[jerf]

Contrary to modestly popular belief, "faith" is not best defined as "choosing to act as if you believe something you actually know is not true". That's a fun definition to smear someone with, but is not useful in understanding very many real phenomena. Faith is much better defined as something like "acting on the truth of a statement that you can not (effectively) 100% prove"; note it does not preclude having a Bayesian probability of greater than 0, it is simply what you act on. When I sit in a chair, I can not 100% prove it will not collapse on me (after all, I have sat in chairs that collapsed on me and my Bayesian probability that a generic chair will support me can not be 100%), but I have faith that it will not; that is, I act as if it will not collapse.

When presented with a perpetual motion device, I can have faith that it will not work. Thanks to the Second Law of Thermodynamics, my Bayesian confidence in that belief is quite high, so it is not as if my faith is stepping out on a limb. But I have not proved that the perpetual motion machine won't work by actually examining it and finding the flaw, I simply have very-well-founded faith that such a flaw exists. Faith is a perfectly appropriate word here.

(I parenthesized "effectively" up there to avoid a massive and irrelevant discussion of exactly what 100% means, though it still pokes through. But is interesting to note that since reaching 100% confidence is very difficult, "faith" comes up in virtually every decision you make.)

[camccann]

I don't really disagree with any of that as such, but it doesn't really mesh with the common usage of the word "faith", especially on something as utterly fundamental as the Second Law of Thermodynamics. From a standpoint of Bayesian confidence, even a cursory understanding of physics would probably lead one to assign the Second Law a likelihood of one minus epsilon, where epsilon is taken to mean roughly "the likelihood that I'm crazy and hallucinating this whole thing".

As an aside, if you're going to be talking about Bayesian probability, you should know better than to call "100%" confidence just "very difficult"...

[jerf]

There's a couple of things I'm willing to assign 100% probability to, in particular something along the lines of "for some reasonably recognizable definition of 'exist', something I can reasonably call 'myself' exists", on the grounds that if I do not in fact exist there's no "my Bayesian probability" to be arguing about in the first place. (It's not quite tautological in the strictest sense, for reasons too long to get into here, but it certainly is close.) So I can't quite go to "impossible". You can't very far on 100%, though; "impossible" is a reasonable approximation.

Oh, and part of my point is that common usage is wrong, on the grounds that the "common usage" is incoherent, meaningless, and information-free. (I basically take it as axiomatic that the worth of the definition of a word can be measured along those axes; I'm generally a descriptive grammarian but that doesn't mean I have to throw all standards out.) Use something more like my definition and you get meaning again.