I use a three part triage model for this sort of thing.
The trigger for going from casual internal reasoning to dumping everything I know onto a single page is if I think the same thought several times, which means the problem is too large for working memory and I’m likely to just spin in circles. (ht Critch and Satvik)
The trigger from going from just a page with everything on it to a simple linear model is if there aren’t clearly dominant considerations in the one page plan.
When making a simple linear model I don’t sweat the details, because in tests even naive adhoc linear models perform great (better than expert intuitive judgement). Rough scale from 1-5 for subjective cognitive difficulty of a task? Sure, not sure what I’m even measuring but it still mostly works.
Worth noting that these sorts of simple models are particularly effective in hiring, where using ad hoc metrics predicted job performance better than experienced hiring managers did.
I use a three part triage model for this sort of thing.
The trigger for going from casual internal reasoning to dumping everything I know onto a single page is if I think the same thought several times, which means the problem is too large for working memory and I’m likely to just spin in circles. (ht Critch and Satvik)
The trigger from going from just a page with everything on it to a simple linear model is if there aren’t clearly dominant considerations in the one page plan.
When making a simple linear model I don’t sweat the details, because in tests even naive adhoc linear models perform great (better than expert intuitive judgement). Rough scale from 1-5 for subjective cognitive difficulty of a task? Sure, not sure what I’m even measuring but it still mostly works.
Worth noting that these sorts of simple models are particularly effective in hiring, where using ad hoc metrics predicted job performance better than experienced hiring managers did.