I've often heard a phrase used in argumentation, but I'm not sure what it means. Does absence of evidence mean evidence of absence in philosophy?
If you look all over your house for your keys, and do not find them, this provides some evidence - not fully conclusive - that you may have left them somewhere else.
The above shows that yes, absence of evidence constitutes some evidence of absence.
First, what is evidence? Evidence for a hypothesis H is an observation O, such that the probability P(H|O) > P(H); in other words, the probability of the hypothesis increases after accounting for the observation.
This is not proof; H may still be false, despite O. H may even be very unlikely. P(H) may be 0.000001 and P(H|O) may be 0.0001, so that H is very unlikely whether we account for O or not. But O is still evidence even in this case, because it increases the posterior probability of H.
We can look at this using Bayes' theorem, which says P(H|O) = P(O|H) P(H)/P(O).
In this case, the observation O is that no keys were found in the house despite a search, and the hypothesis H is that your keys were not in the house.
Suppose your keys are not in the house. Then your search for the keys will not find them. That means P(O|H) = 1.0.
Now, we may say that you usually do leave your keys in the house. So we might say the prior probability that they are not in your house, before considering the fruitless search you performed, is P(H) = 0.05.
What about the probability P(O) that a search for your keys in your house turns up nothing, on any typical day? Well, this also would happen rarely. It could be explained if your keys were not in your house (which we already know has probability 0.05), or if they were in your house in some weird hard-to-find place (which, let's say, has prior probability 0.07). So P(O) = 0.05 + 0.07 = 0.12.
Then, applying Bayes' formula, P(H|O) = P(O|H) P(H) / P(O) = 1.0 * 0.05 / 0.12 = 0.4167. Now, 0.4167 is a slightly less than half, so even after you have done your search, it's more likely the keys are still in your house somewhere and you just haven't found them. But 0.4167 is also a lot higher than 0.05; performing your search has made it a lot more likely that the keys are not in your house. Thus, the fruitless search does provide some evidence that the keys are not there.
The argument "absence of evidence is not evidence of absence" is often used in regards to matters of existence - in particular to the existence of supernatural entities such as a Gods; those possessing the quality/power of invisibility/undetectability - as opposed to matters of presence.
In these cases, the argument essentially states something like "the fact we cannot detect gods (absence of evidence for gods) is not evidence they don't exist (or evidence of absence), because they possess the ability to evade/be impervious to detection".
@causative's answer is good, I only provide a different view at the subject. Suppose we have evidence "types". This corresponds neatly with our typical notion of evidence, for example fingerprints might be one type of evidence, witnesses another, etc.
Absence of evidence does not imply evidence of absence. For suppose there is a death, furthermore the cause of death (healthy victim with happy family and career, and serial killer with mode of operation very similar to the death in question) suggests murder. in this case, suppose we know the killer is very careful. Then, although there are no fingerprints, dna, etc - ie absence of evidence - this constitutes very weak evidence for the prior "there is no murderer". For there is far stronger evidence- namely the m.o. and the fact that the victim should not "otherwise have died".
In general, this is a principle of most proper inductive logics- if we expect that there is some reason that we should not see evidence that X, we should be careful in concluding not X. Again, the turkey has no reason to think it will die and the inductive strength of this conclusion grows stronger until thanksgiving. But if it had known that the farmer has every reason to prevent its death until thanksgiving, it might have thought otherwise. In other words, if we think there might be some reason for absence of evidence, we should be careful concluding that there is absence of evidence. None of this contradicts @causatives answer.
As a recovering ex-engineer, I am personally well-acquainted with that canard. I'd like a nickel for every time a manager refused to pay for a search for evidence of something, and then used the very absence of evidence to argue against the existence of that same thing.
To me, this demonstrates the serious risks of teaching philosophy to people with management potential, because that knowledge is so easily misused. You haven't lived until you've done battle with a radical skeptic manager, or had a manager throw the underdetermination of theory by evidence in your face to make you and the problem you are stuck with go away.
A very simple summary of the Bayesian position:
Absence of evidence is evidence of absence. The strength of said evidence is proportional to the probability that your test would have found something if it was present.
To borrow causative's example of searching for your keys: if you look around the house and don't find them, then that's some evidence that they're not in the house (maybe you left them in the car). If you make a more thorough search, that evidence becomes stronger, and you should be willing to entertain more unlikely alternative theories about where the keys are. But your search may not be perfect — there's always a possibility that you missed the keys that really are in the house.
