Probability on experiential evidence

I received a question via email about Paul Moser's talk of "inferential relations" holding between experiential evidence and beliefs.  I really love his _Knowledge and Evidence_ which I hold to be the most neglected book in all of contemporary epistemology, so I was glad to consider the question.  Also, Williamson raises issues about this in _Knowledge and it's Limits_.  Here is an excerpt from my reply. 

 

a. Linguistic Confusion. It could be that Moser is simply confused, but I find it more likely that he's just using "inferential relations" loosely (and inaccurately in my view) to mean what I would mean by "epistemic support relations."

b. Epistemic Principles. Chisholm's epistemic principles of commonsense (e.g. "If it appears to S that there is an x that is F then it is more reasonable for S to believe there is an x that's F than to suspend judgment" ) are another candidate for the type of thing he could have in mind. They aren't really inferential principles, but the term isn't soooo bad because inferential principles are a form of licenses to move from one state to another. In this case they are admittedly non-propositional but as principles of license they share a genus.

b. Retranslation. I think almost all the tings Moser says can be retranslated in terms of some kind of epistemic support relation such as Chisholm's principles.

c. Proxy Propositions. I hold a phenomenal theory of evidence in that I prefer to consider our experiences. One way of holding this position is to define the inferential or probabilistic over proxy propositions which are the content of the experiences. Unless one is a pretty strong Platonist one should be willing to define inferential and probabilistic over any reasonable field of entities. More nominalistic theorists (Carnap was like this) define such relations over sets of sentences. That's too much for me, but I think a good middle route is to observe (for I think it is a fact) that there exists a proposition corresponding to every conceptualized experience (and so I'm a conceptualist about experiential content) of the form "S has E" and then we can just say that the *event* S's having E makes probable P in virtue of the fact that that event's proxy proposition that S has E more probable.

d. "Unconscious Inference". (i) Using proxy propositions would also allow another kind of interpretation of talk of "inferential relations" between experiential states and doxastic states. Tim McGrew's "old fashioned foundationalism" postulates "tacit beliefs" corresponding to such experiential states (and maybe given that these are doxastic states of a sort he doesn't really need to my proxy propositions). I've had long conversations with him about this--some of them on my blog, so they are observable--and he really believes that if we weren't at *some* level *aware* that we were having the experience in question, then it wouldn't be evidence for anything. This would be what I've called "low-grade" awareness since we certainly don't seem to be aware that we're aware of it (but then again maybe "belief" connotes more conscious awareness than it should, it gets confusing). So such a classical foundationalist as Tim could say that the experiences probabilify in virtue of the propositional content of these "tacit beliefs" states of awareness.

(ii). Just as "inference" can perhaps stretch to cover "license to believe" so "probable" can perhaps work with the more earthy sense of "evidence" you mention, the sense in which it "makes evident" some fact. This is in fact the usage that Earl vastly prefers, eschewing talk of probability. However, "probable" comes from the same root as "probative," "aprobation," "plaudits," and even "applause." So it is surely licit to use the term "epistemic probability" in such a way that when one has "evidence" of the form of a simple vivid phenomenal impression it makes a certain proposition more "probable." That seems well within the linguistic tradition and may well be more common than more formal probabilistic talk (which is more of a mixed bag than people tend to let on).

e. Tethering. Rich and Earl almost always talk about basic evidence being experiential (both in print and in person), yet I've heard Earl say on a couple of occasions he's willing to say something like that OK fine propositions are our basic evidence but they become so by being the contents of experiences which we don't process on any conscious level. So I think perhaps there's not as much riding on it as might first seem.

f. Probability Stuff. Just for the record, many probabilists (including F.P. Ramsey, one of the first modern Bayesians) and Richard Jeffrey and probably Leonard Savage, perhaps the two most influential probabilists of the 20th century, take probability only to govern the explicit doxastic results of sense experience which they either analyzed in terms of reliability--Ramsey--or know-how--Jeffrey. Thus for a certain class of Bayesian there is just no question of the probability of basic beliefs. You take what you get as far as basic beliefs go and the probability calculus just tells you if you're coherent. If you are not, it doesn't tell you what to give up.