They are not "experimenting" on themselves, they are simply engaging in a chemically-mediated attempt at wish fulfillment. There are few, if any, actual experimentation protocols being used and whenever I check out the scene to see if it is any better Instill find a random collection of anecdotes and poorly gathered "data points." I see no baseline data, no blinding or controls, the worst sorts of confirmation bias in data gathering and assessment, and people entering these "experiments" with a random selection of pre-exisiting conditions for which they are attempting self-medication. This is just an exercise in generating noise, not data.
> [offer] "safe self-experimentation sites", complete with medical researchers + monitoring equipment to capture data, and nurses to help with bad reactions?
That's what this whole subthread is about.
You could try teaching these individuals experimental protocols if you like, but if you enforce them, that means you're the experimenter. You're not allowed to be the experimenter. You just have to hope they're doing something useful. Probably you'd point a lot of automated instrumentation at each room, rather than having any human observers giving anyone (costly) individual attention, because most of the experiments would be too mired in sources of error to tell you much.
But keep in mind that the point wouldn't be to capture subtle things like drug efficacy. The point would be to find out about side-effects that might otherwise not be found, by having people around when they suffer adverse reactions to the drug to help the person not-die, and then being able to pull reports directly from the people who helped the person not-die, to find out (in clear medical terms) what the adverse reaction was.