It would be accurate to say that self-directed improvement has a lot of failure modes that are hard to recognize from the inside -- human biases and all. Working with others in a shared environment with scientific ground rules ensures that your biases and their biases form a non intersecting set and you're left with the truth.
I work in open source and it is very often the case that someone new comes to the project with a gigantic, unreviewable pile of changes that they want merged. Almost inevitably, it is 90% bad changes on top of 10% of innovation, and the bad came about because they didn't understand what they were changing or the reason for its existence. The 10% is good but they've got to go back and extract it out which is a long and protracted process. Much better to have been involved in the community from the beginning, where they would have had things they wouldn't have thought of themselves pointed out and learned bits that they wouldn't have thought relevant, but are.
Working with others in a shared environment with scientific ground rules ensures that your biases and their biases form a non intersecting set
I liked your first point but come on here.
LW has a problem. Openly or covertly, many posts here promote the idea that a rational person ought to be able to self-improve on their own. Some of it comes from Eliezer's refusal to attend college (and Luke dropping out of his bachelors, etc). Some of it comes from our concept of rationality, that all agents can be approximated as perfect utility maximizers with a bunch of nonessential bugs. Some of it is due to our psychological makeup and introversion. Some of it comes from trying to tackle hard problems that aren't well understood anywhere else. And some of it is just the plain old meme of heroism and forging your own way.