• Spencer Greenberg

A tool for running informal experiments to improve your daily life


ClearerThinking founder Spencer Greenberg devised the following procedure for running informal experiments designed to help you identify and implement simple improvements for your personal routine. Click here to use a free tool that automatically walks you through this procedure step-by-step.

I recommend running personal experiments every month or two to find new ways to improve your life. The basic procedure is simple:

(1) Think of an important area of your life that you’d like to try to improve (e.g. your fitness, sleep, emotional state, dating life, friendships, productivity, happiness at work, etc.).

(2) Come up with a few ideas for something safe you could start doing that has a reasonable chance of substantially improving your current situation in that area. (Try to avoid tactics that would only make a trivial difference even in the best-case scenario.)

(3) Identify which of these ideas you think is most likely to work.

(4) Put the chosen idea into practice for long enough that you have a reasonably good sense of whether it's causing the desired improvement (usually 1-6 weeks, depending on what you’re trying).

If you'd like a hand while trying out this procedure, click here for a step-by-step guide that'll walk you through the process automatically!

A few example experiments:

I try to run such an experiment on myself every month or two. Some of these flop – they don’t make my life better, or even make it slightly worse for the duration of the experiment. But when that happens I stop, and it’s not a big deal, since I usually avoid risky experiments. On the other hand, some of these experiments have succeeded, which means I’ve learned something new and potentially quite beneficial. I usually try to integrate successful interventions into my daily life as a more permanent change.

To give you some examples of what these self-experiments can look like, here are a few of the experiments I’ve tried for improving my sleep.

(1) Using a night mask when I sleep at night: I found that the mask helps me sleep longer, but I don’t seem to feel better-rested as a result, so overall the experiment left me feeling slightly worse off.

(2) Using ear plugs when I sleep: I found that I was woken up less by random noises, so I considered this experiment a success.

(3) Using a pillow on alternating nights: I found it more pleasant to sleep without a pillow, despite many people believing pillows are mandatory.

(4) Using small pieces of black tape to cover about 80% of every electronic light in the bedroom, reducing background light: I couldn't tell whether this experiment affected my sleep, but since the tape is already there and looks fine, I decided to leave it in place.

More about how to perform self-experiments correctly:

It’s usually smart to limit yourself to experiments that are very unlikely to backfire. Otherwise, if you perform enough of these trials, you face increasingly strong chances of significantly harming yourself sooner or later. Remember, that if you perform N self-experiments, each with a probability p of causing you significant harm, then you're facing roughly an N * p probability of significantly harming yourself at least once during the course of your experimentation — so even if the odds of harming yourself on a given experiment are small, they can add up quickly. So you’ll want to keep p (the probability of significant harm) very small if you make N (the number of experiments you conduct) large!

It’s also usually a good idea to focus your experiments on interventions that could plausibly produce moderate to large improvements in your life. You likely won’t be able to tell the difference between no effect and a slight effect on your day-to-day affairs, given the amount of variance we usually experience in these realms. For instance, if an experiment improves your mood by 2%, you probably won’t notice it unless the effect occurs within seconds. And if something improves your mood by 5%, you may not notice unless it happens within 30 minutes or so. So unless the effects of an intervention are immediate each time you perform it, I suggest you stick to looking for reasonably large effects.

You can try to detect more subtle effects by carefully tracking data on how well the experiment is going, but doing so correctly can be quite tricky and time-consuming — it could take 1-3 months of careful data collection, and even then, you may only be able to tell that your intervention is correlated with some minor changes, rather than establishing a clear causal link between it and the results you're seeing. So I think it’s usually not worth it to take this approach, unless you enjoy tracking data or if the problem you’re working to solve is serious. (For instance, if you get terrible migraines once per week and you want to pinpoint the cause, or if you have started to experience an unpleasant chronic condition that varies in intensity from day to day, and you want to figure out what might be making it worse on some days.)

Tracking data for experiments that deal with serious problems:

In cases where you are trying to track data regarding something serious, consider recording all the relevant variables you can think of every morning and evening. You’ll want to choose a consistent method for measuring outcomes. For instance, if you're running experiments designed to help alleviate migraines, you might want to track the frequency of your headaches along with how bad they are on a “1 = mild” to “5 = severe” scale.

You may also want to consider randomizing behaviors that you think might have causal relationships to these variables (e.g., flipping a coin to decide whether you eat a type of food that you think might be a migraine trigger). Here's why randomizing behaviors can be useful:

Suppose you're running an experiment involving an intervention whose effects you don't expect to last for more than a day. In such a situation, you could can try alternating or randomizing which days you perform the intervention to make it a bit easier to tell if they’re working. Randomizing whether or not to try an intervention on a given day makes it dramatically easier to determine causality, since otherwise there may be confounding variables that correlate both with the outcome you’re measuring and with your performance of the intervention itself.

For instance: it might be confusing to determine whether sleeping late actually make you feel crappy, or whether you just tend to "try out" the intervention of sleeping late when you’ve been drinking the night before, increasing your chances of feeling crappy for reasons that have little to do with sleep. Randomizing when you choose to sleep late will pull apart the two factors so you can tell which one is causing the effect you've observed.

So whenever you run a self-experiment, remember:

(1) Devise several tactics for improving the area of life you're focused on, and pick the one that's most likely to work for your experiment.

(2) Try it for long enough that you can tell whether it's working (typically 1-6 weeks).

(3) Avoid interventions that may backfire and cause you harm.

(4) Look for interventions that could plausibly cause a moderate-to-large improvement in your life.

(5) Consider carefully collecting data on your results if you're trying to solve a serious problem.

(6) Consider randomizing behaviors to determine what's causing the problem at hand.

And don't forget – you can try my experiment-planning tool for an easy walkthrough of this procedure, from start to finish!

Recent Posts

See All