How to decide under low-stakes uncertainty
post by dkl9 · 2023-08-11T18:07:50.133Z · LW · GW · 4 commentsThis is a link post for https://dkl9.net/essays/mental_coinflip.html
Contents
Mental coinflip More than two options None 4 comments
Say you're stuck in uncertainty between two actions you're considering. They seem about equally good, but you suspect one is better, and it's not obvious which. You already have all the information to obviously collect about the problem.
For situations where getting it right really matters, try harder to get more information, and use methods more reliable than those presented here.
For lower-stakes problems:
- Assign one option to 0/tails and the other to 1/heads.
- Flip a coin, i.e. use a one-bit random number generator.
- Start accepting the decision from the coinflip, and observe your revealed unconscious intuition, for you will probably have some.
- If that intuition opposes the coin's decision, go with the intuition. Otherwise, you either actively came to agree with the random decision, or it still seems terribly uncertain, so you go with the random choice.
This is classic advice. I'm just sharing it. The "innovation" here is a way to handle cases where it's inconvenient to flip a coin or do the equivalent.
Mental coinflip
- Assign one option to 0/tails and the other to 1/heads.
- Think about something unrelated.
- Notice the first word that pops into your mind.
- Count letters in that word. If it's even, you get 0; if it's odd, you get 1.
- Use that 0 or 1 to continue at step 3 of the general procedure.
This is very crude randomness, but it doesn't really have to be random, just uncorrelated with the topic of decision. You should not use a mental coinflip to try to run a fair game of chance.
There may be a bias in the length-parity of words you think of. I, so far, haven't noticed one. But that wouldn't necessarily be a problem. Once you notice such a bias, you can exploit it by assigning actions to 0/1 so that the mental coinflip bias opposes a bias you naturally have.
E.g. if you pick odd-length words more often, when the choice is between "doing nothing" and "doing something", assign "doing something" to 1. Laziness may lead you to want to "do nothing" more often than you should.
More than two options
You can use the same methods here for decisions more complicated than a dichotomy.
- List the options you're considering.
- Split that list into two subsets.
- Apply the coinflip-assisted decision procedure to pick one subset.
- If the selected subset has just one item, that's your decision. Otherwise, repeat this procedure on the selected subset.
4 comments
Comments sorted by top scores.
comment by AK1089 · 2023-08-11T22:43:55.493Z · LW(p) · GW(p)
The trick I usually use for this is to look at my watch (which is digital and precise to the second), and taking the seconds counter, which is uniformly distributed across . I then take option 0 if this is even and option 1 if it is odd.
(This also extends nicely to decisions with any small number of options: 3, 4, 5, and 6 are also factors of 60, so seconds modulo X is uniformly distributed across . Plus, I'm not very good at coming up with words.)
Replies from: dkl9, duck_master↑ comment by duck_master · 2023-08-12T01:15:08.425Z · LW(p) · GW(p)
This is an excellent tip! I plan on using it from now on in my day-to-day life.
comment by Olli Järviniemi (jarviniemi) · 2023-08-12T00:04:50.437Z · LW(p) · GW(p)
How I generate random numbers without any tools: come up with a sequence of ~5 digits, take their sum and look at its parity/remainder. (Alternatively, take ~5 words and do the same with their lengths.) I think I'd pretty quickly notice a bias in using just a single digit/word, but taking many of them gives me something closer to a uniform distribution.
Also, note that your "More than two options" method is non-uniform when the number of sets is not a power of two. E.g. with three sets the probabilities are 1/2, 1/4 and 1/4.