A while back, Scott Alexander wrote a blog post about concurrently testing anxiety interventions:
The post covers a template for doing a double blind study on yourself for anxiety supplements and over the counter medications. It’s an interesting general method that’s rarely used.
If you’re trying to solve any problem (in the post’s case, finding effective anxiety supplements for someone), instead of waiting for one solution to work over a long time, you can organize things by whether they can be used together (in the post’s case, whether the anxiety medications have any chemical interactions), and then sorting by the time you expect it to show effect. You can run through medications much, much faster this way, whereas the typical psychiatric routine for anxiety is spread out across several months, with little data, and tests only one medication at a time. The idea should apply to any general intervention. It seems a lot better than building a habit around one optimization. Doing this well is extremely tedious, but likely fast to show substantive results, and I’ve found Beeminder to be the best way to ensure something like that happens.
In terms of the time something should take to work (I like the term “effect window”), ways to organize personal trials might be testing what you expect to work fastest first, or testing one thing that may take a while but show reliable effect alongside other faster alternatives.
What might an anxiety sampling stack include, more broadly than medication? And what might effective ways to use beeminder to aid the testing process?
(Anxiety or neuroticism is sometimes very context-specific, but I find that I have an unusually high average across all contexts. I would like to reduce my general neurotic disposition.)
I’ve listed some ideas to test below. (These are only suggestions!).
Supplements:
- L-Theanine (reported to work within an hour).
- Supposedly interferes with blood pressure medications and stimulants
- Kanna/Zembrin
- Studies done on this, according to the linked post, are not comprehensive.
- Effect window could be assumed 8 weeks, based on duration of study
- No information about interactions
- Ashwaganda
- Effect window assumed 1 month
- Don’t know where to find information about interactions
Diet:
- I’ve heard some people report feeling much better after quitting bread or carbohydrates, and others the reverse effect. I’m not sure what the effect window should be.
- Testing effect of caffeine usage (using/quitting) – ballpark of effect may be a month, to give time to get over withdrawal symptoms.
Exercise:
- I think that it’s hard to be prescriptive with exercise. My psychiatrist once said that a few hours a week is a general mental health recommendation. This doesn’t look that effective in practice, and a serious case of a general correlation (the amount of executive function needed to make the internal decision to go to the gym or get outside to run in the first place is higher than many depressed people have). I think it could be possible that e.g. zone 2 cardio exercise causes better effects than any medication by a staggering amount but only after around 20 hours a week, but not by much in the range used in studies.
- What might be useful to test here is narrowing down what type exercises actually make you feel less anxious and what types of exercise you enjoy. These are the types you are likely to do, which matters more. A ballpark might be “try it 3 times”.
Socialization:
- Like exercise, the goal should be a system that tests many different places and types of socialization as they affect anxiety. Ideas include: sports, board game clubs, attending makerspaces, martial arts.
Misc.:
- Light Therapy
- Unsure about effect window. What might be more useful with something like this is to test it multiple times for a few days but a few weeks apart, and record how you felt, along with time outside, and weather on days where you used it.
- Meditation
- Meditation is too broad to be prescriptive with it. Since there is no good external way to see if someone is meditating right, ineffectively, or in a way that might hurt them, I heavily discount most things in the cluster.
- For trial purposes, I once read in a meditation book that “an hour a day for a week” is a good window to litmus test whether any specific method is effective.