by Kevin Kelly
We know from many classic large population studies that often the medicine we take works because we believe it will work. This is otherwise known as the placebo effect. These quantified-self tricks don’t fully counter the placebo effect; rather they work with it. If the intervention is producing a measurable improvement in you, then it works. Whether this measurable improvement is caused by the placebo effect doesn’t matter since we only care what effect it has on this N=1 subject. Thus a placebo effect can be positive.
In formal studies, you need a control group to offset your bias toward positive results. So in lieu of a control group in an N=1 study, a quantified-self experimenter uses his or her own baseline. If you track yourself long enough, with a wide variety of metrics, then you can establish your behavior outside (or before) the experiment, which effectively functions as the control for comparison.
* * *
• • •
All this talk about numbers hides an important fact about humans: We have lousy mathematical intuitions. Our brains don’t do statistics well. Math is not our natural language. Even extremely visual plots and numerical graphs demand superconcentration. In the long term, the quantification in the quantified self will become invisible. Self-tracking will go far beyond numbers.
Let me give you an example. In 2004, Udo Wachter, an IT manager in Germany, took the guts of a small digital compass and soldered it into a leather belt. He added 13 miniature piezoelectric vibrators, like the ones that vibrate your smartphone, and buried them along the length of the belt. Finally he hacked the electronic compass so that instead of displaying north on a circular screen, it vibrated different parts of the belt when it was clasped into a circle. The section of the circle “facing” north would always vibrate. When Udo put the belt on, he could feel northness on his waist. Within a week of always wearing the north belt, Udo had an unerring sensation of “north.” It was unconscious. He could point in the direction without thinking. He just knew. After several weeks he acquired an additional heightened sense of location, of where he was in a city, as if he could feel a map. Here the quantification from digital tracking was subsumed into a wholly new bodily sensation. In the long term this is the destiny of many of the constant streams of data flowing from our bodily sensors. They won’t be numbers; they will be new senses.
These new synthetic senses are more than entertaining. Our natural senses evolved over millions of years to ensure that we survived in a world of scarcity. The threat of not having enough calories, salt, or fat was relentless. As Malthus and Darwin showed, every biological population expands right to the limit of its starvation. Today, in a world made abundant by technology, the threat to survival is due to an excess of good stuff. Too much goodness throws our metabolism and psychology out of kilter. But our bodies can’t register these new imbalances very well. We didn’t evolve to sense our blood pressure or glucose levels. But our technology can. For instance, a new self-tracking device, the Scout from Scanadu, is the size of an old-timey stopwatch. By touching it to your forehead, it will measure your blood pressure, variable heart rate, heart performance (ECG), oxygen level, temperature, and skin conductance all in a single instant. Someday it will also measure your glucose levels. More than one startup in Silicon Valley is developing a noninvasive, prickless blood monitor to analyze your blood factors daily. You’ll eventually wear these. By taking this information and feeding it back not in numbers but in a form we can feel, such as a vibration on our wrist or a squeeze on our hip, the device will equip us with a new sense about our bodies that we didn’t evolve but desperately need.
* * *
• • •
Self-tracking is much broader than health. It is as big as our life itself. Tiny wearable digital eyes and ears can record every second of our entire day—who we saw and what we said—to aid our memories. Our stream of email and text, when saved, forms an ongoing diary of our mind. We can add the record of the music we listened to, the books and articles we read, the places we visited. The significant particulars of our routine movements and meetings, as well as nonroutine events and experiences, can also be funneled into bits and merged into a chronological flow.
This flow is called a lifestream. First described by the computer scientist David Gelernter in 1999, a lifestream is more than just a data archive. Gelernter conceived of lifestreams as a new organizing interface for computers. Instead of an old desktop, a new chronological stream. Instead of a web browser, a stream browser. Gelernter and his graduate student Eric Freeman define the lifestream architecture like this:
A lifestream is a time-ordered stream of documents that functions as a diary of your electronic life; every document you create and every document other people send you is stored in your lifestream. The tail of your stream contains documents from the past (starting with your electronic birth certificate). Moving away from the tail and toward the present, your stream contains more recent documents—pictures, correspondence, bills, movies, voice mail, software. Moving beyond the present and into the future, the stream contains documents you will need: reminders, calendar items, to-do lists.
You can sit back and watch new documents arrive: they’re plunked down at the head of the stream. You browse the stream by running your cursor down it—touch a document in the display and a page pops out far enough for you to glance at its contents. You can go back in time or go to the future and see what you’re supposed to be doing next week or next decade. Your entire cyberlife is right there in front of you.
Every person generates their own lifestream. When I meet with you, your lifestream and mine intersect in time. If we are going to meet next week, they intersect in the future; if we met, or even shared a photo last year, then our lifestreams intersected in the past. Our streams become richly braided with incredible complexity, but the strict chronological nature of each one means that they are easy to navigate. We naturally slide along a timeline to home in on an event. “It happened after the Christmas trip but before my birthday.”
