There's a post on SSC that sounds a little like that:
https://slatestarcodex.com/2018/04/12/recommendations-vs-guidelines/
Thanks for the link!
The paper I'm thinking of is more about the differences in knowledge on the order of: (a) there's a single, cheap, known fix for your problem, so you should just do x (because we know (almost) exactly what the problem is and know how to solve it, 'mechanically'; versus (b) here's a guideline (because we don't really know what the problem is, in detail, or specifically).
I checked the comments on that post and no one seems to have linked to the paper I'm remembering. I wouldn't be surprised that it's linked in comments on another post on SSC tho as I'm pretty sure I've seen links to it on this site (or maybe Overcoming Bias, before Eliezer stopped blogging there, long ago).
There's a paper (or essay or something similar) I can remember reading that I've been struggling to find for some time now. It described different qualitative types of knowledge and used various examples from medicine, e.g. the difference between various kinds of trauma or emergency care where we know exactly what a problem is and can fix it (almost) exactly/completely versus something like obesity where we basically don't know (at least in the sense of commonly shared knowledge) much of anything about what exactly the problem even is.
Any help in uncovering the paper would be greatly appreciated!