on a PACO server. The issue of privacy looms large over discussions of personal data collection. “It’s your daily ebb and flow,” Evans says of PACO- collected data. “That’s something you need to control.” As PACO is currently built, a user can keep everything private, or she can share data by joining an experiment created by someone else. The information is stored in the cloud, on servers rented from Google. But unlike search terms, data from PACO are not mined by the company for patterns.
Self-tracking tools will probably never catch on with the wider public unless people are confident that their data are safe. “The key is giving individuals more control over their data, yet the flexibility to share it when they need to,” says MIT’s Pentland. To do this, he suggests, data should be protected by a “trust network” that is not a company or government agency. People might then establish their own personal data vaults for which they define the rules of sharing.
Pentland participates in a group called id3, which brings together government officials, academics, and industry representatives to establish guidelines for such networks. He expects the details to be worked out within the next two years. The stakes are high. If secure methods for sharing data anonymously can be developed, it won’t be just individuals taking advantage of the information they gather through self-tracking. Society as a whole could benefit.
in 2009 Matt Killingsworth, a psychology doctoral student at Harvard University, put a call out for people to join a study he called Track Your Happiness. An iPhone app queried participants—ranging in age from 18 to 88, living in 83 countries, and working in 86 job categories—throughout the day about their state of mind, their current activity, and their environment, among other things. At the end of the study, participants were given a happiness report, with graphs illustrating how happy they were and the activities and environment that affected their mood.
In 2010 Killingsworth analyzed responses from more than 2,200 people to see if what they were thinking about affected their happiness. The most striking result was that overall, people’s minds were wandering in almost half the survey responses, and people were less happy when their minds were wandering than when they were not. The findings were unexpected because previous studies, done with small numbers of people in the lab, concluded that people’s minds wander less often.
“The project illustrates that the promise and ability to track things in real time on a mobile phone in the course of your daily life is incredibly powerful,” Killingsworth says. Most previous studies would have been limited to questions asking a small number of people, after the fact, how they had felt at a certain time. Using mobile phones for this sort of study is “incredibly exciting,” Killingsworth says. “It allows us to collect more accurate data from many thousands of people.”
In the same vein as the health-oriented PACO experiments, Ian Eslick, a Ph.D. candidate in the New Media Medicine group at MIT’s Media Lab, is helping online patient communities convert anecdotes about treatments, such as how certain diets affect symptoms, into structured self-experiments. He is building an automated recommendation system that can suggest experiments to people based on their previous symptoms and responses to interventions.
For instance, no studies have uncovered a solid connection between diet and the symptoms of psoriasis, an inflammatory skin condition from which Eslick suffers. Some people find that cutting out sugar alleviates symptoms, while others do not. Eslick hopes that by collecting information on people’s self-experiments over a long period of time, he’ll have enough useful data to warrant the deployment of a traditional clinical trial to investigate the most successful interventions for psoriasis. “It’s a very different model than traditional medical research,” Eslick says. “Trials are expensive and hard to administer. They’re short. They run once and have to get your answer.” Self-experimentation, on the other hand, has the luxury of time. Experiments can run longer and produce more data because they are cheap to administer.
Customizable data collection systems like PACO make it easy to run those experiments, Eslick says. “PACO is cool not so much because it does data collection, but because it’s trying to make it easier to collect just the data you want, and just the stuff that’s relevant.”
Today’s smartphones can collect data such as location, speech patterns, and motion without any active input from the user. This sort of passive sensing of a person’s daily life makes them powerful tools for personal medical and psychological diagnostics.
Data sets of a person’s speech and movement could provide insight into conditions such as depression and Alzheimer’s disease. Some people’s speech and movements slow when they experience severe depression. If phone sensors could effectively measure change in speech or movement over time, then an app could suggest a doctor’s visit when a person’s state of mind declines.
A 2010 study by William Jarrold, a cognitive scientist at the University of California, Davis, suggests that an automated system that analyzes speech patterns on phone calls can potentially pick up on cognitive impairment and clinical depression or determine if someone is in the very early stages of Alzheimer’s. “Machine learning is getting better, the prevalence of cell phones and cloud computing is increasing, and we’re getting more data and doing more studies,” Jarrold says. “When data are collected over the course of years, they can provide relevant information about a person’s cognitive functions, diagnosing a decline before obvious symptoms arise.”
Data tracking could even help monitor infectious disease. Pentland has shown that certain patterns picked up by a person’s phone—such as a decrease in calls and text messages—correspond to onset of the common cold and influenza. If outfitted with software that can intervene when data analysis suggests the early stages of an illness, your next phone could help you figure out you’re sick before you are even aware of a problem.
My PACO experiments ran for about a month. Initially I wasn’t sure I’d like the distraction of a self-tracking app, let alone one that insisted I respond seven to nine times a day. Unexpectedly, I came to appreciate the way the app made me mindful of what I ate and how well I slept.
One thing I learned was that my mother was wrong: It wasn’t sugar that caused my sneezes. The Sneezy experiment told me that my morning meal was the main offender, especially when I drank coffee with cream. Beer also seemed to give me sniffles, though not every time. Thanks to PACO, I have narrowed down the possible culinary culprits. The experiment Happy Work Day was less surprising but also instructive. Twice a day it asked if I was working at my desk, and it often caught me doing something other than work (16 counts for not working to 25 counts for working). It made me more aware of the non-work tasks, like household chores, I spend time on during the day. I’ve since left many of these tasks for after conventional work hours.
The two experiments I hoped would influence my behavior were telling. According to Call Your Mother, I spoke with my mother only three times over the course of the experiment. I can’t say I have radically changed that behavior yet. But Flossy was a complete success. Having PACO ask me every day if I had flossed the day before seemed to do the psychological trick. I’m flossing every day. It’s a small miracle.
My thoroughly nonscientific experiences also suggest that PACO will have widespread appeal. When I explained it to my nontechnical friends, most instantly grasped the possibilities. A social worker imagined using the app to help find the triggers for negative feelings or actions in clients. A teacher wanted to use it to measure how exercise and food affect student engagement in class. A college professor I met thought he could use PACO to get a sense of how students are handling their workload.
It is still early days for the self-tracking movement, and future versions of applications like PACO will, no doubt, be much more powerful. Even if PACO itself doesn’t catch on, the idea of a program that allows people to adjust their behavior and monitor their well-being is too enticing to ignore; someone will make it work. The Bill and Melinda Gates Foundation and the mHealth Alliance, a group that includes representatives from the United Nations and the Rockefeller Foundation, are already encouraging the development of health-related phone apps. They are acting on the premise that a world in which it is easy for anyone anywhere to collect and securely share data with medical researchers could be a healthier place for all of us.
As any self-tracker knows, there is strength in numbers.