The advantage of a lifestream as an organizational metaphor, Gelernter says, is that “the question ‘Where did I put that piece of information?’ always has exactly one answer: It’s in my stream. The idea of a timeline, a chronology, a diary, a daily journal, or a scrapbook is so much older and so much more organic and ingrained in human culture and history than the idea of a file hierarchy.” As Gelernter told a Sun computer representative, “When I acquire a new memory of (let’s say) talking to Melissa on a sunny afternoon outside the Red Parrot—I don’t have to give this memory a name, or stuff it in a directory. I can use anything in the memory as a retrieval key. I shouldn’t have to name electronic documents either, or put them in directories. I can shuffle other streams into mine—to the extent I have permission to use other people’s streams. My own personal stream, my electronic life story, can have other streams shuffled into it—streams belonging to groups or organizations I’m part of. And eventually I’ll have, for example, newspaper and magazine streams shuffled into my stream also.”
Gelernter tried many times since 1999 to produce a commercial version of his software, but it never took off. A company that bought his patents sued Apple for stealing his Lifestream idea and using it in its Time Machine backup system. (To restore a file in Apple’s Time Machine, you slide along a timeline to the date you want and there is “snapshot” of your computer’s content on that date.)
But in social media today we have several working examples of lifestreams: Facebook (and in China, WeChat). Your Facebook stream is an ongoing flow of pictures, updates, links, pointers, and other documentation from your life. New pieces are continually added to the front of the stream. If you care to, you can add widgets to Facebook that capture the music you are listening to or the movies you are streaming. Facebook even provides a timeline interface to review the past. Over a billion other people’s streams can intersect with yours. When a friend (or stranger) likes a post or tags a person in a picture, those two st
reams mingle. And each day Facebook is adding more current events and news streams and company updates into the worldstream.
But even all this is still only part of the picture. Lifestreaming can be thought of as an active, conscious tracking. People actively curate their stream when they snap a photo on their phones, or tag friends, or deliberately check-in to a place with Foursquare. Even their exercise Fitbit data, counting steps, is active, in that it is meant to be paid attention to. You can’t change your behavior unless you pay attention in some capacity.
There is an equally important domain of tracking that is not conscious or active. This passive type of tracking is sometimes called lifelogging. The idea is to simply, mechanically, automatically, mindlessly, completely track everything all the time. Record everything that is recordable without prejudice, and for all your life. You only pay attention to it in the future when you may need it. Lifelogging is a hugely wasteful and inefficient process since most of what you lifelog is never used. But like many inefficient processes (such as evolution), it also contains genius. Lifelogging is possible now only because computation and storage and sensors have become so cheap that we can waste them with little cost. But creative “wasting” of computation has been the recipe for many of the most successful digital products and companies, and the benefits of lifelogging also lie in its extravagant use of computation.
Among the very first to lifelog was Ted Nelson in the mid-1980s (although he didn’t call it that). Nelson, who invented hypertext, recorded every conversation he had with anyone on audio or videotape, no matter where or of what importance. He met and spoke to thousands of people, so he had a large rental storage container full of tapes. The second person was Steve Mann in the 1990s. Mann, then at MIT (now at the University of Toronto), outfitted himself with a head-mounted camera and recorded his daily life on videotape. Everything, all day, all year. For 25 years, if he was awake, he kept the camera on. His gear had a tiny screen over one eye and the camera recorded his first-person viewpoint, foreshadowing Google Glass by two decades. When we first met in July 1996, Mann sometimes called what he did “Quantimetric Self Sensing.” Because there was a camera half obscuring his face, I found it was hard to be natural around Mann, but he is still routinely recording his whole life all the time.
But Gordon Bell at Microsoft Research may be the paragon of lifeloggers. For six years beginning in 2000, Bell documented every aspect of his work life in a grand experiment he called MyLifeBits. Bell wore a special custom-made camera around his neck that noticed a person’s body heat if they were near and photographed them every 60 seconds. Bell’s bodycam also snapped a picture if it detected a change in light of a new place. Bell recorded and archived every keystroke on his computer, every email, every website he visited, every search he made, every window on his computer and how long it remained opened. He also recorded many of his conversations, which enabled him to “scroll back” whenever there was disagreement on what had been said. He also scanned all his incoming pieces of paper into digital files and transcribed every phone conversation (with permission). Part of the intent of this experiment was to find out what kind of lifelogging tools Microsoft might want to invent to help workers manage the ocean of data this lifelogging generates—because making sense of all this data is a far bigger challenge than merely recording it.
