3 ms·

My own view is not that self-experimentation is an appropriate, let alone likely efficacious, substitute for reconciliation to the idea of one's death. I certa

My own view is not that self-experimentation is an appropriate, let alone likely efficacious, substitute for reconciliation to the idea of one's death. I certainly don't endorse interference in others' treatment, however well-intentioned. If you want to say: "you shouldn't treat your terminal illness like a science fair project unless you possess extreme sang froid and are precommitted to the acceptance of your death", I'd find that totally reasonable. But my own view is rather that institutional epistemology is somewhat overrated, and self-experimentation somewhat underrated, relative to the conventional wisdom. (Though some people go too far in this direction.) This leads to general overconfidence in epistemic efficient market hypothesis arguments ("if a protocol were worthwhile, someone would have found it already") and underconfidence in the value of crowd-sourcing trying a bunch of stuff and writing it down. This view was principally informed by developing cancer drugs for a living and coming to appreciate that it's really hard, your knowledge of what's going on during a clinical trial is highly abstracted, and you can't be everywhere at once. It was secondarily informed by watching people do bro science on certain important questions and making interesting progress in large part because they could move much faster than academic or corporate research. If we recast the point of contention as: "what is the largest effect size that could be found by an institution outside of academia or industry?", my position is that it's plausibly non-zero. I'm sorry for your friends and I hope they found peace.

zmgsabst 2y ago
As a pure numbers game, I’d find it more surprising if “broscience” never found a result: - a lot of terminal patients are prone to experimenting - their overall number probably eclipses the total number of trial patients in a given year by at least one order of magnitude and I’d believe two or three - they don’t have institutional barriers to what they can try, eg, they’ll fund non-patentable treatments - a lot of their approaches are taking things from published papers and trying to recreate similar effects (eg, calorie control [1]) That they’ve stumbled across at least one treatment that solved at least one case for at least one patient seems likely. Isolating that from incorrect null results is where the epistemological struggle is. And there’s a good chance that it won’t help you with your particular case. But what’s the harm in trying? — you’re probably going to die anyway. [1] - https://pmc.ncbi.nlm.nih.gov/articles/PMC8749320/ https://pmc.ncbi.nlm.nih.gov/articles/PMC8749320/
- Earw0rm 2y ago
  True. But for the highest-grade nasties, where median life expectancy is unfortunately short and progression near-universal, you don't need much signal to get above the noise. Anyone surviving more than a handful of years with something like that is an outlier such as to merit a full work-back, and at that point it's no longer bro science.
  - zmgsabst 2y ago
    Conversely, those are also the least likely to be solved by random trial and error. Those people largely just die, no matter what you do — that’s what makes it a “highest-grade nasty”.
    - Earw0rm 2y ago
      I think that's partly a survivor (the disease surviving, not the patient) bias effect. Things that could be solved by random discoveries are no longer considered the highest-grade nasties. There were a lot more intractably fatal conditions in 1870 than there are today. So the likelihood of there being answers that could have been randomly discovered by medics with 1870 or 1920 levels of knowledge is tiny. At the same time, the sum of human knowledge has expanded so rapidly since then, it's not impossible for stuff to get missed.
- ryandrake 2y ago
  When you try someone else's "broscience", you're not really experimenting with the unknown, so it's unlikely you're going to stumble into a "result". They know it doesn't work. If it did work, they'd have patented it and licensed it to Merck or Pfizer. Choosing quackery is not experimenting.
  - lambdaphagy 2y ago
    Elsewhere in the thread I argue why efficient market hypothesis arguments are unlikely to fully apply in this case.

zmgsabst 2y ago

As a pure numbers game, I’d find it more surprising if “broscience” never found a result: - a lot of terminal patients are prone to experimenting - their overall number probably eclipses the total number of trial patients in a given year by at least one order of magnitude and I’d believe two or three - they don’t have institutional barriers to what they can try, eg, they’ll fund non-patentable treatments - a lot of their approaches are taking things from published papers and trying to recreate similar effects (eg, calorie control [1]) That they’ve stumbled across at least one treatment that solved at least one case for at least one patient seems likely. Isolating that from incorrect null results is where the epistemological struggle is. And there’s a good chance that it won’t help you with your particular case. But what’s the harm in trying? — you’re probably going to die anyway. [1] - https://pmc.ncbi.nlm.nih.gov/articles/PMC8749320/ https://pmc.ncbi.nlm.nih.gov/articles/PMC8749320/

Earw0rm 2y ago
True. But for the highest-grade nasties, where median life expectancy is unfortunately short and progression near-universal, you don't need much signal to get above the noise. Anyone surviving more than a handful of years with something like that is an outlier such as to merit a full work-back, and at that point it's no longer bro science.
- zmgsabst 2y ago
  Conversely, those are also the least likely to be solved by random trial and error. Those people largely just die, no matter what you do — that’s what makes it a “highest-grade nasty”.
  - Earw0rm 2y ago
    I think that's partly a survivor (the disease surviving, not the patient) bias effect. Things that could be solved by random discoveries are no longer considered the highest-grade nasties. There were a lot more intractably fatal conditions in 1870 than there are today. So the likelihood of there being answers that could have been randomly discovered by medics with 1870 or 1920 levels of knowledge is tiny. At the same time, the sum of human knowledge has expanded so rapidly since then, it's not impossible for stuff to get missed.
ryandrake 2y ago
When you try someone else's "broscience", you're not really experimenting with the unknown, so it's unlikely you're going to stumble into a "result". They know it doesn't work. If it did work, they'd have patented it and licensed it to Merck or Pfizer. Choosing quackery is not experimenting.
- lambdaphagy 2y ago
  Elsewhere in the thread I argue why efficient market hypothesis arguments are unlikely to fully apply in this case.