On Kelly on Evidence and "Bayesianism" (with a sidenote on having evidence against skepticism)

Tom Kelly has a great new SEP entry on "evidence" here.

As a neo-Bayesian evidentialist, though, I want to distance myself from the following kind of statement:

"Within contemporary confirmation theory, a prominent version of Bayesianism is naturally understood as identifying one's evidence with those beliefs of which one is psychologically certain. "

Now I understand that Tom can't give an historical summary of the history of Bayesianism, but such remarks--all too common--do tend to reinforce the lack of appriciation for the fact that many if not most Bayesians have moved on from such ideas and embraced more moderate ideas of evidence. In fact, a not-uncommon view is that of Swinburne that one's evidence consists of the contents of one's basic beliefs. Others might say ones *current* beliefs or one's *justified* beliefs or what have you, but the idea is that Jeffrey's probability kinematics gives us a way to update our credences on less-than-certain input. (It might even be possible to extend Kvanvig and Riggs's idea of including appearance states. Then the views of evidence of Feldman and Conee or of Steup--where basic evidence is non-propositional--are available to Bayesians as well.)

Jeffrey conditionalization is not universally accepted by any means and there are alternative probability kinematics and work remains to be done. But the idea that evidence is to be limited to the certain is actually quite rare from what I can tell. Indeed many, like Kaplan, eschew complete certainty in any context. I'm inclined to say that there is *absolutely nothing* of which I am completely certain (no, not even that statement).

Here's another one I want to be careful about:

"Of course, although the presence of Koplik spots is in fact a reliable guide to the presence of measles, one who is ignorant of this fact is not in a position to conclude that a given patient has measles, even if he or she is aware that the patient has Koplik spots."

There are degrees of awareness and ignorance. I think it doesn't take much awareness of a generalization to have it be true that your belief is supported by the evidence in the right kind of way. In fact, I'm inclined to think that it is sufficient that you have a disposition to endorse the generalization under normal circumstances, but such ideas are *very* hard to regiment. This thought is related to a subject that occupies much of my thought at the moment.

We learn from Lewis Carroll's dialog between Achilles and the Tortoise (sometimes oddly called Carroll's paradox, though it's hard to state it as a paradox exactly) that it *can't* be the case that one must be justified in accepting a rule of inference in order to be justified in making the inference. Yet at the same time it seems true--as Tom Kelly points out--that in the measles case if I have no idea that Koplik spots are a positively correlated with measles then it would be downright irrational for me to infer that someone has measles from observing Koplik spots on them.

In short, I think there is also much room in between depending on the level of generality. For example, the modus ponens case is at a very high level of abstraction and, I think, requires no explicit awareness: it can just seem like the right thing to do at the time and the fact that that is right is sufficient to justify the inference. The Koplik spot case is very specific and requires pretty direct awareness of a correlation. The thesis of my current work is that general empirical reasoning is much more like the deduction case than the Koplik case: I am justified in inferring the existence of green things from having been appeared to greenly so long as it's *true* that being appeared to greenly is positively correlated with there being green things in my visual field.

I think it matters that I think direct realism is *true* but that one can't "prove" it in a way that gives us what Alston calls "full reflective justification" in his work on epistemic circularity.

Finally, I'm very, very *very* glad that Tom is affirming Hacking's idea of the core concept of evidence being the evolution from Ancient/Medieval "natural signs" to a sort of testimony of nature to the explicit idea of positive correlation or as the OED puts it "an indication, mark, sign, token, trace". The sense of "sign" is unfortunately ambiguous and the distinction between natural signs and conventional signs was very important for the development of the concept of evidence. Witness this quotation from Gower from around 1393 "Conf. I. 81 This horse..was to Troie an evidence Of love and pees for evermo.' This is well-contrasted with a quote from Cromwell in 1644 "Orig. Lett. I. 362 III. 300 It had all the evidences of an absolute Victorie." Here the idea of a positive correlation is pretty clear.

It's precisely this concept of evidence that makes phenomenal underdetermination arguments for skepticism so powerful, yet, if I'm right in my analogy with Carroll's case, then at the broadest level of empirical evidence what matters is only that appearances *are* highly correlated with realities and that we *are* directly aware of the world. We don't have to be able to *show* that this holds in order to *know* that this holds.

These last comments are in part the result of thinking about Duncan Pritchard's comments on this post.

Keywords: evidence, bayesianism, confirmation, skepticism, underdetermination