What I often observe on HN is that SlateStarCodex didn't understand how IRBs work, wrote about it, and therefore many people think that IRBs are bad. So, I am wary of IRB arguments here. Not implying that your information on IRBs comes from that one person, just saying that I don't love arguing about IRBs on HN because of a repeated pattern here.
If your point is that "IRBs could be improved in various ways", I'd be happy to agree in many cases. But also, there are some shining examples of enlightened IRBs that make human research at scale quite practical.
I feel ok with supporting Scott Alexander’s position when it is both (a) obviously correct, (b) IRBs regulate sociologists and psychologists doing surveys which is farcical, and (c) Carl E. Schneider, professor of medicine and law at the University of Michigan wrote an entire book on how IRBs fail to work.
IRBs do not have a problem. They are a problem. If we must have IRBs they should be deprofessionalised to the fullest extent possible rather than staffed by people whose job and motivation is to find problems where none exist and add more bureaucracy to prove they are actually doing some work. IRB as jury service seems like it could be acceptable. Twelve people, a three hour course on the Nuremberg code and the scientist presenting their application. And absolutely no bioethicists.
If you’re interested in the book put it on an Amazon wish list and email me the link at barrypcotter at iCloud. I’ll buy it because more doctors and scientists should know IRBs are a cancer on society and science.
If your point is that "IRBs could be improved in various ways", I'd be happy to agree in many cases. But also, there are some shining examples of enlightened IRBs that make human research at scale quite practical.