It seems this claim might itself be amenable to testing on groups of people to see if it "holds on average" across a spectrum of people who vary in the degree to which they are "weird". I suspect such a study would reveal uninformed self experimentation to be less effective than naively expected.

For most of my life I've had a useful heuristic for problems which is "seek out a self help book on the subject that seems relevant and is reasonably well recommended and try things in it if they seem like they might work". I got this heuristic from my mom, though I don't know whether it was her own invention or something she got from someone.

In any case, one of the ways she motivated the advice was to notice that on several distinct occasions she initially thought she was a unique snowflake with an unusual problem and then she found out from a book that lots of people had faced the same problem and were able to articulate surprisingly specific details of the problem that she'd thought where unique to her own circumstances. A sense of a problem being unique was even one of the things people would sometimes bring up as such a detail.

Based on this, its easy to see how people might not talk much about the painful or embarrassing things in their life, but its less obvious to carry that insight through to lowered estimates of one's own uniqueness and therefore a higher estimated value for finding usefully relevant literature.

One nice thing about "non-uniqueness" as a default assumption is that it trivially suggests a method of falsification for a problem: state your problem clearly, work out related keywords, and hit the library. If you don't find anything, then either you need to spend more skill points on library science, you have the wrong key words, or your problem is a genuine counter example and in that case it really would make more sense to deploy self-experimentation techniques instead of library skills.

Personally, having worked on and off according to the above theory for a while, the tricky part seems to be knowing the keywords to search for. On several occasions I've checked the library, found nothing, and only later learned that a literature existed but not where I was searching. Learning about new keywords is something I find LW to be really good for... its one of the concrete benefits I get out of the site. For example, before today, I'd never heard of "bibliotherapy" :-)