6 ms·

I often get irritated when people treat "correlation does not imply causation" as a slam-dunk refutation of a complex argument, although you do rescue yourself

I often get irritated when people treat "correlation does not imply causation" as a slam-dunk refutation of a complex argument, although you do rescue yourself from this by shifting to the use of the word "proof". In the sense of "logically entails", or "proves", then it's trivially true that correlation does not imply causation. In the sense of "is evidence supporting"---which is nearly universally (outside of Math and Philosophy departments) the way that the word "implies" is used in English---correlation absolutely implies causation. It's a good heuristic for deciding where to focus research dollars, if only to uncover a more complex causal relationship. The absence of correlation is one of the most commonly used proofs in science. Surely its presence can be used as good evidence, if only because we have to operate in a universe where the excluded middle isn't self-evident.

mattkrause 6y ago
Sure. Nobody (seriously) believes that correlations can’t be suggestive of a causal relationships....but they do need to be followed up. It’s more that the AZ field has found this particular correlation over and over and over again. Despite millions (possibly closing in on billions) of dollars, however, it’s proven difficult to turn it into a causal link. A-Beta antibodies repeatedly fail, both before and after the plaques appear. Various types of inhibitors don’t work. If a method does something to amyloids, someone has probably tried it already, often in an expensive and ill-advised trial. This is only barely hyperbolic: https://blogs.sciencemag.org/pipeline/archives/2018/06/12/an-alzheimers-statement https://blogs.sciencemag.org/pipeline/archives/2018/06/12/an...
- btilly 6y ago
  Despite millions (possibly closing in on billions) of dollars... Oh, we're definitely into billions. Per year, and not total over time. This year alone, the NIH is spending $2.8 billion on Alzheimer's research. And they are not the only source of research money. Almost all of that goes to research based on the amyloid hypothesis.
  - mattkrause 6y ago
    I thought there's no way that ~10% of the NIH budget would be thrown into that particular money pit, but lo and behold! https://www.nia.nih.gov/sites/default/files/2019-07/FY21-NIA-bypass-executive-summary-508_1.pdf https://www.nia.nih.gov/sites/default/files/2019-07/FY21-NIA...
    - btilly 6y ago
      I actually was surprised to read the following sentence in there, While amyloid continues to be a target, 26 of the 41 pharmacological trials supported by NIA are focused on other targets. That's still far too much being thrown into a theory with a history of failure. But there is a shift there. Perhaps being criticized in places like Nature and Scientific American had an effect? One can hope!
btilly 6y ago
Your irritation notwithstanding, a complex argument that falls apart to "correlation does not imply causation" is a flawed argument. And humans so consistently err on the side of leaping to conclusions in advance of the facts, that we need to consciously counterbalance by reminding ourselves of the fact. Yes, correlation is suggestive of causation. Yes, it suggests an often fruitful line of research. But while pursuing that line of research it is important to remain aware that it isn't proof. And to not shut off other lines of research. In this light, it is worth reading http://calteches.library.caltech.edu/51/2/CargoCult.htm http://calteches.library.caltech.edu/51/2/CargoCult.htm again. Science isn't supposed to be easy. Scientists have had to learn the hard way how to be honest with themselves. Some fields have done better on that than others. Medicine's record on this front has generally been abysmal. And their failure to be honest with themselves about the limits of their own theories in this case has had an opportunity cost measured in trillions of dollars spent on millions of tragically preventable lingering deaths. In the face of that, it is wrong to argue that their mistake was a defensible jump in reasoning that people shouldn't be blamed for. Instead we should notice that a logical error really can have a cost measured in trillions spent on millions of tragically preventable deaths. And figure out how to not do that again.
- ajarmst 6y ago
  You're begging the question. My core point was that "correlation does not imply causation" doesn't "refute" anything but logical postulates---the problem with something that can be so dismissed isn't that it's logically flawed but that it isn't science. Which I think is closer to what Feynman was describing than what you're implying. When you say "science isn't supposed to be easy" (I think most people already know that, but thanks), you're supporting my position: saying "correlation is not causation" isn't helpful. In itself, it's either misleading or trivial. And it deserves a standard response: "Thanks, Professor. What's your experimental design to investigate that? Competing hypotheses? Evidence supporting your position?" Yes, science isn't supposed to be easy, and that applies to disputing theories and evidence, too. I have my own problems with the release in the OP, but are you seriously claiming that it's based on Cargo Cult Science? You know, or should know, that there's more than mere correlation in the evidence and theory around AD research community focus on plaques. You also know or should know that there is good research being done on competing theories---ones treating the plaques, perhaps not as epiphenomena, but as secondary to a more subtle primary mechanism. If you think that more research dollars should go to the second group (and I would likely agree), cool. But if you seriously think that the hypothesized connection between Aβ plaques and AD is based on an easily dismissed "logical error", you're not competent to participate in that discussion.

