Five Suggestions For Rationality Research and Development

post by LoganStrohl (BrienneYudkowsky) · 2021-06-14T01:09:26.532Z · LW · GW · 3 comments
  1. Make field work the bulk of what you do. For every hour you spend engaging with other people's thoughts or reflecting on your own thoughts, spend at least two hours, and ideally two days, trying to directly observe whatever it is you're studying.

  2. Add other people to your sensorium. Make conversation with others a big part of your work, and approach those interactions with the intent to study the experience of others.

  3. Corollary: If you're developing a technique or method, then by all means teach it to people. That's crucial. But don't teach them so they will know your technique, or even so you'll be better at teaching it. Instead, teach them so you can find out what happens for them when they try to use it.

  4. Focus on topics in the intersection of "things you're personally interested in", "things you think might actually matter to someone in particular", and "places where the existing art seems deficient".

  5. If it seems worthwhile to you, go after it. This is not a field with people in the position to tell you what you're allowed to study, or when, or how. This is a frontier. No bus will ever arrive, so you'll have to use your feet. Stop waiting for permission. Just get to work.


Comments sorted by top scores.

comment by DirectedEvolution (AllAmericanBreakfast) · 2021-06-14T03:04:26.120Z · LW(p) · GW(p)

6. Treat the problem of measurement with the same care you treat the intervention. Much of rationality training will have its origins in introspection and intuitive inference. Figuring out what effects your hypothesis ought to predict, precisely enough to come up with a solid form of measurement, might be a lot harder than designing an intervention. Just because you don't have a good way to measure what you're doing doesn't mean it has "no evidence." But if you can find a way to measure, that can be very powerful.

Replies from: ChristianKl
comment by ChristianKl · 2021-06-14T10:15:26.324Z · LW(p) · GW(p)

It's worth noting that inventing good ways of measuring issues is as important for developing the field as developing interventions.

I'd love to see posts on LessWrong that purely focus on the way to measure a particular issue inside of rationality. We had some of that in the past with credence calibration but it would be great if talking about how to measure what we care about would a larger part of the rationalist discourse. 

Replies from: peterson-yook-1
comment by Crackatook (peterson-yook-1) · 2021-06-15T13:23:30.078Z · LW(p) · GW(p)

This conversation reminds about the Measurement Theory. I didn't take it yet, I heard it is an abstract course and applicable to social science. It started from measuring dimensions, and expanded to probability. but for a quick whole picture, it looks very mathematics and no clue how to measure bias, haha.