For the controlled experiment example, it would have to be double blind the entire time. Nobody would ever know who got placebo and who didnt’, and no difference in treatment regimens either way—all on the same diet, same lifestyle, etc etc. I highly doubt the accidental cases fit that profile.
If you want absolute certainty, you need to fulfill koch’s postulates, which is the medical disease variant of experimental physics, and is well grounded.
So yes, in questions of science, we do demand specific evidence for near-certain proof, but naturally when that is not possible for whatever reason, we can still update based on other evidence and reach well grounded conclusions.
First of all, you never have absolute certainty. Proof is for math and alcohol. You may also want to read the Sequences relevant to Bayesian reasoning especially about how 0 and 1 are not probabilities.
Note also that even outside a Bayesian context, Koch’s postulates are a rough framework. For example, sometimes there is great difficulty growing an organism in a culture (Koch’s second postulate requires this), and sometimes only a fraction of a population may be symptomatic. Even Koch was willing to treat the first postulate as a guideline rather than a hard and fast rule; he worked with multiple examples of microorganisms that showed up in some healthy people but didn’t cause disease In fact we know now that this very common. Many causes of minor infections, such as staph, are present on everyone. It then takes some problem, such a wound, or disruption of the immune system to cause the person to become unhealthy. For viral examples, look at asymptomatic herpes or HPV.
Frankly, referring to Koch’s postulates like this are one of the things that makes many people with medical knowledge not take the HIV-AIDS skeptics very seriously. As a general heuristic, cranks like to take old ideas and try to use them to argue against some modern result even when that idea isn’t used in the current form or isn’t as absolute as they make it out to be. This seems for example to occur frequently with creationists who use an oversimplified form of Mendelian genetics that seems to date prior to the work of Hardy, Weinberg, and Fisher. As a way of preventing this signal, referring to such things as Koch’s work can be done, but you need to be clear that you understand the limitations it has. Otherwise, it just sets off alarm bells.
As a general heuristic, cranks like to take old ideas and try to use them to argue against some modern result even when that idea isn’t used in the current form or isn’t as absolute as they make it out to be.
Note also that even outside a Bayesian context, Koch’s postulates are a rough framework. For example, sometimes there is great difficulty growing an organism in a culture (Koch’s second postulate requires this), and sometimes only a fraction of a population may be symptomatic.
It is also difficult to empirically test elements of String Theory or evolutionary history. This difficulty isn’t taken to be evidence for against those theories. Lack of easy clear cut empirical tests is just that—lack of a simple theorem discriminator. I am not claiming that it is some magic pass required for any theory to be credible.
On the other hand, nor can one use lack of said simple theorem discriminator to get a free pass and somehow claim to have extra certainty—that most come from evidence regardless.
However, in this case I do think that the general idea of Koch’s postulate is highly relevant, it is not ‘cranky’ to invoke it, and in fact it is probably the single strongest piece of evidence for the mainstream position—as Vlad M pointed out to me.
I am currently reading:
”
HIV causes AIDS: Koch’s postulates fulfilled”
Guest editorial
Stephen .I O’Brien* and .lames ,i Goederff
It reviews SIV models where SIV can be prepared and injected into monkeys and in some cases does cause AIDS like disease progression (usually more quickly than HIV).
So I don’t think your associations between “referring to Koch’s postulates” and cranks, and your general heuristic are valid at all, not when the mainstream is listing this as a centerpiece of evidence.
I haven’t read the editorial yet, but I’m not sure where the notion that this was a centerpiece of evidence comes from. I was vaguely aware of this result but never have paid it much attention. I suspect that if one talked to active AIDS and HIV researchers they’d say that there was a very strong case even without this. But yes, conservation of expected evidence does come into play here, so you have a valid point. In any event, the upshot that Koch’s postulates are not seen as any sort of absolute is very much relevant.
S.J. O’Brien and J.J. Goedert (1996) (the paper cited by Vladimir_M) specifically references Koch’s postulates. The accidental infection of laboratory workers is only one of five items cited in reference to (I believe) #3, “The cultured microorganism should cause disease when introduced into a healthy organism.”
I’m not sure you are allowed to demand that specific piece of evidence.
If you want absolute certainty, you need to fulfill koch’s postulates, which is the medical disease variant of experimental physics, and is well grounded.
So yes, in questions of science, we do demand specific evidence for near-certain proof, but naturally when that is not possible for whatever reason, we can still update based on other evidence and reach well grounded conclusions.
First of all, you never have absolute certainty. Proof is for math and alcohol. You may also want to read the Sequences relevant to Bayesian reasoning especially about how 0 and 1 are not probabilities.
Note also that even outside a Bayesian context, Koch’s postulates are a rough framework. For example, sometimes there is great difficulty growing an organism in a culture (Koch’s second postulate requires this), and sometimes only a fraction of a population may be symptomatic. Even Koch was willing to treat the first postulate as a guideline rather than a hard and fast rule; he worked with multiple examples of microorganisms that showed up in some healthy people but didn’t cause disease In fact we know now that this very common. Many causes of minor infections, such as staph, are present on everyone. It then takes some problem, such a wound, or disruption of the immune system to cause the person to become unhealthy. For viral examples, look at asymptomatic herpes or HPV.
Frankly, referring to Koch’s postulates like this are one of the things that makes many people with medical knowledge not take the HIV-AIDS skeptics very seriously. As a general heuristic, cranks like to take old ideas and try to use them to argue against some modern result even when that idea isn’t used in the current form or isn’t as absolute as they make it out to be. This seems for example to occur frequently with creationists who use an oversimplified form of Mendelian genetics that seems to date prior to the work of Hardy, Weinberg, and Fisher. As a way of preventing this signal, referring to such things as Koch’s work can be done, but you need to be clear that you understand the limitations it has. Otherwise, it just sets off alarm bells.
This bears repeating.
It is also difficult to empirically test elements of String Theory or evolutionary history. This difficulty isn’t taken to be evidence for against those theories. Lack of easy clear cut empirical tests is just that—lack of a simple theorem discriminator. I am not claiming that it is some magic pass required for any theory to be credible.
On the other hand, nor can one use lack of said simple theorem discriminator to get a free pass and somehow claim to have extra certainty—that most come from evidence regardless.
However, in this case I do think that the general idea of Koch’s postulate is highly relevant, it is not ‘cranky’ to invoke it, and in fact it is probably the single strongest piece of evidence for the mainstream position—as Vlad M pointed out to me.
I am currently reading: ” HIV causes AIDS: Koch’s postulates fulfilled” Guest editorial Stephen .I O’Brien* and .lames ,i Goederff
It reviews SIV models where SIV can be prepared and injected into monkeys and in some cases does cause AIDS like disease progression (usually more quickly than HIV).
So I don’t think your associations between “referring to Koch’s postulates” and cranks, and your general heuristic are valid at all, not when the mainstream is listing this as a centerpiece of evidence.
I haven’t read the editorial yet, but I’m not sure where the notion that this was a centerpiece of evidence comes from. I was vaguely aware of this result but never have paid it much attention. I suspect that if one talked to active AIDS and HIV researchers they’d say that there was a very strong case even without this. But yes, conservation of expected evidence does come into play here, so you have a valid point. In any event, the upshot that Koch’s postulates are not seen as any sort of absolute is very much relevant.
S.J. O’Brien and J.J. Goedert (1996) (the paper cited by Vladimir_M) specifically references Koch’s postulates. The accidental infection of laboratory workers is only one of five items cited in reference to (I believe) #3, “The cultured microorganism should cause disease when introduced into a healthy organism.”