The case goes something like this: A broad conclusion is supported by a large number of inference chains, some inter-related, some independent. The variety and nature of the supporting propositions and/or premises taken with all of their various unstated or implied inferences make the evaluation of the argument as a whole one of measuring inductive strength rather than determining necessary validity or soundness. In cases where the reasoning requires a certain acceptance of a complexity of cause, those making the claim typically choose what they believe to be the handful of supporting arguments that are likely to give their conclusion the greatest amount of inductive strength… they hedge their bets, in other words, by choosing primary and secondary and etc. propositions in proportion to those propositions expected contribution as premises for their conclusion.

In cases where the question of causation is particularly complex, it’s common to discover that the handicappers and oddsmakers miss some important detail which, had it been known before hand, would have changed the relative merits and valuations used in selecting the supporting propositions believed to deliver the greatest inductive strength. Ooops. The temptation is to apply 20/20 hindsight or historical revisionism to ones argument in order to avoid admitting that it turns out that what seemed at the time to be an inductively strong argument was, in fact, NOT a strong argument at all; and that one’s conclusion turned out to be correct mostly by accident.

Say a conclusion, X, is basically supported by premises p1-p9. Now say at least a third of the premises have to be valid for ones conclusion to be passable. If one selects p1-p5 and neglects p6-p9 due to ignorance or etc, and suddenly discovers that three of the five premises ones chosen aren’t any good after the fact… but lucks out in that p4-p9 are ALL good. The temptation is to try to snake out of admitting the original argument with p1-p5 was unsupported… likewise there’s a reverse temptation to refuse to acknowledge that an opponents PARTICULAR argument didn’t support his conclusion, and refusing to conceed the conclusion on these grounds alone, knowing full well in hindsight that there WAS an argument(s) supporting the conclusion in question.

In the Iraqi war case: Bush (and Clinton) can’t deliver WMD’s, which they made their primary justification for the invasion. They do a 20/20 hindsight trick by claiming that “if we hadn’t invaded they WOULD have had WMDs”“, which is unknowable and logically cynical (a fallacious ‘reverse causation’).

If instead the Bushies pointed out that the UN inspectors took bribes from Hussein, and that the Oil for Food payoffs make it unlikely that they’ll be able to find WMDs, then they’d have a better case. As it should be obvious to just about anyone that bribing weapons inspectors is done for a lot of reasons, but that paying them off to NOT find something that’s NOT in Iraq isn’t one of ‘em. This would be an example of including additional extra supporting premises after the fact, rather than subtracting or substantively changing the original primary premises.

Am I being vague?

If I go back in time to shoot at my grandfather (and miss), then there’s a closed time-like curve going from me a few minutes ago, through me at the time of the jump, to the bullet near my grandfather, to his perception of the bullet, to his children, to their children, including me a few minutes ago. The fact that there’s one object (namely me) that has parts right now causally affecting parts in the past suggests that it is reasonable to think of there being a continuous time-like curve going through those parts in that order.

With the rain dance however, the causation is all at a distance of some sort, and no time-like curves are necessary.