August 26, 2006

Fantle and McGrath Against Epistemic Purism

Fantle and McGrath have an excellent new paper on pragmatic encroachment.

The latest version of the Fantle and McGrath argument against epistemic purism goes like this. Suppose that possible subjects S1 and S2 are alike with respect to strength of epistemic position. For instance, we suppose that they share precisely the same evidence regarding a true proposition that p. And we suppose that S1 knows that p. S2 is just like S1 in every respect accept that she differs with respect to stakes. The matter is much more important to S2, and so she is not rational to act as if p. But then by the pragmatic condition on knowledge, S2 fails to know p. And since, ex hypothesi, S1 does know p, it therefore follows that whether one is in a position to know does not supervene on strength of epistemic position. Epistemic purism is false.

The argument harbors some fundamental assumptions. First, subjects S1 and S2 are thought to be the same with respect to position of epistemic strength because they are said to be evidential twins. One implicit fundamental assumption then is this: only evidence effects position of epistemic strength. Second, S1 and S2 are thought to be evidential twins, because it is thought that practical interests do not effect how much evidence one has.

The notion of evidence, for F and M, is meant to be "a broad intuitive concept, that internalists and externalists might analyze in different ways." And in defense of their position, they remark that "it ought to be common ground between theories of evidence that having a lot at stake in whether p is true does not, by itself, provide evidence for or against p." Further, they explain that evidence for p, but not stakes in whether p, affect the probability of p (in some appropriate sense of 'probability').


Here is a possible reply that I discussed at the Epistemic Value Conference. Having a lot at stake does affect evidence. Consider, when stakes are high, evidence previously ignored becomes salient. Such "new" evidence may reduce the probability that p is true. For instance, S1 knows that the train is the Express train based on another traveler's testimony, but had it meant more to him to be right he might have recalled that a small number of travelers are clueless. Weighing in that a small number of travelers are clueless suddenly reduces the probability that the train is the Express. If practical interests can make salient previously ignored evidence, then Assumption 2 is false. Practical interests do affect the amount of evidence one has, and so, by Assumption 1, practical interests (at least indirectly) affect the strength of one's epistemic position.

8 comments:

Mike said...

Joe, couldn't you describe an actual and counterfactual case in which *all* of the evidence is perceptual? Let the actual case differ from the counterfactual one only with respect to what's at stake. Actual case: I see a person leap a fence into your backyard and conclude, "there is someone trepassing". Counterfactual case: I see the same thing and am asked to testify about it. In case two I might not claim to know that someone is trespassing, since (strictly) I can't rule out that it was you that I saw jump the fence.
There is no other evidence available, only my perception in both cases. Would you want to say that my perception in the counterfactual case is less evidential than my (exactly similar) perception in the actual case?

Joe Salerno said...

I don't see the sense in which the stakes are different. If we fill out the example then it becomes harder to see that evidence is not effected. Suppose the stakes are higher in the following way. In case 1 it is more important to be right than in case 2, because if you are wrong then your friend (rather than the evil trespasser) will be shot by your vigilante neighbor. Then the fact that it is dark and that it could very well be your friend becomes an important piece of counterevidence.

pascal engel said...

Dear Joe

Glad to continue the conversation started at Stirling.
You seem to present the F& M argument as if they defended a form of evidentialism, where two evidential twins are in the same epistemic position. As I understand their point, and unless I am mistaken, they argue that what is at stake affects evidence. I agree, but there is an equivocation here: a strong stake "affects" evidence in the sense that the more is at stake, the more one needs to collecti evidence, and saliency raises the demand for evidence. That's quite acceptable. But it's one thing to say this and to say that :

evidence is constained by pragmatic factors

Evidence is constrainted only by epistemic factors. That one suspends judgment or decides to have an opinion on P or not P is constrainted by the stakes. And it may broaden the range of evidence needed. But that does not alter the constitutive connexion between belief and evidence. In other words, and contrary to what F& M seem to say, the pragmatic encroachment has no normative impact on the relationship between belief and evidence, between believing and reasons to believe.

Perhaps they agree. Then I may have misinterpreted them. But they seem to adverstise stronger view, to the effect that evidence becomes a pragmatic matter.

Joe said...

Pascal, thank you for your comment. You mention that F and M seem to argue "to the effect that evidence becomes a pragmatic matter."

I agree that there are at least two senses here. I think that F and M take evidence to be a pragmatic matter only in the following sense. they take practical concerns to affect how much evidence is enough for knowledge. However, they deny that practical concerns affect just how much evidence one in fact has. And I think, among other things, you are agreeing with this point.

I believe that on some conceptions of evidence, however, practical concerns do affect how much evidence one has. Consider a reliabilist conception of evidence. It tells us that a method m is reliable just when m would yield a high proportion of truths in a wide range of typical situations. My vision may be highly reliable, but not for identifying an assailant in the dark. I may be fairly reliable discerning the temperature, but not so on the surface of the sun. Somebody's practical interests play a key role in determining m and its relevant environment, i.e., the range of situations across which m is being evaluated for truth-conduciveness. When the environment is considerably widened or narrowed, then the degree to which m is reliable is decreased or increased, respectively. And so, on a reliabilist conception of evidence, practical interests do affect the amount of evidence one has.

