June 09, 2004

Evidentialism

Via Joe Shieber, Fantl and M. McGrath's paper "Evidence, Pragmatics, and Justification", arguing against evidentialism.

Fantl and McGrath argue that whether a belief is justified can depend on prudential factors*--the familiar Airport-type cases, in which little rides on the correctness of A's belief that p, much rides on the correctness of B's belief that p, so the same evidence may justify A's belief but not justify B's--where by "justified" F & McG. mean "good enough for knowledge, adding in what else is needed." This is meant to refute evidentialism, taken to be the idea that epistemological properties depend only on the evidence that is available.

My first thought here was: I can swallow their whole argument and be as much of an evidentialist as I want to be. Because I have no problem with the idea that being justified tout court depends on factors other than evidence. All I want is that the degree to which a belief is justified depends only on evidence.

Compare: Whether someone is tall tout court depends on contextually set factors other than their height. But how tall they are depends only on their height. This doesn't lead me to be extrinsicist or anything about tallness. It just makes me say that the basic facts about height concern how tall someone is, and the question "How tall is tall?" can rely on extrinsic context-dependent factors without affecting the basic facts concerning someone's height.

Now, the problem I see is that to say that a belief is justified may not be just to draw an arbitrary line at some (evidentially determined) degree of justification. It is to ask the question: Should I just plain believe this? This may seem like a more important question than the degree of justification that our evidence gives to the proposition.

Well, quite possibly it is a more important question. (I tell my students that an argument prefaced "may seem" or "some people think" is like the guy wearing the red shirt on Star Trek--you know it's going to die by the end of the episode. But not this time.) My belief is that perfect information-processing machines wouldn't have (many) categorical beliefs at all but would just keep track of degrees of support. Since we're not perfect i-p machines, we need categorical beliefs; and non-evidential factors will affect which categorical beliefs we should have. There's much more to be said about that, but not here.

*What they're calling pragmatics, but they don't mean implicatures and the like.

Posted by Matt Weiner at June 9, 2004 04:18 PM
Comments

Your thought is that, even if Fantl and I are right in our conclusion that justification simpliciter isn't fixed by evidential factors, evidentialism might still survive as a view about degree of justification. I'd welcome this conclusion if I could see my way around the following worry. If A is tall and B is taller than A, then B is tall. That much is analytic of 'tall' and 'taller than'. But if A is justified simpliciter in believing that p and B is more justified (in your evidentialist sense) than A in believing that p, then assuming evidentialism is false of justification simpliciter, B need not be justified simpliciter (because much more may be at stake for B than for A). What sense of 'more justified than' is this, then? A can be justified, B more justified than A, but B not justified?

Posted by: mcgrath at June 11, 2004 07:01 PM

D'oh! How about this as a new improved analogy?

(1) My cat is big, my car is not big, but my car is bigger than my cat

Still, "bigger than" is a purely intrinsic physical relationship between the two objects, even though "big" is not intrinsic (it seems).

Really working this out would require an understanding of the debate over the semantics of "big" that I don't have, but if you think "big" has a hidden parameter specifying a comparison class, you might also think that "justified" has a hidden parameter specifying a practical environment or some such, usually supplied by the person to whom the belief is attributed. Then your analytic principle would fail, as it fails for "big." I admit that it sounds funny to say "B is more justified in believing p than A is, but A is justified in believing p but A is not," but especially when you're dealing in technical terms like "justified" you sometimes have to stipulate funny results.

--I should say that I don't believe that last sentence at all. It sounds horrible to say that. I think the following solution is better:

I abandon the locution "more justified than" and go with "likelier on the evidence." "Likelier on the evidence" is purely evidentialistic, and to be justified is to have a certain pragmatically determined level of likelihood on the evidence. It's likelihood on the evidence that I'm really interested in.

Posted by: Matt Weiner at June 12, 2004 01:55 PM

We have to hold contextual parameters such as comparative classes fixed if we are to make sentences of the form "A is F, B is F-er than A, so B is F" come out analytic. This holds for 'tall' as well as 'big'. I suppose we should really talk about analyticity relative to an assignment of contextual parameters.

You suggest that maybe 'justified' has a hidden parameter for comparison classes. Two points about this.

First, this is a kind of contextualism, and so is not the view endorsed in "Evidence, Pragmatics, and Justification." So, you can't both accept what Fantl and I say about justification simpliciter and go evidentialist about 'more justified than' without rejecting what seems to be an analytic claim "If A is justified, and B is more justified than A, then B is justified."

Okay, you might say, why not depart from F&M and go for a hidden parameter sort of contextualism? You suggest we could stipulate such a parameter. This is too quick. In the paper, we used 'justified' to mean "has evidence good enough to know." But you can't just stipulate contextual parameters for ordinary locutions such as this. I won't rehash the arguments of Hawthorne and Stanley on the implausibility of a hidden parameter for 'knows'.

By the way, Fantl and I never deny the existence of a dimension of purely epistemic standing (such as a dimension of likelihood on the evidence). We just think that hooking up that dimension to the concepts of knowledge and justification doesn't work in the straightforward evidentialist way. Whether you know or are justified depends not only on your purely epistemic standing but on what's at stake for you.

I don't think we really disagree on anything important. Of course, a lot of traditional epistemologists will have none of it! And sometimes I worry that the costs are too high: can my knowledge that p come and go with changes in my stakes? Such results may be avoidable with some tweaking, but I'm not sure exactly how to doing the tweaking.

Posted by: mcgrath at June 13, 2004 10:06 AM

Matt (Matthew?),
An earlier response got eaten--I may have forgotten to hit "post." Anyway, I think that you're right that there are no big disagreements here. What matters to me is the purely epistemic dimension, and in fact you've convinced me that I shouldn't use the word "justified" for it (see this monster post).

I have a worry about your use of "justified," though--which may not be a worry for you. In the text here I identify

(2) Is this belief justified?

with

(3)Should I believe this?

which seems like an obviously important question. You identify it with

(4) Do I have good enough evidence for knowledge?

which is not obviously the same as (3). Nor, I think, is it obvious that (4) is important, or at least that answering (4) tells us anything about justification that we couldn't have figured out by asking, "Do I know this?" Of course this is all part of my decidedly non-traditional orientation, in that I don't think knowledge is the supreme epistemological concept, or even one that epistemologists should investigate that much! (Except to write long papers about why they shouldn't be investigating it.) Anyway, there's more in the post linked earlier in the comment.

Posted by: Matt Weiner at June 15, 2004 11:58 AM