To borrow Mauro ALLEGRANZA's example (in the question comments) about being in a dark room and not knowing if it's raining outside — this is a test that tells you nothing if it is raining, and nothing if it isn't raining, which is why, when it tells you nothing, that's no evidence of anything.
Of course, sometimes we don't know how reliable our tests are — but a Bayesian approach would be alright with that. As long as we admit some possibility that the test works, then a negative test is some (maybe not much) evidence of non-existence.
The old saw would be better as "absence of evidence is not proof of absence".
If I searched the garage and found no mouse, that would be evidence of absence of the mouse. But Absence of Evidence (of the mouse's existence) isn't limited to cases where I have searched. It includes cases where I haven't looked.
Absence of Evidence includes the case where I haven't looked for something.
The fact I haven't looked in the garage for a mouse -- that I lack any evidence if a mouse is in my garage -- is not evidence that there is no mouse in the garage. This only generates evidence of absence when I do something that would, with some chance, behave differently if the mouse was absent.
Practically this can be run into in real life. Quite often someone won't look for something where proof of its existence would cause a problem. They will have Absence of Evidence of the problem -- then, they'll treat it as Evidence of Absence; the fact they have no proof X exists is used as justification to ignore the possibility X exists.
Take bridge maintenance. If you have nobody inspecting bridges for metal fatigue, you will have an absence of evidence of bridges having metal fatigue. This is not evidence of the absence of metal fatigue; your absence of evidence contributes zero information value to "are bridges about to fail due to metal fatigue".
In a similar situation, you have inspectors looking at bridges for metal fatigue, and not finding it. This is evidence of absence (of metal fatigue on bridges). You could summarize both as "no metal fatigue was found on any bridges", but the evidence provided by the two situations is quite different.
Statistically, this relates to the power of a test, which is the probability that an effect is detected given that it actually exists. Tests with high power will very likely detect an effect if it exists, while tests with low power will likely not detect an effect even if it exists. Absence of evidence is evidence of absence only in situations where you have sufficient power.
Whether absence of evidence is informative depends on the power of the test. If you conduct a high-powered study with a very large sample size, you will likely find evidence of an effect if it is indeed there. In a situation like this, absence of evidence of an effect is indeed good evidence of the absence of an effect - had it been there, you likely would have found it. In the opposite case of a low-powered, small sample size study, you are unlikely to deem an effect significant even if it is truly there. In this case, absence of evidence is not evidence of absence, since you likely would not deem the effect significant whether it was there or not.
Basically, it comes down to "how hard you looked" for evidence. Not finding something is not very informative if you don't look very hard, but can be quite informative if you perform an exhaustive search and find nothing.
The claim you ask after is a caricture of an appeal to ignorance which is a sort of black and white thinking. As an informal fallacy, it has to meet the normal criteria for application. One such criterion is relevance, for instance.
Thus, the phrase itself changes meaning in the context in which is applied. If a prisoner is not in their cell at night, is that evidence they've escaped? Absolutely. The absence of evidence that they are in their cell is certainly evidence they may be missing from the prison system, because inductively prisoners are in their cells at night. Deductively, it's not a sure thing. Perhaps they're in the medical bay.
On the other hand, is the lack of scientific evidence of gods a disproof the gods' existence? No. It's merely a fact to shift the burden of proof on a believer to explain why the gods don't present themselves like bicycles and apple trees.
This represents a type of false dichotomy in that it excludes the possibility that there may have been an insufficient investigation to prove that the proposition is either true or false. It also does not allow for the possibility that the answer is unknowable, only knowable in the future, or neither completely true nor completely false. In debates, appealing to ignorance is sometimes an attempt to shift the burden of proof
After all, the gods may have reasons and use their supernatural powers to keep themselves hidden, and in the future, we must concede, that we will be presented scientific evidence and their existence will become irrefutable.
I've often heard a phrase used in argumentation, but I'm not sure what it means.
I spent some time researching if you've ever sneezed. I did not find anything to confirm this. Therefore, I have proven that you have never sneezed.
I suspect you disagree with that statement. And this is why absence of evidence (not finding evidence of you sneezing) is not evidence of absence (proof that you've never sneezed).
The key flaw in this reasoning is that when you assume absence of evidence to be evidence of absence, you assume that your search was 100% exhaustive, i.e. there was no other possible place still left to look for evidence; and you're 100% certain that you've not overlooked anything either.