Saturday, December 04, 2010

Absence of Evidence

In the comments thread to a post on another blog, a commenter gave the following quotation:

'The simplest explanation for the total absence of evidence for gods is a total absence of gods'.

I've increasingly over the past couple of years come across claims of this sort (the commenter seems to be quoting someone, but I don't know who), and I find it interesting because the underlying idea seems to be pretty clearly wrong, for all its superficial plausibility, even when we ignore the obvious hyperbole of 'total absence of evidence'. (Obviously there's evidence -- religious experiences, etc.; the question is just whether it's adequate to establish the conclusion. It is very rare to find cases of actual disputes where one side literally has no evidence at all, inconvenient as that fact may be. Even believers in house elves have strange occurrences to call to witness. What we usually mean is that the evidence is weak, which is very different from being nonexistent. The rhetorical advantages of conflating weak evidence with no evidence are, of course, obvious; but we should not treat a rhetorical figure as literal speech.) What we mean by "total absence of evidence" is at most a total absence of evidence available to be used in reasoning (otherwise the only way to establish total absence of evidence for X is to prove that X can't possibly exist -- if we aren't talking about evidence available to us, we'd have to take into consideration all evidence available to everyone at every time, including the future, and therefore we would need to establish a guarantee that no real evidence could possibly turn up in the future). And because of this, if all other things are equal, the simplest explanation for absence of evidence is that you've probably just overlooked it or not come across it. It is simplest in at least three different ways:

(1) It involves the weakest supposition about the world. If I commit to the claim that some evidence of X probably exists, I'm not by that committed to any claim for or against the existence of X, just to the existence of something that someone could reasonably classify as evidence for the existence of X, whatever that evidence might be. This is clearly a weaker supposition, with fewer commitments, than the supposition that there is no X.

(2) It is the simplest in that, unlike a categorical rejection or affirmation of something's existence, it allows for the subsumption of the case under an already well-established generalization, that is, the straight psychological fact that people often overlook or fail to come across evidence for things.

(3) It is the simplest in that it provides the least impediment to future inquiry: it closes down the fewest options for further research.

Part of the problem, I think, is that phrases like 'total absence of evidence for the existence of X', despite the literal meaning, actually convey in practical, colloquial speech the idea that there is, overwhelmingly, evidence against the existence of X, and it is indeed true that the simplest explanation for overwhelming evidence against the existence of X is X's nonexistence. And perhaps a failure to recognize that we do not, in actual practice usually judge absence of evidence absolutely but relative to what is accessible to us (the distinction I mentioned above) contributes to this confusion. But what is actually happening is that an entire range of suppositions is being elided. And there's a reason why people often say that absence of evidence is not evidence of absence: it's precisely that actual absence of the thing/event/whatever is not the simplest explanation for absence of evidence, considered on its own.

There are actual cases where an absence of evidence would be evidence of absence, of course; the cases, that is, where all other things are not equal. In practice, all of these are cases where we actually have pre-existing preponderant reason (either through preponderance of evidence or through actual proof) that X's existence is inconsistent with the lack of evidence in question, or else made definitely unlikely by it, and that it is either impossible or unlikely that the reason the evidence is lacking is your fault. But these cases, of course, don't salvage the general principle.