A new framework for continuous learning
One of the key challenges of living and working in the future will be continuous learning and experimentation.
I’d like to propose a framework for this type of learning that is both feasible and focused on the individual: experimental habit formation. I believe it can help resolve one of the fundamental paradoxes of modern life — how to balance our need for stability and routine with our thirst for novelty and exploration.
Experimental habit formation is a gateway to behavior change. The question “How do I change?” is not enough, because it presupposes that you know which behaviors to adopt; even if you do, that these behaviors will lead to the outcomes you expect; and even if they do, that these outcomes will remain personally relevant and meaningful forever. By replacing these risky assumptions with tests, experimental habit formation provides a sandbox to “debug” new behaviors before wider deployment.
The first thing that’s clear is that experimental habit formation cannot be developed top-down like other business and self-improvement frameworks. To be feasible for the average individual, it needs to be built from the bottom up. Concretely, we need to start with people’s actual lived experiences, building from there to communities of practice, and finally to academic theories.
I have a particular community of practice in mind, one made up of particularly well-documented lived experiences: The Quantified Self movement. QS is a global movement of people who measure various aspects of their bodies and lives — from their exercise to their productivity to their diet and far beyond — seeking to better understand themselves and their performance through quantification, broadly defined. QSers, as they’re known, attend meetup groups in cities around the world, where they tell their stories in a “show & tell” format. These short presentations answer three questions: “What did you do?” (i.e. Which aspect of yourself did you measure?), “How did you do it?” (i.e. Which tools or methods or procedures did you use to do so?), and “What did you learn?”
I happen to believe that almost nobody appreciates the true implications of The Quantified Self movement, not even its most avid practitioners. It’s not so much that the implications are bigger, just more interesting: QS is nothing less than a living and breathing blueprint for a community, an ideology, and a toolkit for continuous and substantive lifestyle experimentation.
Minimum Viable Behaviors
The first question we need to address is, “Why are habits good vehicles for behavioral experimentation?” Why not just try new things once or twice when it strikes your fancy?
Essentially, because habits are MVBs — Minimum Viable Behaviors. They have a clear beginning, middle, and end (cue, behavior, reward), making them easy to define and identify when they appear. The good ones tend to be internally coherent and inherently rewarding, thus self-sustaining. They are situated in a physical and social context, which makes them socially acceptable and integrate relatively seamlessly into daily life. Perhaps above all, they align with human neurobiology.
Importantly for our purposes, habits are also well-suited to testing hypotheses. They are complex but can be measured as binary: did it/didn’t do it. They are discrete and disposable, with low barriers to entry and exit. Their time-series, repeating nature lends itself well to teasing out confounding influences: circumstances from willpower, time from location, episodic from continuous. Lastly, habits are famously difficult to create and sustain; yet every person maintains many habits, and they come and go all the time. This paradox is a strong hint that they flourish only as organic, emergent patterns. Since emergence is hard to fake, this gives us a high standard of success in our experiments.
Which brings us to the second question: why is it necessary or beneficial to frame new habits as experiments?
N=1 Science and Beyond
Let’s start by defining terms. An “experiment” in this context is any attempt at measuring any aspect of one’s life. There is no distinction between “observation” and “intervention” when the same person is both the researcher and the subject, because any attempt to measure the behavior inevitably changes it. As I’ll explain later, this is a feature, not a bug.
So the more specific question is, what is the value in using some degree of formal experimental structure in trying new habits, even for laypeople?
There is a temptation to view this question through the lens of a clinical scientist. From this perspective, using “a little science” seems much worse than using none at all. The scorecard does, at first glance, seem bleak: self-experiments involve minuscule sample sizes, radically subjective and non-standardized inputs, widely different measurement criteria and devices, non-normal and non-symmetric distributions that wreak havoc on statistical tests, and hardly a hint of control, blinding, or randomization. By giving such a questionable process the stamp of “science,” aren’t we just inviting people to trust a conclusion that shouldn’t be trusted?
We could look to the well-developed field of Single-Subject Experiment Design to make a case for such studies. But I think the real answer is that the value of scientific thinking extends far beyond strict adherence to the scientific method. There is a scientific sensibility — a subjective, yet dispassionate mode of observing and thinking that lies at the heart of true inquiry. Richard Feynman called it “…scientific integrity, a principle of scientific thought that corresponds to a kind of utter honesty–a kind of leaning over backwards.” The scientific method is useful and necessary in the case of clinical trials affecting the health of millions. But sometimes, achieving statistical significance requires diluting the conclusions so much that their substantive significance is lost, at least at the level of a single individual. Who cares if a weight loss treatment is effective with 99.99999% confidence if the average effect size is 1 pound?
