Lessons From (and For) the Quantified Self Movement

The very first time I heard about Quantified Self I was excited by the world of possibilities contained therein. I pride myself on my self-awareness, and it seemed like adding quantitative rigor to this process would allow me to uncover new patterns below my current awareness – and ideally, to change them. Besides, I liked what I saw from this community: the ethos of self-experimentation and optimization that pervades my own life. You didn’t have to tell me twice, I was already sold. I joined as one of the original members of the NYC QS group, and dove into this world head first.

My Self, Quantified

Seeing some of the projects people were doing was intimidating. Lots of the people involved were the ones building the tools themselves, which was impressive in its own right. Others were gathering lots of esoteric data, combined with stunning visualizations, and I had no idea how to apply this to my own life.

In a lot of ways, my initial foray into QS was unintentional, which is really just a bad idea in general. I was immersed in this culture of self-tracking, it seemed like a cool thing to do, so why not play with it a bit? I was already in the habit of taking substantial notes about my day, so I could already build on top of an existing system incrementally… and that’s exactly what I did. It started off trivially simple, like writing down when I went to sleep and when I woke up in the morning. Pretty soon I was entering into my logs anything that I put in my body: food, supplements, pharmaceuticals, everything. I continued doing this for over half a year.

I did undergo a second, and more directed, period of self-tracking about a year ago, this time directly entered into a spreadsheet instead of my log files. I tracked all of the above variables, and then some: specifically, some readily observable characteristics that I wanted to optimize, things like the appearance of pimples, bleeding gums, and general gut health, as well as things I wanted to encourage myself to do, like exercising. Interestingly enough, this period was even shorter, lasting for about two months in total.

In addition to this logging, I have used a variety of software apps, hardware (like the Zeo or Mindwave), and services like 23andMe. Ultimately a lot of the same lessons apply to these as well. So without further ado…

Lessons Learned

I will begin by talking about the benefits I obtained from this self-tracking, namely a greater degree of self-awareness about my habits. In this regard, I got exactly what I was looking for: I noticed patterns in my sleeping habits, my eating habits, my work habits, etc. This was the biggest impact from my experimentation, and I found that this occurred almost from the very beginning of my tracking. So my takeaway message for you is this: if you want to increase your awareness about anything, track it religiously, even for a brief window of time. This in and of itself is capable of producing behavior change, for instance losing weight after realizing how much soda you’re actually drinking. In my case, by tracking the number of days it had been since my last exercise every single day, I was able to encourage myself into doing it more often.

Beyond that, however, what do I have to show for all of my dedicated hours of self-tracking? Surprisingly little, for a couple of reasons. First and most importantly, I realized that input variables are much easier to track than outcome variables. Think about it this way: whenever something happens, this is a signal for you to go to your log and input some data, e.g. I just ate spinach, or I just woke up, etc. On the contrary, how would one go about tracking, say, happiness levels? If you’re only entering data when you’re particularly happy or particularly sad, or when you just happen to notice it, you’re introducing some pretty serious biases into your data. Instead, you need to be taking readings either regularly or randomly to ensure an unbiased sample.

With my first round of tracking, what could I even evaluate using that data? I could draw conclusions between, say, sleeping and my meal patterns and composition. There might be something there. I could even look at the number of words written in my logs, and see if I had more insights on days I slept more. Certainly there are ways to use data creatively like this, but ultimately I did not know if I were happier, more productive, felt more energetic, or really anything that I clearly cared about… which leads nicely into my next lesson:

Self-tracking is clearly more productive when you have a very clear metric in mind you wish to optimize, and have significant motivation for doing so. Without the former you wind up like me, endlessly tracking a series of useless inputs. Without the latter, you will never survive the QS world for long – in its current incarnation anyway. Only when you have that critical combination does the magic start to happen. Often times this significant incentive comes from serious health problems, or extreme athletic performance, or hell, even vanity! In my particular case, I chose a combination of things I cared about enough to optimize, that were also readily observable inputs. Every day I could brush or floss and see if my gums bled at all, I could look in the mirror and see if I had any pimples or not. And yet, I still did not end up tracking these things for long. Why was that?

I ultimately realized that the best interventions have such a large effect size that you don’t need concepts like statistical significance to verify them. The entire point of gathering a massive number of observations is so that you can tease out an effect that we can barely notice. But why are we aiming for a 3% improvement in some metric when we could be taking much more radical measures and producing obvious results? For example, I used to weigh about 50 lbs more than I currently do, now I have lost that weight and keep it off effortlessly. Was it due to tracking my weight and food intake for years and looking for correlations? No! Instead I researched biochemistry and anthropology and radically overhauled my diet. Similarly, I’ve been unable to fall asleep until the very early morning for most of my life. With an over-the-counter supplement (melatonin), I’m able to shift this back by several hours per night, consistently, which is a drastic change. (That’s not to say you can’t get reasonable marginal improvements to your sleep using QS – for example my friend Matt Bell is an accomplished sleep tracker/hacker.) In the vast majority of cases, people have huge low-hanging fruit they have left to pick. By the time you’re looking for statistical significance, you should consider optimizing another area of your life, or consider yourself very lucky to have made it this far.

The Future of the QS Movement

Despite the skeptical tone of my above lessons, I am actually quite optimistic about the long-term prospects of the QS movement. There are a few critical challenges that will need to be overcome, but if we find solutions to those I predict that QS will take off exponentially in the years to come.

The reason that I ultimately stopped tracking the second time was because having to manually record all of these data perpetually increased my cognitive load. When I was away from my computer, I would have to record it temporarily on my phone or an old-fashioned notepad for later entry. But even with superior mobile apps or when at my computer I still had to undergo the switching costs of changing to a different window or tab, disrupting whatever else I happened to be doing. Indeed, the QS movement will not become mainstream until data collection becomes automatic. There are devices today, like say a Fitbit, that you can wear continuously and it will upload data relatively easily. Hardware peripherals, along with specialized computer programs and mobile apps, are enabling more and more automatic tracking, so this is only a matter of time. Imagine a program that tracks your happiness level by snapping periodic photos of you with your webcam and analyzing your microexpressions…

The next challenge on the horizon is going to be integrating all of that data into a single, comprehensive analysis. Automatic tracking is useless if you can’t liberate that data from its different sources and begin drawing conclusions from the entire dataset. But really, that masks the even greater challenge: analyzing a complex system like the human mind and body at all. Most of our statistical tools are designed to explain and predict very simple physical systems of few dimensions and interactions. While some single variables have enough impact on others to draw a noisy correlation, the true power lies in coming up with a more complete causal diagram, and catching the cumulative or subsequent interactions between many variables. Being able to do complex systems analysis is not just a challenge for the QS movement, but also biology and economics among other fields, so there are lots of allies to tackle this particular problem together.

I don’t think it is going to be easy to solve these problems, but I do think it is possible… and over the long run the possible is the inevitable. The stage is set for QS to expand dramatically under the right conditions. We need automatic data collection first and foremost, and beyond that we need useful data analysis. Providing either of these would be a great service to the QS community, as well as a great business opportunity.

  • Curtis

    For timetracking:

    I use the stickies on my mac (since they allow very quick access) to keep track of my day, then I paste them into workflowy. I can then query for which classes I was working on at which times.

    Example:

    Sunday
    9-11: CS 390
    11-12: Lunch
    12-2: CS 390
    2-4: Nap
    4-6: 390
    6-7: Dinner
    7-9: 390
    9-12: relaxing + workout