The 2012 LW Survey included a few questions that measured standard biases from the heuristics and biases literature. The general pattern was that LWers showed less bias on those questions than the university students in the published studies that the questions were taken from, and that people with closer ties to LW (read more of the sequences, higher karma, attend meetups, etc.) showed less bias than people with weaker ties.

This doesn't necessarily mean that reading the sequences (and getting involved in LW in other ways) causes the reduction in bias on those questions - it could just be that the people who have read the sequences will tend to show less bias on these questions for other reasons. In order to do more rigorous testing with randomization, we'll need smaller/quicker interventions than "read the sequences" (which is something that CFAR is working on).

There was one kinda-scientific test of some LWers' rationality in 2011, described here. But basically yeah, there haven't been scientific tests of LWers' rationality vs. the general population. Of course, that's true for basically all websites. CFAR has some ongoing experiments, though.

Thanks for mentioning this. Many posts in the sequences I've read so far, especially those concerning biases, seemed interesting, but not necessarily useful: I don't really see how to apply that knowledge to my own life. And when debiasing techniques are suggested, they often sound prohibitively expensive in terms of willpower. That said, I've also read quite a few posts of whose eventual usefulness I am reasonably confident. Off the top of my head, the sequence Joy in the Merely Real seemed really beneficial to me - if only because it gave me a strong argument to read more textbooks.