WORK IN PROGRESS

If your self-experiments are to improve your life, they shouldn’t take more time than they make you gain. This means that you should look for measures which take as little time as possible to record. Here I focus on the recording time, and neglect the data analysis / cleaning time. Indeed, the analysis is a fixed cost (compared to a daily recording cost), and has didactic value which makes the time spent during this part more bearable.

COMPLETELY PASSIVE #

Measures which take 0 time to record once set up.

Sleep #

Automatic sleep tracking ($$) In this list, I focus on measures which are time-cheap. But sometimes the time you gain has a monetary price, in which case I put one or several ($) symbol depending on the cost.

Focus / Energy #

Typing quality: using a keylogger, recording your daily error rate and typing speed, while adjusting for the kind of text you’re writing, should provides a good automatic measure of your daily focus. I don’t know of any easy way of doing this right now.
Reading time / speed on e-reader. I think it’s quite easy on Kobo devices, but complicated on Kindle devices (here’s a tutorial, basically you have to setup parental control for yourself, and it only works for books bought on Amazon).

Productivity #

Time spent on different activity on the computer: I use Rescue Time and selfspy.
Github statistics.

Mood #

Moodmap: This films you with your webcam when you’re using your computer, and tries to infer your mood. I tried it but it was taking too much ram for me. Still, it’s a young project and I think it can get much better, though I’m not sure how good a measure you can get with this idea.
Amazon Halo($$) claims to be able to analyse the emotional tone of your voice throughout the day.
In the same kind of idea, you could do a sentiment analysis of your private messages (see https://www.autumn.health/, currently in beta)

Diet #

Continuous Glucose Monitoring ($$) I’m not sure how good of a metric it is for non-diabetic people, and would appreciate any resources on this.

Health #

Automatic heart rate / temperature tracking ($$)

USEFUL ON ITS OWN #

Measures which take some time to record, but for which the recording is useful in itself. If you’re already doing it, it’s a passive measure.

Memory #

Anki: in addition to being an amazing tool for long-term memorization, Anki provides you with a lot of statistics. See the experiment I’m doing.

Focus / energy #

Typing quality / speed: you could export your data from KeyBr. Though it’s less passive than a keylogger, it would probably be easier to analyse.
Programming skills: measure the time you take to solve problems on something like Leetcode, controling for success rate of other participants.

FUN ON ITS OWN #

Measures which take some time to record, but for which the recording is fun in itself. If you’re already doing it, it’s a passive measure.

Focus / cognition #

chess statistics: I haven’t tried but it seems that you can easily export your chess.com games (here or here). You could measure two things: whether you lose or win a game (1), or the mean quality of your moves with something like stockfish(2). (1) would probably be way more noisy, but has the advantage of adjusting for your chess progress if you mainly play online (because you’re matched with opponents of your level), so maybe you could combine the two measures? (Chess puzzle may be easier to analyse if that’s your thing) I think it would be extremely valuable for a regular chess player interested in self-experiments. If that’s you, you can contact me. I can provide help setting up the experiment and statistical analysis. . See Kelsey Piper using her chess statistics for a slightly different reason (quickly assessing her daily cognition to choose which work to prioritize)?
FPS skills with standardized settings, for instance with AimBeast($). Alexey Guzey used this for his sleep deprivation experiment, and you can find his configuration files here.
Whatever game you’re currently addicted to. For instance, Scott Alexander used WordTwist to assess the effect of CO2 on his cognition.
There are also some “games” made especially to test cognitive abilities, like Human Benchmark or Quantified Mind. They don’t seem very fun so I don’t think it fits here, but if your a IQ test fan, it should provide good data.

CHEAP MEASURES #

Measures which take only a little bit of time to record

Mood / Energy tracking: With a few click each day, you can record your daily mood / energy level. For Android phones, I recommend Tap Log.
Weight tracking: You can enter your weight on your phone every morning, or buy a connected weight ($).

MEASURING COVARIATES #

Automatically measure something that impacts your goals

Here I focused on measuring goals, like cognitive functions, health, or sleep. Indeed, if you’re doing randomized experiment, you should probably just randomize some action (e.g “doing sport”, or “eating butter”), and let your passive measures tell you the effect of this action. Still, having passive measures of covariates can be useful for motivation, to make observational studies, which can be valuable in themselves, or provides a good starting point for randomized experiments Though I have seen many people recording a huge amount of observational data, only to find obvious or non-interpretable correlations. This is actually interesting: if it’s so hard to tell if correlation is causation when you’re your own subject, can you imagine how hard it must be when experimenting on other people ? , to provides covariates to add to your analysis as control, and thus improve your statistical power, or to check that the randomization is working well (for instance, if I randomize “doing sport”, does it makes me less eager to take a walk or bike to work ?). Covariates are especially useful to measure automatically when you have good reasons to think that the covariate variations are essentially random, or at least independant of you (air quality would be an example). In this case, if you have good goal passive measures, the experiment is essentially doing itself without any effort on your part! Covariates you can measure automatically includes:

the amount of sport you’re doing ($$)
the amount of sleep your getting ($$)
indoor or outdoor air quality ($$)
indoor temperature / humidity ($)
weather (you don’t need to record that yourself of course).
your social media patterns (e.g like this)
your messaging patterns (for instance, you can export your Messenger messages)
your music listening habits (like this) for Spotify, or using last.fm or Youtube). I think this would be interesting to correlate to one’s mood, though the causality would be hard to guess without randomizing the music listened. Eric Jain has suggested that this could be used as a passive proxy measure for actual mood.
your movie watching habits (e.g this)
where you’ve been (Google, or Moves or something else).
how much you’ve spent (e.g with Mint)