Self-experimentation, as messy and imprecise as it can be at times, is an excellent method for developing a scientific sensibility in the pursuit of self-knowledge. It relies on a behavior that most people already perform in some capacity — a 2013 Pew study found that 70% of Americans track some type of health indicator. It recruits a subject that everyone, no matter their education or training or resources, has access to — themselves. It focuses on things that every person cares about — their personal circumstances and lifestyle.
By relaxing the traditional requirements of population-sized clinical science, we lose universal validity, reliability, and replicability. But we gain a series of powerful benefits in our pursuit of self-improvement.
Five benefits, to be exact.
The first is concrete reflexivity. The best QS presentations tend to conclude with something to the effect of: “I didn’t come to any firm conclusions, and my results raised more questions than they answered, but I’m generally more aware of this aspect of my life.” Curiously, despite their lack of “actionable results,” the presenter usually concludes with a renewed commitment to self-track even more thoroughly.
This highlights an experience many self-experimenters have reported: that the self-awareness they gained in the process of self-tracking was the real reward. It was a reward they received regardless of what their data ultimately showed. Self-tracking enhances self-awareness by providing a concrete mechanism for self-reflection: the act of recording. So-called “active tracking” requires the subject to input something manually — a response to a question, a self-reported evaluation, or a device reading. Instead of self-awareness being something to ponder during intense meditation sessions on Nepalese mountaintops, it is manifested in something much more mundane: manual data entry. Both methods, it turns out, are capable of generating reveries of conscious attention.
As an example, the TrackYourHappiness project out of Harvard seeks to help people measure their moment-to-moment happiness, by sending them questions via a mobile app at random times throughout the day. The goal is to uncover the causes and correlates of happiness through individualized random sampling. I participated for a month, and had questions such as “How happy are you right now?”, “When was the last time you exercised?”, and “Where are you right now?” sent to me a few times a day. By cross-referencing my answers, the app generated reports of which people, places, and activities make me happiest.
As you can see, the reports are not particularly insightful, but I can report that the experience was jarring, in a good way. The prompts, arriving at random times via text message, helped me realize how much of the day I spend on autopilot, somehow barely conscious of what’s going on both outside and inside.
Which makes the conclusion that project creator Matt Killingworth came to after analyzing many thousands of participants’ data (as told on this NPR podcast) especially intriguing and personally relevant: the single factor with the highest correlation with unhappiness across the entire study was mind-wandering. The more someone had their mind on something other than what they were doing, regardless of whether they were thinking about something more pleasant or less pleasant than what they were doing, the more unhappy they were likely to be both while mind-wandering and in general. This is powerful evidence for the importance of what crunchy types would call “presence.” It’s also difficult to imagine how such a conclusion could be reached without random sampling via mobile devices.
When it comes to individual self-experimentation, the Hawthorne Effect is turned on its head: who cares if you change your behavior because you know you’re being watched, when watching yourself continuously is the whole point?
A second benefit to experimental habit formation is that it passes the first and strictest filter we place on all incoming information: is this personally relevant to me?
Paul Lafontaine in this presentation describes his experience tracking his heart rate variability (HRV) continuously throughout his workweek. On this particular day, he was informed 30 minutes ahead of time that he would be briefing six senior executives on a proposal he was developing. One of the executives was attending uninvited specifically to oppose the proposal being discussed. Paul’s first reaction to the news was “This will be a great HRV reading.” Gurus can only dream of such objectivity.
The graph shows the output from his HRV tracking device. Reviewing the data and comparing it with his memory and notes, he realized that the meeting was divided into three parts: his initial presentation (interval 851 to 2976), then a low-key period (2996 to 6000) while the others discussed, followed by a second round of questioning and defense of specific points (6376–8926). Understanding that this is likely a common pattern for his briefing meetings, where he is responsible for both explaining and defending an idea, Paul was able to develop a new strategy: targeting that mid-meeting “break” to regroup, identify which points he wanted to focus on for the second half, and even use his breathing and relaxation techniques to get a second wind.