Jeremy Fantl said...

Hi guys,


Matt and I definitely do not think that evidence is a pragmatic matter. Joe's second paragraph in the preceding paragraph gets it right. (I think Ram Neta may think -- though now I may be misremembering him -- that there is no context-independent spectrum of epistemic evaluation -- evidence or otherwise -- and he uses this point to undermine pragmatic encroachment.)

Re Joe's point that Reliabilism may make evidence determined in part by practical matters: it does seem right that there are cases in which my practical interests plausibly determine how reliable m must for us to know the deliverance of m and cases in which my practical interests determine whether m is reliable and even cases in which my practical interests determine whether it was m that delivered my belief (maybe). If I'm driving and so interested in whether the light is red, my vision may very well count as reliable and sufficiently so that the delivered belief that the light is red counts as knowledge. If I'm a city employee charged with making sure that all the lights are crimson, rather than scarlet, my vision may not be reliable or sufficiently reliable.

But notice that the relevance of my interests is screened off once we specify the proposition in question. That is, in one case I'm interested in the proposition that the light is red and in the other I'm interested in the proposition that the light is crimson. I am reliable about the proposition that the light is red. I am not about the proposition that the light is scarlet. Stakes only determine what proposition I am interested in. We still don't end up with a situation in which two subjects with the same reliability vis a vis the same proposition are such that one knows and the other does not. I don't see why a reliabilist needs to endorse this possibility. There might be a great deal hinging for me (who loves his car) on whether the light is red, whereas, for you (who was intending to commit insurance fraud by trashing his car while still in it anyway), not so much. Reliabilists are more than welcome to say that in this situation your vision is reliable enough for knowledge that the light is red while mine is not (even though all of our eye tests would come out the same). To do this is to endorse pragmatic encroachment. But they need not endorse this view. And if they do endorse the view, it should be just as counterintuitive as if some more internalistically-minded philosopher endorsed it.

Joe said...

Hi Jeremy. Nice to see you at Knowability.

1. I guess I don't see why the "relevance of my interests is screeened off once we specify the proposition in question." Or at least, that is precisely what I'm arguing against. You continue, "Stakes only determine what proposition I am interested in". This confuses me, since it seems to goes contrary to the set-up of your argument for pragmatic encroachment. That is, you begin with possible subjects S1 and S2 believing the same proposition in different practical environments.

2. I'd like to make a point of clarification. You write,

"We still don't end up with a situation in which two subjects with the same reliability vis a vis the same proposition are such that one knows and the other does not. I don't see why a reliabilist needs to endorse this possibility."

I actually agree. The reliabilist it seems can't endorse this possibility, since she takes differences in reliability to be the relevant difference between knowing and failing to know. So my point is different. I claim that to evade the Fantl-McGrath argument for pragmatic encroachment, the reliabilist may accept the following: stakes affect whether we know via m, but stakes also affect how reliable m is. I offered a characterization of how this goes: stakes broaden or narrow the range of possible situations across which one is properly to evaluate the accuracy of m with respect to a given proposition p. Such broadening or narrowing affects how reliable m is relative to p. The line targets the set-up of the argument for pragmatic encroachment, since that argument presupposes exactly what I am rejecting---viz., that evidential twins (for our purposes, twins with respect to reliability) may live in different practical environments.

Jeremy Fantl said...

This is a pretty neat idea, Joe, but maybe I'm still not quite understanding it.

First, re. your 1: I, of course, think that stakes do more than determine the proposition one is interested in. In saying otherwise I was only trying to draw attention to the fact that, in the examples that were then on the table, the stakes only served to change the relevant target proposition. As far as the claim that my interests are screened off once we specify the proposition in question goes, I guess it still seems to me that, in the examples we have so far, the propositions at issue are always different once the stakes have changed. So, the points you are taking issue with in your 1 are really just pleas for an example in which my reliability regarding one and the same proposition changes as my interests change.

Once we make clear that it is the same proposition in question, the examples all look less like my reliability regarding that proposition has changed, and more like the degree of reliability it takes to know has changed. But those sorts of examples implicate the reliabilist in pragmatic encroachment.

I'm not saying there can't be examples of the right sort -- in which reliability regarding p decreases as stakes in whether p go up. But I don't see any obvious possibilities waiting in the wings. And even if there were some examples in which interests could affect reliability, to account for the McGrath/Fantl argument, I think there would have to be some argument that every case in which increasing stakes undermine knowledge, there is a corresponding decrease in reliability. But why would there be? Why would my ratio of truth to error always decrease as stakes rise? (For all I know, empirically, as stakes rise, I get more reliable about one and the same proposition.)

But perhaps when I see a clear example in which it is plausible that my reliability about p decreases as my interests in p go up I'll be able to see better how that example generalizes.

joe said...

Dear Jeremy,

Sorry about the misunderstanding in 1.

As for 2., you are right that to work against the pragmatic encroachment argument, my position must be perfectly general---that is, whenever pragmatic environment properly indicates a shift in whether we know, it must also indicates a shift in reliability. I haven't yet made the general case, but merely waved my hands at it with quick examples. Will try something more ambitious soon.