Hacker News Reader

Comment by nathan_compton | Hacker News Reader

I don't really get what you are saying here. RTCs are designed precisely to allow one to draw statistical conclusions which would be untenable due to confounding effects that would be impossible to disentangle otherwise, particularly in regimes where effect sizes are small and results are sometimes difficult to quantify.

I'm having trouble understanding what you are even getting at with comparisons to astronomy, where the absence of controlled experiments isn't some grand innovation astronomers cooked up but a basic constraint imposed by studying stuff that is light years away. Any decent epistemologist would tell you that the character of knowledge generated by astronomical observations is of a lower quality than that of a RCT. I'm sure some astronomers or cosmologists would give their left arm to do a randomized controlled trial!

View on HN

With astronomy, where the data are mainly derived from observations and simulations, no one is spreading alarms that it is not science. While with RCTs - and specifically RCTs in the filed of human cognitive neuroscience and psychedelics - there is all this monkey circus regarding whether placebos or psychedelic experiences are real. In human neuroscience ~80% of data is derived as well from observations and is effectively non-reverse-engineerable, while the hype regarding pseudoscience is much higher.

You buy aspirin in a pharmacy and the drug's instruction label lists tons of adverse effects - this is obviously a seemingly high quality of knowledge resulting from hard work in RCTs. Yet, there's absolutely no information predicting which exact adverse/beneficial effects will manifest in a specific person in a specific state of consciousness - and this is the actual empirical level where RCT derived information should actually matter and where it is ~50% useless (due to lack of context in RCTs themselves).

I still don't see what you are getting at. It is hard to generate good information about the risks and benefits of drugs and doing RTCs is very difficult, for the reasons to which you refer and others. Are you advocating that we just give up on knowing this stuff? That we do large RTCs that have the statistical power to characterize more "context"? I'm having trouble understanding whether your comment just comes down to "getting knowledge is hard and I'm tired of people trying to do it."

I am not advocating for anything, just exercising my English and typing skills. But you can try measuring more parameters in every person, integrate various findings from science and plan experiment design more carefully. I have heard about only one startup using AI to facilitate RCTs. A more optimal option is of course suicide, as the more of these RCT researchers will be out of the game, the more newer and more flexible brains will come in.

Reply on news.ycombinator.com