Someone in the Slate Star Codex subreddit wasn’t very convinced by my previous post.
I copy here 4bpp’s comment:
I don’t find the Artir post’s section on homeopathy very convincing. He writes:
Against homeopathy: Violation of the laws of physics, the consensus of medical scientists, several meta-analysis finding no effects. Most of the evidence for homeopathy comes from homeopaths.
Except for the laws of physics (and I doubt that Artir has a sufficiently good understanding of physics to arbitrate on that; he may defer to physicists, but then he has to understand which physicists are trustworthy, too), all of these cut both ways. For homeopathy: the consensus of alternative medicine practicioners, several meta-analyses finding effects. Most of the evidence for medicine comes from medics. I strongly get the impression that he just found a fancy way of obfuscating a cognitive process that will never yield a result distinguishable from “keep believing what you already believe” or at most “believe what people with money and status in your society believe”.
4bpp makes a good point, in that I just mentioned weak evidence going for homeopathy rather than ‘meta-analysis finding weak evidence’. I’ve fixed that. But still: meta-analysis done by the homeopath camp find weak effects, while meta-analysis done by the non-homeopath camp finds effects consistent with the null hypothesis of homeopathy being as good as placebo. If homeopaths were finding large effects with high confidence, that would probably make me more interested in the issue.
Regarding the laws of physics, I do know a bit about physics, as I am an engineer. I don’t know much, however, about quantum mechanics, relativistic physics, or the kind of physics that homeopaths invoke to support their claims. They talk about nanobubbles and nanostructures that form in the water. I certainly would have to spend some time to fully understand those papers.
By ‘the laws of physics’ being against homeopathy, I mean that, as everyone acknowledges, in the final homeopathic dilution, there is no, or almost no original substance remaining, just water. And water itself, in the consensus of physicists, doesn’t have the water memory effect homeopaths need. Another weird property that homeopaths need is that the effect gets stronger the more diluted the preparation is. Maybe there is some mechanism that works. Maybe if I read the literature on nanobubbles and nanostructures I’ll be convinced that homeopathy works. But I won’t read it for now, because I predict it will end up not convincing me. Also, if it did work, the effect would be really small, so you wouldn’t be able to benefit from it as you do from conventional medicines. So there is little epistemic or instrumental reason for me to do further research on this. What would I need to believe in homeopathy? As I said in the other post, a low-cost of acquisition source of information: a FAQ that presents homeopathy, deals with criticisms and seems plausible to me. And this would work just because I’m interested in knowing the most truths possible. From a merely instrumental point of view, I wouldn’t even care reading that FAQ, as even if it worked, it wouldn’t be much useful to me.
Next, 4bpp notes, correctly, that by default I defer to the consensus of physicists. I don’t say that if a physicist says that if P is the case, then P is the case. In epistemically relevant fields, there is no person or body of people that makes claims about that field automatically true. That only happens in, say, fiction, where the author says that something is true about a fantasy world. The fact that Brian Josephson disagrees with the consensus is some reason to reduce our confidence in the consensus being right. But, in my case, not enough -and I guess in 4bpp’s case it is the same-to stop believing the consensus.
Next, 4bpp raises the issue of the relevant consensus. The consensus of homeopaths say homeopathy has weak effects. The consensus of medical researchers says it has no effects. Which consensus do you pick? Heuristic: the one that has more endorsements. You can consider each physicist/medical researcher/homeopaths’s opinion as evidence for the conclusion, and the more the better.
Finally, 4bpp says that my heuristics will end up being “believe what you already believe” and “believe what people with money and status believe”.
Perhaps that is a plausible inference, given that in my two examples, homeopathy and Soviet healthcare, I initially began believing a conclusion, and ended up believing in the same conclusion. But this need not be the case. For example, in a forthcoming post on Soviet nutrition, I initially believed that the Soviet Union would be plagued by famines, queues and poor nutrition. But after a literature review, I learned that it wasn’t the case. So I changed my mind. Here, my first heuristic of seeing if it ‘seems plausible’ failed to track the truth. The next level, which was to check the FAO’s statistics on calorie consumption, reduced my confidence in the belief. Next, I couldn’t find relevant critiques of that statistics, which further reduced my confidence. And finally, a literature review changed my mind. I’ll post it here when it’s ready.
That’s an example of changing my mind. Here is an example of believing against the consensus: I believe free banking would work much better than the current monetary system. I came to believe this by reading arguments in favour and against it. At some point, I will do a post on free banking to explain why I think departure from the consensus is justified in this issue.
Where 4bpp has a point, though, is that my low-level heuristics are conservative and lazy. By default you will retain your beliefs, and if you change them, you will adopt that of the consensus. My higher-level heuristics, however, are not conservative and lazy, which is why I rarely apply them, they take some effort. But this is a virtue, not a defect. Usually, going with what seems plausible, or what the biggest group of experts find plausible is a the best idea. Only if you have some special reason to go beyond is the extra effort justified, I think. It would be interesting to apply my heuristics at different points back in time and see what beliefs you end up with, and if there are superior heuristics to mine that consistently yield better truth-tracking.