The point of lifelogging is to create total recall. If a lifelog records everything in your life, then it could recover anything you experienced even if your meaty mind may have forgotten it. It would be like being able to google your life, if in fact your life were being indexed and fully saved. Our biological memories are so spotty that any compensation would be a huge win. Bell’s experimental version of total recall helped increase his productivity. He could verify facts from previous conversations or recover insights he had forgotten. His system had little problem recording his life into bits, but he learned retrieving the meaningful bits needed better tools.
I’ve been wearing a tiny camera that I clip to my shirt, inspired by the one Gordon Bell wore. The Narrative is about an inch square. It takes a still photo every minute all day long, or whenever I wear it. I can also force a shot by tapping on the square twice. The photos go to the cloud, where they are processed and then sent back to my phone or the web. Narrative’s software smartly groups the images into scenes during my day and then selects the most representative three images for each scene. This reduces the flood of images. Using this visual summary, I can flick through the 2,000 images per day very quickly, and then expand the stream of a particular scene for more images to find the exact moment I want to recall. I can easily browse the lifestream of an entire day in less than a minute. I find it mildly useful as a very detailed visual diary, a lifelogging asset that needs to be invaluable only a couple of times a month to make it worthwhile.
Typical users, Narrative has found, employ this photo diary while they attend conferences, or go on vacation, or want to record an experience. Recalling a conference is ideal. The continuous camera captures the many new people you meet. Better than a business card, you can much more easily recall them years later, and what they talked about, by browsing your lifestream. The photo lifestream is a strong prompt for vacations and family events. For instance, I recently used the Narrative during my nephew’s wedding. It includes not only the iconic moments shared by everyone, but captured the conversations I had with people I had not talked to before. This version of Narrative does not record audio, but the next version will. In his research Bell discovered that the most informative media to capture is audio, prompted and indexed by photos. Bell told me that if he could have only one, he’d rather have an audio log of his day than a visual log.
An embrace of an expanded version of lifelogging would offer these four categories of benefits:
A constant 24/7/365 monitoring of vital body measurements. Imagine how public health would change if we continuously monitored blood glucose in real time. Imagine how your behavior would change if you could, in near real time, detect the presence or absence of biochemicals or toxins in your blood picked up from your environment. (You might conclude: “I’m not going back there!”) This data could serve both as a warning system and also as a personal base upon which to diagnose illness and prescribe medicines.
An interactive, extended memory of people you met, conversations you had, places you visited, and events you participated in. This memory would be searchable, retrievable, and shareable.
A complete passive archive of everything that you have ever produced, wrote, or said. Deep comparative analysis of your activities could assist your productivity and creativity.
A way of organizing, shaping, and “reading” your own life.
To the degree this lifelog is shared, this archive of information could be leveraged to help others work and to amplify social interactions. In the health realm, shared medical logs could rapidly advance medical discoveries.
For many skeptics, there are two challenges that will doom lifelogging to a small minority. First, current social pressure casts self-tracking as the geekiest thing you could possibly do. Owners of Google Glass quickly put them away because they didn’t like how they looked and they felt uncomfortable recording among their friends—or even uncomfortable explaining why they were not recording. As Gary Wolf said, “Recording in a diary is considered admirable. Recording in a spreadsheet is considered creepy.” But I believe we’ll quickly invent social norms and technological innovations to navigate the times when lifelogging is appropriate or not. When cell phones first appeared among the early adopters in the 1990s, there was a terrible cacophony of ringers. Cell phones rang at high decibels on trains, in bathrooms, in movie theaters. While talking on an early cell phone, people raised their voices as loud as the ringers. If you imagined back then what the world would sound like in the near future when everyone had a cell phone, you could only envision a nonstop racket. That didn’t happen. Silent vibrators were invented, people learned
to text, and social norms prevailed. I can go to a movie today in which every person in the theater has a cell phone, and not hear one ring or even see one lighted screen. It’s considered not cool. We’ll evolve the same kind of social conventions and technical fixes that will make lifelogging acceptable.
Second, how can lifelogging work when each person will generate petabytes, if not exabytes, of data each year? There is no way anyone can troll through that ocean of bits. You’ll drown without a single insight. That is roughly true with today’s software. Making sense of the data is an immense, time-consuming problem. You have to be highly numerate, technically agile, and supremely motivated to extract meaning from the river of data you generate. That is why self-tracking is still a minority sport. However, cheap artificial intelligence will overcome much of this. The AI in research labs is already powerful enough to sift through billions of records and surface important, meaningful patterns. As just one example, the same AI at Google that can already describe what is going on in a random photo could (when it is cheap enough) digest the images from my Narrative shirt cam so that I can simply ask Narrative in plain English to find me the guy who was wearing a pirate hat at a party I attended a couple of years ago. And there it is, and his stream would be linked to mine. Or I could ask it to determine the kind of rooms that tend to raise my heart rate. Was it the color, the temperature, the height of the ceilings? Although it seems like wizardry now, this will be considered a very mechanical request in a decade, not very different from asking Google to find something—which would have been magical 20 years ago.