Tuesday, August 25, 2009

On the Null Hypothesis

This was a comment elsewhere. I thought it was good enough that it deserved to be a post, though, so I abridged it slightly and pasted.

In the quest for knowledge, many unsupported claims may be made. In the absence of evidence for or against such claims, they may be regarded as 'equally true'. Post-modernism seems to have degraded to the point that this 'equally true' is considered true, but this is logically untenable. As such, a claim with no evidence to back it up may be regarded as provisionally false, and a claim that can never be supported by any evidence is meaningless.

To explain by example: suppose some explorer happened upon a tribe of cave-dwellers. Because they are hypothetical, they have never had to venture outside their cave. If the explorer stayed with them and learned their language, he would be able to make some unsupported claims that we would regard as obvious, for example, that the sky is blue. Given that the best way to explain the 'sky' would be within the context of a larger cave outside the cave, the natives would be justified in doubting that this 'other cave''s walls and roof are blue instead of gray, and even that the 'other cave' truly exists. If they have never seen the sky, it is more reasonable to assume that the outsider is delusional.

The default response to an unsupported claim is doubt, and the default response to an unsupportable claim is derision. Now, to combat doubt, evidence for the claim must be rallied. It is reasonable for a claim that has just been made to have no solid evidence (though this has been getting rarer in some fields), so absence of evidence is not, itself, evidence against the claim, at that point. If repeated attempts have been made to obtain evidence, to no end, then the weight of the fruitless attempts may be considered evidence against the claim. This assumes that, given a prediction of how many attempts would fail before a single success, the number of failures is vastly greater.

In the context of debate, the null hypothesis regarding a given idea is that it is false. Asking for evidence from whoever is making the claim constitutes an attempt to find evidence.