Notice that these directives are both actionable and relevant to this individual’s workplace and physiology. In contrast to the vast majority of online articles full of generic “productivity tips.” They are based on objective records that can be repeated and reinterpreted, or compared with others. Notice that Paul would not and could not draw universal conclusions from this study on what others should do. At the same time, if you decided this topic was relevant to you, you’d have a clear path forward to performing the experiment yourself.
One of the most common accusations leveled against self-tracking is that it lacks context. By isolating one factor (steps, standing minutes, “fuel points,” etc.) and giving it the false authority of numbers, the accusation goes, many of the inter-relationships so important to human behavior are lost.
But in my opinion, as you may have guessed, this contextlessness is actually a major feature. When a particular behavior or health metric is removed from its context, at least temporarily, one has the opportunity to create a new context. Instead of adopting the “meaning” this behavior holds for your social group or the culture at large, you can decide for yourself what it means for you.
In this talk, Anne Wright describes her experience looking for the cause of an unspecified, debilitating condition she suffered from. After years of appointments with specialists offering generic, unhelpful advice, Anne tried an elimination diet that indicated she had problems with bell peppers, tomatoes, and eggplants. It turns out they are all part of the nightshade family, which contains a neurotoxin that inhibits cholinesterase, a vital enzyme. Who knows how many specialists she would have had to see over how many years before she arrived at such a specific answer, if ever?
Anne’s analysis of our medical system is something many of us can relate to: it is like a giant pinball machine, bouncing you around trying to fit you into a predetermined slot. If you don’t fit anywhere, you end up at the bottom with no answers and a stack of medical bills. You are then subtly (or not so subtly) persuaded that it is “just in your head” or otherwise unworthy of serious consideration. If you are extremely tall, you know you are an outlier, and can take measures to compensate for a world designed for the median. But for many things, medical and otherwise, you don’t know where on the distribution you fall. We are all victims, at some point in our lives and especially in our most unique traits, of the Ecological Fallacy — inferences about us made from inferences about a group to which we belong, which are then turned into individual prescriptions presented as objective facts.
Creating one’s own context for a life change is difficult, but crucial, as numerous studies have shown that people are more likely to achieve goals they set for themselves. It allows people to focus on optimization — improving what’s already working to a certain extent — instead of what specialists from medicine to psychiatry to social work to substance abuse tend to focus on — remediation and intervention in extreme cases. As avid QSer Bob Troia says of his experience with self-tracking:
“I thought I felt great, and then you realize that you’ve sort of been going through life for a while with the parking brake on…And when you start fixing all of those areas, you’re like, ‘Wow! I didn’t realize.’ It wasn’t that I felt bad, I just didn’t realize I could, I should be feeling better.”
Perhaps most crucially in making this whole endeavor feasible, contextualization is the key to locating the reward in the process itself, not just the outcome. Explicit rewards have been shown to do more harm than good in anything but routine, repetitive tasks. They reduce not only performance, but also risk-taking and creativity. For people to enjoy the process, they have to do things “in their own way.” And that they do: tracking their farts and sneezes, representing their results via sound and sculpture, and quantifying home births and their cats’ movements.
The fourth benefit that experimental habits provide is the opportunity for vicarious learning — learning through the experiences of others. By documenting the experience in some form — whether it is spreadsheets, graphs, photos, or written accounts — self-tracking provides an artifact around which a community can cohere. This community is the real secret to why QS works.
The fact is, in spite of our illusion of autonomy, most learning is social learning. Even in an especially self-motivated and science-literate group like QS, most new behaviors are picked up by watching others try and fail, and occasionally succeed. I’ve been astounded to discover that this mimicking behavior isn’t just a quirk of human psychology, but seems to be a property of all sorts of networks, from agent-based simulations to macaques and homing pigeons to machine learning tournaments.
The local meetups create the perfect conditions for networked social learning to thrive: meeting people in person facilitates new connections and trust-building among existing ones; informal talks allow amateurs to present their findings in a non-intimidating format, allowing others with similar interests to self-organize into broad areas like fitness, productivity, mindfulness, and diet, while remaining connected enough to the others to benefit from new discoveries. Lastly, posting the presentations online at a central location allows them to spread as far and wide as possible, drawing greater attention in a network effects-driven cycle.