mattkrause 6y ago

Sure. Nobody (seriously) believes that correlations can’t be suggestive of a causal relationships....but they do need to be followed up. It’s more that the AZ field has found this particular correlation over and over and over again. Despite millions (possibly closing in on billions) of dollars, however, it’s proven difficult to turn it into a causal link. A-Beta antibodies repeatedly fail, both before and after the plaques appear. Various types of inhibitors don’t work. If a method does something to amyloids, someone has probably tried it already, often in an expensive and ill-advised trial. This is only barely hyperbolic: https://blogs.sciencemag.org/pipeline/archives/2018/06/12/an-alzheimers-statement https://blogs.sciencemag.org/pipeline/archives/2018/06/12/an...

btilly 6y ago
Despite millions (possibly closing in on billions) of dollars... Oh, we're definitely into billions. Per year, and not total over time. This year alone, the NIH is spending $2.8 billion on Alzheimer's research. And they are not the only source of research money. Almost all of that goes to research based on the amyloid hypothesis.
- mattkrause 6y ago
  I thought there's no way that ~10% of the NIH budget would be thrown into that particular money pit, but lo and behold! https://www.nia.nih.gov/sites/default/files/2019-07/FY21-NIA-bypass-executive-summary-508_1.pdf https://www.nia.nih.gov/sites/default/files/2019-07/FY21-NIA...
  - btilly 6y ago
    I actually was surprised to read the following sentence in there, While amyloid continues to be a target, 26 of the 41 pharmacological trials supported by NIA are focused on other targets. That's still far too much being thrown into a theory with a history of failure. But there is a shift there. Perhaps being criticized in places like Nature and Scientific American had an effect? One can hope!

btilly 6y ago

Your irritation notwithstanding, a complex argument that falls apart to "correlation does not imply causation" is a flawed argument. And humans so consistently err on the side of leaping to conclusions in advance of the facts, that we need to consciously counterbalance by reminding ourselves of the fact. Yes, correlation is suggestive of causation. Yes, it suggests an often fruitful line of research. But while pursuing that line of research it is important to remain aware that it isn't proof. And to not shut off other lines of research. In this light, it is worth reading http://calteches.library.caltech.edu/51/2/CargoCult.htm http://calteches.library.caltech.edu/51/2/CargoCult.htm again. Science isn't supposed to be easy. Scientists have had to learn the hard way how to be honest with themselves. Some fields have done better on that than others. Medicine's record on this front has generally been abysmal. And their failure to be honest with themselves about the limits of their own theories in this case has had an opportunity cost measured in trillions of dollars spent on millions of tragically preventable lingering deaths. In the face of that, it is wrong to argue that their mistake was a defensible jump in reasoning that people shouldn't be blamed for. Instead we should notice that a logical error really can have a cost measured in trillions spent on millions of tragically preventable deaths. And figure out how to not do that again.

ajarmst 6y ago
You're begging the question. My core point was that "correlation does not imply causation" doesn't "refute" anything but logical postulates---the problem with something that can be so dismissed isn't that it's logically flawed but that it isn't science. Which I think is closer to what Feynman was describing than what you're implying. When you say "science isn't supposed to be easy" (I think most people already know that, but thanks), you're supporting my position: saying "correlation is not causation" isn't helpful. In itself, it's either misleading or trivial. And it deserves a standard response: "Thanks, Professor. What's your experimental design to investigate that? Competing hypotheses? Evidence supporting your position?" Yes, science isn't supposed to be easy, and that applies to disputing theories and evidence, too. I have my own problems with the release in the OP, but are you seriously claiming that it's based on Cargo Cult Science? You know, or should know, that there's more than mere correlation in the evidence and theory around AD research community focus on plaques. You also know or should know that there is good research being done on competing theories---ones treating the plaques, perhaps not as epiphenomena, but as secondary to a more subtle primary mechanism. If you think that more research dollars should go to the second group (and I would likely agree), cool. But if you seriously think that the hypothesized connection between Aβ plaques and AD is based on an easily dismissed "logical error", you're not competent to participate in that discussion.