S
Syntax
Guest
I don’t deserve the credit. It’s Dretske’s. But let me be clear we are talking confirmation, not probability, since what scientists (and ourselves) actually do is construct a 2nd hypothesis about the asymmetry of the coin after so many unlikely repeated occurences of the same result have occurred, such as the occurence of nine straight heads after nine flips of the coin.Very good explanation!
So supposing the coin comes up heads nine out of ten flips, we take this empirically observed frequency and construct a new a hypothesis from which we deduce future predictions about the probability of the next flip of the coin turning out heads. So we say something like, “since the ratio of favorable cases to all possible cases has so far been on the order 1, since nine out of nine flips came out heads, then it is much *more likely *that the next flip will turn out heads rather than not.”
But this rests on the assumption that unexamined future cases will **always **resemble past examined cases in any relevent respect for which there is no empirical evidence whatsoever since we have only observed a finite amount of instantances of the alleged generality we are implicitly assuming when we make predictions about the future. So it is the assumption that Nature is governed by a series of general laws, **together with **observable frequencies, that the unexamined future cases are given a probability assignment on the order of approaching 1.
The point to notice, here, is that “confirmation” does not proceed by raising the probability that a given hypothesis is true from one experiment to the next because, as we have seen, this probability is not affected by each experiment or piece of evidence that is encountered. To think otherwise is to be committing the reverse of the Gambler’s Fallacy. No. Confirmation, is raising the probability that the next unexamined case will be like the set of all past examined cases. And this task confirmation alone cannot do without an implicit assumption about natural laws from which predictions about the likelihood of future events, or test scenarious, can be deduced.
But now we’ve arrived at the problem of Induction, namely, the question of whether or not inferences from past to future cases can said to be in any fair sense reliable or justified inferences at all.