A recent example of the power of this flywheel has been the movement to hack glucose-monitoring devices using off-the-shelf hardware and custom software. Apparently, the value of continuous monitoring is only realized when that stream of data can be accessed remotely, by a diabetic child’s parents, for example. The article above portrays this movement as new, but I’ve watched it simmer for years in the QS community, only now bubbling up to the surface. And bubble up it has: the FDA has reclassified these devices partly to encourage user-directed innovation. An open-source group Open APS is adding insulin pumps to hopefully one day create a fully functioning artificial pancreas.
Finally, an experimental approach limits the risk to one’s sense of self-efficacy, which I described previously as the single greatest barrier to behavior change. There is a limit to how many times someone can fail and “try, try again” before their faith in their own abilities starts to erode. Discrete experiments give you more attempts by turning the fundamental attribution error to your advantage: containing failure to a particular experiment, while taking general credit for successes.
Experiments do this by increasing the number of ways to win, while reducing the number of ways to lose. Experiments cannot fail — they can only produce results. At worst, the null hypothesis is confirmed, helping you narrow down the factorial space. Either way, you learn something, making your next try more likely to succeed. By treating changes as situational and temporary, and holding your character constant, you mitigate permanent damage to it. Instead of tying a single habit to a single path-dependent goal — where the failure of the habit is interpreted as the impossibility of the goal — you identify multiple routes. If one doesn’t work, you move on to the next.
I often recommend “habit cycling”: trying one new habit per month, on a regular schedule. Start on the first of the month, even if you feel unprepared. Especially if you feel unprepared, since your expectations will be lower. It can be as easy as trying drinking a glass of lemon water each morning (one of my personal favorites), or as big as starting a new exercise routine. The point is to avoid analysis paralysis and lower the stakes by making new experiments just part of the routine, not some pivotal crossroads. I recommend stopping the habit after 30 days, even if, especially if, it’s going well. Take a few days off and reflect on what you learned, why it worked or didn’t, and what you want to change going forward.
The reason? It is a terrible feeling to fail, and not know why. But in some ways it’s even worse to succeed, and not know why. Was this a fluke, or did you just uncover a fundamental new truth about yourself? Success is a terrible thing to waste, because it’s so rare.
When you change the way you look at things, the things you look at change
Living an experimental lifestyle is an infinite game: the goal is not to win, but to keep playing. These aren’t just n=1 experiments; they are t=∞ experiments.
I once read a haunting sci-fi short story that put a new twist on the idea of “life as an infinite game.” The story centered on a man living in a futuristic, hyper-prosperous civilization. All of society’s problems had been solved, and death was a distant memory. It had been discovered that humans are not meant to live more than about 125 years. It wasn’t a limitation of physics, biology, neurology, or technology — it was a fundamental limit of conscious self-awareness. Each passing year brought more reasons to not do things than to do them, gradually narrowing into a nihilistic tunnel vision that led inevitably to insanity. The main character developed a strategy to cope: every 100 years he dedicated to a personal passion, at the end of which he would have his mind wiped, only to start again on something else. The story recounts the final moments of a century dedicated to the study of insects. Gazing on cases upon cases of meticulously collected and catalogued bugs from every corner of the world, the man reminisces about his vast entomological experience. As the clock winds down and the mind-wipe begins to take effect, he looks forward to his reincarnation with a childlike anticipation that he hasn’t felt in years.
What strikes me about this story is that this man is in exactly the same position as us. Despite the vast technology at his disposal, neither his past selves nor his future selves really matter. There is no passing of information or identity between lives — for the strategy to work, each life has to be completely sealed off from the others. His challenge in a literal infinite game is exactly the same as our challenge in a metaphorical one: keeping things interesting enough to stay motivated.
There is risk and urgency in this challenge, and not just because we’re mortal. As Lewis Hyde explains in The Gift, imagination has a half-life:
“…when possible futures are given and not acted upon, then the imagination recedes. And without the imagination we can do no more than spin the future out of the logic of the present; we will never be led into new life because we can work only from the known.”
This is the deeper promise of experimental habit formation: it provides a way of acting on possible futures without risking too much in the present. It addresses the fundamental tension — between routine and novelty, stability and exploration — by giving us just enough structure to feel comfortable dancing along the frontier between them. Like a deep-sea exploration vessel, it allows us to roam the ocean depths in shorts and t-shirts, and once in a while discover something remarkable.