Quantified Self and False Pattern Recognition

I was watching Episode 3 of the new Cosmos show when Neil deGrasse Tyson said something that reminded me an awful lot of the Quantified Self movement.:

The human talent for pattern recognition is a two-edged sword. We’re especially good at finding patterns, even when they aren’t really there — something known as “false pattern recognition.”

The show was about how our distant ancestors looked up into the night sky and tried to draw meaning from what they saw when a comet passed overhead. But this quote could easily apply to a modern man tracking a few points of data in a spreadsheet and trying to find some hidden meaning.

I tracked my headaches, sleep quality, and coffee intake for 2.5 years trying to find patterns. The single pattern I found was a decrease in headache frequency when I sharply reduced my coffee intake. Look at the chart below.


This was my comet in the sky. And guess what? The pattern was false.

From the moment I stopped doing Quantified Self my headaches plummeted and I didn’t change my coffee intake. Except for the month when I was playing Candy Crush, the headaches have almost all but disappeared. I don’t think I’ve had a single headache this entire year that has woke me up in the middle of the night that was intense enough to prevent me from returning to sleep. During the 2.5-year Quantified Self experiment, I averaged 7.5 bad headaches a month.

Why have the headaches disappeared? And how did they disappear all while consuming high levels of coffee? I don’t have a spreadsheet to tell you the answer, but I’ll speak from the gut. The headaches came from stress. One huge source of the stress was Quantified Self. Tracking something daily that I was failing at publicly clearly played a role.

How did I deal with stress? The dopamine hit of another espresso always made me feel a little bit better. But the fact I couldn’t control my coffee intake also made me feel worse. So when I was able to better manage stress and reduce my coffee levels, my headache levels dropped. Coffee was likely a symptom and not the cause.

Today I am drinking a fair amount of coffee. My sleep is perfect and my headaches seem to be gone. Had I not rejected Quantified Self, I never would have learned that coffee intake was a false pattern recognition. I also suspect a lot of the conclusions others are drawing from their Quantified Self experiments are false pattern recognitions.

The rest of the quote from Neil deGrasse Tyson:

We hunger for significance, for signs that our personal existence is of some special meaning to the universe. To that end, we’re all too eager to deceive ourselves and others.


Add yours

  1. So you can either do Quantified Self or coffee, but not both.

  2. @Garymar – Actually no. My point was that QS could not capture the likely cause of my headaches, which was stress. Coffee intake was increased during times of stress, but it was not the cause. Coffee was a metric of control. Once I grasped this relationship, it lost its power. If you look at the chart, even when coffee was super low, there were still headaches, just at a lower number. Stress and control at the lower bound.

  3. I get it now — the coffee was just a response to the stress.

    I’ve been weighing myself daily for over 4 years, and measuring my waist for almost 3 years. I’m long past the point of getting upset if the numbers increase — it’s pure habit now, so there’s no stress involved. I’m not about to start measuring my biceps and quadriceps though. That way lies madness.

  4. Quantified Self did actually help you solve your headache problem – through the act of quitting it. You started tracking headache data because you already had a problem that you wanted to solve. Coffee looked like a good candidate and may have led to false conclusions, but, had it not been for the added stress of Quantified Self (and then your eventual quitting of it), you may not have landed on stress as the actual cause.

    Do you have less overall stress in you life now than before you started Quantified Self headache tracking (but still had headaches)?

    The big takeaway here is that stress is kind of important. Like, really kind of important.

  5. @Garymar – I did that for a few years myself. I always knew to throw away the highest and lowest number for the week (outliers) and then average the rest. I fell into a very narrow range. During one of my moves, I realized I no longer had a scale and never bothered to replace it. Use the one at the gym now.

    @Aaron – Maybe, but it took me way too long to figure it out. I put out this post as a warning to others that are tempted to use QS to solve problems that can’t be solved via quantification. We need to be developing our intuition.

    I don’t know if I have less stress now, but I do know my response to stress has never been better. Stepping away from the spreadsheet was the first step.

  6. Surprisingly we don’t really recognise our stresses until we look back an identify an improvement when the stress is gone. I had this recently with family visiting and on the night before they arrived, developed fever and on day they were fetched from airport, a streaming head cold. When they left my symptoms improved, and within 24 hours the cold symptoms were gone. I only later recognised that the anticipated stress must have lowered my immune system. But once the cause of stress was gone (all self induced) my cold disappeared. Fascinating stuff. Intuition and hindsight
    awareness helps a lot.

  7. @Pauline – Well said.

  8. Although I agree that false pattern recognition can be a problem for QSers, I think in this case (and in many QS cases), the problem is trying to reinvent the wheel. There’s already been a lot of research into what causes headache, and the very first Google result for “what causes headaches” lists stress as one of the top causes.

    Seth Roberts certainly has the mindset of “look for surprising correlations,” but this method is mostly good for making new discoveries, not solving problems.

    If I were running an experiment on what caused my headaches, I would start out by trying to figure out the previous research, ranking the options by likelihood, then tracking the top three candidates and seeing if any of them correlated. If after a couple months, there was no correlation, move on the next three. Stress could easily be done with a saliva cortisol test, or a simple subjective 1-10 scale.

  9. @Matt – There are many variations of headaches. Mine didn’t fit any of the classic models. Mine woke me up in the middle of the night when I was most relaxed. They didn’t come on during a stressful day.

    The stress didn’t appear stressful at the time. (see Pauline’s comment)

    During the 2.5 years, numerous tests were done with common triggers. If you go back into the older posts (which I don’t advise), you will see MANY comments with what everyone else thought I should do. You name a health practice or vitamin or whatever and you’ll find a comment. Everything from drink more water to Call 911.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.