Ending the IRB nightmare

Something can be done on day 1

Suppose you are running a company that has a website. You are wondering whether marketing campaign A or B works best. Which one do you go with? Why not both? You run an experiment where visitors to your website are shown either ad A or B, and you check which one results in the most sales. Great, you are kind of a scientist. Did you need any permission for this? Of course not. This is an every day thing at thousands of companies worldwide.

Now suppose you are a scientist at a publicly sponsored institution, or God forbid a doctor, with a similar question: does survey question A or B work best for your purposes? Well, just do an experiment. Actually, not really. First you would need to secure permission to ask students or patients some questions. It's a simple matter, right? Oh sweet summer child. Getting permission is the main work of doing a study. It can be near-impossible to do even the simplest research. In 2017, Scott Alexander wrote a post about his own attempt to run a risk-free survey study at the hospital he was working at. You should read it in full here (My IRB Nightmare). Let me give you the summary:

1. There's some screening test (really, a self-report survey) about bipolar disorder and Scott thinks it's fishy.

2. So surely it should be easy to do a study on this with some new alternative right.

3. So let's ask the staff at the hospital’s Research Department.

4. Have you taken the "Pre-Study Training" course yet (teaches you not to build concentration camps)?

5. Have you found a senior staff member to front you (someone who already knows not to build concentration camps)?

6. Have you filled out the 22 page short version of the "New Study Application" (where you promise not to build concentration camps)?

7. Phew, you are now ready.

8. Just kidding, here's a bunch of random and silly questions and complaints about your study that you must address.

9. Repeat the above a few more times.

10. (eventually he gives up after months years of attempting to run a simple survey study)

That 2017 post also inspired me to read more into how this IRB (Institutional Review Board, Ethics board and various other names) and I wrote a review of The Censor’s Hand (by Carl E. Schneider), which you should also read. Later on in 2022, someone more qualified than me (Willy Chertman, an actual medical student) wrote an even longer piece: It’s Time to Review the Institutional Review Boards. As far as I know nothing has really happened because of this. In practice, that probably means things got worse (bureaucracy is like cancer, it always grows when left unchecked).

However, now that Elon and Vivek are going to be put in charge of removing bureaucrats, there is some hope. So what's the simplest, fastest, and move effective thing one can do? The president can issue orders to federal agencies telling them how to interpret various rules. Could Trump just tell the Department of Health and Human Services (HHS, RF Kennedy's gig) to chill out and stop asking for unnecessary paperwork? Yeah well, maybe. The executive orders from the president cannot contradict existing laws, which Congress would have to repeal. However, the president can tell federal agencies how to interpret the laws, within certain limits. So given this, it would appear that Trump could on day 1 issue such an order that says something like:

• All obviously safe research is exempt from IRB and needs for approval. No paperwork needs to be filed for doing such work. This would practically involve telling them how to interpret certain regulations about what research the law covers (I think it's 45 CFR 46).

• Obviously safe research is defined as research that no sane person could think was dangerous, which means that any research that asks questions in questionnaires, and generally doesn't involve putting things into people's bodies (or taking things out of their bodies).

• Also obviously safe is any research merely based on reanalyzing data already on file for other reasons (archival research).

In fact, the office that regulates IRBs already seem to be fine with such a change in interpretation, but the president has to order it:

A similarly narrow reform is implementing an electronic checklist that would allow researchers to self-determine if their research was low-risk and did not require IRB review. A University of Chicago professor, Omri Ben-Shahar, has developed exactly such a tool, and OHRP has no objection, but clear federal guidance would assuage the worries of risk-averse university administrators, who often still require IRBs to approve exempted studies. If universities continued to delay the use of such a tool, Congress could make receipt of government funds conditional on developing and allowing such a tool.

So what are we waiting for? Apparently, nothing.

Anyway, after issuing a such order, the Trump administration should do a simplification of whatever laws regulate IRBs (which seems to be the National Research Act from... 1974). I don't know exactly what needs to be cut, but when it comes to regulations (and firing bureaucrats in the public sector), it is best to remove too much rather than too little (the anti-Chesterton's fence). If one later realizes something was actually important, one can simply add it back (or rehire them). In other words, the prior is that the regulation is not necessary and the person is not needed. Recall how Musk fired over 80% of Twitter staff, and people were predicting it would immediately collapse, and yet Twitter is better than ever (to be fair, he has since hired more people; also financially Twitter is still bad, but Musk can afford that).